Recognition of the boundaries of farmland plow areas has an important guiding role in the operation of intelligent agricultural equipment. To precisely recognize these boundaries, a detection method for unmanned tractor plow areas based on RGB-Depth (RGB-D) cameras was proposed, and the feasibility of the detection method was analyzed. This method applied advanced computer vision technology to the field of agricultural automation. Adopting and improving the YOLOv5-seg object segmentation algorithm, first, the Convolutional Block Attention Module (CBAM) was integrated into Concentrated-Comprehensive Convolution Block (C3) to form C3CBAM, thereby enhancing the ability of the network to extract features from plow areas. The GhostConv module was also utilized to reduce parameter and computational complexity. Second, using the depth image information provided by the RGB-D camera combined with the results recognized by the YOLOv5-seg model, the mask image was processed to extract contour boundaries, align the contours with the depth map, and obtain the boundary distance information of the plowed area. Last, based on farmland information, the calculated average boundary distance was corrected, further improving the accuracy of the distance measurements. The experiment results showed that the YOLOv5-seg object segmentation algorithm achieved a recognition accuracy of 99% for plowed areas and that the ranging accuracy improved with decreasing detection distance. The ranging error at 5.5 m was approximately 0.056 m, and the average detection time per frame is 29 ms, which can meet the real-time operational requirements. The results of this study can provide precise guarantees for the autonomous operation of unmanned plowing units.

Cows mounting behavior is a significant manifestation of estrus in cows. The timely detection of cows mounting behavior can make cows conceive in time, thereby improving milk production of cows and economic benefits of the pasture. Existing methods of mounting behavior detection are difficult to achieve precise detection under occlusion and severe scale change environments and meet real-time requirements. Therefore, this study proposed a Cow-YOLO model to detect cows mounting behavior. To meet the needs of real-time performance, YOLOv5s model is used as the baseline model. In order to solve the problem of difficult detection of cows mounting behavior in an occluded environment, the CSPDarknet53 of YOLOv5s is replaced with Non-local CSPDarknet53, which enables the network to obtain global information and improves the model’s ability to detect the mounting cows. Next, the neck of YOLOv5s is redesigned to Multiscale Neck, reinforcing the multi-scale feature fusion capability of model to solve difficulty detection under dramatic scale changes. Then, to further increase the detection accuracy, the Coordinate Attention Head is integrated into YOLOv5s. Finally, these improvements form a novel cow mounting detection model called Cow-YOLO and make Cow-YOLO more suitable for cows mounting behavior detection in occluded and drastic scale changes environments. Cow-YOLO achieved a precision of 99.7%, a recall of 99.5%, a mean average precision of 99.5%, and a detection speed of 156.3 f/s on the test set. Compared with existing detection methods of cows mounting behavior, Cow-YOLO achieved higher detection accuracy and faster detection speed in an occluded and drastic scale-change environment. Cow-YOLO can assist ranch breeders in achieving real-time monitoring of cows estrus, enhancing ranch economic efficiency.

Harvesting represents the crucial stage in the cultivation process of Agaricus bisporus mushrooms. An important way for the production process of Agaricus bisporus to reduce costs and increase income is to ensure timely harvest of Agaricus bisporus, reduce harvesting costs, and improve harvesting efficiency. There are many disadvantages in manual picking, such as high labor intensity, time-consuming work and high cost. In this study, a set of mushroom picking platform including climbing mechanism, picking robot, and control system was designed and developed. The picking robot consisted of a truss mechanism, an image acquisition device, a mushroom collection device, and a picking actuator. The profile picking actuator could realize the function of constant force clamping. An online size detection algorithm for Agaricus bisporus based on deep image processing was proposed. The algorithm included removal of abnormal noise points, background segmentation, coordinate conversion, and diameter detection. The precision picking system for Agaricus bisporus with coordinate compensation function controlled by Industrial Personal Computer was designed, and the visual control interface was developed based on Labview. Through the performance test, the reliability of machine vision recognition and the overall operating stability of the picking platform were verified. The test results showed that in the process of machine vision recognition, the recognition accuracy rate was higher than 92.50%, the missed detection rate was lower than 4.95%, the false detection rate was lower than 2.15%, and the diameter measurement error was less than 4.50%. The image processing algorithm had high recognition rate and small diameter measurement error, which could meet the requirements of picking operation. The picking platform’s picking success rate was higher than 95.45%, the picking damage rate was lower than 3.57%, and the picking output rate was higher than 87.09%. Compared with manual picking, the recognition accuracy rate of the picking platform was increased by 6.70%, the picking output rate was increased by 1.51%. The overall performance of the picking platform was stable and practical.