Mean Map Kernel Methods for Semisupervised Cloud Classification

摘要: Remote sensing image classification constitutes a challenging problem since very few labeled pixels are typically available from the analyzed scene. In such situations, data extracted other images modeling similar problems might be used to improve accuracy. However, when training and test samples follow even slightly different distributions, is difficult. This known as sample selection bias. this paper, we propose new method combine unlabeled increase reliability A semisupervised support vector machine classifier based on combination of clustering mean map kernel proposed. The reinforces in same cluster belonging class by combining similarities implicitly space. soft version also proposed where only most reliable samples, terms likelihood distribution, used. Capabilities illustrated cloud screening application using MEdium Resolution Imaging Spectrometer (MERIS) instrument onboard European Space Agency ENVISAT satellite. Cloud clear example bias features change great extent depending type, thickness, transparency, height, background. Good results obtained show that particularly well suited for situations information does not adequately describe classes data.