A Simplified Approach for Classifying Urban Land Cover using Data Fusion

Many government agencies and other organisations currently rely on geospatial systems for decision making. Consequently, regular acquisition and updating of geospatial datasets is a priority. Updating geospatial information in changing urban areas is always a challenging task. Geospatial information is usually updated by conventional methods such as surveying and manual editing of geospatial datasets, which are labour intensive, time consuming and costly (Wrzesien et al. 2003). Hence automation of such data updation procedures is of great interest in geospatial community. Conversely high resolution (Hi-Res) remote sensing data offering a detailed view of large coverage areas can be beneficial for automated procedures where information derived by converting pixel based information to meaningful thematic data. Such pixel based land cover extraction procedures are widely used in landscape analysis (Luck and Wu, 2002). However compact juxtaposition associated with urban features and limited number of available spectral information, induces scalar and spectral noise in thematic maps derived from such land-use land-cover (LULC) classification procedures. This crucial problem of urban land cover delineation is targeted in the current study by integrating complementary LiDAR data with Hi-Res WorldView-2 images for downtown Auckland, aiming to develop a simplified procedure for identifying various LULC commonly found in urban areas suitable for geo-spatial applications.

Перевод пока недоступен