基于视觉显著度的皮革表面瑕疵检测.pdf

CN 43—1258／TP ISSN 1007—130X 计算机工程与科学 Computer Engineering Science 第 38卷第 3期 2016年 3月 Vo1．38。No．3。Mar．2016 文章编号：1007—130X(2016)03—0549—07 基于视觉显著度的皮革表面瑕疵检测 朱凌云，严飞华，李汶松 (重庆理工大学计算机科学与工程学院，重庆 400054) 摘 要：针对汽车内饰皮革的瑕疵检测易受皮革 自身纹理干扰、检测难度较大的问题，发现瑕疵存在 于均匀变化 图像 中局部 变化明显的区域 ，符合人眼注意机制 ，故提 出了基于视觉显著模型的皮革瑕疵检测 方法。首先提取皮革图像的颜色和亮度特征，然后利用中心周围差算子分别计算特征显著图，再融合成最 终显著图，最后在此基础上利用区域生长方法对瑕疵区域进行分割，以实现瑕疵的准确定位。实验结果表 明，与 FCM 聚类分割法、阈值分割法及 sVM 分类法相比，本文提出的方法具有较高的检测精度及较快的 检测速度，解决了皮革瑕疵检测过程中受纹理干扰严重等问题，能有效应用于皮革瑕疵的机器 自动检测 中。 关键词 ：皮革；瑕疵检测 ；显著 图；分割 中图分类号：TP391．4 文献标志码：A doi：10．3969／j．issn．1007—130X．2016．03．023 Leather surface defect detection ● 1 ● ’ l■ Dasec 0n vlSUa1sa iency degree ZHU Ling—yun，YAN Fei—hua，LI W en—song (College of Computer Science and Engineering，Chongqing University of Technology，Chongqing 400054，China) Abstract：Leather defect inspection is seriously influenced by its texture．In view that finding the de— fects existing in the local highlight areas of uniform change images conforms to human eye attention mechanism ，we propose an automatic method for detecting leather defects based on the visual saliency mode1．Firstly，the features of color and brightness of the leather’S image are extracted，and then their own characteristic maps are calculated respectively by center—surround operator．Then the fina1 saliency map is formed through merging．On this basis，the region growing method is adopted to locate the flaw detection area．In comparison with the FCM clustering segmentation，the threshold segmentation and the SVM classification，the experimental results show that the proposed method has higher precision and faster detection speed，and it can solve the problems existing in the process of leather defect detection which can be seriously affected by its texture，and be applied to automatic machine detection of leather defects． Key words：leather；flaw detection；saliency map；segmentation 引言 汽车内饰皮革是其必不可少的装饰配件，随着 汽车产业的蓬勃发展，购买者对乘用小汽车内饰皮 革的要求也越来越挑剔，而蚊虫叮咬及加工过程的 精细不足使得原始皮革表面不可避免地产生了各 种缺