计算机与数字工程2025，Vol.53Issue(2)：535-539,544,6.DOI:10.3969/j.issn.1672-9722.2025.02.041
基于改进DeepLabV3+的轮胎缺陷检测
Tire Defect Detection Based on Improved DeepLabV3+
孙英伟 1张岩1
作者信息
- 1. 青岛科技大学机电工程学院 青岛 266061
- 折叠
摘要
Abstract
This paper proposes a tire defect detection model based on the improved DeepLabV3+.The model can not only lo-cate the location of the defect,but also characterize the geometry of the defect.Firstly,the ECA channel attention mechanism is used to focus on the tire defect features,so that the backbone network can effectively extract the feature information of the defects.Then,a Chained Atrous Pyramid Pooling(CAPP)is proposed by improving the Atrous Spatial Pyramid Pooling(ASPP).CAPP adopts a chain connection strategy before the dilated convolution branch to more fully extract the multi-scale information of features.The experimental verification shows that the detection model can deal with defects in various types of tire radiographic images,and has a high detection accuracy.关键词
缺陷检测/注意力机制/多尺度特征提取/轮胎射线图像Key words
defect detection/attention mechanism/multi-scale feature extraction/tire radiographic images分类
信息技术与安全科学引用本文复制引用
孙英伟,张岩..基于改进DeepLabV3+的轮胎缺陷检测[J].计算机与数字工程,2025,53(2):535-539,544,6.
