摄像机与激光测距仪(Camera and laser rangefinder, LRF) 被广泛应用于机器人、移动道路测量车、无人驾驶等领域. 其中, 外参数标定是实现图像与LIDAR 数据融合的第一步, 也是至关重要的一步. 本文提出一种新的基于最小解(Minimal solution) 外参数标定算法, 即摄像机与激光仅需对标定棋盘格采集三次数据. 本文首次提出虚拟三面体概念, 并以之构造透视三点问题(Perspective-three-point, P3P) 用以计算激光与摄像机之间的坐标转换关系. 相对于文献在对偶三维空间(Dual 3D space) 中构造的P3P 问题, 本文直接在原始三维空间中构造P3P 问题, 具有更直观的几何意义, 更利于对P3P问题进行求解与分析. 针对P3P 问题多达八组解的问题, 本文还首次提出一种平面物成像区域约束方法从多解中获取真解,使得最小解标定法具有更大的实用性与灵活性. 实验中分别利用模拟数据与真实数据对算法进行测试.算法结果表明,在同等输入的条件下,本文算法性能超过文献中的算法. 本文所提的平面物成像区域约束方法能从多解中计算出真解,大大提高了最小解算法的实用性与灵活性.%Camera and laser rangefinder (LRF) are widely utilized in various applications, such as robotics, road survey and mapping vehicle, autonomous driving, etc. This paper presents a novel minimal solution based method for extrinsic calibration of a camera and a 2D LRF by introducing a virtual trihedron from 3 input chessboard planes. We formulate the perspective-three-plane (P3P) problem with the virtual trihedron to compute the extrinsic parameters between the camera and the LRF coordinate systems. Compared to the P3P problem in dual 3D space in existing methods, the proposed P3P problem is directly formulated in the primal 3D space, which is more intuitive and straightforward to derive and analyze the solutions. This paper also presents a novel geometric constraint, called restricted area on plane image (RAPI), to derive the unique real solution from up to 8 solutions. Results of both simulation and real data experiment demonstrate that the proposed method outperforms existing methods in terms of the same input image and LIDAR data. Moreover, the proposed method can derive the real solution from multiple solutions, thus making the minimal solution based method more practical and flexible.
展开▼
