Synchronization Error Compensation of Multi-view RGB-D 3D Modeling System

Finding 3D transformation relationship between multiple RGB-D cameras accurately is necessary to generate a complete 3D model from depth images. In this paper, we propose a convenient method to find the 3D transformation between multiple RGB-D cameras using centroids of an independently moving spherical object. In a multiple camera system, we should always consider the synchronization problem. Therefore, we first define how we managed to compensate for this problem. Next, we introduce a method to find the centroids of the moving sphere using the RANSAC algorithm and refine them based on the movement distance values. Finally, we calculate the 3D transformation relationship using the refined centroids and evaluate the robustness of the proposed calibration method by comparing 3D reconstruction results.