A two-stage ALS point cloud segmentation framework for urban areas. Issue 7 (March 2020)

• Record Type: Journal Article
• Authors: Han, Wei; Huang, Daqing; Xu, Cheng
• Is Part Of: IOP conference series. Volume 768:Issue 7(2020)
• Publication Date: 2020-03
• Subjects: Materials science -- Periodicals; 620.1105
• DOI: 10.1088/1757-899X/768/7/072079
• Languages: English
• ISSNs: 1757-8981

Abstract

Abstract: In this paper, a novel methodology for semantic classification of airborne laser scanning point clouds is introduced in this paper. A two-stage framework integrating point-based classification and cluster-based rules is proposed with a general end to end processing scheme, which is from scattered airborne laser scanning (ALS) data to points with semantic labels. In the first stage, energy function consists of a point-wise-based soft labelings term obtained by random forest classifier and a spatially smooth labelling term. Then, a constrained mean-shift-based clustering algorithm combined with sematic rules is proposed to refine the hard labelings of class-wise point clouds, which can be as a post-processing stage. To verify the effectiveness of our method, the classification results derived for an ALS benchmark dataset are shown in Section 3. Base on the semantic classification results, the two-stage framework can be demonstrated with high accuracy and high plausibility in comparison to some classic approaches.