Skip to Main Content
A novel tensor subspace learning algorithm, called tensor subclass discriminant analysis, is presented for inverse synthetic aperture radar target classification. Assuming there are multiple subclasses in each class of inverse synthetic aperture radar data, an objective function, which is based on clustering based analysis criterion and aimed at maximising the distance between subclasses of different classes, is constructed for target classification using inverse synthetic aperture radar images in the form of a tensor. Then the formula of the algorithm is deduced, and the optimal tensor projections for classification are worked out. Simulation demonstrates the effectiveness and robustness of the proposed method.