An introduction to using generalized linear models to enhance satellite-based change detection

A popular satellite-based land cover change detection technique is to use the spectral information to set up a binary “change/no-change” mask. For each pixel, if there is a big enough difference between the reflectance values for two images acquired at different times, the area represented by that pixel is considered to have changed. The different change detection methods are different in how they determine a “big enough difference”. The analyst is left to choose which function of the reflectance values to use and where to set the “change” threshold. This choice is often subjective and effects the accuracy of the change detection. The authors explore the use of generalized linear models (GLM) as a way to enhance satellite based change detection by helping determine the most appropriate function of the reflectance values to use then apply the modeling to select the threshold. The main idea is that reflectance values from satellite imagery can be incorporated into a GLM to predict the probability of change from one land cover to another and that this method of change detection will provide more information than current change detection methods. A major benefit of using generalized linear models is to determining the significance and most useful combinations of the spectral data for predicting changed areas