测绘学报

• 学术论文 • 上一篇    下一篇

面向车载激光扫描点云快速分类的点云特征图像生成方法

杨必胜1,魏征2,李清泉1,毛庆洲2   

  1. 1. 武汉大学测绘遥感信息工程国家重点实验室
    2. 武汉大学
  • 收稿日期:2009-10-19 修回日期:2010-03-17 出版日期:2010-10-25 发布日期:2010-10-25
  • 通讯作者: 魏征

A Classification-oriented Method of Feature Image Generation for Vehicle-borne Laser Scanning Point Clouds

  • Received:2009-10-19 Revised:2010-03-17 Online:2010-10-25 Published:2010-10-25

摘要: 车载激光扫描是空间数据快速获取的一种重要手段。车载激光扫描点云数据的分类和特征提取是目标识别与三维重建的基础。本文以车载激光点云数据为研究对象,提出了一种适合于其快速分类与目标提取的点云特征图像生成方法。该方法首先将扫描区域进行平面规则格网投影,通过分析格网内部点云的空间分布特征(平面距离、高程差异、点密集程度等)确定激光扫描点的定权,从而生成车载激光扫描点云的特征图像。利用生成的点云特征图像,可采用阈值分割、轮廓提取与跟踪等手段提取图像分割的建筑物目标的边界,从而确定边界内部点云数据,实现目标分类与提取。本文以Optech公司的车载激光扫描数据为实验对象,验证了本文提出方法的可行性和实用性。实验结果表明,该方法能快速有效分离出车载激光扫描点云中的地面数据、建筑物数据等。

Abstract: Vehicle-borne laser scanning is a popular and rapid means to capture dense point clouds in urban areas. Automated classification of point clouds is a precondition for further object extraction, segmentation, 3D reconstruction. This paper proposes an efficient method of feature image generation of point clouds to automatically classify dense point clouds into different categories, such as terrain points, building points. The proposed method first uses planar projection to sort points into different grids, then calculates the weights and feature values of grids according to the distribution of laser scanning points, and finally generates the feature image of point clouds. Thus, the proposed method adopts contour extraction and tracing means to extract the boundaries and point clouds of man-made objects (e.g., buildings and trees) in 3D based on the image generated. Experiments show that the proposed method provides a promising solution for classifying and extracting man-made objects from vehicle-borne laser scanning point clouds.