摘要
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。
D u rin g process o f the ice cream stic k p ro d u c tio n , there was a series o f contour flaw s such as the out o f tolerancelength and w id th , la te ra l b e n d in g , e lb o w , fla tto p , oversize stick frin g e , b u rr and rip . A im in g at above-m entioned contourfla w s , th is paper p u t forw ard an on lin e detection m ethod o f ice cream stick contour flaw s based on machine vision , which extractedcandidate pixel of ice cream stick contour via Canny edge detector. Then i t removed contour noise and connected le ftcon tours, thus com pletly closed c o n to u r, w h ich w ould be used fo r the fla w detection. I t calculated barycenter o f ice creamstic k by using zeroth order m om ent and first order moment, and minimum tangent rectangle was obtained by means o f firs t orderm om ent and second order moment, and equal the length and w id th of m inim um tangent rectangle to the length and width of icecream stick. I t detected fla tto p and oversize stick frin g e via c irc le m odel and max offset distance o f ice cream s tick arc zone,and com bined p rojection m ethod and local-thre shold segment fo r detecting rip . The prototype system was developed accordingto the proposed scheme. Test results show th a t the proposed detection scheme can accom plish on line detection of above-m entionedparameters effectively , and achieve hig h precision and accuracy.
作者
苑玮琦
李德健
李绍丽
Yuan Weiqi;Li Dejian;Li Shaoli(Computer Vision Group,Shenyang University of Technology,Shenyang 110870,China)
出处
《计算机应用研究》
CSCD
北大核心
2016年第10期3185-3190,共6页
Application Research of Computers
基金
国家自然科学基金资助项目(61271365)
关键词
雪糕棒
轮廓
在线检测
CANNY算子
ice cream stick
contour
online inspection
Canny edge detector