Next View Planning Policy for Autonomous 3D Reconstruction of Unknown Environments

This paper mainly discusses the problem of next-scan view planning for autonomous scanning of 3D cameras in unknown environments. Based on the fact that the weight of adjacent voxel of boundary hole equals to zero in the truncated distance function (TSDF) model, a boundary extraction algorithm based on TSDF volume is firstly proposed. The boundary point cloud is then divided into multiple clusters by using the edge point elimination and DBSCAN clustering. The connection between boundary point cloud and camera frame is established by creating the centroid frames of point cloud clusters which points to current camera frame. The next scanning poses are finally proposed by using the Rodriguez formula to rotate the current camera frame around the centroid frame. The simulation experiment on the public dataset is conducted and the rendered image of current and final 3D model in transformed camera perspective is compared, verifying the point cloud completion on target object and proving the feasibility of the planning strategy proposed in this paper.