Research On Fire Detection Based On LSTM And Multi-sensor Information Fusion

With the rapid development of smart cities and intelligent fire protection,traditional fire detection methods can no longer meet the current fire protection needs.On the one hand,the acceleration of urbanization has led to an increase in the density of urban buildings,and the problem of fire safety hazards has become increasingly prominent.Once a fire occurs in a concentrated or crowded place,it often causes serious economic losses and casualties.On the other hand,the continuous development of sensor technology and artificial intelligence technology has brought the advantages of high accuracy,short response time and wide detection range to the new fire detection method based on multi-sensor information fusion technology.Therefore,fire detection based on multi-sensor information fusion technology has become one of the research focuses in the field of intelligent fire protection.At present,the research on fire detection based on multi-sensor information fusion mainly focuses on the processing and fusion of the measured values of environmental indicators at a single time.On this basis,there are relatively few studies on fire detection combined with the time series characteristics of the target environment.In view of the above problems,this paper uses Long-short Term Memory(LSTM)to extract the time series characteristic information of the target environment,and studies the fire detection problem based on LSTM and multi-sensor information fusion.The main research contents are as follows:Firstly,this paper proposes a fire detection model based on LSTM and multi-sensor information fusion.This model uses LSTM to model the time series of environmental indicator measurements,and extracts two environmental time series characteristic information: environmental indicator variation information and environmental level information.Then,this paper uses the neural network fusion algorithm in multi-sensor information fusion technology to fuse the obtained environmental indicator variation information,environmental level information and environmental indicator measurement values.The integrated fire state information obtained by fusion will be used for the final fire judgment.The addition of the above two environmental time series characteristic information expands the data source and dimension of information fusion,further improves the consistent interpretation and description of the monitoring environment by the model,and improves the accuracy and reliability of fire detection.Secondly,based on the real-world fire dataset published by the National Institute of Standards and Technology(NIST),this paper compares the proposed LSTM and multi-sensor information fusion fire detection model with other multi-sensor fire detection models,and analyzes its performance in accuracy,confusion matrix,classification performance and alarm time.The results have shown that this model can further control the false alarm rate and false negative rate on the basis of improving the accuracy,showing advanced comprehensive performance.The proposed model also deepens the research on time series characteristic extraction of fire environment,and provides an alternative multi-sensor information fusion fire detection method for fire management department.