现代农业装备2025，Vol.46Issue(2)：98-103,6.DOI:10.3969/j.issn.1673-2154.2025.02.015
融合全局注意力与可形变卷积的玉米病害识别算法研究
A Maize Disease Identification Algorithm Integrating CBAM Attention and Deformable Convolution
摘要
Abstract
Aiming at the problems of low detection accuracy,poor recognition effect,slow detection speed and poor model robustness caused by complex image background and variable lesion scale in maize disease recognition,an improved YOLOv7 maize disease recognition algorithm combining attention module CBAM and deforming convolution was proposed.Firstly,deformable convolution was used in the feature extraction network to optimize the extraction efficiency of irregular features by traditional convolution,so as to improve the location perception ability of the model.Secondly,CBAM modules were added to the feature network and neck network to enhance the model's attention to small target lesions.Finally,the sample data was trained after data enhancement preprocessing to improve the generalization ability and recognition accuracy of the model.Experimental results showed that the accuracy of the improved model for maize disease identification reached 95.6％,the average accuracy was 94.3％,the recall rate was 92.4％,and the detection speed was 48.72 frames/SEC.Compared with the baseline model YOLOv7,the accuracy,average accuracy and recall rate increased by 5.3％,5.2％and 5.2％,respectively.Rapid and accurate identification of maize diseases can be realized.关键词
玉米/目标检测/病害/YOLOv7/可形变卷积/注意力机制Key words
maize/target detection/disease/YOLOv7/deformable convolution/attention mechanism分类
农业科技引用本文复制引用
张焱姝,陈军,孙丽丽,韩璐宇..融合全局注意力与可形变卷积的玉米病害识别算法研究[J].现代农业装备,2025,46(2):98-103,6.基金项目
新疆生产建设兵团财政科技计划项目(1121DB008). （1121DB008）
