Research On Automatic Identification Of Fire Fighting Equipment Based On Deep Learning

In recent years,there are more and more potential safety hazards in the city.All kinds of disasters and accidents have the characteristics of high risk and great harm.The main reason for the increasing security risks is the continuous and rapid development of social economy and the large population gathering caused by the continuous expansion of urban scale.Therefore,the demand for social public security is increasing.With the continuous process of urbanization,the rapid development of high-rise,underground,commercial complex,subway,tunnel and other buildings in the city leads to the serious shortage of fire prevention and extinguishing force in the active service of the public security fire bureau,the long-term fatigue of fire officers and soldiers,and the huge pressure of fire prevention and control.The State Council’s action essentials for promoting the improvement of mega data(2015)indicated that it is necessary to promote the construction and development of intelligent fire-fighting city rescue platform.Guo Shengkun,Secretary of the political and Legal Commission of the Central Committee of the CPC,stressed that "it is necessary to further promote the construction of fire information and actively build" intelligent fire protection ".In this context,the smart fire city rescue platform project emerges as the times require.Using image recognition technology to accurately locate the fire equipment icon on building drawings is an important part of the project of automatic distribution system based on deep learning.Only by identifying and obtaining the location of the fire protection icon,the project can generate the corresponding icon on this basis to realize the automatic distribution function of fire protection drawings.In the project of intelligent fire-fighting city rescue platform,the accurate identification of fire-fighting equipment icon on the building drawing is the premise to locate the position of fire-fighting equipment in the building,so that the rescue platform can provide effective support for fire-fighting and rescue operations.In order to solve the problem of low efficiency of manually inputting fire fighting information of building floors,this thesis proposes an automatic recognition scheme of fire equipment based on Machine learning dependent on YOLOv3 algorithm under Tensor Flow and Keras framework.By collecting data sets,downloading pre training files,and using the YOLOv3 algorithm for self-training,to identify the four kinds of fire equipment icons on the building drawings and export the location information of the fire icons.Then,the function of automatic distribution of fire-fighting equipment icons can be realized through the acquisition of location coordinates.The experimental results show that the scheme of automatic identification of fire-fighting equipment can be significantly It has strong reliability to improve the input efficiency of fire equipment icon on building drawings in the project of intelligent fire city rescue platform.The main novelty of this thesis is that:(1)Using the YOLOv3 Algorithm to identify multiple fire icons at the same time.(2)The trained target detection model can overcome the different sizes of the fire-fighting icons on the building drawings and identify the fire-fighting equipment icons from different angles.(3)The trained target detection model can overcome the interference of other icons on the building drawings to complete the identification of fire icons on the building drawings.