基于优化YOLOv5算法的玉米苗间杂草检测研究

打开文本图片集

中图分类号：S451 文献标识码：A 文章编号：2095-5553（2025）07-0220-06

Abstract：Toaddress the problems of diverse weedspecies，high detection complexity，and slow detection speed in the detection of weedsamong maize seedlings，thispaperproposesa weed detection method based on optimized YOLOv5 algorithm.The SE parameter-freeatentionmoduleis introduced intotheconvolutional layersof theYOLOv5 backbone network，withthe SE一integrated C3module replacing theoriginal C3 moduleto beterfocus ondetection targets. The traditional residual neural network is replaced with the BoTNetmodule，andglobal multi-headself-attention is used to replace 3×3 spatial convolution in thelastthree botleneck blocksof ResNet，thereby improving the accuracy of detecting smalltargets.The improvedtargetdetectionalgorithmisused todetect weeds，withnon-maizeseedlingareas inthe fieldlabeledasweeds.The super-green feature，combinedwith the OTSU threshold segmentation algorithm，is used tosegment thesoilbackgroundandidentifytheforegroundareasof weeds，effectivelysolving theproblemof weed detection inmaizesedling fields.The resultsshowthattheimproved YOLOv5algorithmachievesatarget detection precision of 97.5% for maize seedlings，which is 7.4% higher than the original YOLOv5 algorithm. The detection speed reaches 40ms ，thereby improving the detection accuracy and model robustness to meet the needs of real-time detection.

Keywords：wed detection among maize seedlings；YOLOv5；attention module；BoTNet module；super-green feature; OTSU threshold segmentation

0 引言

与高产作物争夺生长空间、阳光、水分和土壤养分，容易引发病虫害等问题，严重影响作物的健康生长，从而导致农产品的产量和质量下降[1，2]。（剩余9282字）