Search

While previous research has shown the potential of automatic lameness detection by means of a pressure mat, these systems are currently not adopted in practice due to their high cost and low on-farm applicability. Therefore, the aim of this study was to investigate to what level the size (0.61 × 4.88 m) and resolution (0.0127 × 0.0127 m) of the pressure mat can be reduced without significant loss in lameness detection performance. To this end, standard gait variables were calculated based on adapted datasets in which the available data had been reduced to simulate the effects of a decreasing mat length and sensor resolution. These extracted gait variables were then used in a linear discriminant analysis to classify cows as non-lame, mildly lame or severely lame. This analysis indicated that the measurement zone length must be at least 3.28 m to successfully monitor one complete gait cycle, while the size of each individual sensing element should not be larger than 2.58 ∗ 10 −3 m 2 to avoid an increase in the misidentification of imprints. When these limits were taken into account, the obtained overall lameness detection accuracy was not worse than that of the original system. [ABSTRACT FROM AUTHOR]

Changes in the feeding pattern of a pig may indicate disease or welfare issues and can also be strongly related to productivity changes. Automatic measurements of the feeding patterns of group-housed growing-finishing pigs can enable the detection of such feeding related changes. In this paper, a high frequency (HF) Radio Frequency Identification (RFID) system was used to register pigs’ feeding patterns at a commercial-type feed trough. The raw RFID registrations had to be translated into feeding visits, as there were irregular time gaps between the registrations that were inherent to the system for this application. Data from two experiments were used: one used behavioural observations of six pigs during three days with the aim of finding the best method for visit construction (experiment 2, E2), whilst the other experiment used feeding pattern observations of 20 pigs on one day as an independent validation data-set (experiment 1, E1). Three methods for visit construction were proposed: (1) visit criteria (a bout criterion and a minimum duration criterion), (2) a digital filter (using a well-defined time delay and a threshold for feeding) and (3) erosion and dilation operations from mathematical morphology. The best results were obtained using a bout criterion equal to 10 s for two RFID tags per pig or 20 s for one RFID tag per pig. Removal of short visits was found to be unnecessary. This optimal method is equal to a closing operation (a dilation followed by an erosion). For the RFID system with the optimal bout criterion, using two tags per pig, average sensitivity was 83%, specificity 98%, accuracy 97% and precision was 75% for E2. Applying the same bout criterion on the independent test set (E1), resulted in an average sensitivity of 80%, a specificity of 99%, an accuracy of 98% and a precision of 78%. Performance of the method tended to be lower when using one tag per pig. Regression analysis revealed that the observed feeding duration was well-correlated with RFID-based feeding duration (R 2 = 0.86 for two tags per pig, E1 + E2). Other observed variables (number of feeding visits and inter-visit interval) were harder to predict by the RFID system. From these results, there was concluded that feeding patterns can be recorded accurately in an automatic way if the registrations from the RFID system are analysed using an appropriate bout criterion. [ABSTRACT FROM AUTHOR]

Highlights: [•] A HF RFID system was validated for registering feeding pigs. [•] The system was used in simulated commercial farm conditions. [•] Specificity for individual pigs with one or two tags was always above 85%. [•] Sensitivity varied for individual pigs, tags and different time windows used. [•] A sensitivity of 88.58% and a specificity of 98.34% can be reached. [Copyright &y& Elsevier]

Abstract: In this paper, an experiment of using a hyperspectral imaging system for bruise detection on ‘Golden Delicious’ apples is reported. The hyperspectral imaging system was built in the wavelength region between 400 and 1000nm. Chemometric tools such as PCA and PLSDA were used to extract and summarize the spectral information from the hyperspectral images. Image processing methods made it possible to segment the region-of-interest according to the spatial features. Classification algorithms based on PCA and PLSDA results were developed, respectively. Their performance with respect to the classification accuracy and feasibility to implement on-line sorting were compared. The chemometric tools are able to extract and summarize the pixel-based information, while the image processing methods provide region-based analysis to efficiently segment differences of the apple surface. This combination of image processing techniques and chemometric tools provides a very promising approach for studying the quality of apples. [Copyright &y& Elsevier]

• A fiber-optic spatially-resolved spectroscopy (SRS) setup measured milk NIR spectra. • The setup evaluated 30 different illumination-to-detection distances (1.1–4 mm) • Shorter illumination-to-detection distances were optimal for lactose prediction. • Larger illumination-to-detection distances were optimal for fat and protein. • The accuracy of the composition predictions are well within the ICAR requirements. To support in-line quality control of raw milk and ensure the close monitoring of the metabolic and udder health of dairy cows, we propose a fiber-optic spatially-resolved spectroscopy (SRS) setup. This setup allows to vary the interaction of long-wave near-infrared (LW-NIR) light with the milk by submerging a separate optical illumination and detection fiber into the sample and altering their relative distance to optimize measurements for specific milk component(s). We evaluated this approach for predicting milk fat, protein, and lactose content and determined the optimal illumination-to-detection distances for each milk component. The region between 1.1 and 1.8 mm was optimal for lactose, and between 2.2 and 3.8 mm for fat and protein. These distance ranges resulted in a root-mean-square error of prediction (RMSEP) of less than 0.10 % weight/weight (wt/wt) for milk fat, and lower than 0.13 % (wt/wt) for protein and lactose. Integration of these distances into a fiber-optic SRS reflectance probe would allow to simultaneously determine the fat, protein, and lactose content of raw milk in-line with high accuracy. [ABSTRACT FROM AUTHOR]

Ethiopian farmers have been using an ox-drawn breaking plough, known as ard plough – maresha, for thousands of years. Maresha is a pointed, steel-tipped tine attached to a draught pole at an adjustable shallow angle. It has narrow side-wings, attached to the left and right side of it, to push soil to either side without inverting. The aim of this paper is to explore the effect of side-wings on draught using a field soil bin test facility. To this end, a mobile and an in-situ soil bin test system, for online measurements of draught, was designed and developed. This research considered tool geometry ( maresha plough with and without side-wings) and rake angle (shallow – 8°, medium deep – 15°, and deep – 24°, representing primary, secondary and tertiary tillage processes in Ethiopia, respectively). Maresha plough with side-wings has greater contact area, between the moving soil and tool, than its wingless counterpart. When the ploughshare surface and soil slide relative to one another, the draught expected to increase with contact area, as adhesion and friction resistance increases with area. However, experimental analysis indicated that the maresha with side-wings required less draught compared to maresha without side-wings ( ρ < 0.001) . This might be attributed to the effect of side-wings on crack propagation by a wedging effect to enhance and facilitate subsequent ploughing. This paper also dealt with the effect of rake angle on draught. Though the depth setup was getting smaller d 1 < d 2 < d 3 for the successive tillage runs, analysis showed increment in draught force ( ρ < 0.001) with rake angle. This might be attributed to higher soil compaction that comes with depth and downward force resulting from repeated use of maresha every season to the same depth for thousand years. Although more and rigorous studies should be undertaken considering soil, tool, and operational parameters to arrive at conclusive results, this paper gave some insights regarding effect of side-wings on maresha plough and rake angle on draught. This shows that there is still room for improvement of maresha plough geometry for minimum draught requirement and optimum soil manipulation. [ABSTRACT FROM AUTHOR]

• Spectral imaging proves effective in discriminating kernels from material other than kernel in freshly harvested maize forage. • A limited number of appropriate spectra can provide significant improvements in the detection of kernels in maize forage. • A Faster R-CNN model for multispectral images showed improved results compared to its performance on RGB images with the same resolution. In this study, the use of hyperspectral imaging was examined to enhance the kernel processing assessment of maize forage, which is crucial for optimizing the production of dairy cow feed and biogas production. As the visual contrast between the kernels and other crop particles is quite limited, spectral imaging could provide better kernel classification. A pushbroom hyperspectral imaging system was used, scanning samples collected during maize harvesting in the 400–1000 nm range. A PLSDA model for pixel classification was developed to distinguish kernel spectra from the other particles in the forage, achieving a pixel-level classification accuracy of 95.2 %. Next, the most important wavelengths were identified by means of a stepwise procedure using the Wilks Lambda criterion. A Pixel classification using the top five discriminating wavelengths achieved nearly the same accuracy as the full spectrum wavelength model (95.2 % compared to 93.5 %) and did considerably better than the RGB classifier's 86.3 % accuracy. Finally, these top 5 discriminating wavebands were applied in an object detection deep learning model, more specifically a Faster R-CNN model. While object detection based on these 5 wavebands stilled outperformed object detection based on RGB wavebands, detection performance, as measured by AP50, was rather low. This weak performance resulted from the low resolution of the hyperspectral imaging camera. [ABSTRACT FROM AUTHOR]

• Successful classification of flying insects from their optical signatures using machine learning and deep learning. • Practice oriented validation is key in developing insect classification models. • Random data splitting can lead to over-optimistic classification performance estimates. • Classification at the insect genus level is a more feasible option than on the species level. [ABSTRACT FROM AUTHOR]

A requirement for optimising crop processing machinery using DEM simulations is the application of virtual stems that behave realistically during deformation. In this study, data based bending models were developed for virtual segmented crop stems. These models combine realistic bending behaviour with a minimal number of model parameters. Also the effects of plastic deformation and damage were incorporated in the model. The bending model was successfully used to validate the bending behaviour of individual stems through comparison of simulations and validation measurements. It was also shown that the model is suitable for virtual stems with different numbers of segments. Moreover, based on a stem measurement it could be predicted what would happen to the same stem if it would have other dimensions or if it would be supported at different locations. Additional stem measurements were used to validate this. No significant difference ( α = 0.05 ) was observed between measurements and simulations. Finally, pendulum experiments showed that the deformation rate has no significant effect ( α = 0.05 ) on the deformation behaviour of individual crop stems. [ABSTRACT FROM AUTHOR]

The discrete element method (DEM) offers a powerful tool for simulating the interactions of large numbers of particles. Recently, DEM was used to simulate the interactions of tubular particles. While the existing linear elastic and Hertzian contact models can approximate the reversible compression for small deformations, they are inadequate for larger deformations. Here, the force–deformation behaviour is highly non-linear and plastic. In this study, data based contact models were developed for crop stems. These models combine realistic deformation behaviour with a minimal number of model parameters. Furthermore, the effect of plastic deformation and damage was incorporated in the model. The contact models were successfully used to validate, through comparison of simulations and measurements, individual stem and bulk compression. A good agreement was found between both. These validated DEM contact models for compression of crop stems allow to simulate the processing of large numbers of crop stems. [ABSTRACT FROM AUTHOR]

More efficient agricultural machinery is needed as agricultural areas become more limited and energy and labor costs increase. To increase their efficiency, trajectory tracking problem of an autonomous tractor, as an agricultural production machine, has been investigated in this study. As a widely used model-based approach, model predictive control is preferred in this paper to control the yaw dynamics of the tractor which can deal with the constraints on the states and the actuators in a system. The yaw dynamics is identified by using nonlinear least squares frequency domain system identification. The speed is controlled by a proportional–integral–derivative controller and a kinematic trajectory controller is used to calculate the desired speed and the desired yaw rate signals for the subsystems in order to minimize the tracking errors in both the longitudinal and transversal directions. The experimental results show the accuracy and the efficiency of the proposed control scheme in which the euclidean error is below 40 cm for time-based straight line trajectories and 60 cm for time-based curved line trajectories, respectively. [ABSTRACT FROM AUTHOR]

Thinning of pome and stone fruit involves the reduction of tree crop load in order to regulate fruit set and quality. As it is typically carried out through manual labor, thinning comprises a large part of a grower’s production costs. Mechanized thinning has been shown to be a cost-effective alternative but the performance of existing thinning devices needs to be further improved by taking the variation in bearing capacity of the individual trees into account. In this work, a multispectral camera system is developed to detect the floral buds of pear (cv. Conference ) during the growth stages prior to bloom. During a two-year field trial, the multispectral system was used to measure orchard scenes in six distinct optical wavebands under controlled illumination. These wavebands are situated in the visible and near infrared region of the spectrum and were selected based on hyperspectral laboratory measurements described in previous work. The recorded multispectral images were converted to a database containing the spatial–spectral signatures of the objects present in the orchard. Subsequently, canonical correlation analysis was applied to create a spectral discriminant model that detects pixels originating from floral buds. This model was then applied to the recorded data after which an image analysis algorithm was designed and optimized to predict the number of floral buds. In total, approximately 87% of the visible floral buds were detected correctly with a low false discovery rate (<16%). Therefore, it is expected that the multispectral sensor can be used to improve the efficiency of existing thinning devices. Additionally, it could as well be used as a stand-alone sensor for early-season yield estimation. [ABSTRACT FROM AUTHOR]

Monitoring the feeding pattern of a pig enables early detection of diseases and other problems. To monitor the individual feeding pattern of group-housed pigs, it has been suggested to equip the pigs with High Frequency Radio Frequency Identification (HF RFID) tags and the feeding trough with an antenna. The detection range of the HF RFID system is crucial to guarantee that all feeding pigs are detected without detecting the pigs located further from the feeder. The current study examines the factors that influence whether an antenna attached to a round feeding trough (such as those used in group housing of growing–finishing pigs) detects stationary HF RFID tags placed in various orientations and distances from the antenna. Four experiments were performed using a custom-built test set-up that allowed determining the RFID registrations for 70 tag positions, at seven distances from the antenna and for seven orientations of the tags in relation to the antenna. In the first experiment there was determined that which tag side is closest to the antenna had very little influence on the range of registration. The results of the second experiment revealed that all eight HF RFID antennas in the pig house performed similarly, with symmetry observed in their range of registration. In the third experiment the range of the HF RFID system was measured while accounting for tag, tag position and tag orientation, whilst the last experiment was designed to test the effect of interference between tags. Reproducibility between (the order of) the tags and the average agreement between five repetitions of all tests was very high. In total, the sensitivity was 51.0%, with a standard deviation of 43.1 percentage point (pp). The specificity was 87.1% with a standard deviation of 19.4 pp. It was concluded that the performance of the HF RFID system in terms of sensitivity and specificity of the range depends greatly on the height and orientation of the tags. This causes irregular gaps to appear between subsequent RFID registrations of a feeding pig. To improve the performance of the system in practice, it is suggested to adjust the height of the antenna to better match the size of the pigs and to develop algorithms and criteria to merge raw RFID registrations into relevant feeding variables for individual pigs. [ABSTRACT FROM AUTHOR]

This paper presents the nonlinear modeling of the yaw and longitudinal dynamics of a tractor--trailer system. First, the yaw dynamic models of both the tractor and trailer are derived considering the lateral forces and side-slip angles. In order to be able to calculate the side-slips precisely, the relaxation length approach is preferred. Since the obtained yaw dynamic models are nonlinear, a constrained nonlinear optimization problem is formulated for the parameter estimation. Second, the longitudinal dynamic model for the system is derived based-on the static and dynamic responses. The static model consist of two inputs, the hydrostat position and the diesel engine speed, and one output, the longitudinal speed of the system. Afterwards, a dynamic model is proposed to define the dynamic effect between the output of the static model and the actual longitudinal speed. Third, the mathematical models of the steering mechanisms both for the tractor and trailer are identified. Consequently, a complete nonlinear dynamic model for the tractor--trailer system is obtained. The overall resulting model is thought to provide useful physical insight on such a complex mechatronic system, and can serve as the input for model based controller design. [ABSTRACT FROM AUTHOR]

Highlights: [•] A construction of bendable straw particles in DEM is proposed. [•] The particles are calibrated with realistic straw properties. [•] DEM of grain–straw separation with these particles matches with experimental data. [•] A sensitivity study of grain–straw separation w.r.t. crop properties is performed. [•] Crop property influence can be related to straw coverage and grain kernel diameter. [Copyright &y& Elsevier]

Abstract: Combine harvesters play a key role in agriculture as threshable crops take up about two thirds of the worldwide arable land. From an economic perspective there is a continuous demand for increased productivity of these machines, but the limits for the outside dimensions have been reached and the increased processing efficiency should not compromise the quality of the end product. The reduction of the straw length in the threshing section of a combine harvester is considered to be an important indicator for the crop processing quality. An aggressive threshing action or discontinuous flow of the crop material invokes high impact and friction forces on the crop. This breaks up the straw, resulting in a swath of low volume which complicates the subsequent baling process. Therefore, the impact of the settings of the threshing process on the straw quality, expressed in terms of the swath volume has been assessed in this study. An experimental design involving different combinations of machine settings has been applied and for each of these setting combinations the resulting swaths have been scanned with a LiDaR (Light Detection and Ranging)-sensor. By a backward stepwise regression a quadratic relation between the concave clearance and rotor speed setting and the swath height was found (R 2 =0.79). All terms in the model were significant (p value of F-statistic <0.05). The observed relation between machine parameters and straw quality is interpreted as a combination of forces on the crop and residence time of the crop in the threshing and separation section. Insight in the interaction of crop and machine is gained that could be a useful input for optimizing the settings of the threshing and separation section. [Copyright &y& Elsevier]

• An online milk composition sensor is implemented on a farm and tested for 8 weeks. • Milk components are predicted more accurately when models are built post hoc. • Real-time prediction better reflects on-farm challenges for a milk analyzer. • The accuracy of the real-time prediction is well within the ICAR requirements. • A cow-specific bias correction further improved the milk lactose prediction. On-farm monitoring of milk composition can support close control of the udder and metabolic health of individual dairy cows. In previous studies, near-infrared (NIR) spectroscopy applied to milk analysis has proven useful for predicting the main components of raw milk (fat, protein, and lactose). In this contribution, we present and evaluate a precise tool for online milk composition analysis on the farm. For each milking, the online analyzer automatically collects and analyses a representative milk sample. The system acquires the NIR transmission spectra of the milk samples in the wavelength range from 960 to 1690 nm and performs a milk composition prediction afterward. Over a testing period of 8 weeks, the sensor collected 1165 NIR transmittance spectra of raw milk samples originating from 36 cows for which reference values were obtained for fat, protein, and lactose. For the same online sensor system, two calibration scenarios were evaluated: training post-hoc prediction models based on a representative set of calibration samples (n = 319) acquired over the entire testing period, with different cows in the calibration and test set, and training real-time prediction models exclusively on the samples acquired in the first week of the testing period (n = 308). The obtained prediction models were thoroughly tested on all the remaining samples not included in the calibration sets (n respectively 846 and 857). For the post-hoc prediction models, this resulted in an overall prediction error (root-mean-squared error of prediction, RMSEP) smaller than 0.080% (all % are in wt / wt) for milk fat (range 1.5–6.3%), protein (2.6–4.3%) and lactose (4–5.1%), with a coefficient of determination R2 of 0.989, 0.947 and 0.689 for fat, protein, and lactose respectively. For the real-time prediction models, the RMSEP was smaller than 0.092% for milk fat and lactose, and 0.110% for protein, with an R2 of 0.989 (fat), 0.894 (protein) and 0.644 (lactose). The milk lactose predictions could be further improved (RMSEP = 0.088%, R2 = 0.675) by taking into account a cow-specific bias. The presented online sensor system using the real-time prediction approach can thus be used for detailed and autonomous on-farm monitoring of milk composition after each individual milking, as its accuracy is well within the requirements by the International Committee for Animal Recording (ICAR) for on-farm milk analyzers and even meet the standards for laboratory analysis systems for fat and lactose. For this real-time prediction approach, a drift was observed in the predictions, especially for protein. Therefore, further research on the development of online calibration maintenance techniques is required to correct for this model drift and further improve the performance of this sensor system. [ABSTRACT FROM AUTHOR]

• A proximal sensing platform was built and evaluated for in-field hyperspectral imaging. • 750 nm, 550 nm and 680 nm were most discriminating for Alternaria solani detection. • Pixel based accuracy was 0.92, object based precision was 0.22 and recall 0.84. Automatic detection of early blight caused by Alternaria solani could promote a drastic reduction in the consumption of plant protection agents and the related production losses. A proximal sensing platform was constructed and calibrated for acquiring high resolution hyperspectral images in the field, and used to accurately map Alternaria lesions. High resolution canopy reflectance images were obtained for 32 potato plants that had been infected with A. solani and 32 healthy reference plants. Spectral classifiers like partial least squares discriminant analysis (PLS-DA) and support vector machines (SVM) based on PCA scores were tested to discriminate affected and non-affected pixels. Both spectral classifiers performed well at pixel level with accuracies above 0.92. The NIR region (750 nm) was identified as the most discriminative part of the spectrum for detecting the lesions. As the disease pressure is typically expressed as the number of lesions per area, the accuracy was also evaluated at this level. This indicated a considerable number of false detections at the edges of the leaves and the leaf axils. Therefore, a decision tree was designed based on expert knowledge about the shape of Alternaria lesions, and used to post-process the classified images. This reduced the number of false detections, increasing the precision from 0.17 to 0.22 at the expense of a reduction in recall from 0.88 to 0.84. This leaves considerable room for improvement in the classification accuracy at the object level. We learned that (1) few, broad wavelengths are sufficient and (2) spatial context is essential for the detection of lesions caused by Alternaria infection. The application of more powerful object classification techniques such as convolutional neural networks to enhance the model performance by efficiently encapsulating the spatial context in the classifier might further improve the detection performance. This could pave the way to UAV or tractor based Alternaria mapping. [ABSTRACT FROM AUTHOR]