Research Remote Sensing Image Segmentation Based On Cluster Analysis

Cluster analysis of remote sensing image is more important in the field of remote sensing research,it's comprehensive response to the Earth's surface feature information,it has an great significance to the classification of city information,It is an effective reference standard especially for the development plan of the town,the extract of interesting information,etc.So stringent requirements are needed in the classification accuracy and integrity.This paper mainly used remote sensing image in Yangyuan County of Zhangjiakou as the processing data for remote sensing image segmentation cluster analysis,focusing on their own characteristics of GLCM algorithm,principal component analysis and two-dimensional discrete wavelet algorithm combined with three models proposed by k-means algorithm,respectively.Then the experimental results are analyzed and obtained conclusion.The main contents of this paper are:1?Data this paper used was sourced from multispectral images and full-color images in Yangyuan County of Zhangjiakou,before the application of cluster model of remote sensing image,a series of image preprocessing for custom coordinate system,atmospheric correction,data scaling,image fusion were needed.2?This paper did research on the basic principles of K-means algorithm in remote sensing image clustering research,processed careful analysis of the algorithm process and factors,then analyzed the characteristics GLCM Yangyuan County of Zhangjiakou space remote sensing image texture,the theoretical basis of mathematical algorithm was deeply discussed,the k-means clustering based on the description of remote sensing image characteristics described on energy,finally got results.3?Analyzed the influence of principal component on remote sensing images,which had positive effect on dimensionality reduction and noise removal,then the main component band in remote sensing image Yangyuan County was selected for k-means clustering,reducing the impact of discrete points on clustering and improving clustering accuracy.4?Due to existing gaps in image segmentation algorithm based on clustering,and based on advantages and disadvantages of wavelet transform and k-means algorithm,the remote sensing image clustering segmentation algorithm based on wavelet transform was proposed,it can be a good approximation to distinguish image component and detail component,characterized by a few parameters to describe the characteristics of the image,because of this unique advantages,has been widely used in many fields.Deep research on wavelet transform,especially on description of the characteristics of remote sensing images of two-dimensional discrete wavelet transform in space and scale comprehensive,and the advantages of wavelet transform were applied to the cluster of remote sensing images which enabled us dig out the accurate and effective data of remote sensing images we interested.The clustering results of three models were analyzed and compared on the whole and in the details,analyzed heir advantages and disadvantages based on clustering results map and related data of the three models in remote sensing image clustering Yangyuan,which provided references for improving algorithm model and making further study on remote sensing images.