Human observation-based calibration of multiple RGB-D cameras for intelligent space

In this paper, we propose a method to calibrate multiple RGB-D cameras for the Intelligent Space (iSpace) using human observation. The iSpace, that is consisted of many sensors, can understand what is happening in it and can support people in it based on the understand results. In the iSpace, we can add and remove the sensors easily. The proposed method is suitable for calibrating such system and allows us to calibrate the system without stopping the system. The proposed calibration uses human position observations as feature points. In addition, our method consists of two steps for achieving accurate calibration. In the first step, we conduct calibration based on planar surfaces for initial alignments. Then, we calibrate all RGB-D cameras using human position observations. Experimental results show that our calibration method have enough accuracy for recording and merging 3-D information from various information captured by each camera.