Spatial Metrics and Image Texture for Mapping Urban Land Use

The arrival of new-generation, high-spatial-resolution satellite imagery (e.g., Ikonos) has opened up new opportunities for detailed mapping and analysis of urban land use. Drawing on the traditional approach used in aerial photointerpretation, this study investigates an “object-oriented” method to classify a large urban area into detailed land-use categories. Spatial metrics and texture measures are used to describe the spatial characteristics of land-cover objects within each land-use region as derived from interpreted aerial photographs. In assessing how land-use categories vary in their spatial configuration, spatial metrics were found to provide the most important information for differentiating urban land uses. A detailed land-use map with nine categories was derived for the Santa Barbara South Coast Region area. Results from our work suggest that the region-based method exploiting spatial metrics and texture measurements is a potential new avenue to extract detailed urban land-use information from highresolution satellite imagery. Introduction Detailed information on urban land use is essential for applications related to urban management and planning (Jensen and Cowen, 1999). For decades, large-scale air photos have been employed to obtain such information by applying the principles of aerial photointerpretation. Interpretation using texture, context, and spatial configurations of urban landcover features are well documented (Bowden et al., 1975; Haack et al., 1977; McKeown, 1988). The availability of veryhigh-spatial-resolution satellite imagery offers a new avenue to obtain urban information on a very detailed level (Welch, 1982; Donnay et al., 2001; Small, 2001). Traditional human approaches followed the hierarchical relationships of the basic image interpretation elements shown in Figure 1. Tone and color are of fundamental importance and represent primary image elements. For digital data this primary feature is given by the spectral information (on a per-pixel basis) and characterizes the land-cover type of a specific surface object such as a vegetated zone or a built-up area. The spatial arrangement and configuration of the basic elements represent interpretation features of greater complexity such as size, shape and texture, or pattern and association. Higher elements of interpretation usually improve the level of detail and accuracy that can be derived from the remote sensing datasets. Their application, however, commonly requires higher level efforts in the image analysis processes in terms of more sophisticated knowledge of the human interpreter or more complex and customized digital image processing algorithms (Haack et al., 1997). Although an experienced image interpreter can utilize image elements well in visual interpretation, the expert knowledge is not easily translated to the analysis of satellite Spatial Metrics and Image Texture for Mapping Urban Land Use Martin Herold, XiaoHang Liu, and Keith C. Clarke imagery. To explore the rich dataset provided by Ikonos or QuickBird, a bridge needs to be built between the wellestablished approaches of visual interpretation and digital image processing. A key issue is to explore and evaluate the quantitative descriptors of spatial urban form to find distinct relationships between the physical spectral measurement (of radiance) and the land-use, socio-economic, demographic, and ecological characteristics of individual land-cover objects. Techniques for including spatial, textural, and contextual information in digital mapping of urban areas from remotely sensed imagery have been developed and tested in the last three decades (Gong et al., 1992; Barnsley et al., 1993). The approaches vary in terms of their image base (continuous spectral image or discrete land-cover classification), the spatial analysis domain (kernel-based or region-based), and the statistical approach used in describing the spatial and textural components. In the domain of high spatial resolution remote sensing, image analysis has to consider the specific characteristics and limitations of the data for depicting the urban environment. The incorporation of higher order image interpretation elements such as shape and size of land-cover objects requires a clear representation of these characteristics in the remotely sensed imagery. For urban areas, spatial resolutions of better than five meters are usually required for the identification of land-cover objects such as buildings (Welch, 1982; Jensen and Cowen, 1999). The hierarchy of urban image objects (Haack et al., 1997; Zhan et al., 2002) represents the urban landscape as “landcover objects” and “land-use regions” (or objects). Land-use regions are comprised of land-cover objects, e.g., a block (land-use object) consisting of several buildings and vegetated areas (land-cover objects). In contrast to natural environments, man-made structures have been identified as one of the few examples of objects within a landscape that have distinct and crisp boundaries (Couclelis, 1992). This characteristic makes the general approach particularly suitable in urban analysis. Recent developments in “object-oriented” image classification (based on image segmentation) have taken advantage of the detailed spatial characteristics of high-resolution datasets. The research in this area has emphasized the reduction of spectral variability within the objects and the incorporation of additional information from spatial and contextual image/ object characteristics (Johnsson, 1994; Blaschke and Strobl, 2001). Land-use regions or objects follow the concept of “analytical areas” and “photomorphic areas or units” (Peplies, 1974; Haack et al., 1997). This concept of regionalization was developed and is commonly used for aerial photographic interpretation and mapping. Land-use regions are defined as PHOTOGRAMMETR IC ENGINEER ING & REMOTE SENS ING September 2003 991 Department of Geography, University of California Santa Barbara, Santa Barbara, CA 93106 ([email protected]). Photogrammetric Engineering & Remote Sensing Vol. 69, No. 9, September 2003, pp. 991–1001. 0099-1112/03/6909–991$3.00/0 © 2003 American Society for Photogrammetry and Remote Sensing 03-915.qxd 8/7/03 5:29 PM Page 991