Inferring urban land use from IKONOS image

Timely upgrading urban land use/land cover information is a prerequisite for urban planning and management. Nowadays, remotely sensed imageries, especially higher spatial resolution, imageries like IKONOS (1m for the pan data) and Quickbird (0.68 m for the pan data), have become indispensable to provide such a kind of needed information. This study is aimed to produce a land use map using IKONOS imageries obtained for a part of Nanjing City, Jiangsu Province of China. The study area contains seven land use types such as vegetated strips, park, water, road, residential (old & new) and industrial areas according to the urban land use classification system stipulated by the Urban Management Committee of China. In this study, a sequence of methods is developed to obtain urban land-use map and the methods-acquired results are in-situ tested. Specifically, the hierarchy tree classification method is utilized to obtain land cover map with consideration of contextual information. Spatial analytical functions are then applied to improve the accuracy of the land-use classification. The unsupervised ISODATA classification based on characteristic density map using 99×99 moving window is employed to further improve the classification accuracy. The accuracy assessment (total accuracy: 94.54%; Kappa coefficient: 0.93) indicates that methods developed in this study can be confidently used to timely and economically provide needed information for urban planning and management.