卡片划痕检测关键算法设计

Key algorithm design of card scratch detection

针对划痕检测的现有算法大多基于简单背景,对于复杂背景下或者低对比度背景下的划痕检测准确率较低的问题,提出了基于机器视觉的卡片划痕检测算法。首先,通过高分辨率工业相机采集卡片图片,将卡片上的缺陷转化成图片中的检测目标;然后,针对复杂背景,运用引导滤波算法,对复杂背景噪声进行去除的同时保留了任意角度的划痕纹理细节;最后,针对划痕缺陷,提出改进的线状缺陷检测算法,可有效检测低对比度划痕。实验结果表明,该卡片划痕检测算法对于复杂背景低对比度下划痕缺陷识别率达到98%,且具有检测速度快的优点,满足工业场景要求。

Most of the existing algorithms for scratch detection are based on simple background. For the problem of low accuracy of detection in complex or low contrast background, a card scratch detection algorithm based on computer vision was proposed. First, the card images were captured by high-resolution industrial camera, so that the defects on the card could be converted into the detection target on the image. Then, the guided image filter was used to remove the noise on complex background, while scratch details at any angle were preserved. Finally, for the problem of scratches, an improved scratch detection algorithm was proposed, which could effectively detect low contrast scratches. The experimental result shows that the scratch detection algorithm has 98% recognition rate for scratch defect on low contrast and complex background, and has fast detection speed, which satisfies the requirements of industrial scenes.