End-User-Oriented Multisource Information Fusion for Resource Management with Remote Sensing Imagery

We address a new approach to the problem of improvement of the quality of remote sensing (RS) images obtained with several imaging systems/methods as required for end-user-oriented environmental resource management. We present the elaborated end-user-oriented software that provides the necessary tools for numerical implementation/simulation of different RS image reconstruction-fusion paradigms. In this paper, we present the computational methodology and software that performs RS image enhancement/fusion using the recently developed non parametric high-resolution techniques, in particular, regularized constrained least squares method, weighted constrained least squares , robust Bayesian minimum risk , and robust maximum entropy (ME) methods. We develop a modified ME neural network (NN)-oriented technique to perform the reconstruction-fusion tasks in a computationally efficient manner. We also present a quantitative and qualitative characterization of the performance of the developed MENN reconstruction/fusion algorithms evaluated through software simulations, along with their comparison with the previously developed regularized inverse filtering and NN-based image reconstruction techniques that do not accomplish the data/method fusion. Simulation examples are reported to illustrate the good overall performances of the end-user-oriented fussed image reconstruction achieved with the elaborated software in application to the real-world 2-dimensional RS imagery.