Forestry Pests Identification and Classification Based on Improved YOLO v5s

Timely and effectively recognizing forestry pests is helpful for related forestry research. Hence, for solving the problem, we proposed a forestry pests identification method based on improved YOLO v5s. In the method, we use K-means clustering algorithm to cluster the width and height of the marked target box in the training set. And we also improved the screening function of the prediction box of YOLO v5s. Experiments on the self-made dataset which contains 15 common kinds of forest pests show the effectiveness of our method in terms of accuracy, recall and average detection accuracy compared with YOLO v4 and original YOLO v5s method. Experiment results indicate that our proposed method has relatively higher recognition ability and detection accuracy, which can be applied further in the field of forestry pests detection and identification.