New improvements in the ORASIS algorithm

We describe recent improvements to the Optical Real-time Adaptive Spectral Identification System (ORASIS). The NEMO satellite will use ORASIS as a compression algorithm and first pass analysis algorithm. One improvement concerns the quantization aspect of ORASIS. In this step, the hyperspectral data are put through a process that creates a set of spectra (called exemplars) that represent the entire data set within the specified error tolerance. Each exemplar has a high dimensional “sphere” of influence. Any spectra within that sphere are replaced by a reference to that exemplar. However, the spheres tend to overlap. Results are improved by selection of the optimum exemplar. The second improvement is a method that relates sensor noise characteristics to spectral angle. The noise parameters of the sensor can be used to calculate an error tolerance, for each spectrum, dependent upon spectral shape and intensity. This method provides a physical justification for the error tolerance