Three-dimensional videography using omnidirectional cameras: an approach inspired by the direct linear transformation method.
Three-dimensional videography using omnidirectional cameras: an approach inspired by the direct linear transformation method.
复制标题
使用全向相机进行三维摄像：一种受直接线性变换方法启发的方法。
DOI：
10.1016/j.jbiomech.2021.110722
复制
发表时间：
2021
影响因子：
2.4
通讯作者：
A.
中科院分区：
文献类型：
--
作者：
Nagano;A.
直接线性变换(DLT)是运动分析中最常用的方法之一。这种方法是在1971年发展起来的，随后被用于许多研究。然而，由于相机的结构，这种方法带来了不利的挑战。由于DLT方法是基于普通光学相机的结构，其中使用透视投影，因此相机的视角受到限制。在这项研究中，我们提出了一种使用全向摄像机的三维(3-D)标定和重建方法，该方法比传统的摄像机具有更宽的视角。对于兴趣点，从全方位相机拍摄的图像中获得天球的经度和纬度。兴趣点相对于相机参考帧的方向由经度和纬度确定。在标定阶段，全向摄像机相对于全局参考系的位置和方向由具有已知三维坐标的控制点的图像确定。在重建阶段，确定来自摄像机的光线的交点。通过评价实验，验证了该方法的准确性。利用这种方法可以充分利用全向摄像机提供的宽视角。
Direct linear transformation (DLT) is one of the most frequently used methods in motion analysis. This method was developed in 1971 and has subsequently been used in numerous studies. However, owing to the structure of the camera, unfavorable challenges are associated with this method. Because the DLT method is based on the structure of ordinary optical cameras in which the perspective projection is used, the camera’s angle of view is limited. In this study, we propose a three-dimensional (3-D) calibration and reconstruction method using omnidirectional cameras, which have significantly wider angles of view than conventional cameras. For the interest points, the longitude and latitude of the celestial sphere are obtained from the image captured with an omnidirectional camera. The direction of the interest points relative to the camera’s reference frame is determined via the longitude and latitude. In the calibration phase, the position and orientation of the omnidirectional camera relative to the global reference frame are determined from the image of the control points with known 3-D coordinates. In the reconstruction phase, the intersection point of the rays from the cameras is determined. Using an evaluation experiment, we verified the accuracy of the proposed method. The wide angle of view provided by omnidirectional cameras can be fully exploited using this method.