A Novel Extrinsic Calibration Method of ToF Cameras Based on a Virtual Multi-cubes Shaped Object

Extrinsic calibration accuracy for time-of-flight (ToF) cameras is affected by accumulated errors which are generated by recognition of conventional objects with multi-characteristics such as checkerboard. In order to eliminate accumulated errors, in this paper, we propose a novel extrinsic calibration method of ToF cameras with a virtual multi-cubes shaped object. First, we establish the model of a 3-axis translation stage composed of three linear translation stages which are orthogonal to each other. Then, a virtual multi-cubes shaped object with multi-characteristics is generated from an optimized combination of multi-motions of the 3-axis translation stage. After recognizing corner characteristics of this multi-cubes shaped object, the proposed method is accomplished by performing the least square method. Our experimental results show that the measurement accuracy of ToF camera is improved from +/-10mm to 6.85mm, which is much better than that of the conventional method based on a 2D plane checkerboard. The proposed method has the ability to improve calibration accuracy to a high level. It may find great potential applications in many fields.