本文基于三坐标测量机(CMM)设计了一套视觉检测系统, 该系统能够对零件实际空间特征信息进行比较全面地提取。针对位于CMM平台上带有角点的零件, 利用Harris算子提取从CMM三个不同方位获取的零件图像的角点。对于Harris算子提取到的角点, 本文提出一种八链码搜索法和SUSAN区域法相结合的伪角点剔除方法, 最后基于立体视觉的原理, 提出“距离空间图”匹配算法, 将以上3幅图像一一建立匹配关系。实验中多次改变零件在CMM中姿态时, 多次实验数据表明本文的角点提取精度与真实角点间仅存在1~2像素的偏差, 零件的定位误差为1~3 mm。通过实验验证, 角点匹配和定位的稳定性和精度满足要求, 具有一定的抗干扰性和实用性。

In this paper, a set of visual inspection system is designed based on three coordinate measuring machine(CMM), and the new visual system can extract the actual spatial feature information of the parts. For the parts with corner which is on the platform of CMM, the Harris operator is used to extract corners in the images obtained from three different orientation of the CMM. For the corner points extracted by the Harris operator, this paper proposes a method of eliminating the false corners, which combines eight-chain-code false corner search method and SUSAN method. Finally, based on the principle of stereo vision, the “Distance spatial map” matching algorithm is put forward, and then the three images are matched one by one. Although the position and orientation of parts are changed for many times in the experiment, the experiment results show that there are 1-2 pixels difference between extracted corners and the real corner points, and the position error of parts is 1-3 mm. Through the experiment, the accuracy and stability of corner matching and positioning can meet the requirements, and with anti-interference and practicability as well as.