为提高产品零件空间姿态识别的精度和收敛速度,提出基于结构光3D视觉对空间姿态识别的方法。首先,采用投影仪和相机获取产品零件图像信息,利用相移法获取深度信息,通过深度图点云重建获取3D点云数据;然后对3D点云数据进行特征处理和分类时,建立点云网络(Point Network,PointNet)模型;最后,采用改进的迭代最近点(Iterative Closest Point, ICP)算法对3D点云数据配准,从而实现产品零件姿态的识别。实验结果表明,该方法在对产品零件点云特征分类性能上,准确率能达到96%左右,召回率能稳定在92%左右;在配准精度和收敛速度上,较其他两种方法更优越,进一步验证了该方法的有效性和可行性。

In order to improve the precision and convergence speed of spatial attitude recognition of product parts,a method of spatial attitude recognition based on structured light 3D vision is proposed.Firstly,image information of product parts is obtained by projector and camera,depth information is obtained by phase shift method,and 3D point cloud data is obtained by point cloud reconstruction of depth map.Then,when the 3D Point cloud data is processed and classified,the Point Network (PointNet) model is established.Finally,an improved Iterative Closest Point (ICP) algorithm is used to register 3D point cloud data,so as to realize the identification of product part attitude.The experiment results show that the accuracy of the method can reach about 96% and the recall rate can be stable at about 92%.In terms of registration accuracy and convergence speed,it is superior to the other two methods.The effectiveness and feasibility of this method are further verified.