基于双视点特征匹配的激光-相机系统标定方法

Calibration method for integrated camera and laser scanner system based on bi-view feature matching

针对传统的基于棋盘格的标定方法存在耗时且精度不高的问题,提出一种新的基于双视点特征匹配的激光-相机系统标定方法。采用自然场景中的普通物体作为标定物,通过三维激光扫描仪和相机分别采集两个不同视点下关于标定物的激光点云和图像数据,首先利用迭代最近点(iterative closet point,ICP)法对2组激光特征点集进行注册以获取两视点间的位姿转换关系,并结合立体三角化将第1个视点下获得的二维图像特征点转换到三维相机坐标系中,最后通过匹配同一视点下对应的三维激光与相机特征点集,求解出激光与相机坐标系间的外部标定参数。实验结果表明该方法能有效提高标定精度,缩短标定耗时。

Aiming at the problems of time-consuming and low precision existing in traditional checkerboard-based calibration method, a bi-view feature matching-based calibration method for a joint laser/camera system is proposed. This method uses an ordinary object in natural scene as the calibration pattern. The 3D laser scanner and camera are used to acquire the laser point cloud and image data of the calibration pattern from two different viewpoints respectively. The iterative closest point（ICP） algorithm is used to register the laser feature points and obtain the pose transform relation between the two viewpoints. The feature points of the 2D image obtained from the first viewpoint are transformed into 3D camera coordinate system through the stereo triangulation. The corresponding 3D feature point sets of laser and camera from the same viewpoint are matched, and the ultimate extrinsic calibration parameters between laser and camera coordinate systems are obtained. Experimental results indicate that the proposed method can effectively improve the calibration precision and shorten the calibration time.