Generating colored point cloud under the calibration between TOF and RGB cameras

Abstract

3D point cloud is the original format of three dimensional map for the navigation of mobile robotics. Time of Flight camera can directly return 3D points and grayscale image simultaneously, but the color information is lost. In this paper, 3D points are colored by calibrating PMD camera with RGB camera, and then the RGB image is generated with respect to a frame of TOF camera. Point cloud form only one frame is not large enough for navigating the robot or observing the environment, an approach for combining two continuous frames is also proposed. Scale invariant features are detected from the RGB images generated, and features from two images are matched accurately. Iterative Closest Point(ICP) algorithm is used to estimate the relative pose transformation of TOF camera, the initial pose is computed from corresponded 3D feature point pairs. The experiments demonstrate the effectiveness of the approach proposed in the paper.