东华大学学报（自然科学版）

针对多数纽扣生产厂家采用人工方式检测纽扣质量,导致检测效率较低的问题,设计一套基于图像处理的纽扣轮廓瑕疵检测系统。利用最大类间方差法对纽扣图像进行阈值分割,得到二值图像并进行连通区域标记,以进行纽扣区域分割与轮廓提取。再利用轮廓数目判断是否有内孔缺失或颜色瑕疵。同时提出外接圆方差检测算法,这是一种多步轮廓检测算法。利用最小外接圆进行圆形度检测;将圆形轮廓转换为曲线轮廓,利用三次方差进行全局平滑性检测和验证;采用局部最大二次方差对缺口进行检测。试验结果表明,该检测系统不仅能够有效地对纽扣进行瑕疵检测,而且能够满足在线检测对速度的要求。

A set of button contour defect detection system based on image processing is designed for the manual inspection of button quality by most button manufacturers,which leads to the problem of low inspection efficiency.The threshold segmentation of button image was carried out by the maximum interclass method,and the binary image was obtained and connected region was marked for button region segmentation and contour extraction.The number of contour was used to judge whether there was any inner hole missing or color defect.At the same time,the algorithm of detecting the circumscribed circle variance was proposed,which was a multi-step contour detection algorithm.The minimum circumscribed circle was used to detect the roundness;The circular contour was transformed into the curve contour,and the third-order variance was used to detect and verify the global smoothness;The local maximum secondorder variance was used to detect the gap.Experiments show that the detection method can not only effectively detect the defects of buttons,but also meet the speed requirements of online detection.