Multiple Source Remote Sensing Image Fusion Technology Research

Since the 1990s, A lot of remote sensing satellites with different kinds of sensor platform have been launched and put into the commercialization operation one after another, specially the high spatial resolution's commercial smallsat, they are continuously providing the sensing images of different image formation mechanism, different resolution and different phase, this kind of the muti-platforms, the multi-phases, the multi-spectra and the multi-resolution data is welling up fast by a amazing quantity, which have formed multiple source image Pyramid in the identical area. How to obtain more richer, more usefule and more reliable information from these images, to synthesize the different data's respective superiority and the complementarity which is more important and practical, the multiple source sensing image fusion is most valueable technological means in processing the mass data of sensing,which is one of the key researches of the current remote sensing applicatione.Through multiple group experiments, the article takes research and analysis on sensing images's pretreatment and fusion technology of element level, the principal tasks are as follows:Image registration is the the premise and foundation of the multiple source sensing image fusion processing, Contrastive and analysis the three kinds of registration method based on domain transform, the phase correlation method, the logarithm polar coordinate method, the Fourier-Mellin method, show their's characteristic and the applicable scope; We mainly proposed the multi-scale Harris operator of the automatic registration method.For the concrete fusion goal we discussed the the fusion pretreatment technology of the pan sensing image, multi-spectral image and non-remote sensing thematic maps. Processes the pan sensing image by the gradation enhancement and the image peaking; Processe he multi-spectral data by true three-dimensional color enhance and the true color composite technologies; Processes the non-remote sensing thematic maps by the method of imaging geoscience information.Has analyzed 6 commonly used pixel-level fusion method:Including Wavlet, Mutiply, HPF, HIS, Brovey, PCA, through the experiment contrastive and analysis these 6 fusion results, focusing on HIS, Brovey, PCA to carry on the improvement of itself and integration of Wavlet, which realizes the enhancement of the algorithm performance. Studied the subjective 5 point systems and the objective quota evaluation criteria to the fused result, and carried on the quality synthetic evaluation and analysis to the above 12 fusion experimental results, has drawn some useful conclusions.In summary, this article has thoroughly and comprehensively studied the different sensing image registration technology, the period pretreatment technology of different fusion goal, the fusion principle, and several kinds of pixel-level fusion algorithm and the improvement of algorithm, which has a vital mean to the later research and work..