Research On Calibration Algorithm Of Kinect Depth Camera

3D reconstruction is the key technology for computer to express the objective world, it means to build a three-dimensional model which is suitable for computer to handle with. 3D reconstruction technique is the common scientific issue and core technology in computer vision, virtual reality, computer animation and other relevant areas. In recent years, Kinect is very popular among developers due to its advantages in measuring depth, recognition voice and capturing motion, and it has been used on different platforms widely. This paper study on calibration algorithm of Kinect camera, and the main research results obtained are as follows:1. Analyzing relevant calibration algorithms of Kinect camera from domestic and abroad. At present, there are a lot calibration algorithms about Kinect camera, to name a few, Herrera‘s calibration toolbox which is based on checkerboard, Shibo Li and others use several holes to form the checkerboard, in addition, Weihua Liu et al. provide the calibration algorithm based on one dimensional calibration object, and Aaron Staranowicz et al. use the QR decomposition to calibrate the camera based on the relationship between the color and depth camera.2. Studying the calibration algorithm of parameters of Kinect depth camera. Based on analysis of relevant algorithms form domestic and abroad, we propose a simple and effective calibration algorithm based on spatial sphere. This algorithm use the transform relationship between the Euclidean space and projective space of quadric to solve the parameters, the experimental results show the effectiveness and feasibility of the method.3. Studying the calibration algorithm of relative position of multi-Kinect depth cameras. How to compute the external parameters of multi-Kinect simultaneously is a common problem when we use multi-Kinect to reconstruct the scene. The relative position calibration of multiple cameras is mainly using the transformation between corresponding points. However the planar calibration object is generally difficult to be viewed by multiple cameras form different views simultaneously, the spatial sphere can do. This paper presents a relative position calibration algorithm of multi-Kinect based on spatial spheres. We also show this method can be used in internal parameters calibration of multi-Kinect.4. Build a body scanning system based on the algorithms proposed in this paper. Based on the internal and external parameters computed from our calibration algorithm, we developed a body scanning system. The experimental results show that the proposed calibration algorithm based on spatial spheres is practical, accurate and can be applied to various situations.