基于PCS-YOLOv5轻量化模型的布匹外观缺陷检测方法

A Fabric Appearance Defect Detection Method Based on PCS-YOLOv5 Lightweight Model

针对现有布匹外观陷检测模型参数数量大、计算量大、部署于普通工控机检测速度慢等问题,文章提出一种轻量化模型PCS-YOLOv5。首先采用PP—LCNet替换YOLOv5主干网络实现模型轻量化,加快推理速度;在Neck网络引入CBAM注意力模块,抑制干扰信息关注重要特征,提高目标检测精度;修改边界框回归损失函数为SIoU,增强缺陷定位准确率。实验测试结果表明,PCS-YOLOv5相比YOLOv5原模型,在mAP@0.5基本保持一致的情况下,检测速度提高10.2%,参数数量减少56.8%,运算量降低63%,模型权重减小56%,可满足现场布匹外观缺陷在线检测要求。

In response to the problems of large number of parameters,high computational complexity,and slow detection speed when deployed on ordinary industrial computers in existing fabric appearance defect detection models,this paper proposes a lightweight model PCS-YOLOv5.Firstly,PP—LCNet is used to replace the YOLOv5 backbone network to achieve model lightweight and accelerate inference speed.It introduces the CBAM attention module into the Neck network to suppress interference and focus on important features,thereby improving the accuracy of object detection.It modifies the bounding box regression loss function to SIoU to enhance the accuracy of defect localization.The experimental test results show that compared to the YOLOv5 original model,PCS-YOLOv5 performs better in mAP@0.5 under the condition of basic consistency,the detection speed is increased by 10.2%,the number of parameters is reduced by 56.8%,the computational complexity is reduced by 63%,and the model weight is reduced by 56%,which can meet the requirements of online detection of fabric appearance defects on site.