Abstract:Abstract. 3D The three dimensional building modelling has been an interesting topic of research for decades and it seems that photogrammetry methods provide the only economic means to acquire truly 3D city data. According to the enormous developments of 3D building reconstruction with several applications such as navigation system, location based services and urban planning, the need to consider the semantic features (such as windows and doors) becomes more essential than ever, and therefore, a 3D model of buildings as block is not any more sufficient. To reconstruct the façade elements completely, we employed the high density point cloud data that obtained from the handheld laser scanner. The advantage of the handheld laser scanner with capability of direct acquisition of very dense 3D point clouds is that there is no need to derive three dimensional data from multi images using structure from motion techniques. This paper presents a grammar-based algorithm for façade reconstruction using handheld laser scanner data. The proposed method is a combination of bottom-up (data driven) and top-down (model driven) methods in which, at first the façade basic elements are extracted in a bottom-up way and then they are served as pre-knowledge for further processing to complete models especially in occluded and incomplete areas. The first step of data driven modelling is using the conditional RANSAC (RANdom SAmple Consensus) algorithm to detect façade plane in point cloud data and remove noisy objects like trees, pedestrians, traffic signs and poles. Then, the façade planes are divided into three depth layers to detect protrusion, indentation and wall points using density histogram. Due to an inappropriate reflection of laser beams from glasses, the windows appear like holes in point cloud data and therefore, can be distinguished and extracted easily from point cloud comparing to the other façade elements. Next step, is rasterizing the indentation layer that holds the windows and doors information. After rasterization process, the morphological operators are applied in order to remove small irrelevant objects. Next, the horizontal splitting lines are employed to determine floors and vertical splitting lines are employed to detect walls, windows, and doors. The windows, doors and walls elements which are named as terminals are clustered during classification process. Each terminal contains a special property as width. Among terminals, windows and doors are named the geometry tiles in definition of the vocabularies of grammar rules. Higher order structures that inferred by grouping the tiles resulted in the production rules. The rules with three dimensional modelled façade elements constitute formal grammar that is named façade grammar. This grammar holds all the information that is necessary to reconstruct façades in the style of the given building. Thus, it can be used to improve and complete façade reconstruction in areas with no or limited sensor data. Finally, a 3D reconstructed façade model is generated that the accuracy of its geometry size and geometry position depends on the density of the raw point cloud.

Hierarchical attributed graph-based generative façade parsing for high-rise residential buildings

Correlation-Based Facade Parsing Using Shape Grammar

DeepFacade: A Deep Learning Approach to Facade Parsing with Symmetric Loss.

Translational Symmetry-Aware Facade Parsing for 3D Building Reconstruction

Semi-Supervised Representative Region Texture Extraction of Façade

Data-Driven Facade Reconstruction

Intelligent floor plan design of modular high-rise residential building based on graph-constrained generative adversarial networks

3D BUILDING FAÇADE RECONSTRUCTION USING HANDHELD LASER SCANNING DATA

Façade Feature Extraction for Urban Performance Assessments: Evaluating algorithm applicability across diverse building morphologies

Improving facade parsing with vision transformers and line integration

Building Facade Parsing R-CNN

Building façade element extraction based on multidimensional virtual semantic feature map ensemble learning and hierarchical clustering

Adaptive Partitioning of Urban Facades.

Deep learning for detecting building façade elements from images considering prior knowledge

Scene Parsing by Integrating Function, Geometry and Appearance Models

Knowledge-Driven and Diffusion Model-Based Methods for Generating Historical Building Facades: A Case Study of Traditional Minnan Residences in China

Automated building layout generation using deep learning and graph algorithms

Text Semantics to Image Generation: A method of building facades design base on Stable Diffusion model

Progressive Feature Learning for Facade Parsing with Occlusions

A Robust Automatic Method to Extract Building Facade Maps from 3D Point Cloud Data

ZAHA: Introducing the Level of Facade Generalization and the Large-Scale Point Cloud Facade Semantic Segmentation Benchmark Dataset