Applicability Analysis Of Region Multi-center Method For Rs Imagery Classification

Remote sensing technologies can rapidly, accurately and timely provide the latest earth observation data and remote sensing image has become an important data source to obtain land cover information. Remote sensing image classification is an important aspect of RS applications.Due to the same object with different spectrum and different objects with the same spectrum, the difference of intra-class spectral characteristics gets larger and the inter-class spectral characteristics may overlap, and a convenient multivariate statistical model is in general not available for the RS imagery classification. It is difficult to recognize ground objects correctly with the spectral characteristics alone. Region multi-center method (RMC) is based on region characteristics. In this method, the classification cell is region, and classificatory pattern is formed by intra-centers.In this paper, we compare the advantages and disadvantages of the RS classification methods based on sample and parameter selection. And introduce the theory of RMC method. We discuss how different parameters, such as intra-center, classification cell and threshold of RMC features, impact the classification accuracy by applying RMC method to the experiment for LU class. In the experiment, the method also was applied to recognize single class and to classify based on different spatial resolution RS imagery, and we got good classification result maps, consistent with visual interpretation. Finally, the RMC method was employed in the urban expansion change detection.Experiments show that: (1) the RMC method of RS imagery do not ask for sample selection, the intra-center amount effects the classification accuracy in a certain range, the impact of the threshold of percent on the classification is severe and the region cell may be ignored; (2) in the RS classification application, there is a problem that the difference of intra-class spectral characteristics gets larger and the classificatory pattern does not meet multivariate statistical distribution, RMC method can combine the spectral information and region information to solve the problem. And the RMC method can be applied in urban expansion change detection well.