顾及空间上下文关系的JointBoost点云分类及特征降维

随着激光雷达技术的发展及广泛应用，点云数据的分类及理解成为了目前一个研究热点。本文研究了较复杂的电力线路走廊场景的点云自动分类方法，目标类别为地面、植被、建筑物、电力塔、电力线等。本文首先归纳、定义了点云分类所需的关键特征，并利用JointBoost实现地物分类；同时，考虑到点云数据量大，其分类速度较慢，本文结合地物空间上的相互关联关系，提出了一种序列化的点云分类及特征降维方法。该方法在保证分类精度的前提下，使分类所需特征维数降低，缩短了分类所需时间。实际的电力线路走廊的激光扫描点云数据分类实验证明本文研究的分类方法的有效性。

Points Cloud Classification using JointBoost Combined with Contextual Information for Feature Reduction

The requirements of 3D scene classification and understanding have dramatically increased with the widespread using of airborne LiDAR. This paper therefore focuses on complex power-line corridors scenes and presents an approach to automatically classify point clouds in building, ground, vegetation, power-line, and tower classes. Many key features of points cloud are introduced in this paper for classification using the JointBoost classifier. Due to the data of points cloud is “Big Data” and its classification rate is slow, we propose a method of serialized points cloud classification using spatial contextual information between objects for features reduction. The experiments prove that the classification method we study in this paper can be effectively used for points cloud classification in power-line corridors scenes.