Segmentation of urban built up area using Fisher discriminant analysis, intensity and vegetation strength of Landsat images

A series of studies in urban growth involve the segmentation of urban built up area. This segmentation may used as input to geospatial models to quantify and map urban growth. The segmentation may be used as well to study the dynamics of urban growth on a certain time span. We present in this work a method to segment urban built up area from Landsat multi-spectral images. Our method uses the expansion of the image in terms of three variables: intensity, vegetation strength and fisher discriminant analysis. These variables are input into a c-means fuzzy classier. From this classifier, a binary image, named the bitmap, is derived. The bitmap is a logic state image where the state ON is related to city pixels, and the state OFF to non-city pixels. A case of study is presented that comprised one of the largest cities of the world with a high dynamic growth and high urban built up density: México City. A set of images that cover a time span of eleven years is considered. The field data consist of a detailed map and field work. Key-Words: Segmentation, urban built-up, urban morphology