Font Size: a A A

The Research Of Land Use Change Detection Based On Image Segment Information Mining

Posted on:2011-07-04Degree:DoctorType:Dissertation
Country:ChinaCandidate:X LiFull Text:PDF
GTID:1109360305483188Subject:Photogrammetry and Remote Sensing
Abstract/Summary:PDF Full Text Request
Land resources are the base for human survial and sustainable development. With the development of social, economic, scientific and technology, land use/land cover change, which is caused by the human, has become an important components and the main reason for the global environment change. As the land use/land cover change detection is very impotant in many areas, such as agricultural evaluation, environment monitoring, urban planning and so on, the method for large area change detection, which should be efficient, fast and practical, is needed urgently. With the birth and development of remote sensing technology, the method of land use change detection using remote sensing images has become the focus of the research by the domestic and abroad scholars. The research of how to make full use of remote sensing technology to automatically detect changes has great theoretical significance and application value.After the conclution of the result of remote sensing image change detection research, this paper proposed a method of remote sensing image change detection based on image information mining. The analysis object of this method is image segment, and the basic data of this method are remote sensing images and land use vector data. From the perspective of image information processing level, the framework of change detection could be designed. First, image segments are obtained by the conflation of land use vector data and remote sensing images. Second, the features of the image segments are extracted. Third, the sample segments are selected and updated. Last, discriminate the changes and complete the change detection.This paper researchs the key technology, such as the texture feature extraction and correlation analysis, in the application of remote sensing image change detection. A multi-scale texture feature extraction method based on image segment has been proposed. This method obtains texture images from the remote sensing images in different resolution and different direction using wavelet decomposition technology. By measuring the information entropy and information gain ratio in the different texture properties, the contribution index of texture properties is constructed. The contribution index is used to select the texture features, which are significant for the image analysis. This paper analyzes the distribution of the image segment’s features. After revising the algorithm of sample selection and update, a model of remote sensing image change detection based on feature pattern analysis has been proposed. The change discrimination also is the key technology in the remote sensing image change detection. In this paper, the results from different classifiers are analyzed, and the result of the change detection has been optimized by the method of decision fusion. The sequential analysis is a method of decision fusion based on single-classifier. The method of the sequential decision fusion modifies the output results from the single-classifier by feedback iterations. Then, the paper proposed a method of multi-classifiers fuzzy decision fusion based on the classifier’s confidence identification. By constructing the confidence identification of classifiers, the fuzzy decision fusion method integrates the different results from multi-classifiers using the fuzzy set theory.At last, the effect of the historical image auxiliary data in the remote sensing image change detection is studied. After the establishment of the stochastic process model using the historical image, the state trasnsition matrix can be obtained to forecast the rules of the change. The experiment analyzes the conditions on the use of the auxiliary data, compares the accuracy of the state transition matrix obtained by different historical images.
Keywords/Search Tags:Image Information Mining, Change Detection, Land Use, Pattern Recognition, Data conflation, Consistency, Decision Fusion, Random Process, Markov Chain, State Transition Matrix
PDF Full Text Request
Related items