Development Of Embedded Hyperspectral Database And Research On Application Of Target Classification

Hyperspectral remote sensing technology has been developed rapidly in the past30years, which has become a very important means that enable people obtain featureinformation of objects which they’re interested in. It plays a significant role in boththe civil and military fields. Hyperspectral remote sensing technology, with featuresof numerous data bands andinter-band correlation, and so on, makes a higherrequirement for data process. Due to the characteristics of integration of diagram andspectrum of its data, it has always been a basic research work and also indispensablesegment in the field of hyperspectral remote sensing image to collect and accumulateinformation expressed by spectrum response data different kinds of ground objects.However, at present most of the mature spectral database system is too large withpoor portability that limited by the environment. Therefore, it is necessary to do aresearch on creating a lightweight portable embedded database system for spectrumcurve, so that we can classify target of unknown hyperspectral remote sensing datathrough samples stored in the database system. At the same time, we combined theRX anomaly detection algorithm and support vector machine algorithm with the ARMplatform, working with embedded database cooperatively, which realized an aim ofhyperspectral remote sensing target classification and data management on theembedded platform without manual intervention. The main work is as follows:(1) This dessertation introduces the background of hyperspectral remote sensing,and expounds the research content and research purposes. On this basis, it focuses onintroducing the concept of spectral database system and summarizing its currentsituation of the development. Given to the research content, this dessertation alsointroduces some popular embedded database system nowdays.(2) Aimed at the demands of developing the database system, this dessertationelaborates the knowledge of embedded database system. Comparing to commonlyuseing embedded database system, we choose the SQLite database system, andresearch its architecture, interface functions, transactions, and the lock and SQLiteSQL statements. Finally, we transplant SQLite on the ARM development board, andtests SQLite under the circumstance of ARM.(3) With the reference of mature spectral database system home and abroad, the design idea of embedded spectral database system development was introduced in thisdessertation. We built the development environment, cross-compiles Linux GUIinterface Qt/E, and then transplanted to the ARM development board. Thedevelopment process of database was introduceed in detail from three aspects ofinterface design, data table design and function design.(4) Samples were extracted from the established database system and applidanomaly detected using RX. And spectrum angle matching and support vectormachine were applied for target identification for hyperspectral remote sensing data.Then, after analyzing and summarizing the simulation results RX anomaly detectionalgorithm and support vector machine algorithm were transplanted to the ARMembedded platform.