NDVI Optimization Using Genetic Algorithm

Applying ratioing on multispectral images is one of the important techniques used in remote sensing. In this paper a formula containing multiplication, addition and a division at the end is proposed to calculate a measure for classifying land covers. Applying this formula on each pixel of the multispectral image together with a set of thresholds, one can decide if class of the pixel is vegetation, water, soil, etc. NDVI is used to determine water, vegetation and soil areas in a map. In this article Urban Landsat images are used as the experimental dataset. GA optimization is used to derive the optimized coefficients for a kind of NDVI formula. Finally, results are evaluated to see if the optimized formula is more accurate and more robust than the traditional NDVI.