The Study of Object-Oriented Classification Method of Remote Sensing Image

Abstract

The object-oriented classification method of remote sensing image takes the characteristics of the imaging spectrum and differences in geometric characteristics into account, which can extract more accurate image information. Regard Cangzhou City in Hebei Province as the study area, and ASTER remote sensing image as the data source, first of all, the study area has been multi-scale segmented. It can determine the best optimal partition scale and parameters according to different types of ground respectively, and then makes use of the nearest method to object-oriented classification research. Finally evaluate classification accuracy of the classification results via confusion matrix; what's more make the method compared with the traditional pixel-based maximum likelihood classification results. The results have showed that the results of object-oriented classification effectively avoid the salt and pepper phenomenon, and the accuracy of overall classification is 91.5486%, Kappa coefficient is 0.8990, both are higher than the overall accuracy of the maximum likelihood method which the accuracy of overall classification is 61.5134%, and Kappa coefficient is 0.4986. So it has very good application prospects in classification application of ASTER data.