Research On High Resolution Remote Sensing Image Change Detection Based On Object Oriented Classification

With the development of economy and society,the city to speed up the process,people gradually strengthen the breadth and depth of land use,profound changes in land use structure of arable land,water and other natural resources are decreasing,while construction land,roads and other social resources is increasing.Therefore,it is of great significance for China's economic construction and environmental protection to understand the situation of land use in a timely manner.And with the development of sensor and information processing technology,remote sensing technology is developing towards high spatial resolution,high spectral resolution and high temporal resolution,while the traditional change detection methods for high ratio image exposed some drawbacks,so the exploration of new method for high spatial resolution remote sensing image detection of land use is a hot topic of current research.In this paper,we study the problem of determining the change threshold in the change detection,which can effectively avoid the subjectivity and complexity of the change threshold determination,as well as,we can clearly get the change of each category.Using two GF-1 images,overlay category of vector layers based on the theory of superposition analysis,and get new object that contains the change information.Meanwhile,according to the superposition results,define change type directly and classify the type of change through the method of fuzzy classification to complete change detection.In this study,we first need to split the image into the image object;next,extract value and establish of classification rules,and then classify the image object through multi-level fuzzy classification;finally,accomplish the change detection by superposition analysis.By error matrix and consistency indicators of categories to evaluate the change detection,the results show that this change detection method for detection probability is 88.39%,11.61% of false alarm rate,8.04% of false alarm rate,19.65% of total error;most classes of accuracy up to 80%,only a handful lower-accuracy classes and errors,we need to further improve the accuracy.In General,this method in the paper avoids the change threshold to determine and effectively reflects regional variations,as well as,provides a favorable basis for macro monitoring of land use.