改进移动窗口的道路区域识别算法研究

Algorithm for Road Area Identification with Improved Moving Windows

针对车载激光扫描系统进行道路区域识别时出现的边界点云提取的准确性与精度问题,文中提出一种基于扫描线的移动窗口道路区域识别方法。对于无扫描线信息的离散点云,利用点云GPS时间和扫描角度将道路原始点云转换成有序扫描线条带,分析不同目标在扫描线中的空间分布和统计特征,构建多尺度、多窗口组合的移动窗口滤波方法。实验结果表明:采用该方法进行道路区域识别可有效减少边界处非地面点的影响,其综合评价指标在城市区域和复杂街区的准确度分别为98.11%和93.33%,有效提高了道路边界点云提取的准确性。

In order to improve the accuracy and precision of boundary point cloud extraction with the vehicle-mounted laser scanning system for road area recognition,this paper presents a method for road area identification of moving windows based on scan line.For the discrete point cloud without information of scan lines,the road original point cloud is converted into the ordered scan line bands using the GPS time and scanning angle of point cloud.The spatial distribution and statistical characteristics of different targets in the scan line are analyzed to construct the moving window filtering method in which windows of different sizes and combination windows are used.The experimental results show that the influence of non-ground points at the boundary of road area identification can be reduced effectively by the new method,and the accuracy of the comprehensive evaluation indexes in urban areas and complex blocks is 98.11%and 93.33%,respectively.It is concluded that the accuracy of road boundary point cloud extraction is improved by the new algorithm.