基于改进YOLOv5的地下排水管道缺陷检测算法
Defect detection algorithm of underground drainage pipe based on improved YOLOv5
吴谦 1孙丙宇 2房永峰3
作者信息
- 1. 安徽建筑大学 电子与信息工程学院,安徽 合肥 230601
- 2. 中国科学院 合肥物质科学研究院,安徽 合肥 230026
- 3. 中国科学技术大学研究生院 科学岛分院,安徽 合肥 230026
- 折叠
摘要
针对地下排水管道缺陷检测存在效率低和精度差等问题,本文提出了一种基于改进的YOLOv5 模型的地下排水管道缺陷识别方法.首先,为了提高检测精度,引入卷积块注意力机制模块(Convolutional Block Attention Module,CBAM)解决了部分缺陷过小的问题,同时设计了一种新的空间金字塔池化模块,解决不同类型缺陷大小、形状和颜色相差较大的问题.其次,在特征融合层引入加权双向特征金字塔网络结构(bi-directional feature pyramid network structure,BiF-PN)以充分利用不同尺度的特征,提高了特征融合能力.最后,针对图片模糊的问题,对原始图像进行了直方图均衡化、锐化处理.实验结果表明,改进后的算法能很好对障碍物、错口、裂缝、树根和错口等 5 种缺陷进行识别,平均精度(Mean Average Precision,mAP)相比原始YOLOv5 算法提升了2.5%.
Abstract
Aiming at the problems of low efficiency and poor accuracy in the detection of underground drainage pipe de-fects,this paper proposes a defect identification method for underground drainage pipes based on the improved YOLOv5 model.Firstly,in order to improve the detection accuracy,the convolutional block attention mechanism module(CBAM)is introduced to solve the problem that some defects are too small,and a new spatial pyramid pooling module is designed to solve the problem of large differences in size,shape and color of different types of defects.Secondly,a weighted bidirectional feature pyramid net-work structure(BiFPN)is introduced in the feature fusion layer to make full use of features of different scales,which improves the feature fusion ability.Finally,aiming at the problem of blurred pictures,the histogram equalization and sharpening of the original image is carried out.The experimental results show that the proposed algorithm can well identify five kinds of defects such as obstacles,wrong mouths,cracks,tree roots and wrong mouths,and the mean average precision is 2.5%higher than the original YOLOv5 algorithm.
关键词
目标检测/管道缺陷检测/YOLOv5/缺陷识别Key words
object detection/pipeline defect detectiond/YOLOv5/defect identification引用本文复制引用
基金项目
中国科学院合肥物质科学研究院院长基金重点支持项目(YZJJZX202013)
出版年
2024
NETL
NSTL
万方数据