Based On Fuzzy Ⅱ - Type Hidden Markov Models Of Flame Recognition

With the rapid development of computer science and signal processing technology,using a computer to analysis and calculate the video information which captured by thecamera became more and more popular. A variety of video-based flame recognitionalgorithms have been proposed. Though fire could be detected successful by thesemethods but all of this did not take into account the uncertainty in the video data. It is alimit to apply in the complex environment of the large space. In order to solve thisproblem, it is necessary for further study on the flame recognition.Fuzzy sets theory is a powerful tool to deal with uncertain problems. Themembership function of type-1fuzzy sets could not express the uncertainty and the datawhich is destroyed by noise. This thesis introduced two models based on type-2Fuzzysets theory: type-2fuzzy Gaussian Mixture Model and type-2fuzzy Hidden MarkovModel. Finally, a flame recognition experiment was completed by type-2fuzzy HiddenMarkov Model. In general, the main works of this thesis are as follows:1. Expanding researches for type-2fuzzy sets theory and type-2fuzzy logicsystem. A theoretical analysis and simulation experiments by KM, EKMreduced-type algorithms. Designing and implementing an interval type-2fuzzylogic system. Giving the control surface of type-2fuzzy logic system.2. Comparing on type-2fuzzy Gaussian Mixture Model and classical GaussianMixture Model with standard UCI data sets. The results show that fuzzy modelhas higher classification accuracy than classical model in the insufficient traindata or destroyed by noise. Type-2fuzzy Gaussian Mixture Model is moresuitable to the uncertainty problems.3. Programming type-2fuzzy hidden Markov model applied to the real scene ofthe flame. The experimental results show that the model can effectivelyidentify flame based on video.