An empirical review of calibration techniques for the pepper humanoid robot’s rgb and depth camera

Abstract

This paper presents a comparative study of different calibration techniques to align the Pepper Humanoid Robot’s depth camera with respect to its RGB camera. Both cameras are placed at different locations and have different viewpoints and view angles. We kept the image resolution the same for both cameras to avoid scaling issues and retrieved the translation and rotation coefficients. We used an in-house dataset for conducting experiments. The dataset consists of both RGB and depth images of single and multiple objects placed on a table. We used homography, fundamental matrix, and proposed translation estimation techniques to solve the alignment issue. Root mean square error and error variance are used as a measurement to evaluate the efficacy of the system.