Gray scale photographs are an important source of information for studying long-term ecological processes under the severe climate and anthropogenic changes of Mediterranean regions and their hot margins. Classification of Mediterranean and desert-fringe vegetation patterns from gray-scale imagery is a challenging task due to the high spatial variability of patterns representing the organization of trees, shrubs, dwarf-shrubs, herbaceous growth, and bare soil and rock patches. A technique based on generating “pseudo-dynamic’ patterns from single-date imagery was presented earlier based on mathematical morphology principles of gray-scale erosion and dilation. Here we introduce a new parameterization of this “pseudo-dynamics” called Cumulative Distribution Function (CDF). A neural network technique based on the DenseNet Architecture was then implemented for classifying three Mediterranean life-forms at 3 densities based on the CDF and 5 other image types. The results indicate that CDF combination with Red band and Texture yielded classification accuracies comparable and even slightly better than RGB imagery and thus opens the possibilities to the study of ecological changes from historical air-photographs.
