基于3D点云语义地图表征的智能车定位

为提高智能车节点定位准确率, 研究了基于3D点云语义地图表征的智能车定位方法。该方法分为3个部分: ①基于三维激光点云的语义分割, 包括地面分割, 交通标志牌分割和杆状语义目标分割; ②面向智能车的点云语义地图表征, 利用分割的语义目标投影, 生成带权有向图, 语义路, 语义编码, 再以语义编码和高精度GPS的全局位置组成语义地图表征模型; ③基于语义表征模型的智能车定位, 包括基于GPS匹配的粗定位和基于语义编码渐进匹配的节点定位。实验在3种长度不同、复杂度不同的道路场景下进行, 节点定位准确率分别为98.5%, 97.6%和97.8%, 结果表明所提出的定位方法节点定位准确率高、鲁棒性强且适用于不同的道路场景。 

A Localization Method for Intelligent Vehicles Based on Semantic Map Representation Extracted from 3D Cloud Points

An intelligent vehicle localization method based on the semantic-map representation of 3D point clouds is proposed to improve the accuracy of node localization for intelligent vehicles. The method is divided into three parts. ① Semantic segmentation based on 3D laser point clouds includes segmentation of ground, traffic sign, and pole-shaped targets. ② Semantic-map representation for intelligent vehicles uses segmented targets to project. Directional projections with weight, semantic roads, and semantic coding are generated. The coding and global location from high-precision GPS make up the representation model. ③ Localization based on a semantic representation model includes coarse positioning from GPS matching and node localization from semantic coding matching. The experiments are performed in three road scenes with different lengths and complexities, and their localization accuracy is 98.5%, 97.6%, and 97.8%, respectively. The results show that the proposed method has high accuracy and strong robustness, suitable for different road scenes.