Comparison of neuralclassification algorithms applied to land cover mapping
C. Thiel, Ferdinando Giacco, Fredrik Palm · 2008
Abstract. We compared the performance of several supervised classi-fication algorithms on multi-source remotely sensed images. Apart from the Multi-Layer Perceptron, K-Nearest-Neighbour and Radial Basis Func-tion network approaches, we looked more in detail at the Support Vec-tor Machine classifier, which recently showed promising results in our setting. In particular, it is able to provide meaningful answers for the analysis of mixed pixels. They correspond to areas on the ground that comprise more than one distinct class, representing a major challenge for the interpretability of the final land-cover maps. To assess their impact, we performed a rejection-based analysis, allowing classifiers to refuse an-swers on pixels they can not associate mainly with one class. The experimental results lead to the conclusion that the 1vs1 SVM ap-proach with a linear kernel (using Bradley-Terry coupling) has to be pre-ferred over all other classification algorithms examined, both in terms of accuracy as well as ease of visual interpretation. 1