Research On The Fractal Based Remote Sensing Images Fusion And Classification Method

An algorithm of remote sensing object fusing is proposed based on fractal, in this paper.And, a classification algorithm is dragged using the fusion of texture feature classifier and spectral feature classifier based on the fuzzy inference algorithm.In this paper, a new fusing algorithm is dragged based on the theory of fractal used in the common wavelet and IHS algorithm. In the new algorithm, the fractal dim was computed based on every high frequency coefficient of wavelet by adding window. The new algorithm is unlike old fractal remote sensing image fusing method that fractal dim was dragged based on whole image and that ignore the independent function of coefficient. Extracting fractal dim from simulating image, the result proved the fractal dim of the picture can be as an important feature of the texture. The results from the experiment of fusing two real remote sensing images illustrates the algorithm designed in this paper is better than the algorithm of common wavelet pulsed IHS and the old fusing algorithm coinciding the fractal dim in the factor of objective parameter as entropy, spatial frequency and average gradient.In this paper, a classification algorithm is also propped based on fusing the spectral classifier and texture classifier. In the classification algorithm, the spectral of object was dragged from the image frequency domain, then, spectral and texture classifier are designed respectably based on the fuzzy inference. And then, choosing the rule of fuzzy integral merge together the results of the two classifier output. With the characteristics of the spectrum and texture of classification algorithms compared as the algorithm of feature fusing, the algorithm of the paper considers the independence of features of spectral and texture. The overall classification accuracy of the classification method designed in the paper is better than signal classifier, through analyzing the experiment results. At the same time, a fused image and an infused remote sensing image also was tested using the method of the paper, in order to illustrate the fusing of images can get more accuracy classifying result Last, through experiments, the results illustrate the method in the paper is better than the algorithm of feature fusing.