点云分类是机载LiDAR点云应用于城市建模、道路提取等的重要阶段。虽然点云分类的方法有很多,但依然存在如多维特征向量信息冗余、复杂场景下点云分类精度不高等问题。针对这些问题,提出一种基于多基元特征向量融合的点云分类方法。该方法分别基于点基元和对象基元提取特征向量,并结合色彩信息,利用随机森林对点云数据进行分类。实验结果表明,所提的多基元分类方法相较于单基元分类方法能够获得更高的分类精度。为了进一步分析随机森林用于点云分类的有效性,分别使用支持向量机(SVM)以及反向传播(BP)神经网络进行对比分析。实验结果表明,随机森林方法所获得的三组点云分类结果在召回率以及F1得分两个评价指标中均高于另外两种方法。

Point cloud classification is an important stage in the application of airborne LiDAR point cloud in urban modeling and road extraction. Although there are many methods for point cloud classification, there are still some problems such as multi-dimensional feature vector information redundancy and low accuracy of point cloud classification in complex scenes. To solve these problems, a point cloud classification method is proposed based on multi- entity eigenvector fusion. The method extracts the feature vectors based on point entity and object entity and classifies the point cloud data by using random forest combined with color information. The experimental results show that the proposed multi-entity classification method is more accurate than the single-entity classification method. In order to further analyze the validity of random forest for point cloud classification, the support vector machine (SVM) and the back propagation (BP) neural network are used for a comparative analysis. The experimental results show that the three groups of point cloud classification results obtained by the random forest method are higher than those by the other two methods in the recall rate and F1 score.