Feature enhanced radar imaging via neural network computing-based sensor fusion

We address a novel neural network (NN) computing-based approach to the problem of feature enhanced (FE) fusion of remote sensing (RS) imagery acquired with different coherent radar/SAR sensing modalities. The novel proposition consists in adapting the Hopfield-type maximum entropy NN (MENN) computational framework to solving the FE image recovery inverse problems with adaptive multiple sensor data fusion that preserves salient radar/SAR image features. The FE fusion is performed via aggregating the descriptive experiment design and the theoretical informatics inspired maximum entropy regularization frameworks for iterative minimization of the energy function of the multistate MENN with adaptive adjustments of the MENN's synaptic weights and bias inputs. We also feature on the computational implementation issues of the MENN-based multi-sensor radar/SAR data fusion and verify the overall image enhancement efficiency via computer simulations with real-world RS imagery.