Study On Feature Extraction From The Remote Sensing Image Based On Wavelet Analysis

With the rapid development of Remote Sensing technology,extraction and explanation of feature information of interested target, such as spectrum, edge, alteration etc., from the abundant remote sensing information has become an important subject in remote sensing data processing. Moreover, remote sensing image processing method based on wavelet analysis has been come forth as assistant means, which let us extract useful information better and improve the precision of target recognition.Wavelet theory has a good application potential in remote sensing feature information extraction and much superiority than the conventional method. This thesis simply presented wavelet analysis theory at first, and then discussed image enhancing method and principle using wavelet transform. In the end, through the experimental contrast of extraction of remote sensing alterative abnormal information based on different methods taking Jing erquan of East Tianshan in Xinjiang as study area, summarized the most effective way, which is the wavelet adaptive threshold enhancing combined with the principal component transform using"TM1, TM2, TM3, TM4, TM5, TM7"as the feature band. The main content of the paper includes:First, introducing background of the research and the development trend of feature information extraction from the remote sensing image, and simplifying the content, the goal, the significance and the method of the research in this thesis.Second, introducing the wavelet basic knowledge, continuum and discrete wavelet transform, multi-resolution analysis and orthogonal wavelet transform, the application of two-dimensional wavelet transform and wavelet packet in image processing. Third, presenting image enhancing method and principal with wavelet multi-resolution analysis based on studying multi-scale decomposition of remote sensing image according to the feature of the target to be extracted.Fourth, introducing the experiments of extracting the alterative abnormal information using ETM+ data, MATLAB developing flat and remote sensing image processing software as ERDAS,ENVI,PCI, methods of which respectively are the all band principal component extraction, the selective band principal component extraction, the direct extraction with combinative method of principal component and ratio band, and the extraction method of wavelet adaptive threshold enhancing combined with principal component analysis. Among these methods, the extraction model of enhancing based on wavelet analysis not only have a good effect, but also can keep more spectral information, moreover, it is of the features such as rapid, effective and economic etc.At last, the paper contrastively analyzed the alteration information extraction results of different methods, provided primary mineral forming prediction to the area of Jing erquan and confirmed ore-finding targets through geologic explanation of the alteration extraction method proposed in this paper and comprehensive spatial relation analysis between the alteration information and stratum, structure and rock body, which has important referential sense in discovering the big ore or metallogenetic district zoo.