Hybrid Genetic Algorithm (GA)-based neural network for multispectral image fusion

Data Fusion refers to extracting useful information from multisource data such as remotely sensed images and GIS datasets. The emergence of ANN (Artificial Neural Network)-based approaches has greatly promote the data fusion technology in remote sensing classifications. But the traditional ANN-based approach still has some drawbacks, especially in the step of weight training. Hence, this paper proposes an integrated approach, which uses GA (Genetic Algorithm) rather than the gradient descent method for training the connection weights, to address this problem. Design issues for the proposed hybrid fusion methodology are discussed, and then some concluding remarks are presented.