Improved Wheat Spike Recognition Method Based on YOLOv5 with Added Channel Attention Mechanism

Rapid identification and detecting wheat spikes is of great significance for understanding the growth status of wheat, evaluating its growth potential, and estimating future wheat yields. Traditional object detection algorithms often face a trade-off between accuracy and speed. In order to improve the accuracy of wheat spike recognition, this study used an improved one-stage object detection algorithm Yolov5_SE algorithm, which introduced the channel attention mechanism module SE_Block to enhance the focus on local detailed features of wheat spikes and solve the problems of dense and overlapping wheat spikes. This paper continuously changed the position and number of SE_Block in the original network structure when adding the channel attention mechanism to the improved Yolov5 algorithm. The experimental results show that the proposed algorithm in wheat spike recognition tasks achieved better performance, with an average precision (mAP) improvement of 0.8% compared to the original Yolov5 algorithm, while speed was also improved. The results of this study can provide reference and inspiration for the application of machine vision technology in the field of agriculture.