Research On The Hyperspectral Data Processing And Lithology Identification Of The Sichuan-Tibet Railway

Sichuan-Tibet Railway belongs to major projects launched in 2016 of 13 th Five-Year Plan, which passes through Chengdu,Pujiang, Ya’an, Kangding, Litang, Zuo Gong, Bomi, Linzhi and arrives in Lhasa, provincial capital of Tibet.The relief amplitudeof the railway line is larger, and the elevation range is more than fourteen thousand meters.The western region of China is not easy to reach by the conventional transport mode so that it is quite difficult to progress in conventional geological survey research, especially in the cold plateau area that has a bad geographical natural environment and unique alpine Canyon landscape.Therefore, it is epoch-making significance to search for the accurate and convenient detection and recognition technology aimed at western Cold Plateau. Some advantages can be obtained from hyperspectral remote sensing technology, such as fast speed, wide range, unity of image and spectrum, unique spectral feature recognition capability. On the one hand, characterization information of geologic and geological structure can be reflected. Meanwhile, the surface of the earth and even deeper levels of the region can be detected. A large amount of information is contained in the hyperion hyperspectral data of EO-1 satellite, which can be applied in many aspects such as geology, ecology, hydrology, etc. Lithologic mappingachieved by hyperspectral remote sensing technologycan improveefficiency of the investigation and research on the field geological work of the government department and the production unit, which provide favorable conditions for design and production.In this essay, hyperion hyperspectral remote sensing data processing and studying on lithology recognition of the cold plateau of the Sichuan-Tibet Railway are focused on. The Sichuan Tibet Railway in Tibetan Qiang Autonomous Prefecture of Ngawa and Lhasa Province as the research area, Hyperion hyperspectral image data is used to completepreprocessing of image data, data dimension reduction, endmember determination.In the data dimensionreduction process, the kernel method is introduced to construct the kernel principal component analysis method(KPCA) to reducedata dimension and be used for endmember extraction.Finally, lithology mapping and accuracy evaluation of study area are completed. The main achievements and innovations of this paper are as follows:(1) Absolute radiance image of Jinchuan County and Dagze Countyare acquired by Hyperion hyperspectral image acquisition, band rejection, radiometric calibration, bad line repair, striped removal, geometric correction, image cutting processing.(2) Field rock spectral data is obtained to construct lithology spectra database of Jinchuan County and Dagze County by designing spectra database.On the basis of spectral data processing, the lithologycharacteristics of Jinchuan County and Dagze County are identified. In the aspect of waveform characteristic, the waveform characteristics of the fresh surface and weathering surfaceof the same rockare basically the same. Because thesame type of rock material composition is basically the same, the characteristic of spectrum curve is basically the same. In the aspect of reflectivity, the reflectivity of weathered surface is lower than that of the fresh surface.(3) Compared with the traditional MNF and PCA dimension reduction method,KPCA method is more effective to reduce the dimension of hyperspectral data. KPCA method is introduced to the process of Hyperspectral Feature Extraction, which can quickly reduce the dimension and better preserve the nonlinear characteristics of hyperspectral data.(4) Made of Dagze County and Jinchuan County single lithology identification chart. KPCA method was used to reduce the dimension of the data to get endmember and then classify the hyperspectral image data. Respectively gotten Dagze County, Jinchuan County lithology distribution. The classification accuracy were 83.3371% and 75.2853%, the Kappa coefficients were 0.8224 and 0.7281. The overall classification results is good. Usually, the classification accuracy can be higher when the Dagze County little vegetation cover rock exposed obviously.(5) Geological lithology mapping method and technology process aimed at the Sichuan Tibet Railway is completed preliminarily, whichprovides convenience for the Sichuan Tibet railway line selection and investigation.