Research On Optical Remote Sensing Image Change Detection Based On Fuzzy Clustering Algorithm

Remote sensing image change detection is the technique of analyzing and determining the characteristics of surface change in multi-temporal remote sensing images of the same geographical area but from different times.It is a major application of remote sensing technology as well as an important means of national geographic and battlefield conditions monitoring.The image of different phase is easily affected by the difference of environment and sensor performance in imaging process,which may lead to severe "spectral aliasing phenomenon".Therefore,how to remove the influence of these uncertain factors and extract the main changed information in the change detection process is still a difficult problem and research hot spot.Membership is utilized in fuzzy set theory and fuzzy clustering algorithm to describe the fuzziness of the object,which shows great advantage in dealing with the problem of fuzziness and uncertainty.In this paper,on the basis of the characteristics of the optical remote sensing image,unsupervised fuzzy clustering algorithm is used as a main tool to change detection technology of optical remote sensing image systematically,the limitation of classic method is analyzed and the corresponding improvement plan is put forward.The main work and innovation of this paper are as follows:1.The result of the traditional pixel based change detection is discrete,and it's easy to cause "salt and pepper phenomenon".Aiming at solving this problem,an improved type-1 fuzzy clustering algorithm is proposed.Firstly,the change vector analysis method is used to construct the difference image with the well-matched remote sensing image.Then an improved type-1 fuzzy clustering algorithm is used to segment the difference image.By adding neighborhood information to the objective function during iteration,the final changed information is extracted from the image pairs.Comparing with traditional change detection method,this method can effectively reduce the number of broken spots,maintain the structure and shape of object,highlight the changed information and improve the accuracy of change detection.2.There is great uncertainty in change detection procedure such as different objects with the same spectra characteristics or the same object with different spectrums.To solve this problem,an approach based on adaptive interval type-2 fuzzy clustering change detection algorithm is proposed.Due to the serious “spectra aliasing” phenomenon of optical remote sensing image,the traditional type-1 fuzzy clustering algorithm can't describe its high order uncertainty,so it is difficult to accurately model the difference image to extract changed information.Based on interval type-2 fuzzy clustering algorithm,an adaptive interval type-2 fuzzy clustering algorithm is proposed to detect the changed area of remote sensing images.Membership is updated through two kinds of adaptive distance metric that not only the dependency of default parameters of traditional interval type-2 fuzzy analysis is reduced but also spatial neighborhood constraint is incorporated.Finally the change information of land cover is extracted by updating membership iteratively.3.To suppress the "pseudo changed information" in the difference image and utilize the multiple features of the image effectively,a change detection method based on saliency map and multi-kernel clustering algorithm is proposed.Saliency map is incorporated in this chapter and high frequency information is suppressed through multiple Gauss filters.On the basis of saliency map,the regional feature vector is extracted in order to reduce the influence of local registration error.Meanwhile,considering the weight of multi-dimensional features may be inconsistent,a multi-kernel fuzzy clustering method is applied so that the optimal weights of each feature are automatically obtained by minimizing the objective function.The experimental results prove the effectiveness of this algorithm.