WheatNet-CS: A Wheat Ear Detection Algorithm for Complex Background

Being among the most extensively grown crops, wheat planting density is one of the key indicators for yield estimation. In the face of the complex background such as dense shading in the wheat ear planting environment, Traditional detection techniques suffer from issues like poor accuracy and low efficiency. The suggested Improved WheatNet-CS method, which is based on the traditional YOLOv5, aims to increase the detection accuracy. The original NMS approach to get rid of unnecessary boxes is replaced by the CP-cluster method, increase the confidence value of candidate boxes while reducing missed detection in the detection process, and introduce the attention mechanism SE to construct channel correlation and recalibrate features, improve the quality of network learning representation and global information acquisition, and improve the detection performance. The average precision is 94.5%, which, in comparison to the average Yolov5s, is 3.3 percentage points greater.