Deep robust non-negative matrix factorization method based on incremental learning

The invention provides a depth robust non-negative matrix factorization method based on incremental learning, namely, depth incremental non-negative matrix factorization (l2, 1-DINMF) based on l2, 1 norm, which is used for solving the problem that the traditional NMF cannot mine deep features of an SAR (Synthetic Aperture Radar) target. The method comprises the following steps: firstly, constructing a brand new NMF model framework by using a deep learning thought for autonomously learning potential attributes and hidden information of an SAR (Synthetic Aperture Radar) target; meanwhile, an approximate solution of the model is obtained through a gradient descent method, and an increment updating algorithm rule with a higher convergence speed is deduced. Experimental results show that along with the increase of the number of SAR target training samples, the algorithm of the invention can significantly reduce the computational complexity of the SAR ATR model and ensure the stability of the updating of the SAR AT