摘要
为了实现机器视觉检测系统对零件尺寸的微米级自动测量,对随机误差和系统误差进行了研究。改进了双线性插值法,对像素进行细分,用canny算法检测出边缘,并用多项式拟合出轮廓。该轮廓存在随机误差和系统误差。根据轮廓的切向连续性,用期望值和实测值进行最优估计,消除随机误差的影响。分别用不同宽度的量块放在视场的四个象限,测得多组图像,提取轮廓点,结合畸变模型,对畸变进行校正并对系统进行标定。通过细分、卡尔曼滤波、畸变校正和标定,每个像素代表的实际尺寸为4.584μm,精度达到2.5μm。
Random errors and systematic errors of the machine vision system were studied in order to meas- ure micron-sized 2D sizes of parts automatically. The bilinear interpolation algorithm was improved to sub- divide the pixels. The edge was extracted from the digital image with canny algorithm and transformed to a polynomial. This edge has random error and systematic error. The optimal estimation of expected value and the measured value was calculated based on contour tangential continuity, to eliminate the effects of random error. The gauge blocks with different widths were set on the four quadrants of the visual field, their edges were extracted. Distortion correction and system calibration was done combining with the dis- tortion model The actual size of each pixel is 4. 584p, m after using subdivision, Kalman filter, distortion correction and calibration Tests and analyses show that the measurement accuracy is 2.5μm class.
出处
《组合机床与自动化加工技术》
北大核心
2013年第9期108-110,114,共4页
Modular Machine Tool & Automatic Manufacturing Technique
基金
教育部留学回国人员科研启动基金(20100609)
辽宁省教育厅项目(201114126)
关键词
机器视觉
误差分析
像素细分
切向连续性
畸变校正
machine vision
error analysis
pixel sub-division
tangential continuity
distortion correction
