Building Extraction from Airborne Multi-Spectral LiDAR Point Clouds Based on Graph Geometric Moments Convolutional Neural Networks

Li, Dilong; Shen, Xin; Yu, Yongtao; Guan, Haiyan; Li, Jonathan; Zhang, Guo; Li, Deren

doi:10.3390/rs12193186

Cited by 32 publications

(13 citation statements)

References 50 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In recent research, many authors have performed building extraction and roof shape classification using deep learning techniques [48,69]. Although these techniques were able to detect buildings with success, in most cases here there was low planimetric accuracy and individual roof plane extraction was not considered [53].…”

Section: Roof Modelingmentioning

confidence: 99%

Procedural Point Cloud Modelling in Scan-to-BIM and Scan-vs-BIM Applications: A Review

Abreu

Pinto

Matos

et al. 2023

IJGI

View full text Add to dashboard Cite

Point cloud processing is an essential task in many applications in the AEC domain, such as automated progress assessment, quality control and 3D reconstruction. As much of the procedure used to process the point clouds is shared among these applications, we identify common processing steps and analyse relevant algorithms found in the literature published in the last 5 years. We start by describing current efforts on both progress and quality monitoring and their particular requirements. Then, in the context of those applications, we dive into the specific procedures related to processing point clouds acquired using laser scanners. An emphasis is given to the scan planning process, as it can greatly influence the data collection process and the quality of the data. The data collection phase is discussed, focusing on point cloud data acquired by laser scanning. Its operating mode is explained and the factors that influence its performance are detailed. Data preprocessing methodologies are presented, aiming to introduce techniques used in the literature to, among other aspects, increase the registration performance by identifying and removing redundant data. Geometry extraction techniques are described, concerning both interior and outdoor reconstruction, as well as currently used relationship representation structures. In the end, we identify certain gaps in the literature that may constitute interesting topics for future research. Based on this review, it is evident that a key limitation associated with both Scan-to-BIM and Scan-vs-BIM algorithms is handling missing data due to occlusion, which can be reduced by multi-platform sensor fusion and efficient scan planning. Another limitation is the lack of consideration for laser scanner performance characteristics when planning the scanning operation and the apparent disconnection between the planning and data collection stages. Furthermore, the lack of representative benchmark datasets is hindering proper comparison of Scan-to-BIM and Scan-vs-BIM techniques, as well as the integration of state-of-the-art deep-learning methods that can give a positive contribution in scene interpretation and modelling.

show abstract

Section: Roof Modelingmentioning

confidence: 99%

Procedural Point Cloud Modelling in Scan-to-BIM and Scan-vs-BIM Applications: A Review

Abreu

Pinto

Matos

et al. 2023

IJGI

View full text Add to dashboard Cite

show abstract

“…Automatic building semantic segmentation in very high resolution (VHR) remote sensing images has proved used in a range of applications, including emergency management, urban planning, traffic evaluation, and mapping [1]. Segmentation is often used in computer vision [2][3][4] and industrial robots [5][6][7][8], but it has lately been used to remote sensing, which is important in a variety of applications such as environmental monitoring and danger identification [9].…”

Section: Introductionmentioning

confidence: 99%

Weakly supervised building semantic segmentation via superpixel‐CRF with initial deep seeds guiding

Moghalles

Al‐Huda

et al. 2022

IET Image Processing

View full text Add to dashboard Cite

The segmentation of building from satellite and airborne images is necessary for highresolution buildings maps generation and it is still challenging. On annotated pixel-level images, trained deep convolutional neural networks (CNNs) were used to improve segmentation of building. The cost of labelling training data is high, which reduces their usage. Human labelling efforts can be significantly reduced using weakly supervised segmentation techniques. Here, a novel weakly supervised framework is introduced for building semantic segmenting that relies on deep seeds to construct a superpixels-CRF model over superpixels segmentation in order to generate high-quality initial pixel-level annotations, as the initialization step. Then, the segmentation network is trained using the initial pixel-level annotations. Next, the CRF model is used to refine the segmentation masks, and the segmentation network is retrained to achieve accurate pixel-level annotations while iteratively optimizing the segmentation. The experimental results on three public building datasets demonstrate that the proposed framework significantly improved the quality of building semantic segmentation while remaining computationally efficient.

show abstract

“…PU-GCN [15] uses the graph convolutional network (GCN) layers to construct point cloud graphs. In GCN networks, points convey information [21] through graph structure and node features. Compared with the duplicated upsampling method, GCN utilizes more local neighborhood information.…”

Section: Introductionmentioning

confidence: 99%

PU-WGCN: Point Cloud Upsampling Using Weighted Graph Convolutional Networks

Zhang

Wang

et al. 2022

Remote Sensing

View full text Add to dashboard Cite

Point clouds are sparse and unevenly distributed, which makes upsampling a challenging task. The current upsampling algorithm encounters the problem that neighboring nodes are similar in terms of specific features, which tends to produce hole overfilling and boundary blurring. The local feature variability of the point cloud is small, and the aggregated neighborhood feature operation treats all neighboring nodes equally. These two reasons make the local node features too similar. We designed the graph feature enhancement module to reduce the similarity between different nodes as a solution to the problem. In addition, we calculate the feature similarity between neighboring nodes based on both spatial information and features of the point cloud, which is used as the boundary weight of the point cloud graph to solve the problem of boundary blurring. We fuse the graph feature enhancement module with the boundary information weighting module to form the weighted graph convolutional networks (WGCN). Finally, we combine the WGCN module with the upsampling module to form a point cloud upsampling network named PU-WGCN. Compared with other upsampling networks, the experimental results show that PU-WGCN can solve the problems of hole overfilling and boundary blurring and improve the upsampling accuracy.

show abstract

Building Extraction from Airborne Multi-Spectral LiDAR Point Clouds Based on Graph Geometric Moments Convolutional Neural Networks

Cited by 32 publications

References 50 publications

Procedural Point Cloud Modelling in Scan-to-BIM and Scan-vs-BIM Applications: A Review

Procedural Point Cloud Modelling in Scan-to-BIM and Scan-vs-BIM Applications: A Review

Weakly supervised building semantic segmentation via superpixel‐CRF with initial deep seeds guiding

PU-WGCN: Point Cloud Upsampling Using Weighted Graph Convolutional Networks

Contact Info

Product

Resources

About