摘要
针对棉花杂质去除过程中地膜杂质难以去除的问题,研究采用深度学习的方法,提出一种基于改进YOLOv5s的地膜识别模型。首先,对YOLOv5s模型损失函数改进,在保证满足检测速度的基础上,提高对地膜的识别准确率;其次,添加CA(Coordinate Attention)注意力机制,增强算法的特征提取能力和检测精度,进一步提高对棉花地膜的识别准确率,以此得到地膜杂质的精确位置信息。试验结果表明:相比于传统方法,采用深度学习的方法可在原有去除基础上,进一步识别出89.9%的地膜杂质,有效提高棉花质量;且相比于原YOLOv5s模型,改进后模型识别的准确率和召回率分别为88.3%和86.3%,分别提高了7.5%和7.9%,对地膜杂质的检测效果明显提高,有效解决了地膜杂质难以去除的问题。
Aiming at the problem of difficult removal of plastic film impurities in the process of removing cotton impurities,a deep learning method was adopted for removal,and an improved YOLOv5s based plastic film recognition model was proposed.Firstly,the YOLOv5s model loss function was improved to improve the recognition accuracy of plastic film while ensuring the detection speed was met;Secondly,adding CA(Coordinate Attention)attention mechanism enhances the algorithm's feature extraction ability and detection accuracy,further improving the recognition accuracy of cotton plastic film,and obtaining accurate location information of plastic film impurities.The experimental results show that compared to traditional methods,the use of deep learning can further identify 89.9%of plastic film impurities on the basis of original removal,effectively improving the quality of cotton;Compared with the original YOLOv5s model,the improved model achieved recognition accuracy and recall of 88.3%and 86.3%,respectively,with an increase of 7.5%and 7.9%.The detection effect of plastic film impurities was significantly improved,effectively solving the problem of difficult removal of plastic film impurities.
作者
陈东胜
潘江如
董芙楠
董恒祥
CHEN Dong-sheng;PAN Jiang-ru;DONG Fu-nan;DONG Heng-xiang(Kuitun Yinli Cotton Oil Machinery Co.,LTD.,Kuitun Xinjiang 833200,China;Xinjiang University of Engineering,Urumqi Xinjiang 830000,China;College of Transportation and Logistics Engineering,Xinjiang Agricultural University,Urumqi Xinjiang 830052,China)
出处
《林业机械与木工设备》
2024年第11期22-28,共7页
Forestry Machinery & Woodworking Equipment
基金
中央引导地方科技发展专项资金项目“双锯筒锯齿轧花机的研发与应用”。
作者简介
第一作者:陈东胜,工程师,本科,主要从事机械电子工程方面研究,E-mail:13770285829@163.com;通讯作者:潘江如,教授,博士生导师,博士,主要从事内燃机燃烧与排放控制研究,E-mail:pjr1978@126.com。
