摘要
针对新疆红枣产业在加工出售前需要在大量红枣中剔除有裂口、皱皮和变形等缺陷的红枣这一需求,文章提出一种基于改进YOLOv5n模型的红枣缺陷识别方法。该方法首先在YOLOv5n模型目标检测头部引入SEConv通道注意力机制的卷积操作,用于增强模型的特征表示能力,其次使用C3替换SPPF操作加快识别速度,最后通过调整自适应锚定框,更好地适应红枣尺寸和长宽比。实验结果表明,改进后的模型缺陷识别准确度达到了95.8%,相比原模型提升了3.5个百分点,识别速度达到7.2ms,比原模型提升了20.9%。这意味着改进后的YOLOv5模型在保持高准确度的同时,能够更高效地处理大量红枣图像。
In view of the demand that the red dates with defects such as cracks,wrinkled skins and deformation need to be removed from a large number of red dates before processing and selling,this paper proposes a red date defect recognition method based on the improved YOLOv5n model.Firstly,the convolution operation of the SEConv channel attention mechanism is introduced into the target detection head of the YOLOv5n model to enhance the feature representation ability of the model.Secondly,C3 is used to replace the SPPF operation to speed up the recognition speed.The experimental results show that the defect recognition accuracy of the improved model reaches 95.8%,which is 3.5 percentage points higher than that of the original model,and the recognition speed reaches 7.2 milliseconds,which is 20.9%higher than that of the original model.This means that the improved YOLOv5 model can process a large number of date images more efficiently while maintaining high accuracy.
作者
陈星宇
凡玉琪
刘虎涛
蒋培宗
CHEN Xingyu;FAN Yuqi;LIU Hutao;JIANG Peizong(School of Electronic and Information Engineering,Liaoning Technical University,Huludao Liaoning 125105,China)
出处
《信息与电脑》
2023年第14期181-186,共6页
Information & Computer