为了保障无人驾驶车的可靠性与安全性,首先根据多线激光雷达点云数据的特征结合区间共线点特征获取路沿点并利用改进的DPCA 算法对得到的路沿点进行聚类。其次,应用最小二乘法拟合出两侧路沿。最后,通过改进的 DPCA 算法将路面上的障碍物点进行聚类并计算得到障碍物的信息。利用区间共线点特征提取路沿点不易受其他干扰点的影响,而改进的DPCA 算法则能够自动并准确地获取聚类中心,解决了DPCA 算法需要人工选取聚类中心的缺陷。实车实验证明该算法的实时性与准确性。

In order to ensure the reliability and security of the driverless car, firstly, the curb point set is obtained by combining the feature of the multi-layer lidar return data with the interval collinear point features and the resulting road-point clustering is achieved by using the improved DPCA algorithm. Then, both sides of the curb are fitted using the least square method. Finally, the improved DPCA algorithm is used to cluster the obstacle points on the road surface, and the information of the obstacles is calculated. Using the method of interval collinear point features, the curb points will not be affected by other interference points. The improved DPCA algorithm can automatically and accurately obtain cluster centers, which solves the defect that the DPCA algorithm needs to manually select the clustering center. The real vehicle experiment proves the real-time and accuracy of the algorithm.