基于contour let与分形的森林火灾图像分割-信号与信息处理专业论文.docxVIP

、L·· ··t1IBt441 4l， 、 L lle AbstractWith Abstract With the increasingly stringent fire safety requirements，and the promotion of the rapid development in high-tech，the晰de cross—combination of the fire detection technology and modem image processing technology makes the fire detection field into a new development of image-based fire detection period．The fire detection based on the fractal theory is developing in recent year，it is also a new branch of the image-based fire detection field． Aimed at the lack of accuracy and real-time about the existing fire detection I technology，this thesis focuses on the application in forest fire detection algorithm， based on the study of Contourlet transform and fractal theory．Mainly carried out as follows： First of all，the multiple linear interval analysis to forest-fire images，which meet 、Ⅳitll good scale invariance，so the images can be used to characterize multifractal feature analysis． Second，a method of forest·fire image segmentation based on Contourlet transform and fractional Brownian motion is proposed．This algorithm not only inherits the superiority of non-contact itself in the image—based fire detection．Also the rational USeS ● of fractional Brownian motion and the natural self-similarity of the flame to search the forest fire image segmentation，and Can easily get an accurate flame background from a complex environment，and the good image of forest fires to meet the higll accuracy requirements． Finally，to solve the problem of the multifractal spectrum of noise—sensitive，two multifractal spectrum new algorithms have been raised：Four-Region variance method and Multi-Mapping bottom interface method．In this thesis，new methods combine谢m Contourlet transform are used to Image segmentation of forest fire．First use Contourlet ．n． ? ? ? i frequency frequency sub-band for analysis，according to the background and flame multifractal spectrum for image segmentation，in order to get the flame segmentation．Experimental results show that t