Heterogeneous and asynchronous optical tracking system for mobile vehicles surveillance.

Abstract

Given the need for low-cost ground truth in the laboratory with comparable performance to the commercially available motion capture systems, this study proposes off-the-shelf hardware and open-source software for a motion capture system. In the proposed scheme, cameras track passive markers based on their infrared reflection. Four V2 NoIR Raspberry cameras were applied with the goal of subcentimeter accuracy at 100 Hz. Algorithms for clock synchronization, quick contour extraction, and intrinsic camera calibration are ad- dressed. Construction issues of non-uniform lighting or noisy reflections are also discussed. As a result of this study, a correction of the OpenCV extrinsics calibration library, a cal- ibration toolbox for the generic lens model, and an innovative ordering algorithm for the markers were proposed. The system’s performance achieved subcentimeter precision com- pared to industrial arm encoder data. Three drone flights are also included as examples. The code is already public, and further study will be in charge of fine-tuning the available configuration for multiple vehicles.