A major problem in the present flight simulators is that simulated displays such as visual, infra red, synthetic aperture radar etc., do not correlate well with each other. A common data base to drive these displays would facilitate such correlation. A subset of this problem of creating a common data base is the following question: given a mix of input spectra of imagery, is it possible to predict using statistical prediction techniques an infrared image that is not part of the original spectra. This research work is directed towards predicting an infrared image from the best and the smallest subset of the input spectra. Simulations were performed on seven bands of LANDSAT Thematic Mapper images. Three types of predictions were developed and implemented and the results analyzed. The results proved that such predictions were indeed possible and that the statistical prediction may give as good results as those obtained from more complicated neural networks based predictions.
Statistical Prediction of an Infrared Image from Multi-Spectral Imagery for Common Visual Data Base Generation
3 Views