BIM And Computer Vision Based Evacuation Path Generation Under Fire Emergency In Public Buildings

According to report released by the Ministry of Emergency Management of China in 2019,1,045 people were killed by fire in residential and public buildings,accounting for 78.3% of the total.The numbers behind these figures clearly show that much more attention should be paid to the casualties caused by fire inside buildings.The intricate interior space and the rapidly developing fire situation undoubtedly increase the difficulty of evacuation.In order to reduce casualties and property losses,it is particularly important to provide reliable indoor navigation for the survivors in the case of fire.The indoor navigation in the case of fire mainly includes three parts: 1.Building an indoor road network;2.Indoor positioning;3.Path planning under fire situation.At present,indoor positioning and navigation is still an immature field,which is faced with many challenges and problems.Most of the existing path planning revolves around two-dimensional plane plans,ignoring three-dimensional spatial information.The existing positioning methods cannot be well combined with BIM to realize navigation.Aiming at the above problems,this paper studies indoor navigation problems in fire scenarios from the perspectives of BIM and computer vision technology respectively:(1)Indoor navigation network based on BIM.This paper analyzes the overall architecture of BIM's mainstream data exchange standard IFC and the information acquisition methods of entities in IFC,and simplifies the spatial elements and adjacency relationships between spatial elements into nodes and edges by extracting the internal space elements needed for building indoor emergency evacuation network construction.In particular,for the extraction of corridor elements,this paper proposes a method based on Delaunay constraint triangle,generates skeleton lines according to triangle types,and optimizes the intersection and T-junction in the corridor to construct the final network.(2)Indoor positioning method based on computer vision.As for the indoor location of fire evacuees,this paper proposes an indoor location method based on computer vision technology.First,the identification network is trained through the transfer learning method to identify the scene and realize the preliminary location.Then,according to the recognition results,image feature matching and RPn P camera pose estimation algorithm is used to solve the accurate location.(3)Path planning based on BP neural network.In this paper,combined with the graph theory model and through the cost function,the fire location is simulated randomly,so that the neural network model can learn autonomously to generate the path to avoid the fire point under the fire situation.The experimental results show that the method of path planning under fire scenarios based on computer vision and BIM technology in this paper worked effectively.Compared with the shortest path Dijkstra algorithm,the proposed algorithm can provide a more reliable escape path.