| Remote sensing image segmentation is the key technology of remote sensing image processing,which is not only the premise and key of remote sensing image processing task,but also lays the foundation for the subsequent processing of remote sensing image.Remote sensing image is characterized by large amount of data,complex information and strong band correlation.How to accurately segment remote sensing image is one of the key difficulties in remote sensing image processing.Fuzzy clustering algorithm is widely used in remote sensing image segmentation because it is good at processing fuzzy information in the process of clustering and has simple principle and convenient calculation.However,the existing fuzzy clustering algorithm still has some problems such as poor robustness,high time complexity and low precision of edge extraction when it is applied to remote sensing image segmentation.Intuitionistic fuzzy sets are the expansion and development of fuzzy sets.Compared with fuzzy sets,intuitionistic fuzzy sets can express fuzzy information of "neither this nor that" more accurately.Based on this,this paper proposes a high-resolution remote sensing image segmentation algorithm based on intuitionistic fuzzy set.Aiming at the problem that the traditional fuzzy clustering algorithm ignores the spatial information and spectral uncertainty of the image and is insufficient to describe the category fuzzy information,intuitionistic fuzzy set is introduced into the remote sensing image segmentation process.The details are as follows:(1)In order to fully describe remote sensing image in the spatial relationship and spectral measurement uncertainty,fuzzy method is studied in this paper,the existing image,and focused on the task of remote sensing image segmentation,respectively established based on intuitionistic fuzzy image spectral uncertainty model and the uncertainty of image based on spatial relation intuitionistic fuzzy model.The intuitionistic fuzzy model based on the spectral uncertainty of the image analyzes the spectral measure uncertainty of the image through the maximum entropy method,and then calculates the band index to transform the remote sensing image into intuitionistic fuzzy set,so as to model the spectral measure uncertainty of the image.The image intuitionistic fuzzy model based on spatial relation uncertainty describes the influence of neighborhood pixels on central pixels by upper and lower bounds of membership degree function,and then uses interval membership degree to model spatial relation uncertainty of image.Finally,an image intuitionistic fuzzy model is established by combining the two intuitionistic fuzzy models with spatial information and spectral information.This model fully describes the uncertainty of remote sensing image in spatial relation and spectral measure,and achieves intuitionistic fuzzy expression of remote sensing image,which lays a foundation for accurate segmentation of subsequent remote sensing image.(2)In order to eliminate the influence of fuzzy factor values on the clustering results,the upper and lower bound functions of the membership degree and non-membership degree of the interval type II intuitionistic fuzzy sets were constructed by using different fuzzy factor values.Secondly,the category uncertainty is described by combining the category membership degree with the non-membership degree in the form of linear weighting,and the objective function is defined comprehensively by taking the distance between spectral measure intuitionistic fuzzy sets as the dissimilarity measure.By discussing the value of fuzzy factor,the clustering center is recalculated by combining the upper and lower bounds functions of membership degree and non-membership degree.Finally,the upper and lower bounds matrices of clustering center,membership degree and non-membership degree are solved,and the accurate segmentation of remote sensing image is realized by reducing and de-fuzzifying the interval type II intuitionistic fuzzy set.The proposed algorithm and comparison algorithm are used to segment simulated image and real color remote sensing image respectively.Qualitative and quantitative evaluation of segmentation results show that the proposed algorithm can better deal with the uncertainty of spectral measure and spatial relation of image,improve the description of category fuzzy information in the process of clustering,eliminate the influence of fuzzy factors on clustering results,and obtain higher accuracy of image segmentation results. |
