Experimental Study of Artificial Neural Network for Geometric Rectification of Satellite Imagery

In this paper, series of experimental studies about the neural network model for whiskbroom remote sensing imagery geometry correction methods based on BPNN (Back-Propagation Neural Networks) and RBFNN (Radial Basis Functions Neural Networks) with detailed algorithm were raised initially, which were presented on the focus of the establishment of the neural network model for geometric correction and how to improve the performance of the NN. This study shows some experimental results obtained by autonomous procedures developed by the authors based on self-calibrating Collinearity Equation Model (CEM), BPNN, GAimproved BPNN and RBFNN. Comparison among the different methodologies has been conducted taking care of the geometric accuracy from the viewpoint of structure of ANN, GCP number, the resolve of parameters, and the applicability and so on. * Corresponding author. Email addresses: [email protected](H.P.Liu)