偏度特征约束下的机载激光雷达点云数据分类

Classification of Airborne Lidar Point Cloud Data with Skewness Feature Constraint

摘要:

机载激光雷达获得的点云具有密度低、分布不均匀、分支结构不清晰等特点，其动态扫描过程的数据特征动态偏差很小，无法提取有效的数据去噪特征。为此提出偏度特征约束下的机载激光雷达点云数据实时分类方法。该方法将扫描获取的点云大容量实时数据引入在正态分布中，利用衡量对称性正态分布的关键度量偏度特征作为动态特征分界约束，完成数据滤波；提取机载激光雷达点云特征，从中选取优质特征，以此构建SVM分类器。点云大容量数据训练结果即为最终的分类结果。实验结果表明，所提方法分类的准确性与效率较高。

Abstract:

The?point?cloud?obtained?by?airborne?lidar?has?the?characteristics?of?low?density,?uneven?distribution,?unclear?branch?structure,?etc.?The?dynamic?deviation?of?data?features?in?the?dynamic?scanning?process?is?very?small,?and?it?is?unable?to?extract?effective?data?denoising?features.?Therefore,?a?real-time?classification?method?of?airborne?lidar?point?cloud?data?under?the?constraint?of?skewness?features?is?proposed.?In?this?method,?the?large?capacity?real-time?data?of?point?cloud?obtained?by?scanning?is?introduced?into?the?normal?distribution,?and?the?key?metric?skewness?feature?measuring?the?symmetry?of?the?normal?distribution?is?used?as?the?dynamic?feature?boundary?constraint?to?complete?the?data?filtering;?The?point?cloud?features?of?airborne?lidar?are?extracted,?from?which?high-quality?features?are?selected?to?build?a?SVM?classifier.?The?final?classification?result?is?the?result?of?point?cloud?high-capacity?data?training.?The?experimental?results?show?that?the?proposed?method?has?high?accuracy?and?efficiency.