Deep learning-based semantic segmentation of urban-scale 3D meshes in remote sensing: A survey

Adam, Jibril Muhammad; Liu, Weiquan; Zang, Yu; Afzal, Muhammad Kamran; Bello, Saifullahi Aminu; Muhammad, Abdullahi Uwaisu; Wang, Cheng; Li, Jonathan

doi:10.1016/j.jag.2023.103365

Cited by 6 publications

(4 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…While there exists an extensive amount of research on semantic segmentation of urban large-scale data, Ulku et al and Adam et al [3,13] surveys proposed that in the realm of 3D meshes, semantic segmentation pertains to the classification of individual elements into specific categories. Rook et al [14] propose an extension to the existing class definitions of Roof Surface and Wall Surface in CityGML to demonstrate that by expanding these definitions, automatic semantic labeling of a CityGML file with Level of Detail 1 and 2 becomes achievable.…”

Section: Related Workmentioning

confidence: 99%

“…An essential stage in semantic classification involves feature extraction. Deep learning methods tend to perform more effectively when ample training data is available, as observed in studies [3,15,16]. These techniques often rely on contextual information for feature computation or learning.…”

Section: Semantic Classification Of Urban Modelsmentioning

confidence: 99%

“…The 3D scene or shape analysis of a concoction of dense, complex, curved surfaces, or irregular geometries of meshes in large urban environments including buildings, trees, cars, and other elements has been made possible by recent developments in 3D computer vision and photogrammetry [1]. To enable the use of these meshes in a variety of applications, such as smart urban planning, navigation systems, virtual reality, radiation estimation, noise modeling, and photovoltaic perspective, it is essential to extract semantic information from the mesh models [2,3].…”

Section: Introductionmentioning

confidence: 99%

“…Unlike point clouds, which comprise individual data points, textured meshes merge geometric data with surface textures. This fusion enables precise and robust classification in urban environments, capturing intricate details and enhancing feature recognition across various objects and structures within the scene [3]. This comprehensive representation arises from textured urban meshes' capability to integrate both geometric data and high-resolution textures.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

GeoSparseNet: A Multi-Source Geometry-Aware CNN for Urban Scene Analysis

Afzal,

Liu,

Zang

et al. 2024

Remote Sensing

View full text Add to dashboard Cite

The convolutional neural networks (CNNs) functioning on geometric learning for the urban large-scale 3D meshes are indispensable because of their substantial, complex, and deformed shape constitutions. To address this issue, we proposed a novel Geometry-Aware Multi-Source Sparse-Attention CNN (GeoSparseNet) for the urban large-scale triangular mesh classification task. GeoSparseNet leverages the non-uniformity of 3D meshes to depict both broad flat areas and finely detailed features by adopting the multi-scale convolutional kernels. By operating on the mesh edges to prepare for subsequent convolutions, our method exploits the inherent geodesic connections by utilizing the Large Kernel Attention (LKA) based Pooling and Unpooling layers to maintain the shape topology for accurate classification predictions. Learning which edges in a mesh face to collapse, GeoSparseNet establishes a task-oriented process where the network highlights and enhances crucial features while eliminating unnecessary ones. Compared to previous methods, our innovative approach outperforms them significantly by directly processing extensive 3D mesh data, resulting in more discerning feature maps. We achieved an accuracy rate of 87.5% when testing on an urban large-scale model dataset of the Australian city of Adelaide.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Semantic Classification Of Urban Modelsmentioning

confidence: 99%