密集匹配点云的自适应道路提取算法研究
|
摘要
多视倾斜影像密集匹配后能够生成海量点云数据,但数据本身缺乏有效的道路提取信息。针对此问题,提出一种密集匹配点云的自适应道路提取算法。首先对多视影像进行道路感兴趣区特征采样,获取影像区域道路的纹理特征,然后对三维点云进行去噪处理并剔除掉高大建筑物和植被。在点云中选取2个明显的道路点作为初始种子点,搜索种子点一定半径范围内的点,并统计这些点对应的法线与种子点法线的夹角,根据道路边沿法线的突变性提取出道路点。然后利用道路点拟合出道路中心线并根据最小包围矩形优化搜索搜索半径,进而自适应的生成其他道路种子点。实验结果表明,该方法提取的道路能够保持连通性,道路边界清晰,且提取耗时较短,具有较好的稳健性。
After intensive matching of multi-view tilted images,massive point cloud data can be generated,but the data itself lacks effective road extraction information.To solve this problem,an adaptive road extraction algorithm based on dense matching point cloud is proposed.Firstly,thefeatures of road of interest of multi-view images are sampled to obtain the texture features of the road in the image area.Then,the 3 D point cloud is denoised and the tall buildings and vegetation are removed.Two distinct road points are selected as the initial seed points in the point cloud.The points within a certain radius of the seed point are searched and the angle between the seed point normal and the normal corresponding to these points is counted.The road points are extracted in accordance with the mutation of the normal along the road edge.Then,the center line of the road is fitted by the road points and the searching radius is optimized according to the minimum bounding rectangle,which can further adapt to generate other road seed points.The experimental results show that the road extracted by this method can maintain its connectivity,with clear road boundary,less extraction time consumption and perfect robustness.
After intensive matching of multi-view tilted images,massive point cloud data can be generated,but the data itself lacks effective road extraction information.To solve this problem,an adaptive road extraction algorithm based on dense matching point cloud is proposed.Firstly,thefeatures of road of interest of multi-view images are sampled to obtain the texture features of the road in the image area.Then,the 3 D point cloud is denoised and the tall buildings and vegetation are removed.Two distinct road points are selected as the initial seed points in the point cloud.The points within a certain radius of the seed point are searched and the angle between the seed point normal and the normal corresponding to these points is counted.The road points are extracted in accordance with the mutation of the normal along the road edge.Then,the center line of the road is fitted by the road points and the searching radius is optimized according to the minimum bounding rectangle,which can further adapt to generate other road seed points.The experimental results show that the road extracted by this method can maintain its connectivity,with clear road boundary,less extraction time consumption and perfect robustness.
引文
[1]张继贤,林祥国,梁欣廉.点云信息提取研究进展和展望[J].测绘学报,2017,46(10):1460-1469.
[2]陈悦.基于倾斜摄影测量数据的点云划分[D].西安:西安科技大学,2016.
[3]赵明波,何峻,田军生,等.基于改进的渐进多尺度数学形态学的激光雷达数据滤波方法[J].光学学报,2013,33(3):292-301.
[4] Axelsson P.DEM Generation from Laser Scanner Data Using Adaptive TIN Models[J].International Archives of Photogrammetry and Remote Sensing,2000,33(Part B4):110-117.
[5]徐景中,万幼川,赖祖龙.机载激光雷达数据中道路中线的多尺度提取方法[J].红外与激光工程,2009,38(6):1099-1103.
[6] Choi Yun-Woong,Jang Young-Woon,Lee Hyo-Jong,et al.Three-dimensional Lidar Data Classifying to Extract Road Point in Urban Area[J].IEEE Geoscience and Remote Sensing Letters,2008,5(4):725-729.
[7]何培培,万幼川,黄桂平,等.城区LiDAR点云自适应坡度的滤波算法[J].遥感信息,2017,32(5):62-68.
[8]李海波,曹云峰,丁萌,等.基于三维点云聚类的坡度估计方法[J].计量学报,2018,39(3):304-309.
[9]郭志勇.多视影像密集匹配点云分类与测图应用研究[D].北京:中国地质大学(北京),2017.
[10]张爱武,李文宁,段乙好,等.结合点特征直方图的点云分类方法[J].计算机辅助设计与图形学学报,2016,28(5):795-801.
[11]李安福,曾政祥,吴晓明.浅析国内倾斜摄影技术的发展[J].测绘与空间地理信息,2014,37(9):57-59,62.
[12]杨望山,蔡来良,谷淑丹.提取城市道路边线的点云法向量聚类法[J].光子学报,2018,47(6):186-196.
[13]惠振阳.从机载LiDAR点云中提取城市道路网的关键技术研究[D].武汉:中国地质大学,2017.
[2]陈悦.基于倾斜摄影测量数据的点云划分[D].西安:西安科技大学,2016.
[3]赵明波,何峻,田军生,等.基于改进的渐进多尺度数学形态学的激光雷达数据滤波方法[J].光学学报,2013,33(3):292-301.
[4] Axelsson P.DEM Generation from Laser Scanner Data Using Adaptive TIN Models[J].International Archives of Photogrammetry and Remote Sensing,2000,33(Part B4):110-117.
[5]徐景中,万幼川,赖祖龙.机载激光雷达数据中道路中线的多尺度提取方法[J].红外与激光工程,2009,38(6):1099-1103.
[6] Choi Yun-Woong,Jang Young-Woon,Lee Hyo-Jong,et al.Three-dimensional Lidar Data Classifying to Extract Road Point in Urban Area[J].IEEE Geoscience and Remote Sensing Letters,2008,5(4):725-729.
[7]何培培,万幼川,黄桂平,等.城区LiDAR点云自适应坡度的滤波算法[J].遥感信息,2017,32(5):62-68.
[8]李海波,曹云峰,丁萌,等.基于三维点云聚类的坡度估计方法[J].计量学报,2018,39(3):304-309.
[9]郭志勇.多视影像密集匹配点云分类与测图应用研究[D].北京:中国地质大学(北京),2017.
[10]张爱武,李文宁,段乙好,等.结合点特征直方图的点云分类方法[J].计算机辅助设计与图形学学报,2016,28(5):795-801.
[11]李安福,曾政祥,吴晓明.浅析国内倾斜摄影技术的发展[J].测绘与空间地理信息,2014,37(9):57-59,62.
[12]杨望山,蔡来良,谷淑丹.提取城市道路边线的点云法向量聚类法[J].光子学报,2018,47(6):186-196.
[13]惠振阳.从机载LiDAR点云中提取城市道路网的关键技术研究[D].武汉:中国地质大学,2017.