基于点云的堆叠玛钢管件识别定位方法

Identification and positioning method of stacked malleablesteel pipe fittings based on point cloud

针对机器人抓取堆叠玛钢管件识别定位精度较低等问题,提出基于点云信息的堆叠玛钢管件识别定位方法。通过双目视觉获取管件的视差图,采用HSI权重分配及左右一致性校验法对玛钢管件的遮挡点进行校验,得到完整的管件视差图;基于玛钢管件点云的法向量特征,剔除管件接触部分边缘点,利用欧式聚类对相互独立的管件点云进行分割提取,采用ISS算法得到玛钢管件的关键点云数据信息,通过RANSAC算法与ICP算法进行点云匹配,确定玛钢管件的位姿。实验结果表明,该方法对不同工况下堆叠玛钢三通管件的平均识别率为89.5%,机器人抓取精度平均提升了37%。

Aiming at the problem of low recognition and positioning accuracy of robot grasping stacked steel pipe fittings,a recognition and positioning method of stacked steel pipe fittings based on point cloud information was proposed.Binocular vision was used to obtain the disparity map of the pipe fittings.HSI weight distribution and left-right consistency verification method were used to verify the occlusion points of the pipe fittings,and the complete disparity map of the pipe fittings was obtained.Based on the normal vector characteristics of the point cloud of the pipe fittings,the edge points of the contact part of the pipe fittings are eliminated,and the European clustering is used to segment and extract the independent point cloud of the pipe fittings.The ISS algorithm is used to obtain the key point cloud data information of the pipe fittings.The experimental results show that the average recognition rate of the proposed method for stacked Masteel three-way pipe fittings under different working conditions is 89.5%,and the robot grasping accuracy is improved by 37%on average.