纹理无关的轮胎裂纹检测算法  

Texture-Invariant Detection Method for Tire Crack

作　　者：张传海[1] 李雪梅[1] 郭强[2] 余曦晨[1] 张彩明[1,2] 

机构地区：[1]山东大学计算机科学与技术学院,济南250101 [2]山东财经大学山东省数字媒体技术重点实验室,济南250014

出　　处：《计算机辅助设计与图形学学报》2013年第6期809-816,共8页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(60933008;61103150);国家自然科学基金国际重大合作项目(61020106001);教育部博士点基金(20110131130004)

摘　　要：根据从图像中提取的裂纹缺陷特征,提出一种基于线密度投影(PODOL)的轮胎裂纹缺陷检测方法.首先对轮胎图像进行线密度投影,得到它们的PODOL一阶导数绝对值曲线;然后提取裂纹图像的特征曲线,以这些特征曲线作为标准判定轮胎是否含有裂纹.实验结果表明,文中提出的特征曲线检测标准不受图像纹理的影响,可以高效地检测轮胎裂纹缺陷;且该方法速度快、实现简单,已经成功用于轮胎裂纹缺陷的实时在线检测.The paper proposed a new image-based detection method for tire crack-defect according to projection of density of lines （PODOL）. First, the PODOL curves are computed from the tire image and their modulus curves of first order derivative are evaluated. Then, a new feature curve is defined and extracted from the first order derivative modulus curves. It will be employed as a new criterion for detecting the crack-defects. The experimental results show that the proposed method is texture- invariant and can be applied to detect crack-defects in tire images efficiently. The proposed method has been used in real-time online tire detection.

关 键 词：线密度投影 一阶导数曲线 特征曲线 纹理无关 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]