Spectral analysis for automated exploration and sample acquisition

摘要: Future space exploration missions will rely heavily on the use of complex instrument data for determining geologic, chemical, and elemental character planetary surfaces. One important is imaging spectrometer, which collects complete images in multiple discrete wavelengths visible infrared regions spectrum. Extensive computational effort required to extract information from such high-dimensional data. A hierarchical classification scheme allows multispectral be analyzed purposes mineral while limiting overall requirements. The classifier exploits tunability a new type spectrometer based an acousto-optic tunable filter. This image each wavelength passband without spatial scanning. It may programmed scan through range or collect only specific bands analysis. Spectral activities employ artificial neural networks, trained recognize number classes. Analysis networks has proven useful subsets spectral should employed at step classifier. network classifiers are capable recognizing all types were included training set. In addition, major components many mixtures can also recognized. capability prove system designed evaluate strange environment where details composition not known advance.