Modeling the RMS Error in Depth Measurements of Stereo Cameras on Mobile Robots as a function of Distance and Speed
Stereo camera, RMS error, depth, speed
In this work, we propose a method for estimating the RMS error in the depth data provided by a stereo camera arranged on a mobile robot, according to its speed of displacement and distance. We propose to use two sets of 3D points, the first one being an ideal set (or ground-truth) that can be obtained using checkerboards, artificial markers, GPS for interiors, or a LIDAR. The second set is called measured and can be obtained with the camera. By computing the differences (Euclidean distance) between these two sets of points an RMS error value can be obtained for some speeds and distances. After obtaining the RMS error values, we can obtain a mathematical representation of a continuous surface representative of the RMS error, using curve or surface adjustment, probability, stochastic processes, machine learning, etc. This work should become an important contribution to future work that requires, as a parameter, the accuracy of the depth data provided by a stereo camera, as a function of speed and distance.