Research On Classification Algorithms Of High Resolution Remote Sensing Images Based On Fuzzy Clustering

Remote sensing technology is a comprehensive detection technology involving various fields.It can process the electromagnetic wave information radiated by the received features in different ways to observe and identify various ground objects and related phenomena.It is widely used in various fields and has become an indispensable and effective means in the application process of planning and earth environment resource survey.The classification and mapping of remote sensing images provide important technical means for making or updating maps and thematic maps,providing real and reliable information for the development of various industries,and is the basis for geography monitoring and ecological protection.This thesis focuses on the application of fuzzy clustering algorithm and feature selection algorithm in remote sensing image classification.Firstly,this thesis analyzed the development background,research direction,research status and existing problems of image classification at home and abroad,and introduces the segmentation,classification methods and fuzzy clustering algorithm.Secondly,in view of the insufficiency of the fuzzy C-means(FCM)algorithm in the process of operation,further improvement is made to the algorithm.Finally,in order to solve the problem of high correlation and redundancy between features extracted in the feature extraction process in the image classification experiment,which affects the classification efficiency,a remote sensing image classification algorithm combining the mRMR algorithm with the improved FCM clustering algorithm is proposed.The main research contents of this thesis are as follows:1.Aiming at the problems of various image features and mixed features,poor stability of existing FCM algorithm and insufficient utilization of spatial information,an improved FCM clustering algorithm combined with weighted multi-core SVM classifier is proposed for remote sensing image classification.In the computing stage of clustering segmentation objective function,the influence of single pixel on clustering results is considered,at the same time,the influence of neighborhood pixel is measured by attractive model;In the feature extraction stage,spatial pixel template method was used to extract the feature points of image spots,and image classification based on weighted multi-core SVM classifier,so as to obtain the classification information of ground objects.2.To solve the problem of poor classification caused by high correlation and high redundancy between image features,an image algorithm based on mRMR selection and improved FCM clustering was proposed.Firstly,the image was segmented based on the object confidence index(OC),and in order to solve the problem of feature redundancy,the feature selection was realized by using the mRMR algorithm.Finally,the extracted features were clustered by the improving FCM algorithm to obtain the final classification results.Experimental results indicate that the proposed algorithm has the ability to reduce the redundancy and correlation between features,and effectively improve the classification efficiency of images.