Reconstructing compact building models from point clouds using deep implicit fields

Zhaiyu, Chen,; Ledoux, Hugo; Khademi, Seyran; Nan, Liangliang

doi:10.1016/j.isprsjprs.2022.09.017

Cited by 40 publications

(9 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For example, Ref. [26] generates candidate convex hulls by binary space partition, then a deep implicit field is learned from the input point cloud and used as guidance for candidate selection.…”

Section: Related Workmentioning

confidence: 99%

Iterative Low-Poly Building Model Reconstruction from Mesh Soups Based on Contour

Xiao,

Liu,

Zhang

2024

Remote Sensing

View full text Add to dashboard Cite

Existing contour-based building-reconstruction methods face the challenge of producing low-poly results. In this study, we introduce a novel iterative contour-based method to reconstruct low-poly meshes with only essential details from mesh soups. Our method focuses on two primary targets that determine the quality of the results: reduce the total number of contours, and generate compact surfaces between contours. Specifically, we implemented an iterative pipeline to gradually extract vital contours by loss and topological variance, and potential redundant contours will be removed in a post-processing procedure. Based on these vital contours, we extracted the planar primitives of buildings as references for contour refinement to obtain compact contours. The connection relationships between these contours are recovered for surface generation by a contour graph, which is constructed using multiple bipartite graphs. Then, a low-poly mesh can be generated from the contour graph using our contour-interpolation algorithm based on polyline splitting. The experiments demonstrated that our method produced satisfactory results and outperformed the previous methods.

show abstract

Section: Related Workmentioning

confidence: 99%

Iterative Low-Poly Building Model Reconstruction from Mesh Soups Based on Contour

Xiao,

Liu,

Zhang

2024

Remote Sensing

View full text Add to dashboard Cite

show abstract

“…Point clouds carry 3-D spatial information for efficient 3-D reconstruction of objects or scenes. In a recent deep learning based framework [43], authors present a method to reconstruct compact, watertight, polygonal 3-D building models from point clouds with the help of a learnable implicit field using a deep neural network. In this method, the implicit field extracts a smooth surface model of the object by directly learning from the point cloud and MRF extracts the compact surface of the building through combinatorial optimization.…”

Section: Deep Learning Based Algorithmsmentioning

confidence: 99%

“…Due to the unorganized nature of point clouds, training a neural network directly from the point cloud is not an easy task. In a recent work [43], a deep learning framework is presented for reconstructing compact and watertight polygonal building models from point clouds in which learnable implicit fields are used to characterize 3-D surface extracted through MRF. Although this method generates valid models with accurate structural geometry, missing LOD makes such models unsuitable for an interactive urban scene application.…”

Section: F Summary and Remarksmentioning

confidence: 99%

Lightweight Reconstruction of Urban Buildings: Data Structures, Algorithms, and Future Directions

Kamra

Kudeshia

ArabiNaree

et al. 2023

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

Commercial buildings as well as residential houses represent core structures of any modern day urban or semiurban areas. Consequently, 3-D models of urban buildings are of paramount importance to a majority of digital urban applications, such as city planning, 3-D mapping and navigation, video games and movies, and construction progress tracking, among others. However, current studies suggest that existing 3-D modeling approaches often involve high computational cost and large storage volumes for processing the geometric details of the buildings. Therefore, it is essential to generate concise digital representations of urban buildings from the 3-D measurements or images so that the acquired information can be efficiently utilized for various urban applications. Such concise representations, often referred to as "lightweight" models, strive to capture the details of the physical objects with less computational storage. Furthermore, lightweight models consume less bandwidth for online applications and facilitate accelerated visualizations. With many emerging digital urban infrastructure applications, lightweight reconstruction is poised to become a new area of research in the urban remote sensing community. We aim to provide a thorough review of data structures, representations, and state-of-the-art algorithms for lightweight 3-D urban reconstruction. We discuss the strengths and weaknesses of key lightweight urban reconstruction techniques, ultimately providing guidance on future research prospects to fulfill the pressing needs of urban applications.

show abstract

“…The specific surfaces of building models are converted into the CityGML format. Chen et al (119) provided a novel framework that reconstructed urban models by exploiting a learned implicit representation as an occupancy indicator for the extraction of an explicit geometry. This report is the first work in which an implicit field is explored for building reconstruction.…”

Section: Volume-based Reconstructionmentioning

confidence: 99%

A Review of Indoor Automation Modeling Based on Light Detection and Ranging Point Clouds

Cui¹,

Yang²,

Liu³

et al. 2023

Sensors and Materials

View full text Add to dashboard Cite

3D modeling of the indoor environment is essential for urban applications such as indoor navigation, emergency simulations, floor planning, and building construction. With the development of laser scanning sensors, 3D laser scanners can quickly obtain high-density, highprecision 3D coordinates and attribute information, which brings significant advantages in collecting 3D information on indoor scenes. Many studies have been published on the fast reconstruction of 3D models based on point cloud data obtained by various types of laser scanning sensors. In this paper, we review state-of-the-art automated 3D indoor modeling technologies. The 3D modeling standards for indoor environments are introduced, and data acquisition based on laser scanning sensors and characteristics of point clouds are discussed. Indoor object classification and indoor room segmentation are also examined in detail. The 3D indoor reconstruction methods (i.e., line-based, plane-based, and volume-based) are systematically introduced and the advantages and disadvantages of these methods are presented. Future research directions in this field are discussed and summarized. This review can help researchers improve current approaches or develop new techniques for 3D indoor reconstruction.

show abstract

Reconstructing compact building models from point clouds using deep implicit fields

Cited by 40 publications

References 31 publications

Iterative Low-Poly Building Model Reconstruction from Mesh Soups Based on Contour

Iterative Low-Poly Building Model Reconstruction from Mesh Soups Based on Contour

Lightweight Reconstruction of Urban Buildings: Data Structures, Algorithms, and Future Directions

A Review of Indoor Automation Modeling Based on Light Detection and Ranging Point Clouds

Contact Info

Product

Resources

About