Research On The Cloud Classification Technology Based On The Texture Analysis

With the development of satellite remote sensing technology, the use of the information processing technology of satellite cloud of information extraction for relevant and processing has become the moment one of the most important means of satellite cloud. Satellite cloud remote sensing imaging information up to the most one of the data source, the use of image processing methods to analyze and extract useful information of cloud images, in order to determine the change of the weather situation and the changes of atmosphere has become a mainstream mode of operation of the weather forecasting community. Therefore, the satellite cloud information processing, analysis and application, has become the main research directions of the satellite cloud letter processing.Obtained after a previous study concluded:GLCM able to make the image intensity statistical laws brought into full play, while the wavelet analysis method can make the image multi-scale characteristics of a good play, Based on the multi-scale point of viewproposed a new method based on wavelet scale co-occurrence matrix texture extraction.The first use of the wavelet transform multi-scale decomposition of the image, the different cloud types scale wavelet high-frequency and low frequency diagram, then the image of the multi-scale GLCM texture extraction and then its combination to form a feature vector, the final texture image classification.The main research results are as follows:(1) From the gray level co-occurrence matrix texture analysis method and wavelet transform analysis method, combining the two methods, proposed the use of le wavelet scale co-occurrence matrix texture analysis method on the satellite image texture feature extraction.(2) For the satellite cloud classification, selection of appropriate kernel function, the training algorithm, presents many types of SVM (Support Vector Machine) model, based on the multi scale texture feature extraction, participate in the multi-class support vector machine classification in order to achieve different cloud types of classification.(3) From aspects of the engineering practice, the whole system design framework and the various function module were introduced, combined with scale symbiotic matrix and many kinds of support vector machine research, design and implement the "satellite cloud pictures cloud classification software system".