A dynamic matrix equation solution method based on NCBC-ZNN and its application on hyperspectral image multi-target detection.

For effectively solving dynamic matrix equation (DME) problems, a non-convex and bound-constrained zeroing neural network (NCBC-ZNN) model is designed to relax the convex constraint on the demand of activation function and achieve robustness against various noises. Moreover, rigorous mathematical analyses and proofs are presented under the circumstances of noise-free and noise-perturbed to theoretically investigate the global convergence and robustness of the proposed NCBC-ZNN model. In addition, simulative experiments are further provided to substantiate the superior performance of the proposed model in solving the DME problem compared with the existing state-of-the-art models. Finally, the proposed model combines the constrained energy minimization (CEM) algorithm to solve the hyperspectral image multi-target detection (HIMTD) problem. Compared with existing schemes, our method is competitive in terms of accuracy. [ABSTRACT FROM AUTHOR]