Semi-supervised techniques based hyper-spectral image classification: A survey

Now a days hyperspectral image processing is gaining attraction to the researchers because of its ability to detect and recognize the unique land cover types with high accuracy. The challenging task in hyperspectral image processing is classification. Image classification is complex process in hyperspectral imaging because of huge amount of data, limited labeled samples. Researchers have evolved many ways of hyperspectral image classification. Based on nature of training samples used in classification, Supervised, Unsupervised and Semi-supervised are methods of classification. This review paper presents various semi-supervised techniques used for hyperspectral image classification and research challenges are identified from comparison between them.