雪糕棒轮廓质量视觉在线检测方法

摘要

During process of the ice cream stick production,there was a series of contour flaws such as the out of tolerance length and width,lateral bending,elbow,flattop,oversize stick fringe,burr and rip.Aiming at above-mentioned contour flaws,this paper put forward an online detection method of ice cream stick contour flaws based on machine vision,which ex-tracted candidate pixel of ice cream stick contour via Canny edge detector.Then it removed contour noise and connected left contours,thus completly closed contour,which would be used for the flaw detection.It calculated barycenter of ice cream stick by using zeroth order moment and first order moment,and minimum tangent rectangle was obtained by means of first order moment and second order moment,and equal the length and width of minimum tangent rectangle to the length and width of ice cream stick.It detected flattop and oversize stick fringe via circle model and max offset distance of ice cream stick arc zone, and combined projection method and local-threshold segment for detecting rip.The prototype system was developed according to the proposed scheme.Test results show that the proposed detection scheme can accomplish online detection of above-men-tioned parameters effectively,and achieve high precision and accuracy.%雪糕棒加工过程中会产生长度和宽度超差、侧弯、偏头、平头、头部多肉、劈裂等轮廓问题。针对上述轮廓质量问题，提出了一套基于机器视觉的雪糕棒轮廓质量在线检测方法，使用了Canny算子提取轮廓候选像素，通过去噪和间隙连接形成的完整闭合轮廓用于各项缺陷检测。其中运用零阶矩和一阶矩获得雪糕棒质心，并通过一阶矩和二阶矩获得最小外接矩形，由最小外接矩形的长宽表示雪糕棒的长宽；通过建立的圆模型与雪糕棒圆弧区的最大偏离距离判断多肉和平头缺陷；通过投影与局部阈值分割相结合的方法检测劈裂缺陷。根据提出的方案研制了样机，现场测试结果表明，该检测方案能够有效地实现上述参数的在线检测，并具有较高的检测准确率和精度。