SHREC 2021: 3D point cloud change detection for street scenes. (October 2021)

• Record Type: Journal Article
• Authors: Ku, Tao; Galanakis, Sam; Boom, Bas; Veltkamp, Remco C.; Bangera, Darshan; Gangisetty, Shankar; Stagakis, Nikolaos; Arvanitis, Gerasimos; Moustakas, Konstantinos
• Is Part Of: Computers & graphics. Volume 99(2021)
• Page Start: 192
• Page End: 200
• Publication Date: 2021-10
• Subjects: SHREC 2021 -- 3D Point cloud change detection -- Graph convolutional networks -- Siamese networks; Computer graphics -- Periodicals; 006.6
• DOI: 10.1016/j.cag.2021.07.004
• Languages: English
• ISSNs: 0097-8493

Abstract

Highlights: We provide a unique classification-based 3D change detection dataset from a complex street environment. There are no other open 3D point cloud datasets released for our purpose. We evaluate different algorithms on the dataset and help finding solutions for 3D point cloud change detection tasks. The results show that the proposed siamese graph convolutional networks (SiamGCN) are good at extracting representative geometric features and can hereby outperform compared algorithms on the 3D change detection dataset. Graphical abstract: Abstract: The rapid development of 3D acquisition devices enables us to collect billions of points in a few hours. However, the analysis of the output data is a challenging task, especially in the field of 3D point cloud change detection. In this Shape Retrieval Challenge (SHREC) track, we provide a street-scene dataset for 3D point cloud change detection. The dataset consists of 866 3D object pairs in year 2016 and 2020 from 78 large-scale street scene 3D point clouds. Our goal is to detect the changes from multi-temporal point clouds in a complex street environment. We compare three methods on this benchmark, with one handcrafted (PoChaDeHH) and the other two learning-based (HGI-CD and SiamGCN). The results show that the handcrafted algorithm has balanced performance over all classes, while learning-based methods achieve overwhelming performance but suffer from the class-imbalanced problem and may fail on minority classes. The randomized oversampling metric applied in SiamGCN can alleviate this problem. Also, different siamese network architecture in HGI-CD and SiamGCN contribute to the designing of a network for the 3D change detection task.