Accurately calibrate an embedded 3-D depth perception system

Nowadays, structured light systems are widely used in numerous applications such as robotic assembly, visual inspection, vision-aid surgery, etc. Among these systems, small-sized structured light systems have received significant attention recently. Miniaturization craft and embedded technology used for this kind of system can remarkably reduce system size to make it portable, convenient and real-time. However, the crucial issue for calibrating small-sized projector-camera system is still not well solved yet. In this paper, we present an improved systematic calibration method focusing on accurate calibration of a small-sized 3-D depth perception structured light system. In this system, the projector and camera are fixed at a short baseline. We identify that the degradation of precision mainly results from the limitation of pixel discrimination of depth data. Due to the inherent problem, we describe an improved calibration method which incorporates Phase-Shifting mechanism into Gray Code structured light patterns. With this method, a sub-pixel corresponding point pair between image plane and projector plane can be established with the help of an easy planar board. For each sub-pixel point pair, a 3-D Depth Data Box (DDB) is created via triangulation method. The accurate depth points in DDB are obtained using maximum likelihood estimation method. Then we can calibrate the camera and projector simultaneously using these 3-D target points and corresponding sub-pixel points on the camera/projector image plane. Some experiments results are presented to demonstrate the performance of this method.