Study On The Hyper-spectral Image Classification Algorithm Application Based On Wavelet Transform

Because of the hyper-spectral remote sensing technology, including the charac-teristics of large amount of data information, hyper-spectral remote sensing technologyhas been widely applied to the investigation of vegetation, remote sensing, remotesensing, agriculture, environmental monitoring and atmospheric research, the analysisand processing for hyper-spectral remote sensing image is faced with greatopportunities and challenges. In the image processing and analysis process, howto retain more information based on effective removal of redundancy is the core of thecurrent study, and this paper is focused on the application research of classification ofhyper-spectral remote sensing image.In the past twenty years, wavelet transform has obtained the rapid development inthe field of experts and scholars efforts. Discrete wavelet transform (DWT) isthe frequency analysis method of signal processing and development basedon Fourier transform.This method can simultaneously performance characteristics intime domain and frequency domain, it was play a decisive role of the algorithm inimage coding.Hyper-spectral remote sensing image classification method and classificationof existing without any algorithm can be applied to all types of images, and there willbe a mistake, omission and error in the classification process. Hyper-spectral remotesensing image data can be read as a three-dimensional cube data, this paper attemptsto do the Fourier transform and wavelet transform of the signal processing onpixel spectrum curve,and then to classify the hyper-spectral remote sensing image. AtFirst, put forward the classification of hyper-spectral image by Fourier transform basedon MATLAB programming.The Fourier transform are the methods of signalprocessing can be used for classification of hyper-spectral images of the simulated data,but if in the face of mixed pixel or actual complex The Fourier transform cannot meetthe requirements. On this basis, further puts forward algorithms for hyperspectralimage classificationusing wavelet transform, this algorithm presents a method forwavelet feature extraction, after the discrete wavelet transform of pixel spectralcurves,, and in accordance with this algorithm on simulated data and AVIRIS data for MATLAB programming.For the same AVIRIS data, this paper will also in output image and quantizationprecision two aspects to contrast two algorithm——hyper-spectral image classificationbased on wavelet transform and the traditional classification method classificationalgorithm based on spectral angle mapping, prove that discrete wavelet transform caneffectively applied to hyper-spectral image classification applications in.