Automatic Classification of Central Italy Land Cover: Comparative Analysis of Algorithms

The specific objective of this paper was to provide a comparative analysis of three automatic classification algorithms: Quinlan’s C4.5 and two robust probabilistic classifiers like Support Vector Machine (SVM) and AdaBoost (a short for “adaptive boosting”). This work is part of a wider project whose general objective is to develop a methodology for the automatic classification, based on CORINE land-cover (CLC) classes, of high resolution multispectral IKONOS images. The dataset used for the comparison is an area of approximately 150 km comprising both urban and rural environments. Input data are basically constituted by multispectral (red, green, blue and infrared bands), 4m ground-resolution images. In some classifications they are integrated by the NormalizedDifference-Vegetation-Index (NVDI), derived from the red and infrared bands, a Digital Terrain Model (DTM) of the area and pixel by pixel gradient values, derived by the DTM. All the above algorithms had to perform full data classification into four classes: vegetation, water bodies, bare soil, and artificial cover. The output is constituted by an image with each pixel assigned to one of the above classes or, with the exception of C4.5, let unclassified (somehow a better solution than a classification error). In addition, a confusion matrix for control data is produced to evaluate the accuracy of each algorithm, by computing the percentage of correctly classified pixels with respect to the total number of pixels, the user’s and producer’s accuracy indexes and the Cohen’s coefficient to evaluate global accuracy. Even if an optimal distribution of the samples in the training set has a great influence, results demonstrate the suitability of supervised classifiers for high resolution land cover classification. In particular, all the proposed approaches work fine, so that we are now exploring the use of more classes, that is at the second level of the CORINE legend. * Corresponding author.