Research On Automatic Classification Algorithm Of High-resolution Image Based On SVM And Its System Realization

With the implementation of "high resolution earth observation" great project in China,the rich geographic information data obtained by high resolution images provide more services for social development and human progress.Because of huge amount of information conveyed by image,invest expensive costs and long time workloads are ineluctable to manual segmentation of images for statistical data.In order to provide convenience for subsequent application of information on land resources,automatic classification of high resolution images by computer is considered the better choice compared with manual segmentation,it is more conducive to the extraction and recognition of surface information of earth.Compared with the traditional classification algorithm,the advantages for high resolution image classification are provided by the nonlinear classification function of SVM(Support Vector Machines)algorithm.This kind method can be applied to the high resolution image classification of small samples with fast classification speed and high accuracy,and solves the current situation of visual interpretation and manual segmentation of images.Therefore,it is important for automatic classification of high resolution images to be fast and autonomous by designing and implementing automatic classification system of high resolution images based on SVM.The thesis mainly includes two parts: algorithm research and system implementation.The algorithm research is carried out around SVM theory,and the system implementation is completed in the VS2010 environment using C# programming language.In the algorithm section,the mathematical principle of SVM algorithm and the principle of multi-class classification are studied.The purpose of study is to obtain the spatial features which are more conducive to the classification of high resolution images.The SVM classifier model is designed and simulated for classification using different eigenvalues,and the classification results are visualized.The availability of simple color texture in high resolution image classification is verified.In the system part,after understanding the system requirements,the designing process of the classification system is programmed,and the system design scheme is formulated,including the overall design idea,the interface design and the module division.The algorithm design and system implementation of three important modules are completed,including feature extraction,sample training and data prediction.Finally,compared with other traditional classification algorithms,the performance of SVM algorithm in high resolution image classification is verified.The results show that the SVM algorithm combining with the high resolution image data characteristics makes full use of the advantages of SVM algorithm in the case of small sample classification,including high accuracy,fast calculation speed and strong generalization ability.Therefore,automatic classification system of high resolution images has the characteristics of extendibility,rapidity,practicability and autonomy,which provides important reference value for technology research and software development of high resolution image classification.