Discrimination of vegetation from the background in high resolution colour remote sensed imagery

Abstract

Different transformations of RGB colour space were compared to develop the best method for discrimination of vegetation from the background in open pure cork oak stands in southern Portugal in high-resolution colour imagery. Normalised difference index, i1i2i3 colour space and other indices developed for classic band imagery were recalculated for near infrared imagery and tested. A new method for fully automated thresholding was developed and tested. The newly developed index shows the equal accuracy performance but provides the smallest overestimation error and retains the largest scale of grey levels for a subsequent shape analysis.