An adaptive method for registration of RGB-D data

In this paper an adaptive method for indoor mapping and registration of environment with texture less regions using RGB-D data is presented. Basically, four main points involved in the method are here discussed: the calibration of the sensors integrated in the Kinect device; the registration of 3D point clouds pairs acquired successively by the imaging sensor; the inclusion of weights in the point-to-point ICP algorithm; and a global adjustment for the refinement of transformation parameters. The sensors calibration is realized using the Zhang method. The 3D rigid body model is employed to calculate the transformation parameters. The weights are defined based on the theoretical accuracy of depth measurements. A linear model is proposed for global adjustment. Experiments were conducted and the results showed the efficiency and potentially of the proposed method.

Keywords :
RGB-D cameras; camera calibration; indoor mapping; registration of RGB-D data; global adjustment, ICP.