The Study Of Change Detection Of Remote Sensing Image Based On Contourlet

Change detection ofremote sensing image is a method that could detect the change information in the area over different time through the analysis of remote sensing images in different periods. Change detection in hyperspectral remote sensing technique development is based on the rapid spread of satellite and other satellite series launch and normal operation of the Foundation. In order to get the regional differences and changes in the Earth’s surface, change detection has become one of the most important research directions. Along with the rapid expansion of the scale of urban,land surface of urban changes more and more quickly. Predict the development of a region according to the change detection is of practical significance. Andchange detection of remote sensing imagetechnology also has active meaningin earthquake victims monitoring field, usually earthquake disaster has the features of short and spread range wide, and disaster of rescue work need to understand earthquake victims situationin short time, General ways are hard do this, but change detection methodprovides a reliable way for earthquake victims information extraction and the disaster monitoring.This method is designed based on nonsamplingcontourlet transform and fuzzy c-means clustering method. Method in this paper is an unsupervised method, to avoid the subjectivity of some classification. Nonsamplingcontourlet transform is a method based on contourlet transform, with the advantages of noise both on the basis of the translation invariance, provides a good basement for the clustering method. Improved fuzzy c-means clustering method considered the neighborhood information in the traditional method of fuzzy c-means clustering, reduces the error rates that the pixels in the same areadivided into different categories. Unsupervised method is designed in this article, including the construction of difference image, nonsampling contourlet transform filter bank design and fuzzy c-means clustering parameters selection. This article chose five experimental zone for experiments, suburb of Huairou, respectively, as well as different areas of the earthquake-hit ludian County, for monitoring urban land development and earthquake monitoring is of positive significance. Paper compared experiment get changes detection resultsthrough maximum class between variance method and design method, through qualitative analysis and quantitative confusion errors matrix including detection success rate, wrong check rate, missed rate and the Kappa coefficient, changes detection evaluation index on two species method multiple experiment district of detection effect for has evaluation, founding paper detection method has strong of stability and good of detection effect, changes features profile of detection effect also better, has strong of application value.