Litchi Detection in a Complex Natural Environment Using the YOLOv5-Litchi Model

Detecting litchis in a complex natural environment is important for yield estimation and provides reliable support to litchi-picking robots. This paper proposes an improved litchi detection model named YOLOv5-litchi environments. First, we add convolutional block attention module each C3 the backbone of network enhance ability extract feature information. Second, small-object layer enable locate smaller targets performance small targets. Third, Mosaic-9 data augmentation increases diversity datasets. Then, accelerate regression convergence process prediction box by replacing target loss function with CIoU. Finally, weighted-boxes fusion bring boxes closer reduce missed detection. An experiment carried out verify effectiveness improvement. The results study show that mAP recall were 12.9% 15%, respectively, comparison those unimproved YOLOv5 network. inference speed detect picture 25 ms, which much better than Faster-RCNN YOLOv4. Compared network, increased 17.4% large visual scenes. best five models. Therefore, achieved good balance between speed, size, accuracy, can meet needs agriculture technical