基于三维变径螺旋线的路侧激光雷达和相机联合标定的方法

Roadside LiDAR-Camera Calibration Based on 3D Variable Diameter Spiral

车路协同技术是车联网发展新阶段迫切需求的一个方向,进行路侧激光雷达和相机设备的联合标定成为实现智慧道路的必要工作。鉴于路侧与单车智能传感器部署场景的不同,本文设计了一种适用于路侧联合标定的特殊标记物,主体结构为三维变径螺旋线。首先,在激光雷达和相机的共视区分别采集标记物的点云数据和图像数据;对于点云,采用改进的随机采样一致性算法提取螺旋线结构并进行点云补全;对于图像,采用语义分割技术获取标记物的像素信息,进行距离变换形成激励掩码。我们将标定计算流程分为两个阶段。在粗标定阶段,构建pnp问题的模型获得联合标定参数旋转矩阵R和平移向量t的初始值;在细标定阶段,构建目标函数,最大化标记物点云在图像掩码上的投影覆盖率来优化标定参数。实验结果验证了所提方法相比传统方法在标定效率、鲁棒性和精度上具有显著优势,为车路协同系统中传感器的空间校准提供了有效的解决思路。

Cooperative Vehicle Infrastructure System (CVIS) is a crucial direction in the development of Internet of Vehicles, making the spatial calibration of roadside LiDAR and camera essential for implementing intelligent roadside systems. Given the differences in deployment scenarios between roadside and onboard intelligent sensors, a unique marker is designed specifically for roadside spatial calibration, featuring a three-dimensional variable-diameter helical structure. Initially, point cloud and image of the marker are collected respectively in the common field of view of the LiDAR and camera. For the point cloud, an improved Random Sample Consensus (RANSAC) algorithm is used to extract the marker and complete the point cloud. For the images, semantic segmentation technology is employed to obtain pixel information of the marker and perform distance transformation to create an excitation mask. The calibration process is divided into two stages. In the coarse calibration stage, a model based on the perspective-n-point (PNP) problem is constructed to obtain initial values for the spatial calibration parameters, including the rotation matrix R and translation vector t. In the fine calibration stage, an objective function is constructed to optimize the calibration parameters by maximizing the projection coverage of the marker’s point cloud on the image mask. Experimental results demonstrate that the proposed method significantly outperforms traditional methods in terms of calibration efficiency, robustness, and accuracy, providing effective technical support for spatial calibration of sensors in Cooperative Vehicle Infrastructure System.