Recognition of building group patterns using graph convolutional network

Zhao, Rong; Ai, Tinghua; Yu, Wenhao; He, Yakun; Shen, Yilang

doi:10.1080/15230406.2020.1757512

Cited by 35 publications

(18 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As a data‐driven solution, neural networks have been applied widely in image data processing. In recent years, there are also increasing attempts to introduce the neural network model to solve geospatial problems (Janowicz et al, 2020; Li, 2020), such as road extraction (Wang et al, 2016), urban land‐use mapping (Huang et al, 2018), classification (Guo & Feng, 2020; Zhao et al, 2018), pattern recognition (Zhao et al, 2020), and spatial interpolation (Zhu et al, 2020). Since the neural network has outstanding performance in feature extraction (Lecun et al, 2015), Yan et al (2020) proposed to use an autoencoder structure to extract the features of vector building data.…”

Section: Related Workmentioning

confidence: 99%

Data‐driven polyline simplification using a stacked autoencoder‐based deep neural network

Chen

2022

Transactions in GIS

Self Cite

View full text Add to dashboard Cite

Automatic simplification of polylines is an important issue in spatial database and mapping. Traditional rule‐based methods are usually limited in performance, especially when the man‐made rules have to be adapted to different polylines with different shapes and structures. Compared to the existing neural network methods focusing only on the output layer or the code layers for classification or regression, our proposed method generates multi‐level abstractions of polylines by extracting features from multiple hidden layers. Specifically, we first organize the cartographic polylines into the form of feature vectors acceptable to the neural network model. Then, a stacked autoencoder‐based deep neural network model is trained to learn the pattern features of polyline bends and omit unimportant details layer by layer. Finally, the multi‐level abstractions of input polylines are generated from different hidden layers of a single model. The experimental results demonstrate that, compared with the classic Douglas–Peucker and Wang and Muller algorithms, the proposed method is able to properly simplify the polylines while representing their essential shapes smoothly and reducing areal displacement.

show abstract

Section: Related Workmentioning

confidence: 99%

Data‐driven polyline simplification using a stacked autoencoder‐based deep neural network

Chen

2022

Transactions in GIS

Self Cite

View full text Add to dashboard Cite

show abstract

“…The morphological patterns of building distributions are either simple or complex. Simple patterns, such as collinear, curvilinear, or grid-like alignments have been extensively studied to facilitate map generalization in cartography [17][18][19][20] or to identify functional neighborhoods in spatial indexing [21][22][23]. These simple patterns are mostly described by geometric variables, such as distance and angles, and some morphological descriptors, such as proximity, similarity, closure, and continuity, which are based on visual variables proposed by the Gestalt theory [24].…”

Section: Complexity Of Urban Morphologymentioning

confidence: 99%

“…GCN models have proved to be very capable of detecting spatial and attributive features in many fields, such as traffic flow prediction and social spammer detection [44,45]. Recent studies also show that GCN models can help to classify the regular or irregular patterns of building distributions [19,46]. Moreover, GCN embedded with an autoencoder framework has recently been applied to encode individual building shapes [47].…”

Section: Graph Convolutional Networkmentioning

confidence: 99%

A New Graph-Based Fractality Index to Characterize Complexity of Urban Form

Seipel

Brandt

et al. 2022

IJGI

View full text Add to dashboard Cite

Examining the complexity of urban form may help to understand human behavior in urban spaces, thereby improving the conditions for sustainable design of future cities. Metrics, such as fractal dimension, ht-index, and cumulative rate of growth (CRG) index have been proposed to measure this complexity. However, as these indicators are statistical rather than spatial, they result in an inability to characterize the spatial complexity of urban forms, such as building footprints. To overcome this problem, this paper proposes a graph-based fractality index (GFI), which is based on a hybrid of fractal theory and deep learning techniques. First, to quantify the spatial complexity, several fractal variants were synthesized to train a deep graph convolutional neural network. Next, building footprints in London were used to test the method, where the results showed that the proposed framework performed better than the traditional indices, i.e., the index is capable of differentiating complex patterns. Another advantage is that it seems to assure that the trained deep learning is objective and not affected by potential biases in empirically selected training datasets Furthermore, the possibility to connect fractal theory and deep learning techniques on complexity issues opens up new possibilities for data-driven GIS science.

show abstract

“…This method underlines the advantages of using Breps in digital building-envelope reconstruction, mainly because Breps help define roof and facades within the same model. Furthermore, by using a convolutional network (ConvNet) 35 that uses raw inputs rather than hand-crafted features, 36 the work of Huang develops a method for retrieving door and window openings from planar facades. The problem of topology or non-coplanar facades is not addressed by Huang’s work.…”

Section: Related Workmentioning

confidence: 99%

Parametric point-cloud slicing for facade retrofitting

Bohórquez

Derigent

El-Haouzi

2021

International Journal of Architectural Computing

View full text Add to dashboard Cite

This work presents a method for retrieving 3D building contours usable in facade retrofitting projects, which uses a parametric modeling workflow that utilizes a point-cloud slicing method to retrieve such 3D contours. Since current commitments by European governments seek to reduce energy consumption as a means to reduce carbon emissions from building stock by 2050, facade retrofitting appears as an alternative for addressing operational and embedded building emissions. Within such a context, the main contribution of this work consists of a workflow and a 3D reconstruction solution that uses a parametric environment for capturing building topology and bypassing ground-level occlusions. A real case study and a strategy for converting 3D building contours into Industry Foundation Classes entities, directly from the parametric modeling environment, served as a scenario for testing the capabilities of a Grasshopper solution and open new perspectives for this approach.

show abstract

Recognition of building group patterns using graph convolutional network

Cited by 35 publications

References 27 publications

Data‐driven polyline simplification using a stacked autoencoder‐based deep neural network

Data‐driven polyline simplification using a stacked autoencoder‐based deep neural network

A New Graph-Based Fractality Index to Characterize Complexity of Urban Form

Parametric point-cloud slicing for facade retrofitting

Contact Info

Product

Resources

About