Calibration and 3D Reconstruction of TOF Camera Based on Marker
Wentao Liu, Bingyu Sun · Proceedings of the 2021 5th International Conference on Electronic Information Technology and Computer Engineering · 2021
The pointcloud registration is the first step of 3D reconstruction. However, the classical iterative nearest point (ICP) algorithm and its variants have some problems, such as unable to deal with lowoverlapping pointcloud data, relying too much on normal vector, curvature and other features, and slow speed. To solve these problems, this paper proposes a multi-camera calibration algorithm based on marker. Firstly, Azure-Kinect devices from different perspectives detect the corners of the marker and identify the ID. After converting the 2D corners into 3D corresponding points, combined with the manually entered world coordinates, the rigid transformation parameters of the corresponding camera are calculated through Procrustes analysis to complete the global registration. Secondly, the local registration is performed by using the Colored-ICP algorithm to optimize the gap between adjacent view pointclouds. In order to intuitively perceive the quality of reconstructed model, Hololens2 is deployed for holographic display. The experimental results have shown that the proposed algorithm framework improves the accuracy and speed of 3D reconstruction, and simplifies the operation process of traditional offline calibration algorithm to a certain extent.