Study on the Detection of Defoliation Effect of an Improved YOLOv5x Cotton

Abstract

In order to study the detection effect of cotton boll opening after spraying defoliant, and to solve the problem of low efficiency of traditional manual detection methods for the use effect of cotton defoliant, this study proposed a cotton detection method improved YOLOv5x+ algorithm. Convolution Attention Module (CBAM) was embedded after Conv to enhance the network’s feature extraction ability, suppress background information interference, and enable the network to focus better on cotton targets in the detection process. At the same time, the depth separable convolution (DWConv) was used to replace the ordinary convolution (Conv) in the YOLOv5x model, reducing the convolution kernel parameters in the algorithm, reducing the amount of calculation, and improving the detection speed of the algorithm. Finally, the detection layer was added to make the algorithm have higher accuracy in detecting small size cotton. The test results show that the accuracy rate P (%), recall rate R (%), and mAP value (%) of the improved algorithm reach 90.95, 89.16, and 78.47 respectively, which are 8.58, 8.84, and 5.15 higher than YOLOv5x algorithm respectively, and the convergence speed is faster, the error is smaller, and the resolution of cotton background and small target cotton is improved, which can meet the detection of cotton boll opening effect after spraying defoliant.

Keywords

• defoliating agents
• cotton detection
• deep learning
• defoliation effect