基于室内结构特征的二维激光雷达与相机标定方法

Calibration Method of 2D LIDAR and Camera Based on Indoor Structural Features

针对激光雷达与相机数据融合问题,提出了一种基于室内结构特征的二维激光雷达和彩色相机之间的标定方法.该方法利用室内场景的柱特征,根据点在线上的位置关系推导激光雷达与相机坐标系间的投影矩阵方程.采用Canny算子和Hough变换提取墙角图像中直线特征,利用RANSAC方法拟合点云中的角点特征.通过奇异值分解法求解投影矩阵,在剔除重投影误差较大的点线特征数据后,进一步优化标定结果.实验结果表明:采用优化后的投影矩阵,柱子特征点的平均重投影误差从0.3755个像素降低至0.0459个像素.与两步标定法相比本文提出的算法具有更好的重投影效果.同时该方法不需要特定的标定物体,仅用较少的时间即可实现较高的投影精度.

Aiming at the problem of LIDAR and camera data fusion,a calibration method of 2D LIDAR and color camera based on indoor structural features was proposed.According to the position relationship between the point and lines,pillar features of the indoor scene are used to derive the projection matrix equation between the 2D LIDAR and the camera coordinate systems.The Canny operator and Hough transform are employed to extract the line features of the corner image.Afterwards,the RANSAC method is used to fit the corner features from the point cloud.The projection matrix is solved by the singular value decomposition method.Finally,the calibration results are further optimized after removing the data with large reprojection errors.The experimental results show that the average reprojection error of the pillar feature point is reduced from 0.3755 pixels to 0.0459 pixels by adopting optimized projection matrix.Compared with the two-step calibration method,the algorithm proposed in this paper has a better reprojection effect. At the same time,this method does not require a specific calibration object,and canachieve high projection accuracy in less time.