Design And Implementation Of Spatio-temporal Smoke And Fire Detection And Fire Behavior Analysis System

Fire has always been a major safety problem in human society,which has caused a serious threat to people’s life and property safety.At present,pyrotechnic sensors are the main means of fire risk monitoring,which is very dependent on the monitoring range and response speed of sensors,and a large number of deployment of sensors will also bring high costs.With the rapid development of the IoT,computer vision and artificial intelligence technologies,intelligent monitoring systems are gradually penetrating into every field of society,playing an important role in improving the level of public security management and ensuring the safety of people’s lives and property.Therefore,studying how to apply artificial intelligence and computer technology to the fire abnormal monitoring system is of great help to improve the intelligent fire protection system and promote the informatization and intelligence in the field of fire safety.This paper firstly conducted an in-depth study on the current research direction in the field of pyrotechnics detection,and concluded that the current research work on smoke and fire detection has certain limitations:the existing smoke and fire detection methods can only simply predict whether a picture or video clip contains smoke and flame,but cannot determine the time interval of the occurrence of smoke and fire,so they are usually used to realize the fire alarm function in real-time scenes.At the same time,existing methods lack the means to assess fire risk.Therefore,a time-space smoke and fire detection and fire analysis algorithm is proposed in this paper.This paper divide smoke and fire detection task into three sub-tasks:object target detection of smoke and fire,temporal smoke and fire detection and fire risk evaluation,so as to realize all-round smoke and fire detection in both time and space dimensions.On the other hand,most of the existing smoke and fire detection models are trained on public datasets,which are inconsistent in type,small in data volume,and lack of temporal boundary annotation of smoke and fire video fragments.To this end,a temporal smoke and fire detection dataset is constructed in this paper.Videos of various fire scenes are adopted,and frame level temporal boundaries and segment categories are used as labels.Subsequently,a series of experiments are carried out on the three methods using the public dataset and the self-built dataset respectively to verify the effectiveness of the proposed method.Finally,this paper designs and implements a spatio-temporal smoke and fire detection and fire behavior analysis system.This paper firstly analyzes the requirements of the system in detail,and divides the system into data access module,fireworks detection and analysis module,business processing module,data persistence module and system visualization module.The smoke and fire detection and analysis module adopts the spatio-temporal smoke and fire detection and fire analysis algorithm proposed in this paper.Finally,through the design system test,it is verified that the system can meet the requirements of scene use,in line with the system design expectations.