Search

Livestock farming is ever increasing in production scale and automation in recent years. Accurate and efficient monitoring of individual livestock can greatly contribute to the growth and health status of animals. Behavioral and health information can often be involved in the individual livestock. Among them, the lameness behavior of a cow can indirectly represent the health status of its legs. Accurate leg tracking is very critical to detect the lameness of a cow. However, the existing intelligent detection of cow lameness is often used at the milking aisle of the farm. There are the occlusion, blurring, and deformation of the leg in this viewpoint. In this study, leg tracking and lameness detection were proposed for dairy cows in the occlusion environment. Firstly, the video of cows walking was recorded and then pre-processed in an actual farming setting. 50 video segments were chosen as the samples for the multi-target tracking dataset of cow legs. Each segment was featured as a walking cow with a duration of 8-20 s. The dataset included a total of 27 healthy and 23 lame cows. In the dataset samples, the training and testing sets were randomly divided in the ratio of 8:2. 40 videos for model training, and 10 videos were used for model testing. Secondly, the improved FairMOT target tracking model of the cow leg was designed to obtain the cow leg trajectory. Specifically, the coordinate attention (CA) module and Squeeze-Excitation attention (SE) module were introduced into the FairMOT Iterative Deep Aggregation (IDA) and Hierarchical Deep Aggregation (HDA), respectively. The important features of the cow's leg were focused on reducing the interference of redundant information under shape mutation and blurring along the spatial and channel directions. NSA Kalman filtering was introduced into the trajectory state for the updating phase, in order to obtain more accurate motion states of leg targets. Tracking performance was improved to reduce the measurement weights of low-confidence detection targets when the leg occlusion occurred; the AFLink linking mechanism was introduced into the tracking post-processing stage, in order to improve the tracking performance of the model. The curve-fitting of cow movement trajectory was also designed in the cow gait cycle under the occlusion environment. Finally, the individual thresholds were used to classify the phases of the cow's gait movement. Six gait features were extracted, namely, stride length, total gait period, support phase gait period, swing phase gait period, support phase share, and swing phase share. The Navie Bayes model was then utilized for the detection and classification of limping behavior. The experimental results show that the improved model shared better localization accuracy and tracking, compared with the original FairMOT model. Specifically, the MOTA (Multiple Object Tracking Accuracy) was improved by 3.79 percentage points, whereas, the MOTP (Multiple Object Tracking Precision) and IDs (Total Target ID Switching The total number of identity switches) decreased by 0.022 and 35 times, respectively. The improved model outperformed the existing mainstream multi-target tracking models. In addition, CowMOT shared the more accurate target localization more suitable for tracking the movement of a cow's leg; The classification accuracy of lameness detection was 86.67%, indicating better defection on the lameness in dairy cattle. The finding can also provide a strong reference to intelligent monitoring of cow lameness behavior. [ABSTRACT FROM AUTHOR]

Lameness is a common issue on dairy farms, with serious implications for economy and animal welfare. Affected animals may be overlooked until their condition becomes severe. Thus, improved lameness detection methods are needed. In this study, we describe kinematic changes in dairy cows with induced, mild to moderate hindlimb lameness in detail using a “whole-body approach”. Thereby, we aimed to identify explicable features to discriminate between lame and non-lame animals for use in future automated surveillance systems. For this purpose, we induced a mild to moderate and fully reversible hindlimb lameness in 16 dairy cows. We obtained 41 straight-line walk measurements (containing > 3 000 stride cycles) using 11 inertial measurement units attached to predefined locations on the cows’ upper body and limbs. One baseline and ≥ 1 induction measurement(s) were obtained from each cow. Thirty-one spatial and temporal parameters related to limb movement and inter-limb coordination, upper body vertical displacement symmetry and range of motion (ROMz), as well as pelvic pitch and roll, were calculated on a stride-by-stride basis. For upper body locations, vertical within-stride movement asymmetry was investigated both by calculating within-stride differences between local extrema, and by a signal decomposition approach. For each parameter, the baseline condition was compared with induction condition in linear mixed−effect models, while accounting for stride duration. Significant difference between baseline and induction condition was seen for 23 out of 31 kinematic parameters. Lameness induction was associated with decreased maximum protraction (−5.8%) and retraction (−3.7%) angles of the distal portion of the induced/non-induced limb respectively. Diagonal and lateral dissociation of foot placement (ratio of stride duration) involving the non-induced limb decreased by 8.8 and 4.4%, while diagonal dissociation involving the induced limb increased by 7.7%. Increased within-stride vertical displacement asymmetry of the poll, neck, withers, thoracolumbar junction (back) and tubera sacrale (TS) were seen. This was most notable for the back and poll, where a 40 and 24% increase of the first harmonic amplitude (asymmetric component) and 27 and 14% decrease of the second harmonic amplitude (symmetric component) of vertical displacement were seen. ROMz increased in all these landmarks except for TS. Changes in pelvic roll main components, but not in the range of motion of either pitch or roll angle per stride, were seen. Thus, we identified several kinematic features which may be used in future surveillance systems. Further studies are needed to determine their usefulness in realistic conditions, and to implement methods on farms.

Objective gait analysis provides valuable information about the locomotion characteristics of sound and lame horses. Due to their high accuracy and sensitivity, inertial measurement units (IMUs) have gained popularity over objective measurement techniques such as force plates and optical motion capture (OMC) systems. IMUs are wearable sensors that measure acceleration forces and angular velocities, providing the possibility of a non-invasive and continuous monitoring of horse gait during walk, trot, or canter during field conditions. The present narrative review aimed to describe the inertial sensor technologies and summarize their role in equine gait analysis. The literature was searched using general terms related to inertial sensors and their applicability, gait analysis methods, and lameness evaluation. The efficacy and performance of IMU-based methods for the assessment of normal gait, detection of lameness, analysis of horse–rider interaction, as well as the influence of sedative drugs, are discussed and compared with force plate and OMC techniques. The collected evidence indicated that IMU-based sensor systems can monitor and quantify horse locomotion with high accuracy and precision, having comparable or superior performance to objective measurement techniques. IMUs are reliable tools for the evaluation of horse–rider interactions. The observed efficacy and performance of IMU systems in equine gait analysis warrant further research in this population, with special focus on the potential implementation of novel techniques described and validated in humans. [ABSTRACT FROM AUTHOR]

Cow lameness is a complex locomotion behaviour, and effective characteristics are the key to identifying early lameness. Machine learning methods were used to quantify the movement of the hooves and head of cows, and they showed the possibility of using the head- and back-hoof linkage features extracted from motion analysis to classify early lameness. Firstly, the key points of the body in the side-view walking videos were tracked by DeepLabCut (DLC). Based on the curves of hoof movement and head swing, the head-hoof linkage pattern was analysed. Second, the curve of cow's back was extracted and the back-hoof linkage feature was built when the hoof touched the ground. At the same time, the movement curves were analysed to extract five other gait characteristics, including asymmetry of stride length, tracking up, asymmetry in the proportion of supporting phase, landing speed, and moving speed. Multiple models were trained on a dataset containing 212 videos, and 10-fold cross-validation was used to verify the performance of the algorithm. The classification performance was improved by optimising the model and algorithm. Overall accuracy and recall were 89.2% and 90.7%, respectively. The scores of the importance of features calculated by the Chi-square test showed that the head- and back-hoof linkage features play an important role in the classification of early lame cows. The hooves and their linkage features with the head and back of cows proposed in this paper can thus effectively detect early lameness in cows. • DeepLabCut was used to track four hooves and head of walking cow. • Head- and back-hoof linkage features was analysed. • Seven lameness indicators were used to build the model. • The recall of classification for early lame cows was 90.7%. [ABSTRACT FROM AUTHOR]

This study proposed a method for detecting lameness in dairy cows based on machine vision, addressing the challenges associated with manual detection. Data from a dairy farm in Taigu, Shanxi, China were collected and divided into two parts. The first part was utilized to precisely position the cow's back by employing a dedicated deep learning model named GhostNet_YOLOv4, which can be implemented on mobile or embedded devices. The second part was used with the Visual Background Extractor (Vibe) algorithm, incorporating additional morphological processing techniques. Enhancing the Vibe algorithm, a widely used background subtraction algorithm for image sequences, achieved more accurate recognition of the specific pixel areas of cows. Subsequently, cow shape-related feature parameters were extracted from the back area using the combined approach. These parameters were used to calculate the average curvature, which describes the degree of curvature of the cow's back contour during walking. The differences in curvature values were employed for classification to detect lameness. Through extensive experimentation, distinct average curvature ranges of [-0.025, -0.125], [-0.025, +∞], and [-∞, -0.125] were established for normal cows, early lameness, and moderate-severe lameness, respectively. The algorithm's effectiveness was validated by processing 600 image sequences of dairy cows, resulting in a lameness detection accuracy of 91.67%. These findings can serve as a reference for the timely and accurate recognition of lameness in dairy cows. [ABSTRACT FROM AUTHOR]

A 2D stick model was developed to simulate sow walking on concrete floors. The model consisted of thirteen connective body joints to form a simple stick figure that approximates a sow's body skeleton. The video images of sows with reflective markers on their bodies walking along a corridor of concrete floor were analysed and translated into a 2D coordinate dataset. The motion of each landmark, including displacement, velocity and acceleration were calculated. A step of walking was modelled as three consecutive phases of motion: acceleration of foot leaving the ground; constant speed of foot swinging; and deceleration of foot touching the ground. One gait cycle was separated into several time steps and the movements of each joint at a time step in the horizontal (x) and vertical (y) directions were described by a combination of first and 2nd order polynomial functions. The coordinates of all joints were calculated in the x- and y-directions and these joints were then linked by synchronizing the motions of each joint at a given time step. The joint coordinates at each time step were then used to calculate a selected set of characteristic gait parameters, including the stride length and stance time for fore- and hindlimbs, diagonality, back angle and walking speed. Based on the comparisons of the predicted gait parameters between the lame and non-lame sows, the criteria were established for using the predicted gait parameters to detect sow lameness. The results showed an accuracy greater than 92% for lameness detection. • A 2D stick model was developed to simulate sow walking on concrete floors. • The model consisted of 13 joints and 12 sticks to represent a sow skeleton. • Joint movements were modelled by a 2nd order polynomial or a linear function. • The model predicted a set of parameters to characterize sow gait. • The model was used to detect lame sows with an accuracy over 92%. [ABSTRACT FROM AUTHOR]

• The DeepLabCut model is used to detect 17 skeleton key points in cows. • Key frames are extracted based on the moment the hoof touches the ground. • 16 motion features are extracted to detect lameness in dairy cows. • This method improves the robustness of lameness detection. Lameness is one of the common problems on cattle farms and it seriously affects the health and welfare of dairy cows. The existing lameness detection methods fail to accurately locate the key frame of the hoof landing, which makes it difficult to extract the motion features of the dairy cow in a specific posture, such as the front hoof release angle. In this work, the key point detection model is used to locate the key frame, and the multiple motion features in a specific posture are extracted to detect lameness in dairy cows. First, the backbone network of the DeepLabCut model is replaced with a lightweight convolutional neural network EfficientNet-B0 to detect the key points of dairy cows. Second, the backtracking analysis method is used to optimize the key point prediction results, and the video frames of the hoof landing are selected as the key frames according to the coordinate changes of the hoof. Third, according to the key point information contained in the key frames and hoof movement period, the motion features such as front hoof release angle, back arch height, neck tilt angle, support phase and overlap are extracted. Finally, the above features are normalized and input into an SVM model to realize lameness classification. The lameness detection effect of the model was verified by the ten-fold cross-validation method with the videos of 330 different dairy cows. The average accuracy, sensitivity and specificity of lameness detection were 93.64%, 92.74% and 96.89%. In addition, the proposed method was verified in different cattle farm environments, and the experimental results showed that the proposed method has strong robustness. [ABSTRACT FROM AUTHOR]

Simple Summary: Lameness has high economic and welfare cost to the U.K. dairy industry; accurate and early detection of lameness minimises this cost. Thermal imaging devices can be used as a method of detecting lameness; however, these devices are typically high-cost and fragile, limiting their usefulness in a farm setting. This study looked at the effectiveness of low-cost thermal imaging devices when used as lameness detection aids, by comparing one to a research-specification thermal imaging device. Thermal images were taken of cattle feet, and each cow was assessed for lameness. Both devices tested were able to determine whether the cattle were lame; however, the research-specification device performed marginally better at this function. This minimal difference in effectiveness between these devices suggests that low-cost thermal imaging devices could be used as a lameness detection aid; increased use of these devices by farmers may increase lameness detection rates and benefit animal welfare. Lameness has a high economic cost to the U.K. dairy industry; accurate and early detection of lameness minimises this cost. Infrared thermal imaging (IRT) devices have shown promising results for use as a lameness detection aid in cattle when used in research settings; these devices are typically high-cost, limiting their adoption. This study analysed the effectiveness of low-cost IRT devices (LCDs) as lameness detection aids, by comparing both maximum environmentally adjusted temperature values and hindfeet temperature difference collected by an LCD to the mobility score of the cow; this test was repeated for data collected by a research-specification device. Data collection occurred during routine milking of 83 cattle; each cow's mobility was scored afterwards. Significant differences were found between lame and sound cows with the LCD, upon analysis of both methods. There was no significant difference between the data captured by differing devices. The maximum sensitivity and specificity values for the LCD were calculated as 66.95 and 64.53, respectively, compared with 70.34 and 70.94, respectively, for the research-specification device; optimum threshold values for these were equivalent for both devices, suggesting IRT lameness identification is not device-dependent. It was concluded that a minimal difference in effectiveness between tested devices suggests that LCDs could be used as a lameness detection aid; consequently, there is potential for widespread adoption as on-farm detection aids. [ABSTRACT FROM AUTHOR]

Lameness, a wellknown issue in sport horses, impedes performance and impairs welfare. Early detection of lameness is essential for horses to receive needed treatment, but detection of hindlimb lameness is challenging. Riding instructors and trainers observe horses in motion in their daily work and could contribute to more efficient lameness detection. In this cross-sectional and prospective study, we evaluated the ability of riding instructors and trainers to assess hindlimb lameness. We also evaluated different feedback methods for improved lameness detection. For the cross-sectional part, n = 64 riding instructors and trainers of varying level and n = 23 high-level trainers were shown 13 videos of trotting horses, lameness degree: 0–3.5 (test 1) and tasked with classifying the horses as sound, left hindlimb lame, or right hindlimb lame. For the prospective part, the riding instructors and trainers of varying levels were randomly allocated to three different groups (a, b, c) and given 14 days of feedback-based, computer-aided training in identifying hindlimb lameness, where they assessed 13 videos (of which three were repeated from test 1) of horses trotting in a straight line. Participants in groups a-c received different feedback after each video (group a: correct answer and re-viewing of video at full and 65% speed; group b: correct answer, re-viewing of video at full and 65% speed, narrator providing explanations; group c: correct answer and re-viewing of video at full speed). After computer-aided training, the participants were again subjected to the video test (test 2). Participants also provided background information regarding level of training etcetera. Effects of participants' background on results were analyzed using analysis of variance, and effects of the different feedback methods were analyzed using generalized estimation equations. On test 1, 44% (group a), 48% (b), 46% (c), and 47% (high-level trainers) of horses were correctly classified. Group a participants significantly improved their test score, both with (p < 0.0001) and without (p = 0.0086) inclusion of repeated videos. For group c, significant improvement was only seen with inclusion of repeated videos (p = 0.041). For group b, no significant improvement was seen (p = 0.51). Although test 2 scores were low, computer-aided training may be useful for improving hindlimb lameness detection.

• The spatiotemporal features are important for lameness detection. • The motion trajectory of cows is stacked into images. • The temporal gait and spatial post features of cows are extracted. • This method improves the robustness of lameness detection. Lameness is one of the common diseases in dairy farms, which seriously affects the health and welfare of dairy cows. The existing lameness detection methods based on computer vision have poor ability to extract spatiotemporal features, which reduces the lameness detection accuracy in dairy cows. In this paper, a spatiotemporal energy network (SEN) is proposed to detect lameness in dairy cows. It compresses the cow walking video into history energy image (HEI) and gait energy image (GEI), and then accurately extracts the temporal gait and spatial pose features of dairy cows. First, the YOLOv8n algorithm is used to remove background information from the video and segment the cow's body. Second, the body contour sequence of the dairy cow is stacked based on the time series, and then different body parts of dairy cows are stacked on the channel to form the HEI. The HEI preserves the movement trajectory of the dairy cows. The first lameness classification model is obtained by training the MobileNetv2 network with HEI dataset. Third, the body contour sequence of the dairy cow is cropped according to the center of mass and body proportion and then stacked into an image, which is named GEI. The second lameness classification model is obtained by training the MobileNetv2 network with GEI dataset. Finally, the prediction results of the two models are fused according to the weight ratio to detect lameness in dairy cows. To verify the lameness detection effect of this method, 264 samples were randomly selected from 330 samples as the training set and the remaining 66 samples as the test set. The proposed method obtained competitive lameness detection results with classification accuracy of 96.22 %, sensitivity of 96.35 % and specificity of 98.14 %. By combining HEI and GEI, the spatiotemporal features of the cow walking are extracted more comprehensively. This method provides guidance for the lameness detection of dairy cows and reference for the gait analysis of animals. [ABSTRACT FROM AUTHOR]

• Lameness is common in dairy cows and causes decreased welfare and production loss. • Visual lameness detection is complicated; reliable automatic methods are needed. • Equine lameness is extensively studied; we used that knowledge to study lame cows. • We identified several kinematic features which may be used to detect lameness in cows. • This knowledge can be used to develop automatic lameness detection systems. Lameness is a common issue on dairy farms, with serious implications for economy and animal welfare. Affected animals may be overlooked until their condition becomes severe. Thus, improved lameness detection methods are needed. In this study, we describe kinematic changes in dairy cows with induced, mild to moderate hindlimb lameness in detail using a "whole-body approach". Thereby, we aimed to identify explicable features to discriminate between lame and non-lame animals for use in future automated surveillance systems. For this purpose, we induced a mild to moderate and fully reversible hindlimb lameness in 16 dairy cows. We obtained 41 straight-line walk measurements (containing > 3 000 stride cycles) using 11 inertial measurement units attached to predefined locations on the cows' upper body and limbs. One baseline and ≥ 1 induction measurement(s) were obtained from each cow. Thirty-one spatial and temporal parameters related to limb movement and inter-limb coordination, upper body vertical displacement symmetry and range of motion (ROMz), as well as pelvic pitch and roll, were calculated on a stride-by-stride basis. For upper body locations, vertical within-stride movement asymmetry was investigated both by calculating within-stride differences between local extrema, and by a signal decomposition approach. For each parameter, the baseline condition was compared with induction condition in linear mixed−effect models, while accounting for stride duration. Significant difference between baseline and induction condition was seen for 23 out of 31 kinematic parameters. Lameness induction was associated with decreased maximum protraction (−5.8%) and retraction (−3.7%) angles of the distal portion of the induced/non-induced limb respectively. Diagonal and lateral dissociation of foot placement (ratio of stride duration) involving the non-induced limb decreased by 8.8 and 4.4%, while diagonal dissociation involving the induced limb increased by 7.7%. Increased within-stride vertical displacement asymmetry of the poll, neck, withers, thoracolumbar junction (back) and tubera sacrale (TS) were seen. This was most notable for the back and poll, where a 40 and 24% increase of the first harmonic amplitude (asymmetric component) and 27 and 14% decrease of the second harmonic amplitude (symmetric component) of vertical displacement were seen. ROMz increased in all these landmarks except for TS. Changes in pelvic roll main components, but not in the range of motion of either pitch or roll angle per stride, were seen. Thus, we identified several kinematic features which may be used in future surveillance systems. Further studies are needed to determine their usefulness in realistic conditions, and to implement methods on farms. [ABSTRACT FROM AUTHOR]

Accurate objective, automated limb lameness detection and pose estimation play an important role for animal well-being and precision livestock farming. We present a wearable sensor-based limb lameness detection and pose estimation for horse walk and trot locomotion. The gait event and lameness detection are first built on a recurrent neural network (RNN) with long short-term memory (LSTM) cells. Its outcomes are used in the limb pose estimation. A learned low-dimensional motion manifold is parameterized by a phase variable with a Gaussian process dynamic model. We compare the RNN-LSTM-based lameness detection method with a feature-based multi-layer classifier (MLC) and a multi-class classifier (MCC) that are built on support vector machine/K-nearest-neighbors and deep convolutional neural network methods, respectively. Experimental results show that using only accelerometer measurements, the RNN-LSTM-based approach achieves 95% lameness detection accuracy and also outperforms the feature-based MLC or MCC in terms of several assessment criteria. The pose estimation scheme can predict the 24 limb joint angles in the sagittal plane with average errors less than 5 and 10 degs under normal and induced lameness conditions, respectively. The presented work demonstrate the successful use of machine learning techniques for high performance lameness detection and pose estimation in equine science. Note to Practitioners—Automation technologies are increasingly used for precision agriculture but few have focused on monitoring individual animals in open field for precision livestock farming. Limb lameness detection and pose estimation in open field is labor-intensive, unsafe for farmers, and inefficient. The presented machine learning-enabled, wearable inertial sensor-based design provides an effective and efficient approach for horse limb lameness detection and pose estimation applications. We present an RNN-LSTM for lameness detection and an integrated manifold learning model is used to predict the horse limb joint angles in walk and trot gaits under normal and induced lameness conditions. We also present a systematic analysis and experiments to demonstrate the impacts of the wearable sensor locations and signal information on lameness detection and pose estimation performance. Several other machine learning-based lameness detection methods are also presented and compared. The extensive multi-horse testing results are presented to demonstrate the superior accuracy and higher performance than other types of machine learning methods. One attractive feature of the proposed design lies in its high performance and fast computational capability for potential real-time applications in open field. [ABSTRACT FROM AUTHOR]

Lameness is common in dairy cows. Methods that used RGB-based images to detect lameness in dairy cows always have low accuracy and poor robustness because of the complex background environment involved. In this study, a lameness detection method for dairy cows based on multiple features, including RGB, optical flow and skeletons, was proposed. The network was divided into three branches according to different inputs: for Branch1 and Branch3, a convolutional neural network (CNN) was used to predict lameness according to the input images and optical flow, and for Branch2, a spatial temporal graph convolutional network (ST-GCN) was used to predict lameness according to the cows' skeletons. Finally, the weight was adjusted, and the prediction scores of these three branches were fused to complete the lameness detection process. In this research, 680 different videos were used for training and testing. The segmentation ratio of the train set and test set was 6:1. The best ACC was 97.20% when the weight of the RGB, optical flow and skeletons was 1:0.5:0.5. To verify the robustness of the method, the gamma transform was used to adjust the brightness of the image to simulate a change in illumination. Under different illumination settings, the maximum error of the method was 2.65%, which was significantly lower than other methods without skeletons. The results showed that the proposed method was effective for the detection of early-stage lameness, severe lameness, and non-lameness of dairy cows. • Fusion of RGB, optical flow and skeleton features for the detection of lameness in dairy cows. • Cow skeletons are used to increase network attention to cows. • A spatial temporal graph convolutional network (ST-GCN) was used to predict lameness according to the cows' skeletons. • This method is more stable in the antagonism of environmental robustness. [ABSTRACT FROM AUTHOR]

Lameness in dairy cattle could cause significant economic losses to the dairy industry. Detection of lameness in a timely manner is critical to the high-quality development of dairy industry. The traditional method is visual locomotion scoring by dairy farmers, which is low efficiency, high cost and subjective. The demand for automated lameness detection is increasing. The review was conducted to find out the current state and challenges of automatic lameness detection technology development and to learn from the latest findings. The current automatic lameness detection systems were reviewed in this paper mainly rely on five technologies or combinations thereof, including machine vision, pressure distribution measuring system, wearable sensor system, behavior analysis and classification; the principle, function and features of these technologies were analyzed. Machine vision technique is to extract feature variables (e.g. back arch, head bob, abduction, stride length, walking speed, temperature, etc.) from video recordings of cattle movement by image processing. Pressure distribution measuring system contains an array of load cells to sense gait variables, when dairy cattle are walking by. By using accelerometer with high frequency data collection, the gait cycle parameters can be extracted and used for lameness detection. By using wearable devices, the number of lying/standing bouts and their duration, the total time spent lying, standing and ruminating per day can be recorded for individual cattle. The lameness can also be detected by behavior analysis. Currently, most of these studies were in the stage of sensor development or validation of algorithm. A few studies were in the stage of validation of performance and decision support with early warning system. The challenges to apply automatic lameness detection system in dairy farm includes the difficulties of acquiring high quality data of lameness features, lack of techniques to detect early lameness, identification errors caused by individual gait differences among dairy cattle, difficulties to function well in unstructured environment and difficulties to evaluate the benefits. To accelerate the development of automatic lameness detection systems, recommendations are proposed as follows: ①promoting lameness data sharing and data exchange among dairy farms; ②developing individual-based lameness classification model; ③developing multifunctional smart station which can detect lameness, measure body condition score, weighing, etc; ④evaluating the significance of automatic lameness detection to the dairy industry from the perspective of animal welfare, environment and food safety.

Bovine lameness has relatively large negative economic and welfare implications on the U.S. dairy industry. Due to the ramifications, early lameness detection will aid in assisting dairy producers to mitigate downstream effects through early treatment. The objective of this study was to determine the minimum standing time required among 2-, 3-, 4-, 5-, and 10 min time intervals to obtain an accurate weight distribution estimate for each leg when attempting to detect lameness. An embedded microcomputer-based force plate system was developed to measure vertical forces from individual cow limb weight distribution to detect bovine lameness when utilizing an induced synovitis lameness model. The force plate has four quadrants, with each load cell quadrant measuring the force placed on it from a single limb. The force plate recorded weight (kg) every second from each load cell quadrant, after which, a 60 s moving average for weight distribution was calculated. A sequential study design was employed to evaluate non-lame and induced lameness to ensure time requirements were consistent. Prior to induction, the force plate system was used to measure weight distribution every second for 15 min. After lameness induction, additional 15 min increments were recorded every 24 h for seven days. Lameness was induced by injecting the left hind distal interphalangeal joint in three cows with amphotericin B, 12 h prior to the start of the study. Data were analyzed using a linear mixed effect that included the fixed effects of day relative to lameness induction, time period, foot and injected foot. Cow within replicate was included as a random effect. Cumulative minutes were assessed up to 15 min by comparing the least square rolling 60 s cumulative means expressed as a percentage of each animal’s BW percentage placed on each leg for 2-, 3-, 4-, 5-, and 10 min intervals. Results indicate that the minimum time needed for accurate lameness detection in cows was 2 min.

Lameness has a high economic cost to the U.K. dairy industry; accurate and early detection of lameness minimises this cost. Infrared thermal imaging (IRT) devices have shown promising results for use as a lameness detection aid in cattle when used in research settings; these devices are typically high-cost, limiting their adoption. This study analysed the effectiveness of low-cost IRT devices (LCDs) as lameness detection aids, by comparing both maximum environmentally adjusted temperature values and hindfeet temperature difference collected by an LCD to the mobility score of the cow; this test was repeated for data collected by a research-specification device. Data collection occurred during routine milking of 83 cattle; each cow’s mobility was scored afterwards. Significant differences were found between lame and sound cows with the LCD, upon analysis of both methods. There was no significant difference between the data captured by differing devices. The maximum sensitivity and specificity values for the LCD were calculated as 66.95 and 64.53, respectively, compared with 70.34 and 70.94, respectively, for the research-specification device; optimum threshold values for these were equivalent for both devices, suggesting IRT lameness identification is not device-dependent. It was concluded that a minimal difference in effectiveness between tested devices suggests that LCDs could be used as a lameness detection aid; consequently, there is potential for widespread adoption as on-farm detection aids.

Claw lesions are a serious problem on dairy farms, affecting both the health and welfare of the cow. Automated detection of lameness with a practical, on-farm application would support the early detection and treatment of lame cows, potentially reducing the number and severity of claw lesions. Therefore, in this study, a method was proposed for the detection of claw lesions based on the acoustic analysis of a cow's gait. A panel was constructed to measure the impact sound of animals walking over it. The recorded impact sound was edited, and 640 sound files from 64 cows were analyzed. The classification of animal-lameness status was performed using a machine-learning process with a random forest algorithm. The gold standard was a 2-point scale of hoof-trimming results (healthy vs. affected), and 38 properties of the recorded sound files were used as influencing factors. A prediction model for classifying the cow lameness was built using a random forest algorithm. This was validated by comparing the reference output from hoof-trimming with the model output concerning the impact sound. Altering the likelihood settings and changing the cutoff value to predict lame animals improved the prediction model. At a cutoff at 0.4, a decreased false-negative rate was generated, and the false-positive rate only increased slightly. This model obtained a sensitivity of 0.81 and a specificity of 0.97. With this procedure, Cohen's Kappa value of 0.80 showed good agreement between model classification and diagnoses from hoof-trimming. In summary, the prediction model enabled the detection of cows with claw lesions. This study shows that lameness can be detected by machine learning from the impact sound of hoofs in dairy cows. [ABSTRACT FROM AUTHOR]

Lameness is a common problem in breeding sows, which often goes undetected for long periods with severe consequences for animal welfare and farm productivity. Automatic lameness detection could help pig farmers to recognise and treat lameness sooner. The SowSIS (Sow Stance Information System) is a device consisting of four force plates and providing non-invasive force measurements per leg of the sow. In this study, the SowSIS was built into electronic sow feeders and validated for lameness detection in group-housed gestating sows. Data was automatically collected for 71 sows. Visual gait scoring was performed twice a week using a 150-mm tagged visual analogue scale to determine the sows' lameness status. Only data from 32 gait scoring days were included, adding up to 674 sow days. A sow was classified as lame using >60 mm as the cut-off value for the visual gait scores. Stance variables were calculated from the SowSIS data per sow per day. First, a multivariable linear mixed model was used to detect lameness, using stance variables with significant influence on the gait score. The model's performance was 78.5% sensitivity, 81.4% specificity, 80.7% accuracy, 57.4% lame predictive value and 92.2% non-lame predictive value. Second, five types of classification models were tested to determine the lame leg on a sub-dataset. The random forest model could predict the lame leg correctly 90% of the time. The SowSIS shows great promise as an on-farm lameness detection system, as it allows continuous non-invasive data collection in a practical setting. • The Sow Stance Information System is robust and well adapted for on-farm use. • Daily non-invasive stance measurements of individual sows during feeding visits. • A linear mixed model using stance variables can correctly detect lameness in sows. • The lame leg can be identified using machine learning models. [ABSTRACT FROM AUTHOR]

Lameness has a considerable influence on the welfare and health of dairy cows. Many attempts have been made to develop automatic lameness detection systems using computer vision technology. However, these detection methods are easily affected by the characteristics of individual cows, resulting in inaccurate detection of lameness. Therefore, this study explores an individualized lameness detection method for dairy cattle based on the supporting phase using computer vision. This approach is applied to eliminate the influence of the characteristics of individual cows and to detect lame cows and lame hooves. In this paper, the correlation coefficient between lameness and the supporting phase is calculated, a lameness detection algorithm based on the supporting phase is proposed, and the accuracy of the algorithm is verified. Additionally, the reliability of this method using computer vision technology is verified based on deep learning. One hundred naturally walking cows are selected from video data for analysis. The results show that the correlation between lameness and the supporting phase was 0.864; 96% of cows were correctly classified, and 93% of lame hooves were correctly detected using the supporting phase–based lameness detection algorithm. The mean average precision is 87.0%, and the number of frames per second is 83.3 when the Receptive Field Block Net Single Shot Detector deep learning network was used to detect the locations of cow hooves in the video. The results show that the supporting phase–based lameness detection method proposed in this paper can be used for the detection and classification of cow lameness and the detection of lame hooves with high accuracy. This approach eliminates the influence of individual cow characteristics and could be integrated into an automatic detection system and widely applied for the detection of cow lameness. [ABSTRACT FROM AUTHOR]

The growing world population has increased the demand for animal-sourced protein. However, animal farming productivity is faced with challenges from traditional farming practices, socioeconomic status, and climate change. In recent years, smart sensors, big data, and deep learning have been applied to animal welfare measurement and livestock farming applications, including behaviour recognition and health monitoring. In order to facilitate research in this area, this review summarises and analyses some main techniques used in smart livestock farming, focusing on those related to cattle lameness detection and behaviour recognition. In this study, more than 100 relevant papers on cattle lameness detection and behaviour recognition have been evaluated and discussed. Based on a review and a comparison of recent technologies and methods, we anticipate that intelligent perception for cattle behaviour and welfare monitoring will develop towards standardisation, a larger scale, and intelligence, combined with Internet of things (IoT) and deep learning technologies. In addition, the key challenges and opportunities of future research are also highlighted and discussed.

Intelligent lameness detection is important for improving cow welfare. A method based on YOLOv3 deep learning algorithm and relative step size characteristic vector is proposed to classify lame and non-lame cows. Videos were decomposed into sequence frames, and leg targets of cows in each frame were detected by YOLOv3 algorithm. Relative step sizes of cow's front and rear legs were calculated based on leg coordinates, and the relative step size characteristic vector was constructed. Finally, a trained Long Short-Term Memory (LSTM) classification model was used to classify lame and non-lame cows based on the characteristic vector. A total of 210 videos were selected for verification using LSTM, support vector machine (SVM), K-Nearest Neighbour (KNN) and decision tree classifier (DTC) algorithms. Results showed that accuracy of lameness detection based on LSTM was 98.57%, which was 2.93%, 3.88%, and 9.25% higher than SVM, KNN, and DTC, respectively. True positive rate of the LSTM was 0.97, which was 0.03, 0.04 and 0.06 higher than SVM, KNN and DTC, respectively. False positive rate of LSTM was 0.03, which was 0.03, 0.06 and 0.11 lower than SVM, KNN and DTC, respectively. A bidirectional LSTM performed slightly better than LSTM but would be more demanding on hardware. Comparison of LSTM with a purely deep learning method showed the latter performed slightly better, but was less conducive to interpretation and diagnosis of lameness. The relative step size characteristic vector proposed was effective for classification of lame and non-lame cows, and could lead to intelligent detection of lameness. • Realising cow leg target detection based on YOLO v3 algorithm. • Proposing the relative step size characteristic vector for lameness detection. • Realising lame and non-lame detection based on the characteristic vector by LSTM. • The proposed method contributes to the intelligent detection of cow lameness. [ABSTRACT FROM AUTHOR]

The computer vision technique has been rapidly adopted in cow lameness detection research due to its noncontact characteristic and moderate price. This paper attempted to summarize the research progress of computer vision in the detection of lameness. Computer vision lameness detection systems are not popular on farms, and the accuracy and applicability still need to be improved. This paper discusses the problems and development prospects of this technique from three aspects: detection methods, verification methods and application implementation. The paper aims to provide the reader with a summary of the literature and the latest advances in the field of computer vision detection of lameness in dairy cows.

To tackle the high prevalence of lameness, techniques to monitor cow locomotion are being developed in order to detect changes in cows’ locomotion due to lameness. Obviously, in such lameness detection systems, alerts should only respond to locomotion changes that are related to lameness. However, other environmental or cow factors can contribute to locomotion changes not related to lameness and hence, might cause false alerts. In this study the effects of wet surfaces, dark environment, age, production level, lactation and gestation stage on cow locomotion were investigated. Data was collected at Institute for Agricultural and Fisheries Research research farm (Melle, Belgium) during a 5-month period. The gait variables of 30 non-lame and healthy Holstein cows were automatically measured every day. In dark environments and on wet walking surfaces cows took shorter, more asymmetrical strides with less step overlap. In general, older cows had a more asymmetrical gait and they walked slower with more abduction. Lactation stage or gestation stage also showed significant association with asymmetrical and shorter gait and less step overlap probably due to the heavy calf in the uterus. Next, two lameness detection algorithms were developed to investigate the added value of environmental and cow data into detection models. One algorithm solely used locomotion variables and a second algorithm used the same locomotion variables and additional environmental and cow data. In the latter algorithm only age and lactation stage together with the locomotion variables were withheld during model building. When comparing the sensitivity for the detection of non-lame cows, sensitivity increased by 10% when the cow data was added in the algorithm (sensitivity was 70% and 80% for the first and second algorithm, respectively). Hence, the number of false alerts for lame cows that were actually non-lame, decreased. This pilot study shows that using knowledge on influencing factors on cow locomotion will help in reducing the number of false alerts for lameness detection systems under development. However, further research is necessary in order to better understand these and many other possible influencing factors (e.g. trimming, conformation) of non-lame and hence ‘normal’ locomotion in cows.

The objective of this study was to evaluate if a multi-sensor system (milk, activity, body posture) was a better classifier for lameness than the single-sensor-based detection models. Between September 2013 and August 2014, 3629 cow observations were collected on a commercial dairy farm in Belgium. Human locomotion scoring was used as reference for the model development and evaluation. Cow behaviour and performance was measured with existing sensors that were already present at the farm. A prototype of three-dimensional-based video recording system was used to quantify automatically the back posture of a cow. For the single predictor comparisons, a receiver operating characteristics curve was made. For the multivariate detection models, logistic regression and generalized linear mixed models (GLMM) were developed. The best lameness classification model was obtained by the multi-sensor analysis (area under the receiver operating characteristics curve (AUC)=0.757±0.029), containing a combination of milk and milking variables, activity and gait and posture variables from videos. Second, the multivariate video-based system (AUC=0.732±0.011) performed better than the multivariate milk sensors (AUC=0.604±0.026) and the multivariate behaviour sensors (AUC=0.633±0.018). The video-based system performed better than the combined behaviour and performance-based detection model (AUC=0.669±0.028), indicating that it is worthwhile to consider a video-based lameness detection system, regardless the presence of other existing sensors in the farm. The results suggest that Θ2, the feature variable for the back curvature around the hip joints, with an AUC of 0.719 is the best single predictor variable for lameness detection based on locomotion scoring. In general, this study showed that the video-based back posture monitoring system is outperforming the behaviour and performance sensing techniques for locomotion scoring-based lameness detection. A GLMM with seven specific variables (walking speed, back posture measurement, daytime activity, milk yield, lactation stage, milk peak flow rate and milk peak conductivity) is the best combination of variables for lameness classification. The accuracy on four-level lameness classification was 60.3%. The accuracy improved to 79.8% for binary lameness classification. The binary GLMM obtained a sensitivity of 68.5% and a specificity of 87.6%, which both exceed the sensitivity (52.1%±4.7%) and specificity (83.2%±2.3%) of the multi-sensor logistic regression model. This shows that the repeated measures analysis in the GLMM, taking into account the individual history of the animal, outperforms the classification when thresholds based on herd level (a statistical population) are used.

Highlights • Walking speed is suitable for the detection of lame cows. • Lame cows have a significantly slower walking speed than non-lame cows. • The longer lameness lasts the slower the animals walk. • Cows walk significantly slower after morning milking than after evening milking. Abstract Lameness detection is performed based on deviations from the normal gait pattern of cows. Thereby each cow must be individually considered and thoroughly assessed. This process makes lameness detection a time-consuming job for the farmer, especially since herd sizes keep increasing. A sensor-based lameness detection could save the farmer time and complies simultaneously with advancing technological developments. The goal of this study was therefore to determine, whether there is a close causal relationship between walking speed and lameness, respectively lameness degree, so that walking speed could be integrated as a feature into a prediction model. To survey walking speed, we performed repeated time measurements on 53 lactating dairy cows of the Holstein Friesian breed while, passing through a 8.40 m long, straight test track. After morning and evening milking, the run time and the current locomotion score (LMS) were recorded for each cow. We found that lameness was significantly (p < 0.05) associated with walking speed. Lame cows (LMS ≥ 3) covered the test track at a significantly slower walking speed than non-lame cows (LMS ≤ 2). Also, the longer cows had suffered from lameness before, the slower the cows cover the distance. We conclude that such simple walking speed measurements can efficiently support lameness detection and thus contribute to the well-being of cows as well as the farmers' benefit in time. [ABSTRACT FROM AUTHOR]

Lameness diseases could cause high elimination rate of the worthless dairy cow, and early detection of dairy cow lameness disease is a significant research field of dairy farming. In this research, a lameness detection method of dairy cows with the fusion of LCCCT (local circulation center compensation tracking) and DSKNN (distilling data of k-nearest neighbor) was proposed. By using the double normal background statistical model (DNBSM), the dairy cow videos were decomposed into image sequences, and segmented to target region and backgrounds. Then, the obtained pixel area of the upper contour of the cow body was extracted by the LCCCT. In the detected region, DSKNN were used to extract the slope data of the head, neck and the back connected to the neck region. Firstly, the DNBSM model is used to separate the target dairy cow’s pixel area and background from the dairy cow’s sequence image. Since the frontal movement of dairy cow body is greater than the back movement of the dairy cow body, the DNBSM algorithm is used to lead to a better detection of the front body pixels for the dairy cow. Then, the LCCCT model is used to track and extract the pixel area of the front dairy cow body, and the DSKNN model is used to extract the target’s head and neck for slope data of contour line fitted. The changes in the slope data of the fitted straight line of the dairy cow’s head and neck were used as the basis for the detection of lameness of dairy cows. The characteristics of the slope data of head-neck fit lines of different dairy cows are different. Compared with other body parts of dairy cows, the head and neck area data of dairy cows are relatively easy to obtain. When dairy cows stand or walk, the head and neck characteristics can be extracted stably, which is the basis for judging lameness. The upper contour of the dairy cow's body starts from the tip of the dairy cow’s head and ends beyond the neck line. If the extracted contour line belongs to a defined area range, it is considered to be a valid contour line. Outside this range, it is considered that the extracted contour line is abnormal or has an error. The videos were divided into 3 slope data sets with 3 kinds of labels including slight lameness, moderate-severe lameness, and normal based on large sample video sequence frame data. In order to verify the validity of the algorithm, a total of 18 videos of dairy cows were processed. The slope data set was obtained from fitting line at the junction of the head, neck and back region. On the original data set, SVM, Naive Bayes algorithm and KNN classification algorithm were used to test the accuracy of dairy cow lameness detection, and the detection correct rates of SVM and Naive Bayes algorithm were both 82.78%, which were higher than that of KNN algorithm which was 81.67%. Test result illustrates that the slope feature of fitting line at the junction of the head, the neck and the back can be used to detect dairy cow lameness diseases. After cleaning the original data set, the correct detection rate of lameness classification by KNN classification algorithm was 94.44%, and the correct detection rates of SVM classification and Naive Bayes classification algorithm were 91.11%, and 86.11%, respectively. The results show that the KNN classification algorithm can get better results using cleaned data set. The results of this research can provide the reference for the prevention and diagnosis of lameness disease of dairy cows. [ABSTRACT FROM AUTHOR]

Lameness has become a frequent and serious problem for herd productivity and animal welfare in the dairy industry. As the most significant characteristic of lameness, gait characteristics have been used to estimate lameness by human experts. The objective of this study was to analyze leg swing using computer vision techniques and to develop an automatic and continuous system for scoring the locomotion of cows to detect and predict lameness with high accuracy and practicability. The focus was on quantifying the movement pattern of cows and demonstrating the possibility of classifying lameness using the features extracted from movement analysis. Side-view videos were recorded after the cows were milked. Cows were scored by an expert on a scale from 1 (sound) to 3 (severely lame). The data set included 621 videos from 98 cows. The motion curve was plotted by extracting the position of the moving leg by image processing, and the motion curve was analyzed to generate six features referring to the gait asymmetry, speed, tracking up, stance time, stride length, and tenderness. A box-plot of the features within 3 classes showed that the dataset was nearly linear and separable under the six features and that the cows had different lameness indicators in different lameness stages. The Decision Tree classifier was applied to the dataset, and 2-, 3-, and 10-fold cross validation was used to verify the performance of the algorithm. The accuracy of the classification was 90.18%, and the averages of sensitivity and specificity were 90.25% and 94.74%, respectively. This research demonstrates the feasibility of classifying dairy cow lameness based on the six motion features extracted by leg swing analysis. [ABSTRACT FROM AUTHOR]

Lameness in dairy herds is traditionally detected by visual inspection, which is time-consuming and subjective. Compared with healthy cows, lame cows often spend longer time lying down, walk less and change behaviour around feeding time. Accelerometers measuring cow leg activity may assist farmers in detecting lame cows. On four commercial farms, accelerometer data were derived from hind leg-mounted accelerometers on 348 Holstein cows, 53 of them during two lactations. The cows were milked twice daily and had no access to pasture. During a lactation, locomotion score (LS) was assessed on average 2.4 times (s.d. 1.3). Based on daily lying duration, standing duration, walking duration, total number of steps, step frequency, motion index (MI, i.e. total acceleration) for lying, standing and walking, eight accelerometer means and their corresponding coefficient of variation (CV) were calculated for each week immediately before an LS. A principal component analysis was performed to evaluate the relationship between the variables. The effects of LS and farm on the principal components (PC) and on the variables were analysed in a mixed model. The first four PC accounted for 27%, 18%, 12% and 10% of the total variation, respectively. PC1 corresponded to Activity variability due to heavy loading by five CV variables related to standing and walking. PC2 corresponded to Activity level due to heavy loading by MI walking, MI standing and walking duration. PC3 corresponded to Recumbency due to heavy loading by four variables related to lying. PC4 corresponded mainly to Stepping due to heavy loading by step frequency. Activity variability at LS4 was significantly higher than at the lower LS levels. Activity level was significantly higher at LS1 than at LS2, which was significantly higher than at LS4. Recumbency was unaffected by LS. Stepping at LS1 and LS2 was significantly higher than at LS3 and LS4. Activity level was significantly lower on farm 3 compared with farms 1 and 2. Stepping was significantly lower on farms 1 and 3 compared with farms 2 and 4. MI standing indicated increased restlessness while standing when cows increased from LS3 to LS4. Lying duration was only increased in lame cows. In conclusion, Activity level differed already between LS1 and LS2, thus detecting early signs of lameness, particularly through contributions from walking duration and MI walking. Lameness detection models including walking duration, MI walking and MI standing seem worthy of further investigation.

An important factor for animal welfare in cattle farming is the detection of lameness. The presented study is part of a project aiming to develop a system that is capable of an automated diagnosis of claw lesions by analyzing the footfall sound. Data were generated from cows walking along a measurement zone where piezoelectric sensors recorded their footfall sounds. Locomotion of the animals was scored and they were graded according to a three-scale scoring system (LS1 = non-lame; LS2 = uneven gait; LS3 = lame). Subsequently, the cows were examined by a hoof trimmer. The walking speed across the test track was significantly higher in cows with LS1 compared to those with LS2 and LS3 and thus, they were showing a smoother gait pattern. The standard deviation of volume (SDV) in the recorded footfall sound signal was considered as a factor for the force of a cow’s footsteps. Cows with non-infectious claw lesions showed lower SDV than healthy cows and those with infectious claw diseases. This outcome confirmed the hypothesis that the evaluated cows affected by non-infectious claw lesions have a greater sensitivity to pain and demonstrate a less forceful gait pattern. These first results clearly show the potential of using footfall sound analysis for detecting claw lesions.

• Improve the Siamese network by combining attention mechanism for cow legs tracking, and promoted the success rate. • Detecting lameness by using step sizes of the front and rear legs as feature vectors for the classification algorithm. • The proposed method can deal with the partial occlusion of the cow's legs during tracking. The analysis of a cow's motion behavior is based on the positioning and tracking of its legs. This paper proposes a Siamese attention model (Siam-AM) fusion attention mechanism to obtain the automatic tracking and monitoring of cow legs in large-scale farms. Firstly, the features of the cow's leg are extracted using the first frame of motion image data. The search area in the subsequent frames is also input into the network to extract additional features. Then an attention module is used to assign weights to the features. The similarity between different regions and image features of the first frame was compared. The image with the most significant similarity was taken as the predicted position of the cow's leg. Finally, the relative step size of the cow's front and rear legs is calculated using the leg's position coordinates. A Support Vector Machine (SVM) classifier is constructed to realize the lameness detection of the cow. The experimental results show that the proposed Siam-AM algorithm has an average tracking accuracy of 93.80 %, an average frame rate of 57 fps, and a lameness detection accuracy of 94.73 %. These results provide an effective method for accurate tracking and lameness detection of a cow's leg. [ABSTRACT FROM AUTHOR]

To tackle the high prevalence of lameness, techniques to monitor cow locomotion are being developed in order to detect changes in cows' locomotion due to lameness. Obviously, in such lameness detection systems, alerts should only respond to locomotion changes that are related to lameness. However, other environmental or cow factors can contribute to locomotion changes not related to lameness and hence, might cause false alerts. In this study the effects of wet surfaces, dark environment, age, production level, lactation and gestation stage on cow locomotion were investigated. Data was collected at Institute for Agricultural and Fisheries Research research farm (Melle, Belgium) during a 5-month period. The gait variables of 30 non-lame and healthy Holstein cows were automatically measured every day. In dark environments and on wet walking surfaces cows took shorter, more asymmetrical strides with less step overlap. In general, older cows had a more asymmetrical gait and they walked slower with more abduction. Lactation stage or gestation stage also showed significant association with asymmetrical and shorter gait and less step overlap probably due to the heavy calf in the uterus. Next, two lameness detection algorithms were developed to investigate the added value of environmental and cow data into detection models. One algorithm solely used locomotion variables and a second algorithm used the same locomotion variables and additional environmental and cow data. In the latter algorithm only age and lactation stage together with the locomotion variables were withheld during model building. When comparing the sensitivity for the detection of non-lame cows, sensitivity increased by 10% when the cow data was added in the algorithm (sensitivity was 70% and 80% for the first and second algorithm, respectively). Hence, the number of false alerts for lame cows that were actually non-lame, decreased. This pilot study shows that using knowledge on influencing factors on cow locomotion will help in reducing the number of false alerts for lameness detection systems under development. However, further research is necessary in order to better understand these and many other possible influencing factors (e.g. trimming, conformation) of non-lame and hence 'normal' locomotion in cows. [ABSTRACT FROM AUTHOR]

The objective of this study was to evaluate if a multi-sensor system (milk, activity, body posture) was a better classifier for lameness than the single-sensor-based detection models. Between September 2013 and August 2014, 3629 cow observations were collected on a commercial dairy farm in Belgium. Human locomotion scoring was used as reference for the model development and evaluation. Cow behaviour and performance was measured with existing sensors that were already present at the farm. A prototype of three-dimensional-based video recording system was used to quantify automatically the back posture of a cow. For the single predictor comparisons, a receiver operating characteristics curve was made. For the multivariate detection models, logistic regression and generalized linear mixed models (GLMM) were developed. The best lameness classification model was obtained by the multi-sensor analysis (area under the receiver operating characteristics curve (AUC) = 0.757 ± 0.029), containing a combination of milk and milking variables, activity and gait and posture variables from videos. Second, the multivariate videobased system (AUC = 0.732 ± 0.011) performed better than the multivariate milk sensors (AUC = 0.604 ± 0.026) and the multivariate behaviour sensors (AUC = 0.633 ± 0.018). The video-based system performed better than the combined behaviour and performance-based detection model (AUC = 0.669 ± 0.028), indicating that it is worthwhile to consider a video-based lameness detection system, regardless the presence of other existing sensors in the farm. The results suggest that Θ2, the feature variable for the back curvature around the hip joints, with an AUC of 0.719 is the best single predictor variable for lameness detection based on locomotion scoring. In general, this study showed that the video-based back posture monitoring system is outperforming the behaviour and performance sensing techniques for locomotion scoring-based lameness detection. A GLMM with seven specific variables (walking speed, back posture measurement, daytime activity, milk yield, lactation stage, milk peak flow rate and milk peak conductivity) is the best combination of variables for lameness classification. The accuracy on four-level lameness classification was 60.3%. The accuracy improved to 79.8% for binary lameness classification. The binary GLMM obtained a sensitivity of 68.5% and a specificity of 87.6%, which both exceed the sensitivity (52.1% ± 4.7%) and specificity (83.2% ± 2.3%) of the multi-sensor logistic regression model. This shows that the repeated measures analysis in the GLMM, taking into account the individual history of the animal, outperforms the classification when thresholds based on herd level (a statistical population) are used. [ABSTRACT FROM AUTHOR]

Lameness is a critical issue on dairies with an impact on production and animal welfare. Early lameness detection followed by effective treatments could improve prognosis and cure rate of lame cows. Current methods for lameness detection are based on locomotion score (LS) that requires observation of cows walking, preferably at the exit of the milking parlor. This is a time-consuming task that is difficult to implement on large dairies. Therefore, a common methodology for lameness detection is based on milkers’ and cow pushers’ observations of cows walking to the milking parlor or standing at the milking stall (MPP). Observation of postural abnormalities predictive of lameness while cows are locked at stanchions (S) can be used as an alternative detection method. The objective of this research was to study the association between postural and gait abnormalities observed with S and MPP methodologies and lameness using LS ≥ 3 as the reference method, as well as to evaluate the epidemiological characteristics of those methods as a diagnostic test for lameness. A secondary objective was to describe the type of hoof lesions observed with postural and gait abnormalities detected with LS, MPP, and S methodologies. A cross-sectional study design was performed on 2274 cows from one farm in California (US). Arched back, cow-hocked, wide-stance, and favored-limb postures as well as uneven gait were observed. Both lameness detection methodologies, S and MPP, indicated that arched back and favored-limb were postural abnormalities associated with lameness. However, the epidemiological test characteristics for each of the postures evaluated as a diagnostic test for lameness indicated that both detection methods, S and MPP, had good specificity (> 0.91) but poor sensitivity (0.04–0.39). A convenience sample of 104 cows, selected based on LS > 3, favored-limb, presence of two or more abnormal postures, and gait anomalies with either S or MPP methods, received a hoof examination. Lesions were observed on cows selected by LS (17/24), MPP (21/30), and S (33/60) criteria, suggesting a lack of concordance between lameness detection methodologies and visible hoof lesions. Nevertheless, due to the lack of acceptance of LS as the lameness detection method on large commercial dairies in California, it is imperative that future research evaluates modifications of S and MPP lameness detection techniques, considering hoof lesion as reference method. [ABSTRACT FROM AUTHOR]

Lying behavior is an important measure of comfort and well-being in dairy cattle, and changes in lying behavior are potential indicators and predictors of lameness. Our objectives were to determine individual and herd-level risk factors associated with measures of lying behavior, and to evaluate whether automated measures of lying behavior can be used to detect lameness. A purposive sample of 40 Holstein cows was selected from each of 141 dairy farms in Alberta, Ontario, and Québec. Lying behavior of 5,135 cows between 10 and 120 d in milk was automatically and continuously recorded using accelerometers over 4 d. Data on factors hypothesized to influence lying behavior were collected, including information on individual cows, management practices, and facility design. Associations between predictor variables and measures of lying behavior were assessed using generalized linear mixed models, including farm and province as random and fixed effects, respectively. Logistic regression models were used to determine whether lying behavior was associated with lameness. At the cow-level, daily lying time increased with increasing days in milk, but this effect interacted with parity; primiparous cows had more frequent but shorter lying bouts in early lactation, changing to mature- cow patterns of lying behavior (fewer and longer lying bouts) in late lactation. In barns with stall curbs >22 cm high, the use of sand or >2 cm of bedding was associated with an increased average daily lying time of 1.44 and 0.06 h/d, respectively. Feed alleys ≥350 cm wide or stalls ≥114 cm wide were associated with increased daily lying time of 0.39 and 0.33 h/d, respectively, whereas rubber flooring in the feed alley was associated with 0.47 h/d lower average lying time. Lame cows had longer lying times, with fewer, longer, and more variable duration of bouts compared with nonlame cows. In that regard, cows with lying time ≥14 h/d, ≤5 lying bouts per day, bout duration ≥110 min/bout, or standard deviations of bout duration over 4 d ≥70 min had 3.7, 1.7, 2.5, and 3.0 higher odds of being lame, respectively. Factors related to comfort of lying and standing surfaces significantly affected lying behavior. Finally, we inferred that automated measures of lying behavior could contribute to lameness detection, especially when interpreted in the context of other factors known to affect lying behavior, including those associated with the individual cow (e.g., parity and stage of lactation) or environment (e.g., stall surface). [ABSTRACT FROM AUTHOR]

Lameness in dairy herds is traditionally detected by visual inspection, which is time-consuming and subjective. Compared with healthy cows, lame cows often spend longer time lying down, walk less and change behaviour around feeding time. Accelerometers measuring cow leg activity may assist farmers in detecting lame cows. On four commercial farms, accelerometer data were derived from hind leg-mounted accelerometers on 348 Holstein cows, 53 of them during two lactations. The cows were milked twice daily and had no access to pasture. During a lactation, locomotion score (LS) was assessed on average 2.4 times (s.d. 1.3). Based on daily lying duration, standing duration, walking duration, total number of steps, step frequency, motion index (MI, i.e. total acceleration) for lying, standing and walking, eight accelerometer means and their corresponding coefficient of variation (CV) were calculated for each week immediately before an LS. A principal component analysis was performed to evaluate the relationship between the variables. The effects of LS and farm on the principal components (PC) and on the variables were analysed in a mixed model. The first four PC accounted for 27%, 18%, 12% and 10% of the total variation, respectively. PC1 corresponded to Activity variability due to heavy loading by five CV variables related to standing and walking. PC2 corresponded to Activity level due to heavy loading by MI walking, MI standing and walking duration. PC3 corresponded to Recumbency due to heavy loading by four variables related to lying. PC4 corresponded mainly to Stepping due to heavy loading by step frequency. Activity variability at LS4 was significantly higher than at the lower LS levels. Activity level was significantly higher at LS1 than at LS2, which was significantly higher than at LS4. Recumbency was unaffected by LS. Stepping at LS1 and LS2 was significantly higher than at LS3 and LS4. Activity level was significantly lower on farm 3 compared with farms 1 and 2. Stepping was significantly lower on farms 1 and 3 compared with farms 2 and 4. MI standing indicated increased restlessness while standing when cows increased from LS3 to LS4. Lying duration was only increased in lame cows. In conclusion, Activity level differed already between LS1 and LS2, thus detecting early signs of lameness, particularly through contributions from walking duration and MI walking. Lameness detection models including walking duration, MI walking and MI standing seem worthy of further investigation. [ABSTRACT FROM PUBLISHER]

• We designed a deep learning algorithm for lameness detection. • We constructed a dimension-reduction spatiotemporal model for cow gait. • We reduced the impact of irrelevant information on lameness classification. Lameness has a negative effect on the welfare and production of dairy cows. In previous studies of lameness detection in cows based on computer vision, cow walking videos contained considerable irrelevant information, lowering detection accuracy. In this study, we proposed a dairy cattle lameness detection method, namely, the Dimension-Reduced Spatiotemporal Network (DRSN), to reduce the impact of irrelevant information, consisting of video dimensionality reduction and deep learning algorithms. The YOLOv4 algorithm was used to detect the cow's hooves from the video. The video was dimensionally reduced to a spatiotemporal image based on the locations of the cow's legs according to the previous detection, retaining gait information while removing much irrelevant information. Finally, the DenseNet algorithm was used to classify the lameness degree into a locomotion score according to the spatiotemporal image. To evaluate the performance of the algorithm, videos of 456 cows were used as the dataset for testing. After comparing different target tracking and classification algorithms, the YOLOv4 object detection and DenseNet classification algorithms demonstrated the best performance, with target detection and classification accuracies of 92.39% and 98.50%, respectively. The result of video-based lameness classification was compared with our method, and the experimental results showed that our method was more accurate. The newly proposed approach can effectively remove irrelevant information and improve the detection accuracy for lameness in dairy cow walking videos. The method can be further integrated into a system for automatically detecting cow lameness. [ABSTRACT FROM AUTHOR]

Highlights: [•] An automatic lameness detection system that uses 3D cameras is proposed. [•] Algorithm to extract the back posture automatically from 3D camera is presented. [•] Result of the algorithm are compared to the visual locomotion scores of an expert. [•] Classification result of lameness obtained from 3D and 2D cameras are comparable. [ABSTRACT FROM AUTHOR]

The hypothesis was that sensors currently available on farm that monitor behavioral and physiological characteristics have potential for the detection of lameness in dairy cows. This was tested by applying additive logistic regression to variables derived from sensor data. Data were collected between November 2010 and June 2012 on 5 commercial pasture-based dairy farms. Sensor data from weigh scales (liveweight), pedometers (activity), and milk meters (milking order, unadjusted and adjusted milk yield in the first 2 min of milking, total milk yield, and milking duration) were collected at every milking from 4,904 cows. Lameness events were recorded by farmers who were trained in detecting lameness before the study commenced. A total of 318 lameness events affecting 292 cows were available for statistical analyses. For each lameness event, the lame cow's sensor data for a time period of 14 d before observation date were randomly matched by farm and date to 10 healthy cows (i.e., cows that were not lame and had no other health event recorded for the matched time period). Sensor data relating to the 14-d time periods were used for developing univariable (using one source of sensor data) and multivariable (using multiple sources of sensor data) models. Model development involved the use of additive logistic regression by applying the LogitBoost algorithm with a regression tree as base learner. The model's output was a probability estimate for lameness, given the sensor data collected during the 14-d time period. Models were validated using leave-one-farm-out cross-validation and, as a result of this validation, each cow in the data set (318 lame and 3,180 nonlame cows) received a probability estimate for lameness. Based on the area under the curve (AUC), results indicated that univariable models had low predictive potential, with the highest AUC values found for liveweight (AUC = 0.66), activity (AUC = 0.60), and milking order (AUC = 0.65). Combining these 3 sensors improved AUC to 0.74. Detection performance of this combined model varied between farms but it consistently and significantly outperformed univariable models across farms at a fixed specificity of 80%. Still, detection performance was not high enough to be implemented in practice on large, pasture-based dairy farms. Future research may improve performance by developing variables based on sensor data of live-weight, activity, and milking order, but that better describe changes in sensor data patterns when cows go lame. [ABSTRACT FROM AUTHOR]

Abstract: Lameness research in cattle focuses on daily or weekly variation of several typical gait features, i.e. stride time, stride length or arching of the back. In human gait research however, the variation of typical gait characteristics within one measurement is taken into account and serves as a sensitive and clinically relevant parameter in the evolution of diseases and the response to therapeutic interventions. In this paper the utility of such within measurement variation of gait variables was investigated for lameness research in cattle by comparing variables between lame cows, cows that are scored mildly lame by a trained observer but not yet noticed lame by the farmer, and cows that were scored severely lame and were noticed lame by the farmer in two case-control studies. Additionally, the repeatability of all variables was tested using eight non-lame cows. The measurements with the GAITWISE were found to be highly repeatable and are therefore suited for cow gait characterization. The new set of inconsistency variables was able to distinguish between non-lame and becoming lame (mildly lame but not yet noticed by the farmer) cows. These results suggest the usefulness of such variables of stride-to-stride-fluctuation for early detection of lameness. Whether they are more sensitive for early lameness detection compared to the more conventionally used averaged gait variables should be confirmed in longitudinal studies with non-lame cows and cows that develop lameness. All cows that were selected in the case control studies were mildly or severely lame in the left hind leg. Cows that were severely lame on their left hind leg put less force and take shorter steps with that sore leg increasing stance times on the right side. In addition, their stride to stride fluctuations of ‘stance times on the left front leg’ and ‘step length between left hind and right front leg’ significantly increase for the cows that were mildly lame but not yet noticed lame by the farmer. This strong relationship between the localization of the lameness and the significantly changed variables, suggest that the variables measured by the GAITWISE are promising in defining which leg is developing lameness in cows. [Copyright &y& Elsevier]

This investigation analysed the applicability of principal component analysis (PCA), a latent variable method, for the early detection of mastitis and lameness. Data used were recorded on the Karkendamm dairy research farm between August 2008 and December 2010. For mastitis and lameness detection, data of 338 and 315 cows in their first 200 d in milk were analysed, respectively. Mastitis as well as lameness were specified according to veterinary treatments. Diseases were defined as disease blocks. The different definitions used (two for mastitis, three for lameness) varied solely in the sequence length of the blocks. Only the days before the treatment were included in the blocks. Milk electrical conductivity, milk yield and feeding patterns (feed intake, number of feeding visits and time at the trough) were used for recognition of mastitis. Pedometer activity and feeding patterns were utilised for lameness detection. To develop and verify the PCA model, the mastitis and the lameness datasets were divided into training and test datasets. PCA extracted uncorrelated principle components (PC) by linear transformations of the raw data so that the first few PCs captured most of the variations in the original dataset. For process monitoring and disease detection, these resulting PCs were applied to the Hotelling's T2 chart and to the residual control chart. The results show that block sensitivity of mastitis detection ranged from 77·4 to 83·3%, whilst specificity was around 76·7%. The error rates were around 98·9%. For lameness detection, the block sensitivity ranged from 73·8 to 87·8% while the obtained specificities were between 54·8 and 61·9%. The error rates varied from 87·8 to 89·2%. In conclusion, PCA seems to be not yet transferable into practical usage. Results could probably be improved if different traits and more informative sensor data are included in the analysis. [ABSTRACT FROM PUBLISHER]

Currently, diagnosis of lameness at an early stage in dairy cows relies on visual observation by the farmer, which is time consuming and often omitted. Many studies have tried to develop automatic cow lameness detection systems. However, those studies apply thresholds to the whole population to detect whether or not an individual cow is lame. Therefore, the objective of this study was to develop and test an individualized version of the body movement pattern score, which uses back posture to classify lameness into 3 classes, and to compare both the population and the individual approach under farm conditions. In a data set of 223 videos from 90 cows, 76% of cows were correctly classified, with an 83% true positive rate and 22% false positive rate when using the population approach. A new data set, containing 105 videos of 8 cows that had moved through all 3 lameness classes, was used for an ANOVA on the 3 different classes, showing that body movement pattern scores differed significantly among cows. Moreover, the classification accuracy and the true positive rate increased by 10 percentage units up to 91%, and the false positive rate decreased by 4 percentage units down to 6% when based on an individual threshold compared with a population threshold. [ABSTRACT FROM AUTHOR]

Abstract: The aim of this study was to explore wavelet filtering for early detection of mastitis and lameness. Data were recorded at the Karkendamm dairy research farm between January 2009 and October 2010. In total, data of 237 cows with 46,427 cow days were analysed. Mastitis was specified according to three definitions: (1) udder treatment, (2) udder treatment and/or somatic cell count with more than 100,000cells/ml and (3) udder treatment and/or somatic cell count exceeding 400,000cells/ml. Lameness treatments were used to determine two definitions of lameness. They differed in the length of the corresponding disease block: (1) day of treatment including three days before treatment, (2) day of treatment including five days before treatment. Milk electrical conductivity and cow activity were utilised as indicator parameters for detection of mastitis and lameness, respectively. These values were filtered by wavelets. Filtered values of cow activity and the residuals between the observed and filtered values of milk electrical conductivity were applied to a classic and a self-starting CUSUM chart to identify blocks of disease (days of disease). Regarding performance of mastitis detection, the classic chart showed better results than the self-starting chart. The block sensitivity ranged between 72.6% and 76.3% (self-starting chart between 72.1% and 74.5%) while the obtained error rates were between 69.2% and 94.4% (self-starting chart between 73.4% and 95.7%). For both charts, block sensitivity and error rate improved from definition (1) to (3). In the case of lameness detection, the block sensitivity of the classic chart varied between 40.4% and 48.3%, which was lower than the block sensitivities of the self-starting chart (47.2% and 63.5%). The error rates of lameness detection were also high (90.6% to 93.3%). In conclusion, wavelet analysis seems to be applicable to mastitis and lameness detection in dairy cows. Results could probably be enhanced if more traits for a multivariate consideration are used. [Copyright &y& Elsevier]

This paper describes a synchronized measurement system combining image and pressure data to automatically record the angle of the metacarpus and metatarsus bones of the cow with respect to a vertical line, which is useful for lameness detection in dairy cattle. A camera system was developed to record the posture and movement of the cow and the timing and position of hoof placement and release were recorded using a pressure sensitive mat. Experiments with the automatic system were performed continuously on a farm in Ghent (Belgium) for 5 wk in September and October 2009. In total, 2,219 measurements were performed on 75 individual lactating Holstein cows. As a reference for the analysis of the calculated variables, the locomotion of the cows was visually scored from recorded videos by a trained observer into 3 classes of lameness [53.5% were scored with gait score (GS)1, 33.3% were scored with GS2, and 9.3% were scored with GS3]. The contact data of the pressure mat and the camera images recorded by the system were synchronized and combined to measure different angles of the legs of the cows, together with the range of motion of the leg. Significant differences were found between the different gait scores in the release angles of the front hooves, in the range of motion of the front hooves, and in the touch angles of the hind hooves. The contact data of the pressure mat and the camera images recorded by the system were synchronized and combined to measure different angles of the legs of the cows, together with the range of motion of the leg. With respect to the classification of lameness, the range of motion of the front hooves (42.1 and 42.8%) and the release angle of the front hooves (41.7 and 42.0%) were important variables. In 83.3% of the cows, a change in GS led to an increase in within-cow variance for the range of motion or the release angle of the front hooves. In 76.2% of the cows, an increase in GS led to a decrease in range of motion or an increase in release angle of the front hooves. [ABSTRACT FROM AUTHOR]

Abstract: In this paper results on utilizing image analysis techniques towards early lameness detection in dairy cattle are presented. Data from two different dairy farms in Belgium were gathered. Preprocessing on raw data is required because of non-predictable behaviours of cows such as stopping for a while in front of the camera or non-uniform walking behaviour during experiments. Prelocalization of cow in each frame has been done based on two steps separation: (1) A coarse estimation of moving objects was obtained through background subtraction, (2) second statistical analysis of intensities in gray-scale image along with binarization was utilized to detect moving object in video. A common problem in on-farm collected videos is the similarity of the background and the cow''s body colour since the use of classic algorithms for segmentation purposes does not work. Here a hierarchy background/foreground exaggeration is proposed to segment the cow in each frame and track it in video. The combination of logarithm and exponential, background subtraction as well as statistical filtering are used to find the accurate shape of the cow. Furthermore, the back posture of each cow during standing and walking was extracted automatically. It was done by detecting the arc of back posture and fitting a circle through selected points on the spine line. The average inverse radius of four frames displaying the hind hoofs in contact with the ground (two frames for each hoof in a row) was assigned to the cow. Based on this curvature value, a score representing the status of lameness in the individual cow was given automatically. Experimental results from two different databases show promising results in automatic lameness detection based on back posture information. [Copyright &y& Elsevier]

• A new deep learning framework to detect dairy cow lameness was proposed. • This method was more stable in the antagonism of environmental robustness. • This method improves the detection accuracy of dairy cow lameness. Manual detection of lameness poses several problems, such as difficulty in finding sudden, severe or early lameness behavior. A dairy cow's lameness is closely related to curvature of the cow's back. Focusing on the curvature features of dairy cows' backs, this study proposes a lameness detection method that combines machine vision technology with a deep learning algorithm. Firstly, the FLYOLOv3 algorithm was used to construct a Cow's Back Position Extraction (CBPE) model to realize the extraction of the dairy cow's back position coordinates. Simultaneously, a First to Last Frame Image Difference (FLFID) algorithm was used to construct a Cow's Object Region Extraction (CORE) model to separate the dairy cow from the image background and obtain the pixel region of the dairy cow. Then, a Cow's Back Curvature Extraction (CBCE) model was used to extract the dairy cow's back curvature data from the acquired dairy cow's back position and pixel region of the dairy cow. Finally, a Noise+ Bilateral Long Short-term Memory (BiLSTM) model was used to predict the curvature data and match the curvature features of the dairy cow's lameness, so as to classify and detect dairy cow lameness. To verify the effectiveness of the algorithm, 567 videos were used to train the network model in a Long Short-term Memory (LSTM) model, a BiLSTM model, Noise+LSTM model, and the model proposed in this paper, respectively, and 243 videos were used for verification and testing. According to the fitting curvature data of the dairy cows' back obtained by the algorithm used in this paper, it was found that the average classification accuracy of the model proposed in this research was 8.04%, 2.09%, and 5.78% higher than the average classification accuracy of the LSTM, BiLSTM, Noise+LSTM models, respectively. In the parallel experiment that classified the detection of dairy cow lameness, the average classification accuracy of the model proposed in this paper was 96.61%. The above results show that the lameness of dairy cows can be correctly detected through analysis of the curvature features of dairy cows' backs. The proposed method is a novel, deep learning-based method for dairy cow lameness early detection which may have significant economic impact on the dairy industry, and the proposed method provides an innovative means for detecting dairy cow lameness. [ABSTRACT FROM AUTHOR]