ESA       
   workshops       
envmail@esa.int

 

An Autocorrelation-based Technique for Built-up Area Detection in ALOS PALSAR Imagery

Mattia Stasolla(1) and Paolo Gamba(1)

(1) University of Pavia, via Ferrata, 1, 27100, Italy

Abstract

Global mapping is the key issue of many international projects [1],[2], with a particular focus on urbanization and infrastructure monitoring [3]. It is then clear that, due to its proportions, Remote Sensing is the only way to address this task. Unfortunately, the actual knowledge that can be derived from this irreplaceable technology is subordinate to the effectiveness of the procedures employed to extract the desired information. The most common approach is human visual interpretation, but it is very time consuming and in the last years many (semi-)automated techniques have been proposed that can substitute for experts, or at least be powerful support tools. As can be easily seen, the basic RS source is optical imagery, simpler to understand and characterized by higher spatial resolution; nevertheless, SAR sensors, especially thanks to their all-weather acquisition capabilities, can be a strong alternative to conventional imaging [3]. In particular, the new generation sensors, from ALOS PALSAR to the very recent TerraSAR-X and COSMO/SkyMed, have sensibly improved spatial resolution and therefore widened the application fields, so that they can be now employed in place of optical ones and provide comparable results. In this paper we present an automated technique exploiting local autocorrelation for the detection of built-up areas in ALOS PALSAR images. The method is based on the computation of spatial association indicators, a family of indexes [4] that provide non-overlapping information about the spatial patterns throughout the image. The algorithm starts with the retrieval of built-up area seed points (‘hotspots’), obtained as a suitable combination of Moran's Index and Geary's Index, and the extraction of a urban extent mask from Getis-Ord Index. Finally, the built-up area map is derived from the previous steps by considering the density of the hotspots within every single object of the urban extent mask. For a wide assessment of its potentialities and capabilities, the algorithm has been tested over several locations, showing an effective portability and achieving very high accuracies.

[1] http://www.epa.gov/geoss/ [2] http://www.gmes.info/ [3] http://sedac.ciesin.columbia.edu/gateway/guides/grump.html [4] F.M. Henderson, and Z.-G. Xia, “SAR applications in human settlement detection, population estimation and urban land use pattern analysis: a status report”, IEEE Trans. Geosci. Remote Sens., vol. 35, no. 1, pp. 79-85, 1997. [5] Anselin, L, "Local indicators of spatial association – LISA". Geographical Analysis, 27, 93-115, 1995

 

 

  Higher level                 Last modified: 07.05.06