Research On Processing Software Of Supervised Point-source Spectrometer’s Mixed Pixels

With the rapid development of information technology and computer hardware,remote sensing technology has begun to enter the era of hyper spectral. Hyperspectral datawith high latitude, high resolution characteristics, brought the problem of the huge amountand easily affected by the noise interference that hard to deal with.To deal with this kindof problem, this paper proposes an unmixing software for solving non imaging spectraldata based on database, effectively improved the accuracy of spectral matching withoutlosing the spectral information and provides a convenient and efficient solution ofunmixing tool.First of all this paper analyzed the research background of the spectral unmixingsystem. Then the significance of the research status of hyperspectral unmixing solutions isanalyzed. After all we have figered out the demand of the software, and set its variousfunctional modules. This research chosed MATLAB simulation environment as theexperimental platform to select the algorithm. First, the effectiveness、the efficiency andaccuracy of the algorithm were verified. Under the condition of only one set of data in thespectral the target spectral database, we proved the feasibility of unmixing algorithms forthe solution. Then when a plurality of spectral are in the database,we found that we canuse the mean variance as the judgement of the best set of spectrum, and the variance canbe caculated by the abundance we get. It has been proved that the best set of spectrum’smean variance is much smaller than the other set, so that we can find out with set ofspectrum is corresponding to the target spectrum.Finally, to achieve all the argorithm and spectral processing steps, we finished thesoftware of this system. Some of the software’s functions and interface are demonstrated.It’s been proved that our system is convinence and stable, and the argorithm was improvedthan before.