Study Of Multiple Kernel Support Vector Machine In The Application Of High-Resolution Remote Sensing Image Classification

In recent years,with the continual development of space and remote sensing technology,earth observation has entered a new era.The high-resolution remote sensing images are becoming the main data sources for obtaining ground information increasingly,which has been widely used in many fields such as the general survey and monitoring of geographical conditions,urban planning,resource survey and large-scale mapping and so on.In addition to providing spectral features,the high-resolution remote sensing images also have rich shape and texture features,which bring new challenges to the intelligent classification of images.Support vector machine(SVM)as a classification algorithm based on kernel function which has good generalization performance in pattern recognition and so on,has been widely used in various fields.But because of the different kernel function in SVM has different classification performance and the complexity of the distribution of land information in high-resolution remote sensing images,which contributes to the classification research of high-resolution remote sensing images based on SVM remains to be further.Therefore,the application of SVM in the classification of high-resolution remote sensing images is studied in this article.Firstly,the background and significance,current research and existing problems of the research are analyzed.The basic principles,characteristics and existing problems of support vector machine are introduced.The characteristics of high-remote sensing images and some classification and segmentation methods are introduced.Secondly,in view of the problems of SVM in the classification of high-resolution remote sensing images,this article proposes to use multiple kernel function weighted combination to construct multiple kernel SVM for classification.Finally,in order to further improve the generalization ability of classifiers,an ensemble learning model based on improved RBaggSVM algorithm is proposed,which adds multiple kernels information to RBaggSVM algorithm.This method improves the classification performance and generalization ability of single SVM classifier.The main contents of this thesis are:1.In view of the complex land information and different kernel function and kernel parameters has great influence to classification results.In this article,to segment the image firstly.Extracting the spectral,texture and shape features.And feature selection and weighted processing is proposed;secondly,combinating the global and local kernels to construction multiple cores SVM and using artificial bee colony algorithm to optimize the kernel parameters of multi core SVM.The multi-core SVM classification model is applied to the high-resolution remote sensing image classification experiment.2.To further improve the multiple kernel SVM's performance in high-resolution remote sensing image classification,a multiple kernel SVM ensemble classification model of improved RBaggSVM algorithm is proposed in this article.On the basis of RBaggSVM algorithm,multiple kernel functions are used to replace the single kernel function in the algorithm,and the membership classification results are combined by weighted voting.