Search

• This paper presents a Bayesian multi-task learning model for pest detection. • The model estimates size of aphids, which is an indicator of infestation severity. • It also quantifies the predictive uncertainties enabling confident decisions. • It performs thorough audit studies that offer insight for future model refinements. Aphids are persistent insect pests that severely impact agricultural productivity. The detection of aphid infestations is critical for mitigating their effects. This paper presents an artificial intelligence approach to detect aphids in crop images captured by consumer-grade RGB imaging cameras. In addition to detecting the presence of aphids, the size of the aphid is an important indicator of infestation severity. To address these, we present a Bayesian multi-task learning model to detect the presence of aphids and estimate their size simultaneously. Our model employs a joint loss function, combining a classification loss and a customised size loss. The classification component aims to identify images containing aphids, whilst the customised size loss function estimates the size of the aphids. The latter is specifically designed to account for discrepancies between the estimated and actual ground truth sizes, enhancing the accuracy of the size estimation. The model utilizes a ResNet18 backbone, ensuring robustness and adaptability across various conditions. The proposed model was evaluated using an agricultural pest dataset consisting of images of corn, rape, rice, and wheat crops. It achieved aphid presence detection accuracies of 75.77%, 66.39%, 70.01%, and 59% for corn, rape, rice, and wheat images, respectively. An in-depth evaluation of predictive uncertainties revealed areas of high confidence and potential inaccuracies for both size and presence of aphids in images, offering insight for future model refinement. We also conducted an ablation study to thoroughly analyse the contributions of each component in proposed model. Our model offers a valuable tool that can be used in pest management strategies for facilitating more sustainable and efficient agricultural practices. [ABSTRACT FROM AUTHOR]

• This paper establishes an image dataset of sika deer obtained from different perspectives. • This paper proposes an AD-YOLOv5 model to detect key parts of the sika deer. • This paper improves the loss function of the model to enhance both the training and inference ability of the model. • This paper proposes an adaptive channel attention mechanism for better extracting shallow features of the key parts of sika deer. • By altering the feature fusion network and combining it with the channel attention mechanism, the feature information is balanced across different scales, which enhances the accuracy of detection. Accurate and rapid fixation of key parts of sika deer is essential for the mechanized and intelligent collection of both deer blood and velvet antler. To solve the low detection accuracy problem of key parts of deer caused by the diversity of deer farm breeding environment and the rapid movement of deer, this paper proposes an AD-YOLOv5 algorithm for detecting key parts of sika deer based on YOLOv5s. Specifically, a new weighted bidirectional feature pyramid network S-BiFPN is first proposed, which both simplifies the structure and effectively performs skip feature fusion to better capture object information at different scales, so that the adaptive feature fusion capability of the model is optimized. Second, this paper proposes to add a SENet module after the C3 module of the 9th Layer in the backbone part of the network, so that the network can adaptively readjust the channel weights of feature maps to enhance the important features for key parts of sika deer and improve the expression ability of the model for key parts. Furthermore, to make the model more stable in the training process, a novel bounding box regression loss function SIoU is led into, which can better learn the position and size of the bounding box by fusing the orientation information between the ground truth box and the predicted box. The effectiveness of the proposed algorithm is demonstrated through ablation experiments. Experimental results show that the mAP @0.5 of the proposed algorithm reaches 97.30 % for the detection of key parts of sika deer, which is 12.85 %, 9.65 %, 9.18 %, 7.13 %, 3.59 %, 4.6 %, 0.90 % and 0.10 % higher than SSD, EfficientDet, YOLOv3, YOLOv4, Faster R-CNN, YOLOv5s, YOLOv7 and YOLOv8s, respectively. This research provides a new idea for detecting key parts of animals in complex environments and lays a theoretical foundation for the development of intelligent fixation devices for sika deer. [ABSTRACT FROM AUTHOR]

Crop counting is a crucial step in crop yield estimation. By counting, crop growth status can be accurately detected and adjusted, improving crop yield and quality. In recent years, with the rapid development of convolutional neural networks, deep learning-based object detection methods have been widely used in crop counting. By summarizing the research related to crop counting, this paper reviews the development status of object detection and crop counting. It then compares deep learning-based object detection counting methods with other counting methods. The paper also introduces public datasets and evaluation metrics commonly used for algorithmic models and provides a more in-depth analysis of the application of object detection in crop counting. Finally, the current problems that need to be solved, such as the lack of datasets, difficulties in small object counting, occlusion in complex environments, and some future directions are summarized. We hope this review will encourage more researchers to use deep-learning object detection methods in agriculture. • The paper compares the advantages and disadvantages of the methods currently used for crop counting tasks. • It is the first summary of object detection-based methods' application in crop counting tasks, covering commonly used models, issues, and challenges. • Public datasets utilized for crop counting tasks through object detection-based methods have been summarized. [ABSTRACT FROM AUTHOR]

Against the backdrop of a global labour shortage, the cost of agriculture has risen rapidly. Using robots to replace manual orchard maintenance tasks has attracted more attention. To avoid collisions between the robot and the tree canopy, a reliable vision system that can detect tree branches under natural occlusions is critical for robot navigation. In this paper, a regression deep learning based vision model, HOB-CNNv2, is proposed for the detection of continuous tree branches under natural occlusions in summer. The model is tested under two occlusion conditions, heavily occluded and extremely occluded. The experimental results show that HOB-CNNv2 can accurately detect tree branches in both occlusion conditions and outperforms the state-of-the-art semantic segmentation model, U-Net, in terms of branch position and thickness accuracy. To our best knowledge, this is the first vision model designed to deal with extremely occluded trees. In addition, we investigated five different methods using paired winter tree images to improve the performance of HOB-CNNv2. The results show that the best method is to use the winter and summer images of the same tree as two inputs to the neural network to extract features separately. This suggests that the model potentially acquires distinct knowledge or patterns from the winter and summer images. • Proposed HOB-CNNv2 to detect branches of occluded summer trees. • Investigated 5 approaches using winter data to improve summer tree branch detection. • Built an orchard database linking winter and summer images of each tree. [ABSTRACT FROM AUTHOR]

• The SY-Track algorithm combines improved YOLOv7-tiny and improved StrongSort. • The improved YOLOv7-tiny integrates SAGConv, ELAN-SA, and SIOU for efficiency. • The StrongSort with optimized Kalman filter is used for chicken flock tracking. • SY-Track calculates the Unrest Index to reflect the activity level of chickens. Poultry farming plays a vital role in providing animal protein for human consumption. Implementing efficient automation in chicken farming while ensuring the welfare and health of the birds is crucial. Monitoring changes in chicken activity can reflect their welfare and overall health, and provide a basis for agricultural production systems. Thus, maintaining appropriate levels of activity is vital for successful poultry farming. This paper proposes a method for detecting the activity levels of chicken flocks based on object detection and tracking. Firstly, a dataset for chicken flock detection and tracking was established. Secondly, an improved algorithm named SY-Track was proposed, leveraging optimized YOLOv7-tiny and improved StrongSort to ensure accurate chicken flock detection, tracking, and calculation of activity indices. In the optimized YOLOv7-tiny algorithm, a new lightweight convolution named Spatial Separable and Ghost Convolution (SAGConv) was proposed, complemented by the incorporation of the Efficient Long-range Spatial Aggregation Network (ELAN-SA). SAGConv effectively reduced both model parameters and computational complexity. The proposed ELAN-SA significantly improved the performance of YOLOv7-tiny. Additionally, the adoption of SIOU improved the convergence of bounding box regression. In the improved StrongSort algorithm, the optimized Kalman filter with predicted width and height was employed to enhance tracking performance. Compared to the original model, the improved YOLOv7-tiny reduced the model size by 48.75 %, increased the FPS by 44.15 %, decreased the computational load by 44.62 %, while achieving an AP50 of 96.30 %. The improved StrongSort algorithm achieved impressive MOTA values of 84.33 %, 79.00 %, and 71.62 % on three videos with different camera angles and flock disturbances. Finally, the SY-Track algorithm was utilized for calculating the Unrest Index to reflect activity level of chicken flock. The lightweighting SY-Track algorithm demonstrates superior competitiveness compared to mainstream detection and tracking algorithms when applied to the chicken flock dataset. It effectively calculates the Unrest Index of the chicken flock, providing a foundation for intelligent agricultural production and automation in chicken farming worldwide. [ABSTRACT FROM AUTHOR]

Automatic fruit blossom detection plays a crucial role in agricultural intelligence to predict fruit yield. Existing deep learning methods for vision-based fruit blossom detection oftentimes rely on large labeled samples and are tailor for a single category of fruit blossoms, limiting their flexibility and adaptability in real applications. In this paper, a fast and data-efficient framework is proposed to achieve multi-class fruit blossom detection using few labeled samples, which not only enhances the flexibility of model training by reducing the need for extensive samples but also extends the model's applicability to a wider range of blossom categories. Specifically, the proposed framework incorporates the paradigm of few-shot object detection into a lightweight two-stage detector named CenterNet2. To improve the recall of blossom proposals, we introduce a location guidance module (LGM) to highlight foreground regions in query images that bear resemblance to the blossom reference provided by support images. Moreover, a contrastive learning scheme (CLS) is introduced to further distinguish fruit blossom categories with similar appearances, enhancing classification accuracy. Experimental results demonstrated that our framework achieved comparable performance to existing state-of-the-art models, with a mean average precision of 74.33% over 12 categories of fruit blossoms and a frame per second rate of 47. In summary, our proposed framework can reduce data dependency and enhance flexibility of existing deep learning detectors, making it a promising solution for automatic monitoring in real-world agricultural applications. • The proposed framework can detect multi-class fruit blossoms with few labeled samples. • A location guidance module is introduced to guide the framework to produce more accurate blossom proposals. • A contrastive learning scheme is introduced to distinguish between different fruit blossoms with similar appearances. • The proposed framework achieves comparable performance to existing state-of-the-art models and faster inference speed. • The proposed framework can reduce data dependency and enhance flexibility of existing deep learning detectors. [ABSTRACT FROM AUTHOR]

With rich annotation information, object detection-based automated plant disease diagnosis systems (e.g., YOLO-based systems) often provide advantages over classification-based systems (e.g., EfficientNet-based), such as the ability to detect disease locations and superior classification performance. One drawback of these detection systems is dealing with unannotated healthy data with no real symptoms present. In practice, healthy plant data appear to be very similar to many disease data. Thus, those models often produce mis-detected boxes on healthy images. In addition, labeling new data for detection models is typically time-consuming. Hard-sample mining (HSM) is a common technique for re-training a model by using the mis-detected boxes as new training samples. However, blindly selecting an arbitrary amount of hard-sample for re-training will result in the degradation of diagnostic performance for other diseases due to the high similarity between disease and healthy data. In this paper, we propose a simple but effective training strategy called hard-sample re-mining (HSReM), which is designed to enhance the diagnostic performance of healthy data and simultaneously improve the performance of disease data by strategically selecting hard-sample training images at an appropriate level. Experiments based on two practical in-field eight-class cucumber and ten-class tomato datasets (42.7K and 35.6K images) show that our HSReM training strategy leads to a substantial improvement in the overall diagnostic performance on large-scale unseen data. Specifically, the object detection model trained using the HSReM strategy not only achieved superior results as compared to the classification-based state-of-the-art EfficientNetV2-Large model and the original object detection model, but also outperformed the model using the HSM strategy in multiple evaluation metrics. • Proposing the hard-sample re-mining (HSReM) training strategy for practical plant disease detection methods to enhance the fine-grained diagnosis performance when diseases and healthy data appear with high similarities. • HSReM strategically selects training samples that have a positive performance impact on both diseases and healthy, resulting in substantial improvement in diagnostic on practical large-scale cucumber and tomato datasets (42.7K and 35.6K images). • Detection models trained with the HSReM strategy show superior diagnostic performance compared to not only the classification-based EfficientNetV2-Large model but also other strong detection baselines. [ABSTRACT FROM AUTHOR]

• Constructed a dataset of early growth stages of clam seedlings under the microscope. • Four different attention mechanisms were employed in ablation experiments for discussion. • The K-means++ algorithm was used to enhance anchor accuracy, and Soft-NMS was applied to solve missed detections due to dense stacking of clam seedlings. • Heatmaps were constructed for visualization, analysis the effects of attention mechanism. • Software for automatic detection of clam seedlings biological features was developed. In order to monitor the early growth status of clam seedlings and meet the demands of precision seedling cultivation in a factory setting, this study improves the YOLOv5s model and proposes a high-precision and lightweight method for detecting biological features of clam seedlings. Firstly, dataset of clam seedlings under the microscope was constructed. Secondly, we established a clam seedling detection model based on four different attention mechanisms and analyzed the differences in the features focused on by different attention mechanisms in clam seedling detection. The accuracy of the anchors was optimized using the K-means++ algorithm, and the Soft-NMS method was employed to address the issue of missed detections caused by dense stacking of clam seedlings. The proposed approach has the optimal comprehensive performance by comparing various lightweight networks. The results show that the AP-health, AP-death and mAP of this approach are 98.15%, 93.87% and 96.01%, which can meet the standard for high precision. With a transmission speed of 86 FPS, an average response time of 11.54 ms, and a model size of 27.59 MB, the approach satisfies the requirements for efficiency and portability. Finally, a biological feature detection software of clam seedlings was developed, which can automatically calculate phenotypic information such as the number of clam seedlings, survival rate, and average size. This paper provides a basis for real-time and accurate assessment of clam seedlings biological features while offering technical support for the automation and intelligence of clam seedlings production in factories. [ABSTRACT FROM AUTHOR]

• Rotated bounding box detector significantly outperforms horizontal detector in densely-housed pigs. • Encoding the angle parameter with direction vector enables the learning of pigs' orientation for subsequent computer vision tasks. • A sequential non-maximum suppression to improve the performance of video object detection. • A super lightweight model achieved satisfying performance that can be deployed in edge device. Pig detection in real production environments is a challenging task due to the variations of housing system and dynamic background. Though considerable progress has been made, for practical settings, there still existing challenges for densely housed pigs as they are often arbitrarily arranged at varying orientations in presence of lens distortion, overlap, occlusion, and motion blur. In this paper, we propose a rotated and oriented bounding box detector for fast and accurate predict the location and orientation of each animal. The key point is to parameterize pigs' geometric parameters (body centre, body length, body width, orientation) with an orientated bounding box (box centre, long edge, short edge and direction vector). To further improve the performance on video object detection, a fast sequential non-Maximum Suppression (FastSeq-NMS) method is proposed by making used of the orientation and temporal information. To quantitative evaluate the proposed method, 3123 images from 27 different pens were selected as training and validation sets, video clips from three new environments were selected as test set. Our lightweight model (1.7 M) achieves 99.21 Average Precision (AP@0.5) on validation set, and 96.54 AP@0.5 on test set, further improved to 97.41 AP@0.5 with the proposed NMS method. The experiments show the effectiveness of the proposed method. More information available online: https://gitlab.kuleuven.be/m3-biores/public/m3pig. [ABSTRACT FROM AUTHOR]

Fusing color (RGB) images and range or depth (D) data in the form of RGB-D or multi-sensory setups is a relatively new but rapidly growing modality for many agricultural tasks. RGB-D data have potential to provide valuable information for many agricultural tasks that rely on perception, but collection of appropriate data and suitable ground truth information can be challenging and labor-intensive, and high-quality publicly available datasets are rare. This paper presents a survey of the existing RGB-D datasets available for agricultural robotics, and summarizes key trends and challenges in this research field. It evaluates the relative advantages of the commonly used sensors, and how the hardware can affect the characteristics of the data collected. It also analyzes the role of RGB-D data in the most common vision-based machine learning tasks applied to agricultural robotic operations: visual recognition, object detection, and semantic segmentation, and compares and contrasts methods that utilize 2-D and 3-D perceptual data. • Review of public colored point cloud datasets available in agri-robotics. • Overview of perception algorithms employing RGB-D datasets. • Outline of challenges and missing data in the literature. [ABSTRACT FROM AUTHOR]

• This paper proposes an automatic system of pest detection and counting in granaries. • The automatic system has 2 parts, a car and YOLOv4 model. The car with a camera and a supplementary light can take clear photos of the granary surface. And YOLOv4 model can process photos synchronously and achieve high accuracy in pest detection and counting. • The system solves the randomness and insufficient accuracy of pest traps and human eye recognition. And it can be used to the early warning of pests in granary, which has high accuracy and minimal manual effort. Stored-grain pests cause serious economic losses during grain storage. Therefore, it is important for us to know the accurate number and species of stored grain pests as soon as possible so that appropriate measures can be taken to reduce economic losses. However, current research on grain pest detection has two problems. The background of the dataset does not contain any grain, so the results cannot be applied to detect pests in actual situations. The other problem is that methods based on pest traps cannot reflect situations on the surface of the whole granary. This paper proposes an automatic system of pest detection and counting to solve these problems. This system consists of two main parts: a deep learning object detection model called YOLOv4 and a small car with a camera and a supplementary light. And YOLOv4 model that has been trained is embedded in the car. The car can run on the surface of the granary. The camera on the small car can take photos of the pests for YOLOv4 to identify the species and number of pests. In the experiment, a few hundred kilograms of wheat were used to lay a simulated granary. And 2 typical stored-wheat pests, namely, the red flour beetle and the rice weevil, were taken as the research objects. The experimental results demonstrated that the mean average precision (mAP) of the proposed method reached 97.55%, which can meet the accuracy requirements of the detection and counting of pests in the granary in practical applications. The system solves the randomness and insufficient accuracy of pest traps and human eye recognition. And the system can be used to the early warning of pests in granary, which has high accuracy and completely release manual labor. [ABSTRACT FROM AUTHOR]

• A novel deep learning model is proposed for critical emerging frost detection. • The detection accuracy has been improved and real-time is satisfied. • It provides technical support for other researches related to plant frost detection. Natural radiation frost is a potential threat or even a serious disaster during the growth of agricultural crops. To avoid or reduce the risk of frost damage to crops, frost detection especially the timely and real-time detection of emerging frost crystals, is of great importance in agricultural frost protection. Therefore, this paper proposes an object detection model Frost-YOLOv5 based on microscopic imaging of emerging frost crystals on the leaf surface. Firstly, a YOLOv5 initial detection model was established by observing frost under natural conditions and using microscopic cameras to acquire images of emerging frost crystals formed on the leaf surface. From these images, the channel attention mechanism and spatial attention mechanism were combined to construct a small frost crystal feature depth extraction network. Secondly, the RGB three-channel Wiener filtering algorithm was used to recover the noise source images from motion blur in the frost crystal dataset. Through experimental validation on different light intensities and different frost crystal cluster datasets, the Frost-YOLOv5 model all showed better robustness compared with various object detection algorithms, and its frost crystal detection accuracy P and average accuracy AP values were 98.89% and 81.72%, respectively. Ultimately, the Frost-YOLOv5 model can effectively detect emerging frost crystal targets and provide theoretical principles and technical support for agrometeorological monitoring research. [ABSTRACT FROM AUTHOR]

[Display omitted] • Deep learning automates precise cereal plant head detection in agriculture. • Maize, rice, wheat, and sorghum case studies evaluated. • Comparison of deep learning approaches for object detection and image segmentation. • Challenges and opportunities for future research outlined. • Emphasis on how deep learning can contribute to increasing crop yields and food security. Cereal plant heads must be identified precisely and effectively in a range of agricultural applications, including yield estimation, disease detection, and breeding. Traditional methods that rely on manual feature extraction and thresholding take a lot of time and work, and they are also impacted by crop variability. Deep learning algorithms can be used to automate this procedure because they can directly extract complicated information from images and produce cutting-edge outcomes. This review provides a comprehensive overview of recent research on deep learning-based head detection in cereal plants, emphasizing object detection and image segmentation. We also discuss the major benefits and drawbacks of different deep learning architectures and training methods, as well as examples of their application in maize, rice, wheat, and sorghum. Developing robust image processing algorithms, using deep learning in other domains like unmanned aerial vehicles, and utilizing large and diverse datasets are all challenges outlined in our study as future research directions. Through the integration of advanced computer vision techniques with precision agriculture, this paper attempts to promote further research and innovation in this intriguing field. We provide a thorough analysis of current developments in deep learning-based head detection for cereal plants and emphasize how this technology can contribute significantly to precision agriculture. [ABSTRACT FROM AUTHOR]

• Proposing an end-to-end noise-tolerant feature fusion network to integrate outdoor RGB-D information. • Designing an attention-based fusion module to eliminate the negative influence of depth noise and make full use of the complementary multi-modal features. • Learning object position from multi-scale features from color images and fusion modules, further improving the network noise immunity. • Achieving accurate fruit detection in outdoor scenes with a real-time speed. In the process of farm automation, fruit detection is the basis and guarantee for yield prediction, automatic picking, and other orchard operations. RGB images can only obtain the two-dimensional information of the scene, which is not sufficient to effectively distinguish fruits that are dense growth and occlusion by branches and leaves. With the development of depth sensors, using RGB-D images with more complementary information can boost the performance of fruit detection. However, due to the nature of sensors and scene configurations, the quality of outdoor depth images is poor, posing a challenge when fusing RGB-D features. Therefore, this paper proposes an end-to-end RGB-D object detection network, termed as noise-tolerant feature fusion network (NT-FFN), to utilize the outdoor multi-modal data properly and improve the detection accuracy. Specifically, the NT-FFN first uses two structurally identical feature extractors to extract single-modal (color and depth) features, which is the base of the subsequent feature fusion. Then, to avoid introducing too much depth noise and focus the perception on the important part of the features, an attention-based fusion module is designed to adaptively fuse the multi-modal features. Finally, multi-scale features from the color images and the fusion modules are used to predict object position, which not only improves the network's ability to detect multi-scale objects but also further enhances the noise immunity of the network. In addition, this paper constructs an RGB-D citrus fruit dataset, which contributes to comprehensively evaluating the proposed network. Evaluation metrics on the dataset show that the NT-FFN achieves an AP50 of 95.4% with a real-time speed, which outperforms single-modal methods, common multi-modal fusion strategies, and advanced multi-modal detection methods. The proposed NT-FFN also achieves excellent detection results in other fruit detection tasks, which verifies its generalization ability. This study provides the possibility and foundation for performing multi-modal information fusion in outdoor fruit detection. [ABSTRACT FROM AUTHOR]

• An image restoration method based on convolutional autoencoder is proposed. • Use the area growth method to divide and code the obstruction according to the contour. • Predict the overall outline of citrus based on the unobstructed citrus outline. • Fill the area to be repaired with the selection method of similar pixels. • The repair rate and image recognition rate have been effectively improved. At present, machine vision and deep learning theories have been widely used in fruit recognition and picking. However, in the process of identification and picking, there are often situations where the target is blocked, existing methods cannot accurately identify, or the identification accuracy rate is low. For improving the recognition rate of fruits under occlusion, this paper proposes an image restoration method based on a convolutional auto-encoder. This method first encodes the obstruction and then determines the general shape of the occluded fruit and compares the general shape with the coding part is combined to determine the area to be repaired. Finally, the area to be repaired is filled with pixels to realize the image repair and recognition of the occlusion map. The average repair rate of the method proposed in this paper is 95.96%, the restoration rate is 3.52% higher than the traditional convolution method, the L2 loss value is 0.63% lower than the traditional convolution method, the average detection accuracy of the restored fruits is 94.77%. [ABSTRACT FROM AUTHOR]

• This paper reviews the research status of typical machine-vision-based weeding robots. • This paper has evaluated the performance of eight baseline methods that are based on deep learning on a public dataset. • This paper presents the application of object detection algorithms based on deep learning to weed identification. Due to its obvious advantages in saving labor and pesticides, weeding robots are one of the key technologies for modern and sustainable agriculture and have attracted increasing attention from researchers and developers. Some papers on machine-vision-based weeding robots have been published in recent years, yet there is no clear attempt to systematically study these papers to discuss the components of a robotic weed control system, such as visual navigation, weed detection and directional weeding. In this paper, typical machine-vision-based weeding robots proposed or constructed in the last 30 years, together with a few open datasets for weed detection, are reviewed. Key technologies such as image preprocessing, image segmentation, navigation line extraction, and weed recognition based on machine learning (ML) or deep learning (DL) for weeding robots are discussed. To illustrate the application of DL algorithms to weed detection, this paper provides weed object detection results and a comparative analysis of eight baseline methods based on DL using a public dataset. The study found that there are still many issues that need to be addressed in each part of the robotic weeding control system. Because of environmental variation and system complexity, machine-vision-based weeding robots are still in their early stages. The results of the systematic review provide an understanding of innovative trends in the use of machine vision in weeding systems and references for future research on weeding robots. [ABSTRACT FROM AUTHOR]

In recent years, Light Detection and Ranging (LiDAR) technology has been one of the most innovative subjects in laser scanning, remote sensing, and object detection systems. This technology may be popular because it can pinpoint structures or zones of interest in millimetre detail. It can also highlight variations and irregularities, such as surface degradation and vegetation growth. This paper presents a review of the specialised literature on LiDAR systems applied to precision agriculture; specifically, in cultivating crops. First, some preliminaries of LiDAR systems according to the mode of transport used, considering terrestrial, mobile, and aerial laser scanners, are given. Subsequently, a well-organised taxonomy of recent LiDAR applications based on the activity being performed is presented. Here, the following four categories are considered: (1) crop-related metric estimation, (2) tree and plant digitisation, (3) vision systems for object detection and navigation, and (4) planning and decision support. Lastly, we discuss some current trends and research challenges in applying LiDAR technology to cultivation activities in accordance with the state-of-the-art literature. • Compared with photogrammetry, LiDAR technology often obtains more precise 3D models. • Voxelisation is reasonably accurate when applied to tree digitisation. • The crop's health status is often monitored by estimating the NDVI, LAI or height. • The sunlight index and wood volume are adequate for evaluating pruning structures. • The most used LiDAR sensors are VLP-16, LMS400/111/511, UTM-30LX, and Focus X330. [ABSTRACT FROM AUTHOR]

• From 2017 to 2022, the area and number of PGs in Zhoukou increased by 41.38% and 47.08%, respectively. • A model YXLD + was proposed that can capture both the area and the quantity of PGs. • New CSPLayer method can fully extract the spatial feature of images. • A multi-temporal greenhouse map based on VHR remote sensing imagery was plotted. • The spatial distribution of PGs was mainly affected by the slope. Plastic greenhouses (PGs), as safeguard agricultural facilities, have changed the cultivation structure of agriculture, to a certain extent, and can seriously impact the environment. Exploring the temporal characteristics of greenhouses in terms of quantity and area is significant for sustainable agricultural development and environmental protection. Pixel or object-based large-scale greenhouse mapping has difficulty meeting the demand for the simultaneous extraction of greenhouse quantity and area. Multibranch task model can provide the number and area information of greenhouse. However, there are various labeling methods and model between different branches, which lead to the deviation between quantity and area results. This paper proposes a model that is based on improved you only look once X (YOLOX) and kullback–leibler divergence (KLD), called YOLOX and KLD plus (YXLD +), to obtain spatiotemporally consistent greenhouse mapping. The new model still has stable performance in dense scenes, with a 5.7 % improvement in average precision compared to BBAVectors. Finally, a multitemporal greenhouse map of Zhoukou City was obtained using imagery provided by Google Earth (GE). Overall, the distribution of greenhouses in Zhoukou City remained stable between 2012 and 2017. However, the number and area of greenhouses in local areas has increased rapidly in the past five years, with the number of greenhouses increasing by 47.08 % and the area expanding by 41.38 %. Greenhouses' distribution and changing characteristics concentrate on the gently sloping areas from the northwest to the southeast. These results provide an objective perspective on the development of the local agricultural cultivation structure, which is of significant importance for the promotion of agricultural modernization. [ABSTRACT FROM AUTHOR]

• A detection approach for late-autumn shoots of litchi tree based on UAV remote sensing is provided. • A remote sensing dataset of late-autumn shoots of litchi tree is constructed. • A deep learning model based on the combination of YOLOv5 and Swin-transformer is proposed. • The detection accuracy of late-autumn shoots in remote sensing images is improved. • It will provide technical support for other researchers on late-autumn shoots and orchards manager. Litchi is one of the most common economic fruits in southern China, however, the growth of late-autumn shoots of litchi hinders flower bud differentiation and reduces yield of fruit. The early identification of the late-autumn shoots is of great significance for orchard management to control shoots and then increase fruit yield. At present, the identification of late-autumn shoots still relies on manual methods, which is not suitable for smart orchard management in a large area due to low recognition efficiency and high subjectivity. Therefore, a convenient, fast and cost-effective method is urgently needed. In response to this problem, the paper proposes a method based on the combination of unmanned aerial vehicle (UAV) remote sensing and object detection algorithm to detect late-autumn shoots. For this purpose, a remote sensing dataset of late-autumn shoots of litchi is first constructed by UAV. An improved YOLOv5 algorithm called YOLOv5-SBiC is then developed for late-autumn shoots identification. In the YOLOv5-SBiC algorithm, the transformer module is introduced to speed up the convergence of the network and improve detection accuracy, the attention mechanism module is employed to help the model extracting details, and BiFPN is used to better solve the multi-scale problem in detecting and then improve the recognition effect of small-sized objects. In addition, CIOU is selected as the loss function of bounding boxes regression to achieve high-precision localization of the boxes. The test results demonstrate that the recognition accuracy of YOLOv5-SBiC reaches a relatively high value of 79.6%, which is 4.0% higher than that (75.6%) of the original YOLOv5 algorithm and 15.9% higher than that (63.7%) of the pure transformer algorithm. It's also demonstrated that YOLOv5-SBiC is more competitive than the mainstream target detection algorithms in the dataset of late-autumn shoots. [ABSTRACT FROM AUTHOR]

• Maize tassel pose estimation method for precise maize detasseling. • Leaf vein pose matching with oriented bounding box pose. • Look Twice method for pose information extraction. • Lightweight model for maize tassel pose estimation. The maize detasseling process was gradually automated with the continuous promotion of the maize detasseling machine. Nevertheless, some problems are gradually becoming more prominent for it. Most maize detasseling machine has a simple maize tassels detection system. That is, the height of the crop was identified, and not the specific information on the maize tassels, leading to low detasseling precision and a high leaf injury rate during maize detasseling machine working. Aiming at these issues, this paper proposed a novel method for the maize tassel pose estimation based on computer vision and oriented object detection. Specifically, a two-step framework for maize tassel pose estimation is developed. Firstly, the maize plant is captured from above, then the maize tassels and the second leaf's vein are posed in the horizontal plane, and their poses are matched to the oriented bounding box. Second, an oriented bounding box is generated according to the Oriented R-CNN model detection. Since maize tassels in the field differ in size and the morphological color of different growth stages, estimating the maize tassel pose accurately by the angle of the oriented bounding box alone is unfeasible. Therefore, these pixels of the oriented bounding box were put into the Look Twice module to extract the critical information about maize tassels, which can accurately determine the final pose of the maize tassels. Finally, evaluation metrics on the test set indicate the proposed method performed with correct maize tassels pose estimation rate of 88.56% and 29.57 Giga Floating-point Operations (GFLOPs), which indicated that the feasibility of maize tassel poses estimation using the proposed method. This study provides the possibility and foundation for precise maize detasseling in maize detasseling machines. [ABSTRACT FROM AUTHOR]

• The algorithm for the detection of weeds in paddy field was proposed. • A dataset containing rice and weeds was established. • The method for rice crop localization algorithm was proposed. • The proposed model achieves a high accuracy of 94.1% and a high speed of 24.3 fps. Weeds are one of the main hazards affecting the yield and quality of rice. In farmland ecosystem, weeds compete with rice for resources such as light, water, soil and space, and also cause plant disease by providing habitats for pests, resulting in a decline in yield of rice. Spraying herbicides is used most frequently due to their lower cost than manual weeding and they can effectively control a varieties of weeds. Precise spraying application may utilise intelligent technology to precisely limit the weed growth at specific locations and control the consumption of herbicides, which not only has lower cost and risk for the environment but also increases the economic benefits of agricultural products. The target of this work is to use deep learning model to perform weed detection in rice crop images and to achieve accurate real-time detection and low machine cost resulting in widely used in practice. For this purpose, a dataset of rice and weeds for object detection is established by means of on-site shooting and crawling of images on the web, which contains rice and eight categories of weeds. However, the overlap between crops and weeds poses a great challenge to weed detection. To overcome this challenge, this paper proposed a model named WeedDet based on RetinaNet, which improves the feature extraction ability of the backbone, feature pyramid network and detection head to deal with the complex information in the image, respectively. And the structure of feature pyramid and detection head are lightened to improve speed of detection. The specific implementation methods are as follows: Firstly, we propose Det-ResNet to reduce the loss of detailed information and improve the feature extraction ability of detailed textures by modifying the initial convolution structure of ResNet. In addition, the improved feature pyramid network fuses the features extracted by Det-ResNet with higher efficiency without loss of accuracy. Secondly, we propose Efficient Retina Head (ERetina-Head) with thin feature maps (small channel number feature maps) and large separable convolution, which not only saves memory and computation, but also pooling more powerful feature maps during training and inference. Finally, we combine Smooth L 1 loss and GIoU loss to calculate the regression loss. In addition, Varifocal loss is used to unify the classification and quality evaluation branches in training. Our network achieves a high mAP of 94.1% and the frame rate of 24.3 fps, which is 5.5% mAP and 5.6 fps higher than the baseline RetinaNet. Experiments show that fps of WeedDet (24.3) is second only to YOLOv3 (26.3) while the mAP is 9.9% (0.941 vs 0.842) higher than YOLOv3, and the mAP is also higher than other models, verifying the effectiveness and efficiency of the model. [ABSTRACT FROM AUTHOR]

• A novel deep learning strategy for very small objects based on YOLOv4 backbones. • An image mosaicking and a fusion of hypothesized objects to classify whiteflies. • An annotated database of soybean leaf images with whiteflies and development stages. • Detection and classification results, with comparison with a state-of-art CNN. This paper presents a novel strategy to detect and classify adult whiteflies and five important related stages on images of detached soybean leaves. The whitefly Bemisia tabaci is a major pest in soybean crops, and by detecting, counting and differentiating its related life stages in field collected leaves control management decisions can be made. The proposed solution is based on a deep learning object detection algorithm (YOLOv4), modified into an specific new learning strategy with innovations on data augmentation, image mosaicking, and fusion of hypothesized object categories. A real and annotated dataset of images is provided from a controlled experiment infected with whitefly eggs having 121 images and 973 annotated objects. The experimental results showed a promising performance of the proposed system, reaching an f1-score of 0.87, in comparison with a single YOLOv4 algorithm that reached f1-score of 0.80. The overall strategy could be extended to work in other similar tasks for image based pest management. [ABSTRACT FROM AUTHOR]

• Complete and accurate counting of fruit on the tree. • High performance single stage deep Convolution Neural Network YOLOX used. • Superior detection rate (≥99% when R ≤ 1.2 m) was obtained. • Excellent robustness in complex orchard scenarios. Fruits counting is important in management of orchard and plantation since better decision for labor and logistic can be made based on complete and accurate counting of fruits. Computer vision-based fruits counting has been research focus as it's an automatic way for recognition of dense fruit on the branch. However, complete fruits counting of a whole tree hasn't hardly been studied. And there is a lack of robust and accurate fruits counting method in complex orchard scenarios, like covering, shadow, clustering in image. In this paper, a panoramic method for fruit complete yield counting based on deep learning object detection is proposed, and was validated on a holly tree with dense fruits. Firstly, images were taken surrounding the fruit trees using UAV, and SIFT based images matching were performed to form a complete panoramic unfolding map of the fruit tree surface. Then, a YOLOX object detection network was built and trained with novel samples augmentation and composition strategies. Finally, fruits counting YOLOX was performed on the panorama to count the whole plant fruits number. The accuracy and effectiveness of this method were tested at different scales and scenarios. The results show that: (1) high-quality panoramic images can be built for an accurate fruit number counting. (2) The Statistical Rate (SR) between detected number and actual number is as high as SR > 96% when the ring shot parameter of Holly tree is R ≤ 1.2 m, SR > 95% when R ≤ 1.6 m. The Detection Rate between detected number and captured number in the panorama image is over 99% when R ≤ 1.2 m and over 97% when R ≤ 2.0 m. The result is superior to previous researches. (3) it has good robustness against shading, covering, incomplete contour. Comparisons between the proposed method and other methods has been done and the result show the proposed method is the most effective in fruits counting. Moreover, we proposed and verified the positive effects of Gaussian convolution kernel and γ-component control on fruit detection rate. The YOLOX-based fruit counting method can be extended to a wide range of fruits, like apples, lychee and so. Moreover, YOLOX has excellent inferencing efficiency which makes it a good potential for real-time application in orchard and plantation management. [ABSTRACT FROM AUTHOR]

• Proposing a robust and effective DSE-YOLO network. • Proposing a detail-semantics extraction module for feature enhancement. • Proposing enhanced loss functions for class imbalance. • Validation of an effective framework for multi-stage strawberry fruits detection. Multi-stage strawberry fruits detection is one of the important clues to estimate crop yields and assist robotic picking in modern agricultural production. However, it is difficult for detecting strawberries due to their small size, foreground-foreground class imbalance, and complex natural environment. Many works focus on how to detect fruits while ignoring multi-stage fruit detecting problems. In this paper, we propose DSE-YOLO (Detail-Semantics Enhancement You Only Look Once) to detect multi-stage strawberries. In DSE-YOLO, DSE (Detail-Semantics Enhancement) module is designed for detecting small fruits and distinguishing different stages of the fruit with higher accuracy, which utilize pointwise convolution and dilated convolution to extract various detail and semantics features in the horizontal and vertical dimensions. Exponentially Enhanced Binary Cross Entropy (EBCE) and Double Enhanced Mean Square Error (DEMSE) loss function are constructed to focus on small fruits, which can deal with foreground-foreground class imbalance problem. Experiments conducted on datasets demonstrate the superiority of DSE-YOLO over state-of-the-arts. The detection results had a mAP value of 86.58% and an F 1 -Score value of 81.59%, which demonstrates the effectiveness of the proposed model. Especially, DSE-YOLO can almost detect every stage of strawberry fruit accurately in the natural scene, which can provide an important theoretical basis and premise for automatic picking and monitoring system. [ABSTRACT FROM AUTHOR]

• A novel object detection network model is proposed for forest fire smoke detection. • Build a UAV-IoT system to capture remote sensing images to monitor forest fires. • The proposed method can achieve precise positioning of forest fire smoke. The occurrence of forest fires can lead to ecological damage, property loss, and human casualties. Current forest fire smoke detection methods do not sufficiently consider the characteristics of smoke with high transparency and no clear edges and have low detection accuracy, which cannot meet the needs of complex aerial forest fire smoke detection tasks. In this paper, we propose Adjacent layer composite network based on a recursive feature pyramid with deconvolution and dilated convolution and global optimal nonmaximum suppression (ARGNet) for high-accuracy detection of forest fire smoke. First, the Adjacent layer composite network is proposed to enhance the extraction of smoke features with high transparency and no clear edges, and SoftPool in it is used to retain more feature information of smoke. Then, a recursive feature pyramid with deconvolution and dilated convolution (RDDFPN) is proposed to fuse shallow visual features and deep semantic information in the channel dimension to improve the accuracy of long-range aerial smoke detection. Finally, global optimal nonmaximum suppression (GO-NMS) sets the objective function to globally optimize the selection of anchor frames to adapt to the aerial photography of multiple smoke locations in forest fire scenes. The experimental results show that the ARGNet parametric number on the UAV-IoT platform is as low as 53.48 M, mAP reaches 79.03%, mAP50 reaches 90.26%, mAP75 reaches 82.35%, FPS reaches 122.5, and GFLOPs reaches 55.78. Compared with other mainstream methods, it has the advantages of real-time detection and high accuracy. [ABSTRACT FROM AUTHOR]

• Machine-vision system for animal identification. • Deep learning algorithms were used for sheep identification. • ArcFace loss function used for animal identification. • Sheep identification based on facial images. This paper describes a sheep biometric identification system based on facial images. A machine vision system and deep learning model were developed and applied for animal identification. The system included two 8-MegaPixels cameras installed in a controlled water trough adapted to work with NVIDIA Jetson Nano-embedded system-on-module (SoM). Data from 81 Assaf breed sheep, aged two to three months, from two different groups of sheep, were collected over a period of two weeks. The biometric identification model included two steps: face detection and classification. In order to locate and localize the sheep's face in an image, the Faster R-CNN deep learning object detection algorithm was applied. The detected face was provided as input to seven different classification models. Different transfer learning methods were examined. The best performance was obtained using a ResNet50V2 model with the state-of-art ArcFace loss function. The identification system resulted in average accuracies of 95% for the two groups tested. When applying transfer learning methods, average identification accuracies improved to 97% in both groups, and the training process was accomplished in half the time. The newly developed system proves the feasibility of individual biometric identification of sheep on commercial farms. [ABSTRACT FROM AUTHOR]

• Mechanical weeding module for precise hoeing in vegetable farming. • Real-time precise stem detection in RGB images on embedded computer. • Deep learning object detection based neural network for precise stem location. • Tracking algorithm for temporal aggregation of stems in successive frames. • F1-score of 94.74% and 93.82% for maize and bean with sub-centimetric location accuracy. Developing alternatives to the chemical weeding process usually carried out in vegetable crop farming is necessary in order to reach a more sustainable agriculture. However, a precise mechanical weeding requires specific sensors and advanced computer vision algorithms to process crop and weed discrimination in real-time. In this paper we propose an algorithm able to detect, locate, and track the stem position of crops in images which is suitable for precision actions in vegetable fields such as mechanical hoeing within crop rows. The algorithm is twofold: (i) a deep neural network for object detection is first used to detect crop stems in individual RGB images and then (ii) an aggregation algorithm further refines the detections taking advantage of the temporal redundancy in consecutive frames. We evaluated the pipeline on images of maize and bean crops at an early stage of development, acquired in field conditions with a camera embedded in an experimental mechanical weeding system. We reported F1-scores of respectively 94.74% and 93.82% with a location accuracy around 0.7 cm when compared with human annotation. Moreover, this pipeline can operate in real-time on an embedded computer consuming as little power as 30 W. [ABSTRACT FROM AUTHOR]

• Automatic date palm tree detection algorithm in drone imagery. • Detecting crowded date palm trees using YOLO-V5. • Comparison of one-stage object detection methods for date palm tree detection. Date palm trees are an important economic crop in the Arabian Peninsula, Middle East, and North Africa. Counting the numbers and determining the locations of date palm trees are important for predicting the date production and plantation management. In this paper, we exploit the effective use of the state-of-the-art CNN, YOLO-V5, in detecting date palm trees in images captured by a camera onboard of a drone flying 122 m above farmlands in the Northern Emirates of the United Arab Emirates (UAE). In the dataset preparation process, we randomly selected 125 captured images and divided them into three datasets: training (60%), validation (20%), and testing (20%). The images of date palm trees in the training and validation datasets were manually annotated and those in the training dataset were used to train the four sub-versions of YOLO-V5 CNNs. The validation dataset was used during the training process to assess how well the network was performing during training. Finally, the images in the test dataset were used to evaluate the performance of the trained models. The results of using YOLO-V5 for date palm tree detection in drone imagery are compared with those obtainable with other popular CNN architectures, YOLO-V3, YOLO-V4, and SSD300, both quantitatively and qualitatively. The results show that for the amount of training data used, YOLO-V5m (medium depth) model records the highest accuracy, resulting in a mean average precision of 92.34%. Further it provides the ability to detect and localize date palm trees of different sizes, in crowded, overlapped environments and areas where the date palm tree distribution is sparse. Therefore, it is concluded that the method can be a useful component of an automated plantation management system and help forecast the quantities of date production and condition monitoring of the date palm trees. [ABSTRACT FROM AUTHOR]

• An improved active obstacle separation method for strawberry-harvesting robots. • A faster vision system that used deep learning for object detection and tracking. • The method combined push and drag-push to precisely separate the obstacles. • The separation was controlled in a hybrid visual loop based on the vision system. • The closed visual loop and continuous "look-and-move" made the system more precise. Selectively picking a target fruit surrounded by obstacles remains a challenge for fruit harvesting robots. This paper presents improvements to the active obstacle separation method for strawberry picking in clusters. A faster and more accurate vision system was developed that combined two neural networks and color thresholding for real-time detection, tracking and localization of strawberries. We propose an improved active obstacle separation method that used a push and a drag-push operation to separate the obstacles from the target in three stages. The push and drag vectors were simplified and precisely calculated based on the exact locations of obstacles. Also, different from many systems that only "looked" once for the entire picking process, the new system used a hybrid vision-based control method. In stage 1, the push operation was controlled by a simple closed-loop vision at two key points. In stages 2 and 3, the vision system re-perceived the environment to update the target and obstacle information before each round of drag-push movements. Field evaluation showed that the proposed method was more precise to separate the obstacles without reducing the speed, increasing the whole process success rate to 62.4% in clusters on the "Murano" strawberry cultivator that was 36.8% higher than the previous work. [ABSTRACT FROM AUTHOR]

• An image synthesis and semi-supervised learning pipeline is proposed. • The pipeline constructs an image library of plant instances with pixelwise masks. • Images can be synthesized using the library for site-specific weed detection. • Without manual labels, the pipeline achieves precision close to supervised learning. • Ablation study provides insights into the construction of a weed image database. Precise and efficient weed detection in agricultural fields is the key for robotic weed control. Recent developments in convolutional neural networks (CNNs) have achieved significant success in this regard. CNNs that simultaneously localize and classify objects in images are the predominant forms that have been widely used to detect crops and weeds. However, the use of CNNs in agriculture, particularly for weed detection, has been impeded by a lack of large training dataset with ground truth annotations. Cut-and-paste image synthesis approach and semi-supervised learning are popular methods to alleviate the training data deficiency problem. In this paper, we propose a novel image synthesis and semi-supervised learning pipeline to train site-specific weed detection models without the need for manually labeled images. The CNN models trained by this pipeline achieve performance levels close to that of the supervised models. We investigated the behavior of the proposed pipeline by varying several key components and showed that color match between the training and testing images, training-time color augmentation, and iterative semi-supervised learning largely improve the model performance. These promising results can be used to guide the construction of a weed image database applicable to different weed detection scenarios. [ABSTRACT FROM AUTHOR]

• Object detection benchmark on seven crop types by three leading networks. • An algorithmic method for handling scale problems using a multi resolution image pyramid. • Comparison of detector and human accuracy. • Highlight and quantification of the main causes for detection errors: object scale and partial occlusion. • Suggestion of additional detection metrics for the agricultural context. Visual object detection is an important component in several applications of automated agriculture. In this paper we consider how to properly apply modern deep networks for detection tasks in agricultural contexts, benchmark their performance, and compare their accuracy to human performance. Seven diverse datasets were collected for the benchmark, with three recent networks tested. Experiments have revealed that handling small objects and large scale variance are important failure points, and hence a multiple-resolution approach for network usage was developed, which significantly increased detection accuracy on most datasets. Detection results were compared to human accuracy, judged based on the consistency of multiple annotators. Quantitative analysis shows that for large unoccluded objects accuracy of both algorithms and humans is near perfect, and quantifies the degradation of both due to occlusion and scale difficulties. Finally, application-specific accuracy metrics were suggested based on the needs of several agricultural tasks, and used for estimating the best performing detectors. [ABSTRACT FROM AUTHOR]

• Graph based method to detect and remove duplicated cattle using multiple UAV's images. • Competitive results with significant reduction of runtime. • Novel benchmark image collection for cattle detection and counting in large areas. Among the production areas with largest impact on global economy, agriculture and livestock play a prominent role. Technologies have been developed in order to automate and increase the efficiency of these fields. The use of Unmanned Aerial Vehicles (UAVs) has been extensively investigated to improve the efficiency of agricultural production and in the monitoring of animals. One of the most important and challenging tasks in animal monitoring is cattle counting. In this paper, we propose a method for detecting and counting cattle in aerial images obtained by UAVs, based on Convolutional Neural Networks (CNNs) and a graph-based optimization to remove duplicated animals detected in overlapped images. We show that maximizing the degree of matching between animals is a suitable strategy to reduce duplicate counting. We also offer a dataset of real images, obtained from large pasture areas, both for training as well as for testing/benchmarking of cattle counting techniques. Our results show that the proposed method is very competitive, outperforming the state of the art in detecting duplicated animals, while significantly reducing the computational cost of the overall counting task. [ABSTRACT FROM AUTHOR]