纹理无关的轮胎裂纹检测算法

Texture-Invariant Detection Method for Tire Crack

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.