基于随机森林的全波形点云数据分类研究

针对机载全波形数据,本文提出了一种基于随机森林法的点云分类算法,首先通过全波形分解获得振幅,回波次数,以及回波宽度3个特征,再结合本文中提出的特征提取方法,构建一个多维特征向量并进行特征选择,最后利用随机森林法将激光点云分为植被,地面部分以及建筑物三类.对比支持向量机分类方法,实验证明该方法具有很好的稳定性以及高效性,能够在城市区域取得很好的分类精度.

Research on Classification for Airborne Full Waveform Lidar Data Based on Random Forest

Aiming at airborne full waveform data classification,the paper proposes a random forest method based on point cloud classification algorithm.In this method,the amplitude of the full-waveform echo and echo times,as well as the echo width are extracted.Then extracting and selecting the features using the method proposed in this paper,building a feature vector.Finally,using the random forest method,the laser point clouds are divided into three types of vegetation,ground and building.By comparative analysis of random forests method and the support vector machine,the results presented in the paper that the extracted features show good stability and efficiency,and the random forest classification method can achieve good classification effect in the urban classification applications.