自适应光照变化的高频工件图像识别算法  

High-frequency Workpiece Image Recognition Algorithmwith Self-adaption Illumination Change

作　　者：孙成龙 李节 李柏林[1] 王逸涵 欧阳 SUN Chenglong;LI Jie;LI Bailin;WANG Yihan;OU Yang(School of Mechanical Engineering,Southwest Jiaotong University,Chengdu 610031,China;School of Mechanical Engineering,Chengdu University,Chengdu 610106,China)

机构地区：[1]西南交通大学机械工程学院,四川成都610031 [2]成都大学机械工程学院,四川成都610106

出　　处：《机械制造与自动化》2025年第2期227-232,共6页Machine Building & Automation

基　　金：四川省科技厅重点研发项目(2021YFN0020)。

摘　　要：针对工业场景中不同光照条件下高频工件识别准确率低的问题,提出一种自适应光照变化的高频工件图像识别算法。采用通道注意力模块增强图像中对光照变化不敏感的特征,以减少光源照度对识别结果的影响;以ResNet50网络为基础,构造了两个分支网络,分别从整体图像中提取全局信息以及从有效的区域中提取局部信息,其中有效的区域由弱监督区域检测模块得到;融合两个分支网络的识别结果,实现了高频工件的图像识别。通过对多种类别高频工件的实验结果表明:该算法能够自适应光照的变化,提高了复杂光照条件下的高频工件识别性能,识别准确率达到了94.8%。To improve the low recognition accuracy of high-frequency workpieces under different illumination conditions in industrial scenes,an image recognition algorithm of high-frequency workpieces with self-adaptive illumination change is proposed.The channel attention module is used to enhance the features in the image that are insensitive to illumination change,so as to reduce the influence of light source illumination on the recognition results.Based on ResNet50 network,two branch networks are constructed to extract global information from the whole image and local information from the effective region,where the effective region is obtained by the weak supervised region detection module.The recognition results of the two branch networks are fused to achieve the image recognition of high-frequency workpieces.The experimental results of multi-type high-frequency workpieces show that the algorithm can adapt to the change of illumination and improves the recognition performance of high-frequency workpieces under complex illumination conditions,with 94.8%recognition accuracy.

关 键 词：光照变化 工件识别 注意力机制 区域检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]