Application Specific Research On Fire Detection Based On Data Fusion Of Artificial Intellgence

Fire not only deprives humen's lifes and estates, but also destroy entironment and nature resources which support all persons in the world. Therefore,it is a forever task for humen to prevent fire breaking out and reduce losing owning to fire.So the study of fire detecting technology shows more important. According to detecting parameter of certain single fire detector, the comparative law of threshold is adopoted to alarm fire in traditional fire detecting technology. But it is very usual to misalarm due to the contradiction between diverse fire singals and single detector. In order to advance alarm and reduce misalarm, at present, many researchers are devoted to using multisensor instead of single sensor to distinguish non-fire singal and genuine fire singal.If multisensor's singals are processed only by "and-not" judge, it's equal to be a combination of plus and minus.That can't make full use of the function of multisensor. Multisensor data fusion is a new developing technology in recent years, which can make full use of the function of multisensor. It utilizes various kinds of sensors information to draw the overall, correct understanding of the environment or target's characteristics. The project which the thesis based on is to do some researches on the multisensor data fusion technology, then to design and implement fire detecting system. The theoretic in this thesis consists of three parts: basical principle and system structure of data fusion technology; Principle and method of fire detecting and fire detecting system based on data fusion systeraAfter analyzing the three layer structure of data fusion system and the fire's non-structure character, then a kind of fire detection's three-layer structure based on data fusion is proposed, this paper proposes a kind of algorithm using data fusion technology in view of artificial intelligence. The input analog signals are provided by temperature, smoke, gas and ultrared light sensors, and the algorithm synthesizes those signals by use of neural network and fuzzy logic technology. The aim is to design a kind of quick, exact and effective fire detecting system, which can reach the aim of early alarm and reducing misalarm.The paper simulates all kinds of fire and disturbing environment. Simulating model can give fire probability in different environment and proves the feasibility and validity of the fire detecting system proposed in this paper.