中文摘要：

针对利用Lidar数据进行桥梁自动识别中存在的检测效率较低,检测结果与桥梁形状有关以及检测精度受植被影响等问题,提出了一种改进的Lidar数据桥梁提取算法.通过对Lidar数据格网化,将求点的邻域点问题转化为先求网格的邻域网格,达到快速获取离散点邻域点的目的;构造了一种新的三维离散点形态学算子,滤除植被对桥梁检测的影响;利用并查集优化剖面分析算法中最小生成树求解和连通域处理,可滤除建筑物等大物体;利用优化的剖面分析方法并结合桥梁的拓扑特点提取桥梁,解决算法仅能检测特定形状桥梁的问题;为解决Lidar数据量大引起的检测效率问题,采用OpenMP实现算法并行.通过桥梁提取实验验证了算法的有效性和高效性.

英文摘要：

To deal with the existing problems of studies on bridge detection based on Lidar data：slow speed,extraction result is related to bridge shapes,detection accuracy is influenced by vegetation and so on.An improved bridge extraction algorithm is given.Firstly,with gridding the Lidar data,the neighborhood of the grid instead of the grid point is obtained,resulting in a fast way to get the neighborhood of discrete points.Together with a new morphological operator of three dimensional discrete points,vegetation is filtered effectively.Secondly,union-find sets is introduced to optimize two parts of the profile analysis method,MST and the connected area searching,and the buildings are filtered.Finally,with the optimization profile analysis method and the bridges＇ topology characteristics,the bridges are extracted and this method is independent of bridge shape.Meanwhile,in order to solve the efficiency problem caused by large Lidar data,OpenMP is introduced to realize the parallel processing successfully.In the last part,the feasibility and effectiveness are showed by bridge detection experiment.