The Urban Construction Information Extraction Based on New Structure Decision Tree

摘要

In this paper, the information of construction land is extracted by decision tree and mixed pixel decomposition model. The decision tree has a new structure and is designed as two layers. The top layer emphasizes the synthetically application of spectrum and texture, and bottom layer emphasizes the use of knowledge. The decision rules comes from spatial characters of common city and KT transform .It has been tested with TM data in Nanjing and the accuracy of the result rises from 82.7% to 91.7%. At last, the urban spatial expansion direction and temporal change gradation between 1994 and 2002 is analyzed based on multi-temporal data.