RGB-D Sensors Extrinsic Calibration In Controlled Environment

Abstract

RGB-D sensors provide rich information about the scene including color and 3D data. This information is valuable for the forthcoming analysis on the object of interest or the scene as a whole. Depending on the task, multiple sensors might be needed in order to provide details about process under observation. In this case, calibration between the sensors is necessary in order to provide consistent information related to the 3D world. A lot of work has been done related to RGB-D sensor calibration, since the introduction of Kinect in 2010, however it is still an active area of research. In this paper we propose an algorithm for RGB-D sensors extrinsic calibration, in a scenario where each sensor is located in closed and known environment. This knowledge of the environment helps to exploit its structure in the most optimal way for the calibration process. It includes planes locations and planes relative position. We propose two stage procedure for extrinsic calibration, in pairwise and global manner. The experimental results validate our approach for real applications.