Development And Application Of Typical Land Cover Remote Sensing Scale Analysis System

Scale as a basic feature of remote sensing images, representing the richness of the information contained in the image. In the research and production practice, there will be a contradiction that existing remote sensing images may not be able to meet the actual needs of the application, such as :For example: equally spaced, scale continuous image can not be obtained by the sensor, particular scale remote sensing image may express one object accurately,but not the other. Therefore, it’s very meaningful to make full use of the data by obtaining the optimal scale to express particular object through researching the scale effect of the object, In this paper, Based on Pleiades and Landsat images, the research of remote sensing image feature extraction, optimal scale analysis and scale effect in image classification were carried out in this paper. The methods used in this paper and the conclusions are as follows:(1) remote sensing image feature extraction includes two aspects: spectral feature extraction and texture feature extraction. Spectral feature extraction includes image gray mean, variance, etc. This paper used three methods: histogram statistics, high-pass filtering, gray co-occurrence matrix to obtain texture images of samples, and then create the corresponging feature database.(2) This paper analysised the optimal scale of several typical objects by using the local variance, variogram methods, then optimal scale to express different objects were obtained, during the calculation of the optimal scale, optimal scale calculation and scale expand methods are both have influence on the results.(3) In remote sensing image classification, this paper studied image classification with the auxiliary of texture features. select an area in the original remote sensing image as an experimental zone, use GLCM’s four characteristic parameters: contrast, correlation, energy, homogeneity to implement texture feature extraction, then combine the band of resulting texture image with the original image, result shows some feature has been enhanced after the combination, and then did image classification, result shows the classification accuracy improved with the help of texture images. In addition, this paper studied how the accuracy of landcover classification varies smoothly with the change of scale, there is a peak or small pieces of recovery near the optimal scale.Finally, build the scale effect analysis software system based on.Net Framework, ArcSDE, Oracle and other technologies The system include several modules: Remote sensing data feature extraction, management system of multi- sources characteristics, optimal scale analysis, remote sensing image classification, backup and recovery of database etc. Facilitate the study of scale issues.