RANGE CAMERA CALIBRATION

Abstract

Range cameras are used in applications such as robot navigation, object recognition, automated inspection and so on. Camera calibration is the process of establishing the relationship between 30 world coordinates and their corresponding image frame coordinates. This project is based on two-stage method proposed by Tsai. Only linear equations are involved in the first stage but non-linear optimization is needed in the second stage. Both the intrinsic and extrinsic parameters are calibrated. Non-coplanar and coplanar methods are implemented. Point features which are the corners of squares are used as calibration points. An algorithm is implemented for detecting these corners to sub-pixel resolution. A triangulation method is developed for assessing the accuracy of calibrated parameters. In this triangulation method, two line-of-sights are intersected in 30 space to calculate the world coordinates of the object. From the results, it is found that the accuracy is within the expected range.