基于机器视觉的三维点云匹配算法在散货物料实时检测中的应用与研究

Research and Application of 3D Point Cloud Matching Algorithm Based on Machine Vision in Real Time Detection of Bulk Materials

面对杂乱堆叠的散货物料实时检测场景,传统机器视觉会受物体遮挡限制,使得识别结果存在偏差。为解决此问题,本文运用三维点云算法,设计基于点对特征的位姿估计算法,同时调整场景点云中法线方向使其朝向点云成像设备视点,以优化其一致性,最后基于该算法构建实际物体抓取系统。结果显示,此次研究设计的位姿检测算法的平均重合率为98%,平均内点均方根误差为0.0003 mm,点云匹配成功率为100%,抓取成功率为97.1%。结果表明,该基于点对特征的三维点云算法在散货物料场景中的检测准确率较高,且在抓取场景中也有较理想的表现,具有一定的实际应用意义。

Face of real-time detection scenarios of disorderly stacked bulk materials,traditional machine vision is limited by object occlusion,which may result in biased recognition results.To solve this problem,a 3D point cloud algorithm is studied and designed for pose estimation based on point-to-point features.At the same time,the normal direction in the scene point cloud is adjusted to face the viewpoint of the point cloud imaging device to optimize its consistency.Finally,an actual object grabbing system is constructed based on this algorithm.Results show that the average coincidence rate of the pose detection algorithm designed in this study is 98%,the average internal point root-mean-square deviation is 0.0003 mm,the point cloud matching success rate is 100%,and the capture success rate is 97.1%.Results show that the 3D point cloud algorithm based on point-to-point features has high detection accuracy in the scene of bulk materials,and also has ideal results in the scene of grasping,which has practical application significance.