基于点云聚类评估的激光雷达鲁棒定位方法CSCD

基于先验地图的激光雷达定位方法在封闭工业场景下得到了广泛应用,然而环境变化、行人和车辆等动态物体的干扰会影响激光雷达与先验地图的匹配精度。提出了一种动态环境下基于点云聚类评估的三维激光雷达鲁棒定位方法:通过设定角度和距离双阈值,对点云深度图像进行分割聚类,相较于传统分割方法,分割结果对点云噪声更为鲁棒;通过对原始点云进行分割聚类,在粗匹配结果下评估聚类的匹配度,剔除误匹配聚类进行二次匹配以提高匹配精度;通过聚类评估的结果判断匹配成功和失败的点对,进而对点云整体匹配结果的正确性进行评估,相较于传统仅基于距离阈值的判断准则,具有更高的准确性;最终,分别通过公开数据集和实际试验验证了该算法的有效性。试验结果表明,相较于传统匹配方法,该方法有效提高了动态场景下的定位精度和匹配结果评估的准确度,定位误差可以维持在10cm以内。

Robust 3D Lidar Localization Based on Point Cloud Clustering Evaluation

Lidar localization based on priori map has been widely used in closed industrial scenes. However, environmental changes and interference from dynamic objects such as pedestrians and vehicles can affect the matching accuracy of Lidar and prior maps. A robust 3D lidar localization method based on point cloud clustering evaluation in dynamic environment is proposed. The range image of the point cloud is segmented and clustered by setting the dual thresholds of angle and distance, with the results more robust to noises compared with traditional segmentation method. After segmentation and clustering of the raw point cloud, the matching degree of the clusters is evaluated under the rough matching result. The mismatched clusters are eliminated to achieve a more accurate matching result in the second matching process. The result of clustering evaluation is used to judge whether the corresponding point pairs are correct, and then the correctness of matching result of the whole point cloud is evaluated. Compared with the traditional judgment criterion based only on distance threshold, the method proposed in this paper achieves higher accuracy. Finally, the effectiveness of the algorithm proposed in this paper is verified through the public dataset and practical experiment, respectively. The experiment results show that compared with the traditional method, the method proposed in this paper effectively improves the localization accuracy and the evaluation accuracy of the matching result in dynamic scenes. The localization error can be maintained within 10cm.