One-Class Risk Estimation for One-Class Hyperspectral Image Classification

Hyperspectral imagery (HSI) one-class classification is aimed at identifying a single target class from the HSI by using only knowing positive data, which can significantly reduce requirements for annotation. However, when meets HSI, it difficult classifiers to find balance between overfitting and underfitting of data due problems distribution overlap imbalance. Although deep learning-based methods are currently mainstream overcome in multi-classificaiton, few researches focus on classification. In this paper, weakly supervised classifier, namely HOneCls proposed, where risk estimator—the xmlns:xlink="http://www.w3.org/1999/xlink">One-Class Risk Estimator —is particularly introduced make full convolutional neural network (FCN) with ability one case Extensive experiments (20 tasks total) were conducted demonstrate superiority proposed classifier.