Abstract: With the assistance of high resolution remote sensing image data, an object-oriented recognition model for river system information is proposed. This model summarizes the characteristics of water information in image, and defines the rules and its sub-characteristics in detail. Moreover, an experiment for water system information extraction is conducted for Yuan'an County, Yichang City as a case study. Result demonstrates that the proposed model could extract the river system features on remote sensing images effectively, and therefore is more accurate in identifying water system in comparison with conventional classification methods. The present research is helpful in improving the automation and accuracy of river information extraction and verification for the promotion of river chief system.

Key words: river water system, information extraction, characteristics and rules, water body recognition, river chief system