利用Faster R-CNN自动识别建筑平面图功能构件的方法

Automatic Recognition Method of Building Functional Components Using Faster R-CNN

高精度建筑内部三维模型在地理信息应用领域需求旺盛,但建模的自动化程度和智能化程度仍然不高,以建筑平面图为基础的建模方式仍然是当前的主流。其中,首要解决的是建筑平面图上各类建筑构件的识别和提取问题。从建筑平面图的栅格图像出发,设计了一种基于Faster R-CNN目标检测模型的建筑功能构件自动识别方法,通过该方法识别并剔除床、桌子等无效的建筑功能构件,获得以墙体为主的室内结构图,为三维模型的快速生成提供无噪声的基础数据。经过实验验证,模型在功能构件检测中具有较好的准确率和召回率,能够满足多种场景下的建筑构件自动识别需求,提高建筑平面图分析处理的效率。

High-precision three-dimensional model of building interior is in great demand in the field of geographic information application, but the degree of automation and intellectualization of modeling is still not high. Modeling based on building plan is still the mainstream at present. Among them, the first problem to be solved is the identification and extraction of all kinds of building components on the building plan. Starting from the raster image of building plan, the automatic recognition method of building functional components based on Faster R-CNN object detection model was designed. Through this method, invalid building functional components such as beds and tables were identified and eliminated, and indoor structure maps with walls as the main body were obtained, which provided basic data without noise for rapid generation of three-dimensional models. The experimental results show that the model has good accuracy and recall rate in component detection, and can meet the needs of automatic identification of building components in various scenarios. It improves the efficiency of analysis and processing of building plan.