基于改进YOLO v5s的水稻害虫识别研究  

Rice Pest Identification Based on Improved YOLO v5s

作　　者：王泰华[1,2] 郭亚州 张家乐 张晨阳 WANG Taihua;GUO Yazhou;ZHANG Jiale;ZHANG Chenyang(School of Electrical Engineering and Automation,Henan Polytechnic University,Jiaozuo 454003,China;Henan Key Laboratory of Inteigent Detection and Control of Coal Mine Equipment,Jiaozuo 454003,China)

机构地区：[1]河南理工大学电气工程与自动化学院,焦作454003 [2]河南省煤矿装备智能检测与控制重点实验室,焦作4540032.河南理工大学电气工程与自动化学院,焦作454003

出　　处：《农业机械学报》2024年第11期39-48,共10页Transactions of the Chinese Society for Agricultural Machinery

基　　金：国家自然科学基金项目(61903126);河南省科技攻关项目(212102210503)。

摘　　要：水稻害虫识别时,受稻田环境影响易出现目标被遮挡、与背景颜色相似、多目标相邻等问题导致识别精度降低。为此本文提出了一种基于改进YOLO v5s的水稻害虫识别模型YOLO v5s-Coordslimneck,通过替换主干网络中的普通卷积为CoordConv,增强了模型对目标位置信息的获取能力;引入CBAM注意力机制,提升了模型对目标区域的关注度;采用Slim-neck减少了计算量并增强了特征处理能力;引入Soft-NMS算法优化相邻目标边框筛选,减少漏检。实验结果表明,改进后的YOLO v5s模型在水稻害虫数据集上的平均精度均值达到94.3%,相比原模型提升3.8个百分点,比其他主流模型YOLO v3、YOLO R-CSP、YOLO v7、YOLO v8s提升1.5、12.7、13.6、1.9个百分点。消融实验进一步验证了改进模型中各个组件的有效性。热力图分析也体现了改进模型能够更好地学习害虫特征。综上,本文提出的改进YOLO v5s模型在提高水稻害虫检测精度方面取得了显著成效,为防控水稻虫害提供了一种精确的识别方法。When identifying rice pests, issues such as targets being obscured, similarity to the background color, and proximity of multiple targets due to the rice field environment can lead to reduced identification accuracy. To address this, a rice pest identification method was proposed based on an improved YOLO v5s. The method enhanced the model’s ability to capture target location information by replacing ordinary convolution in the backbone network with CoordConv. It introduced the CBAM attention mechanism to increase the model’s focus on the target area. The Slim-neck architecture was adopted to enhance feature processing capabilities and reduce computational load. The introduction of the Soft-NMS algorithm optimized the selection of adjacent target bounding boxes, reducing missed detections. Experimental results showed that the improved YOLO v5s model achieved an mAP of 94.3% on the rice pest dataset, which was an increase of 3.8 percentage points over the original model and 1.5, 12.7, 13.6 and 1.9 percentage points higher than that of the other mainstream models such as YOLO v3, YOLO R-CSP, YOLO v7, and YOLO v8s, respectively. Ablation experiments further validated the effectiveness of each component in the improved model. Heat map analysis also demonstrated that the improved model can better learn pest features. In summary, the improved YOLO v5s model proposed achieved significant results in improving the accuracy of rice pest detection, providing a more precise identification method for the prevention and control of rice pests.

关 键 词：水稻 害虫识别 CBAM Slim-neck YOLO v5s 

分 类 号：S435.16[农业科学—农业昆虫与害虫防治]