Research Of Multi-sensor Information Fusion And Application In Fire Prediction

Sensor is a detection device which is a kind of information of the object to be tested,usually based on the amount of the number and types of sensors in the system,divided into the single sensor system and the multi-sensor system.Multisensor system is superior to the single sensor greatly in such aspects as the effectiveness of the system,the integrity of information,geographical universality,the credibility of information,so the application value greatly enhanced.But following is the problem of multi-sensor information processing.As a result of the multisensor location distribution,data redundancy,collected more information characteristic,make information fusion and decision problems become the new hot research topic.This article first elaborated the research field of multi-sensor information fusion,and introduced.hierarchical structure of multi-sensor information fusion,architecture and fusion method.From feature level to the decision-making level hierarchical in-depth study based on the theory of intelligent information fusion method,focuses on the characteristics of the magnitude of the BP neural network classification and prediction algorithm and ELM extreme learning machine,through the experimental simulation performance comparison of two algorithms are given,at the same time discusses the decision-making level of D-S evidence theory algorithm,mainly illuminates the application fields of the method.So as to put forward a shorter training time,small error,the prediction method of decision clear.This article expounds the multi-sensor information fusion method based on the theory of the intelligent used in fire prediction First analyzes the basic characteristics of the fire signal,points out the shortage of early fire signal recognition algorithm,based on this,puts forward fire prediction system.model;Second,hierarchical in turn from the shop information layer,characteristic layer and the design for fire forecast method;Finally,the simulation verify the method through the experiment on time performance and prediction accuracy is better than existing methods.This article will improve the extreme learning machine algorithm,using online sequence extreme learning machine method as a feature layer of the algorithm,greatly improve the performance of the algorithm,making the prediction results more quickly and accurately.In the process of fire rescue,each save a second time will reduce a lot of economic losses and casualties.Therefore,this method is worth promoting.