Target Detection Method of Apple Harvesting Robot Based on Improved YOLO v5

A target detection method for picking robot based on depth-separable convolution YOLO v5 is proposed. After the apple sample image was collected and the experimental data set was made, model training and testing were carried out. The deep separable convolution YOLO v5 network was introduced to extract features from the apple image, which solved the problem of parameter redundancy in the network and improved the recognition speed of the picking robot. The CIoU-Loss function and DIoU-NMS non-maximum suppression method were used to optimize the loss function and improve the positioning accuracy of the robot vision system on Apple.