基于色彩分割及信息熵加权特征匹配的刹车片图像分类算法  

Brake pad image classification algorithm based on color segmentation and information entropy weighted feature matching

作　　者：赵磊[1] 张文[1] 孙振国[1,2] 陈强 ZHAO Lei;ZHANG Wen;SUN Zhenguo;CHEN Qiang(Key Laboratory for Advanced Materials Processing Technology of Ministry of Education, Department of Mechanical Engineering, Tsinghua University, Beijing 100084, China;Yangtze Delta Region Institute of Tsinghua University, Jiaxing 314006, China)

机构地区：[1]清华大学机械工程系、先进成形制造教育部重点实验室,北京100084 [2]浙江清华长三角研究院,嘉兴314006

出　　处：《清华大学学报（自然科学版）》2018年第6期547-552,共6页Journal of Tsinghua University(Science and Technology)

摘　　要：在基于机器视觉的汽车刹车片外观检测中,刹车片图像前景分割及形状特征的提取和匹配算法是刹车片分类的关键。为实现汽车刹车片外观高精度检测分类,该文提出并实现了一种包括前景分割、几何特征提取及特征模板匹配的图像处理算法。该算法首先将工业相机拍摄得到的红、绿、蓝（red green blue,RGB）格式图像转换到色调、饱和度、明度（hue saturation value,HSV）色彩空间,利用其中的饱和度S通道从暗色传送带背景中分割刹车片;然后提取刹车片区域的多维几何特征进行特征融合;最后采用基于信息熵加权的改进特征匹配算法,通过刹车片特征的信息熵对特征空间的Manhattan距离进行加权。对98张、54类刹车片图像进行了分类实验,结果表明：算法准确率为95.00%,每张平均耗时110ms,可以应用于刹车片生产过程中的实时自动分类。In appearance inspection of car brake pads based on machine vision, the segmentation of the brake pad image and the extraction and matching of the shape features are keys of brake pads classification. In order to realize high-precision shape classification, this paper proposes a brake pad image classification algorithm including foreground segmentation, geometric feature selection and template matching. First the RGB image captured by the camera isconverted to HSV color space and using saturation channel the brake pad is segmented from the dark belt background. Then the multi-dimensional geometric features of the area are extracted. Finally, an improved feature matching algorithm based on information entropy weighting is proposed, in which Manhattan distance of the feature space is weighted by the information entropy of the brake pad features. The experimental results show that the accuracy of the algorithm is 95.00%, and the average processing time is 110 ms. It can be applied to the real-time automatic classification procedure during brake manufacturing.

关 键 词：刹车片 机器视觉 色调、饱和度、明度(huesaturation value HSV)色彩空间 几何特征 信息熵 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]