基于点云中心约束的点对特征三维物体识别算法

A Fast 3D Object Recognition Algorithm Based on Point Cloud Center Constrained Point Pair Feature

目标姿态识别系统中能够快速、准确在复杂场景里识别目标物体是实现机器人在线抓取的关键,针对传统点对特征(PPF)算法相邻物体点云干扰的问题,提出了一种基于点云中心约束点对特征的三维物体识别算法。首先通过法向量相交约束为条件提取物体中心,然后根据模型的大小筛选出相关的点对特征,最后利用共面点对和物体中心特征,实现计算三维物体的姿态,以此来提高识别速度。实验结果表明,该算法显著减少特征描述符数量,提高了三维物体在复杂场景中的识别速度。与原始的点对特征算法相比,该方法对三维目标识别具有较好的性能与速度。

In the target pose recognition system,the ability to quickly and accurately identify the target object in complex scenes is the key issue to realize the robot′s online grasping.Aiming at the problem of point cloud interference of adjacent objects in the traditional point pair feature(PPF)algorithm,a 3D object recognition algorithm based on point cloud center constraint point pair feature is proposed.Firstly,the center of the object is extracted based on the normal vector intersection constraint,and then the relevant point pair features are screened according to the size of the model.Finally,the coplanar point pair and the object center feature are used to calculate the pose of the three-dimensional object,so as to improve the recognition speed.Experimental results show that the algorithm significantly reduces the number of feature descriptors and improves the recognition speed of 3D objects in complex scenes.Compared with the original point pair feature algorithm,this method has better performance and speed for 3D target recognition.