基于估计曲率的带钢边裂检测算法

Steel strip edge split inspection based on curvature estimation

边裂是严重影响冷轧带钢质量的缺陷之一,是引起断带的主要原因。介绍了在基于机器视觉的检测系统中用图像处理技术检测带钢边裂的方法。首先通过边界追踪得到带钢边界的曲线链码,然后应用距离累积曲率算法估计链码重构曲线的曲率分布,以曲率的大小和方向为判断条件检测边裂缺陷,进而计算边裂的位置及开口深度。实验结果表明,该方法应用到实际检测系统中能够可靠地检出带钢边裂。

Edge split is one of the main reasons which affect the quality of cold-strip steel which makes the strip steel broken. A method of detecting the steel strip edge split was proposed based on image processing technology in machine vision system. The curve chain code of steel strip edge was obtained by the way of edge tracking estimation. The curvatures distributions of chain code reconfiguration curves were calculated by means of point-to-chord distance accumulation. The steel strip edge split disfigurement was detected based on the curvature by which the position and depth of the split was obtained. The results show that the proposed method is reliable for strip edge split inspection and can be applied in strip defect online inspection.