Weighted Decision Fusion for Supervised and Unsupervised Hyperspectral Image Classification

Many studies have been undertaken to develop and analyze the combination of results from different classifiers for a better result than using each individual classifier In general, a supervised classifier can provide better classification than an unsupervised classifier. In addition to training data limitation, a supervised classifier may result in overclassification for some homogeneous areas. An unsupervised classifier, although it may be less powerful, it can generally well classify those spectrally homogeneous areas. Thus, fusing supervised and unsupervised classification may yield better performance since the impact from trivial spectral variations may be alleviated and the subtle difference between spectrally similar pixels may not be exaggerated. Although individual classifiers are pixel-based, the final fused classification has a similar result to an object-based classifier In our previous work, we presented a majority-voting (MV)-based decision fusion for support vector machine (SVM) and Kmeans clustering results, which are typical supervised and unsupervised classifier, respectively Spatially adjacent pixels grouped by the Kmeans clustering are re-classified using the majorityvoting rule by considering the SVM classification result. In other words, all the pixels in each local segmented region are classified into the same class, which is the class that most pixels belong to using the SVM-based decision. The same idea was independently proposed by other researchers i