Urban information extraction for remote sensing images considering the human cognitive characteristics - A case study of central urban area of Guangzhou

Abstract

Remote sensing images are playing an increasing important role in the research of urban ecological environment, especially the extraction of the urban information. However, previous approaches towards processing the remotely sensed images required a lot of manual work. To try to promote the auto-processing of RS images, this paper proposed a novel approach based on the shape adaptive neighborhood (SAN) for RS images, considering the characteristics of cognitive psychology. Firstly, like the mechanism of attention in the psychological process, the heterogeneity based on the color characteristics was employed to determine the SAN of each pixel. Then all the color features, texture features and shape features were extracted from each SAN, and were implemented a feature level data fusion in the classified feature space of the RS image. Finally, the features were used to extract the information. As a study case, a central area of Guangzhou city was selected to test the accuracy of the SAN-based information extracting approach. The area was classified by five types of land use type, and the proportion of each type of land was calculated. A sampling schema based on the cluster sampling method was used to do the assessment of the accuracy. Experiment results showed that, the total precision of the classification was 0.96873. ©2009 IEEE.