Registration Of Depth Maps Based On Irls-Icp-Tps
2016 9TH INTERNATIONAL CONGRESS ON IMAGE AND SIGNAL PROCESSING, BIOMEDICAL ENGINEERING AND INFORMATICS (CISP-BMEI 2016)(2016)
摘要
In the problem of three-dimensional reconstruction from depth maps, the objects may produce deformation; the camera calibration error may lead to slight deformation of depth images. We propose a new approach named IRLS-ICPTPS (Iteratively Re-weighed Least Squares Iterative Closest Point Thin-Plate Splines) to align the depth images. It can be used to calibrate the slight deformation of the depth images and align the point sets of deformable objects. In the first step, we convert the depth images into point cloud, then, we use the Spin-image coarsely match the point sets. Thirdly, we make use of the IRLS-ICP( Iterative Closest Point) algorithm to register the point clouds on the basis of coarse matching. In the end, we deform the point cloud using TPS(Thin-Plate Splines) and joint the point sets. The experiments show that the proposed method can get better results in registration of depth maps, both for rigid and non-rigid objects, than original registration methods.
更多查看译文
关键词
depth maps,point clouds,non-rigid,deformation,IRLS-ICP,TPS
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络