Abstract:

SAR images vary with different aspects greatly even for the same target, which makes it difficult to identify targets from SAR images based on a single aspect. To solve the problem, an SAR image recognition system based on multipleeigenspaces of ICA and fuzzy minmax neural network ensemble is proposed. Some SAR image eigenspaces are constructed for different aspects based on independent component analysis. An independent neural network is trained for each of eigenspace of different aspects. The trained neural networks are used to recognize SAR targets, and their results are combined through D-S evidence reasoning. The simulation results show that the recognition accuracy is higher than that of using a single neural network with aspect estimation preprocessing. Moreover, it does not need aspect estimation preprocessing.