Research of space target segmentation algorithm based on ToF camera

In order to make robots perceive and explain the scene, and plan its object grabbing and processing behavior, 3D information becomes increasingly important. For obtaining 3D expression of objects from the ToF camera data, this paper has carried a three-step process: a) data pre-processing, including the definition of neighborhood, the establishment of point cloud topology, and outliers removal; b) using of RANSAC algorithm to detect the background; c) separating the candidate targets point cloud into different clusters. Finally, the experiment verifies that the approximated neighborhood reduces the calculation complexity of surface normal vector and curvature, RANSAC is a robust plane fitting algorithm, which can get a better plane parameter estimates steadily.