正交双目视觉长轴类零件装配端位姿检测方法

发布时间：2018-05-16 10:12

本文选题：位姿检测 + 立体视觉　；参考：《哈尔滨工业大学学报》2017年01期

【摘要】：为解决非正交双目立体视觉在长轴类零件三维位姿检测中需要多个特征点进行匹配,计算量大以及视野中难以获取整个零件的图像等问题,提出一种基于正交双目立体视觉检测长轴类零件装配端三维位姿方法.采用光轴正交的两个相机同时采集长轴类零件装配端的图像,分别采用亚像素法对所采得的图像进行处理,可以得到被检测长轴类零件装配端在两个相机图像坐标系内的位置和姿态,再依据正交双目立体视觉模型对两组位置信息进行融合得到其三维位姿.实验结果表明:利用两个CMOS相机构成的正交双目立体视觉能够快速、精确检测长轴类零件装配端的三维位姿.该方法只需要一个特征点进行匹配,计算量小、简单易行,适用于通过机器视觉引导工业机器人进行精密长轴类零件的快速装配.
[Abstract]:In order to solve the problem that non-orthogonal binocular stereo vision needs many feature points to match in 3D position and pose detection of long-axis parts, it is difficult to obtain the image of the whole part in the field of vision and so on. A method based on orthogonal binocular stereo vision is proposed to detect 3D position and pose of long axis parts. Two cameras with orthogonal optical axis are used to collect the image of the assembly end of the long axis parts simultaneously, and the sub-pixel method is used to process the image. The position and attitude of the assembly end of the detected long axis parts in the two camera image coordinate systems can be obtained, and then the position and pose of the two groups of position information can be obtained by fusion of the two sets of position information according to the orthogonal binocular stereo vision model. The experimental results show that the quadrature binocular stereo vision composed of two CMOS cameras can quickly and accurately detect the 3D position and pose of the assembly end of long axis parts. This method only needs one feature point to match, the calculation is small, the method is simple and easy to carry out, and it is suitable for the rapid assembly of the precision long axis parts of the industrial robot guided by machine vision.
【作者单位】：哈尔滨工业大学机电工程学院;
【分类号】：TP391.41

【相似文献】