Building Interior Modeling Based On Three-dimensional Line Structure

With the growth of urban population and the amount of large-scale buildings,the de-mand for the detailed spatial information in the indoor environment is increasing.In re-cent years,3D modeling and reconstruction of the interior of buildings provide essential 3D models for applications,such as building foundation protection,building maintenance,disaster rescue,and building renovation planning.The major requirement of these appli-cations is the 3D indoor building models,which are composed of the primitives of the building interiors,such as the ceilings,floors,walls,windows,and doors,but not the movable objects in the indoor spaces,such as furniture.This paper presents,for indoor environments,a novel semantic line framework-based indoor building modeling method using backpacked laser scanning point cloud data.The method that we proposed semantic labels the original point clouds into the ceilings,floors,walls and other objects firstly.Then extract the line structure from the labeled point cloud.To optimize the detected line structures caused by furniture occlusion,a conditional Generative Adversarial Nets(cGAN)deep learning model is constructed.The line framework optimization model includes structure completion,extrusion removal,and regularization.In the end,an indoor building frame model with semantic information was obtained.This method can also be used to evaluate the quality of the original point cloud and integrate indoor and outdoor modeling tasks.The main research content of this article include:1.For indoor 3D modeling problems,a line framework is proposed to represent the structure of the indoor building model based on the semantic labeling result.In the pro-posed method,no assumption of the specific structure of the building is required,such as the vertical wall or horizontal floor hypothesis.2.For incomplete line framework extraction problem caused by incomplete point cloud data,a conditional Generative Adversarial Nets(cGAN)-based deep learning model is de-veloped to optimize the line framework against the clutter background and heavy occlusion in indoor environment.3.For the case of inconsistent quality of indoor and outdoor point clouds,a method of integrated registration of indoor and outdoor scenes based on line structure is proposed,the door and window line structure extracted on the wall is used for scene registration.In this paper,we verified the above three research contents on different data sets and made quantitative and qualitative analysis of the experimental results.In the line frame-work extraction,a comparative experiment was performed with the similar indoor line structure extraction methods.The results show that the method proposed in this paper performed better than other methods on line structure extraction.