FITTING A NORMAL PROBABILITY DISTRIBUTION TO DEPTH ESTIMATIONS OF THREE REALSENSE™ RGB-D CAMERAS TESTED IN SCENES WITH TRANSPARENCY

In the last decade, various companies have released different versions of RGB-D sensors, improving their performance at various levels (resolution, frame rate, robustness). These devices can measure depth using one of the following optical technologies: Structured-Light, Active Stereoscopy or Time-of-Flight / Lidar. This paper aims to evaluate and compare the performance of three low-cost RGB-D cameras in the estimation of depth of a wall when transparent elements such as glass and water are added to the field of view. We propose an experimental setup for data acquisition involving an aquarium (empty and filled with water). The evaluation is based on the statistical distribution and dispersion of the Normal Probability Distribution estimated for each case