基于双目视觉的零件多尺寸在线测量系统

Binocular Vision Based Multi-dimensions On-line Measuring System for Workpieces

针对零件的自动化测量环节,搭建了一套可用于工业流水线生产的零件多尺寸在线测量系统。首先,对经典的张正友相机标定方法进行了改进;其次,把边缘聚焦思想用于Canny算子实现了高精度的零件轮廓提取;然后根据相机的极线对齐理论,提出了基于极线阈值约束的灰度相关边缘匹配算法,获得零件轮廓点云图;最后,对目标点云数据拟合得到零件关键尺寸。整个过程基于高斯金字塔多分辨率技术来实现,在满足测量精度的前提下进一步提高了检测速度。经过多次实验测试,该系统具有测量精度较高、重复性误差小的优点,可以较好地完成零件的在线检测任务。

Aiming at automatic measurement of workpieces,a binocular vision based multi-dimensions on-line measuring system used in the industrial assembly line for workpieces was constructed.Firstly,an improvement was made on the classic Zhang s camera calibration.Secondly,the edge focusing was employed in Canny detector to accurately extract structural boundaries.Furthermore,according to epipolar alignment theory,the gray correlation matching algorithm based on the epipolar and threshold constraint was proposed to obtain point cloud of contours.Finally,the target point cloud was processed and fitted to yield the key workpieces dimensions.The whole process was based on Gaussian pyramid multi-resolution technology,and the detection speed was further improved under the premise of satisfying the measurement accuracy.After several experiments,the system has the advantages of high measuring accuracy and small repeatability error,and can complete the on-line testing of parts well.