Classifying landsat thermal data to detect patterns of urban sprawl with the multilayer level set approach

Urban sprawl is a multifaceted issue concerning the expansion of auto-oriented development. For Taipei City in Taiwan, the price of real estate is increasing rapidly, such that people have been forced to move to suburban areas. This phenomenon should be monitored and observed closely. Land surface temperature (LST) can be used to reflect the population accumulation and distribution in an area. Landsat thermal data provides information about the LST of the Taipei Metropolitan Area. However, it is difficult to analyze LST because the temperature difference is unusually small. In this paper, a multilayer level set approach is introduced to segment the Landsat thermal data such that the segmented regions can be approximated by regional constants according to preselected level values. In doing so, the pattern of urban sprawl can be extracted.