基于户型图实例分割的室内场景建模研究

Study on indoor scene modeling based on instance segmentation of house plan

为满足房产室内场景在线三维建模与展示的需求,提出了一种基于户型图实例分割与ThreeJS的室内场景建模方法。该方法以户型图栅格数据和户型结构矢量数据作为输入,将Mask R-CNN算法应用于户型图栅格数据的实例分割,实例分割结果与户型结构矢量数据进行配准,最后,通过家具模型库自动适配,基于ThreeJS完成室内场景三维在线快速建模。实验结果表明:该方法建模过程简单、高效、灵活,模型效果符合预期。

Aiming at the needs of online 3D modeling and display of real estate indoor scenes,an indoor scene modeling method based on instance segmentation of house plan and ThreeJS is proposed.This method takes the grid data of the house plan and the vector data of the house structure as input,and applies Mask R-CNN algorithm to implement the instance segmentation on the grid data of the house plan.The instance segmentation results are matched with the vector data of the house structure.Finally,through the automatically matching in the furniture model library,three-dimensional online rapid modeling for indoor scenes is completed based on ThreeJS.The experimental results show that the method is simple,efficient and flexible,and the effect of the model has reached expectations.