The Research Of Object-based High Resolution Remote Sensing Land Use Change Detection

With the rapid development of modern society, the exploitation of land resources is growing faster, while the change period of land use is shorter, the change range is wider and the update of data falls behind. The remote sensing technology with advantages of obtaining data quick, covering large area and costing less manual work is the best means for change detection. The improving resolution of remote sensing image data on the one hand enhances the accuracy of the image information extraction; on the other hand it demands the image processing technology smarter. Optimization of high resolution remote sensing image processing technology is of great significance for land use change detection. This article has applied the object-oriented technology to detect land use change and has made specific researches on how to choose the optimal segmentation scale and how to apply the object features for classification.First, this article has proposed an improved evaluating index and a calculation model of the optimal segmentation scale using experiments to prove both of them effective. The evaluating index includes intra-homogeneity index and the inter-heterogeneity index. The intra-homogeneity index is calculated by combining the standard deviation, the area and the account of objects. The inter-heterogeneity is calculated using the neighborhood total variation within the border region between objects. The calculation model is the non-linear regression model coming after analyses of experiment data by using the mathematical statistics method.Secondly, this article has designed the feature-rule base of land types against the classification problem after segmentation, providing effective storage and management for the main object features and classification rules of land types and offering dynamic link of classification rules for the classification of land data. The classification rules have combined the fuzzy logic method and the decision tree method making structured fuzzy rules for the land types. The features selection extracts the optimal feature set by using the information theory to calculate the amount of information and correlation between features.Finally, this article has experimented on the land use change detection with the method introduced above. Using the typical land types, the experiments has calculated the optimal segmentation scale, selected the optimal feature set and made the classification rules, and also established the corresponding feature-rule base instance. This article has processed the image data of different times by following the steps of segmenting image with optimal scale, classifying image with classification rules, comparing the classification results and finally extracting the land use change information. The experimental result has proved that the proposed method is effective and practical.