| The shortage of mineral resources prompts the government to involve in a largenumber of human and material resources to probe them. Obviously, the conventionalmethods of searching mineral resources have not been far more useful to confirm thesafety of mineral resources, so it is an extremely urgent need to develop a new meansof surveying minerals. The appearance of hyperspectral remote sensing technologyprovides a new path to detect mineral resources. The history of remote sensingdetection of resources in china begins in the1980s of the last century. During theshort20-year period, the remote sensing detection technology has experienced thestart from scratch to the-state-of-the-art.This thesis has deeply learnt the theory of hyperspectral remote sensing detectionand based on it, through the aspect of technology flow, to study form obtaining datafrom sensors to accomplishing the valuation of accordant results, and till to finish thewhole tech flow. The main processes which need to realize are as follows:(1) Preprocess the original hyperspectral data from sensors. The Preprocessingprocedure contains removing the bad wave range and noise wave range, scalingdata radiation, recovering bad lines, removing vertical stripes, getting rid of smileeffect, atmospheric correction and geometric correction. Then, clean spectral datacan be got through the process of preprocessing data.(2) Extract the geometric features hidden in the data. Due to the characters of largeamount of hyperspectral data, the high redundancy and low signal-to-noise ratio,directly using the tech of finding data will cause the problem of low efficiency.Extracting the geometric features hidden in the data appears to a significantmeaning. The geometric features related to the thesis include spectral absorptionindex, spectral absorption width, spectral absorption area, spectral absorptiondepth, spectral slope, trough’s locations and numbers and peak’s locations andnumbers.(3) Clean up the data affecting the accuracy of matching. Hyperspectral data can onlyreport the information of the surface but can’t report the information ofunderground vegetations and waters.while Data masking techniques can be used to cover these distubing data so that the matching accuracy can be improved. Inthis thesis, I mainly used the two techniques of Vagetation Masking Techniquesand Water Masking Techniques.(4) Research the methdologies of spectral matching. Many classic algorithms arerelated to the methdologies of spectral matching, including Spectrum AngleMatching, Spectral Correlation Coefficient Matching, Waveforms Macthing,Information Divergence Macthing and spectral matching method which has beendevelopped basing on considerable experiments and multi-character decisiontrees.(5) Realise and integrate the software of hyperspectral remote sensing resourcesdetection. In term of developing software, the aspects of operation environment,technical precesses and structure design which are ralated to two self-developpedsoftwares of hyperspectral remote sensing resources detection systems have beenintroduced.what’s more, analysis on metallogenic prospect areas has mainly beenintroduced.Although hyperspectral remote sensing resources detection technologyhas experienced a several-decade development, there are not so much successfulexamples, so pulling geological experts’ rich knowledge and experices in helpingto find minerals can efficiently increase the manching accuracy and objectivelyappraise the matching results.Apart from the the basic research of theory, the writer has worked in the KeyLaboratory of Mathematics and Geology of Sichuan as a member of the hyperspectralteam who reserched the softwares of AutomaticRecognition System for Distributionand Types of Tailings of Hyperspectral Remote Sensing as well as HyperspectralRemote Sensing of Mineral Resources Subsystems.So the writer verified the accuracyand feasibility by the above methos in practice. The thesis has used the two softwaresto present the treatment effect on data in every step and judged the effect. |
PDF Full Text Request