Light Engine Scanning Hyperspectral Image Edge Radiation Correction And Parallel Optimization Research

Hyperspectral remote sensing owns a high spectral resolution which can be utilized to give diagnostic partitions to the subtle surface objects. As a frontier of remote sensing, hyperspectral remote sensing has a promising application potential. Hyperspectral remote sensing with optical mechanical scanner inevitably brings in edge radiation distortion in the process of imaging due to its working principles and the influence of attenuation effect of atmosphere. The quality and reliability of hyperspectral images is seriously affected by the distortion. Calibration of edge radiation lays a foundation for all kinds of subsequent quantitative and semi-quantitative analyses. The high spectral resolution of hyperspectrum brings not only abundant details but also the challenge of mass data processing at the same time. Conventional approaches of serial processing can’t meet the basic requirements of performance and practical applicability and it’s significant to study parallel optimization of edge radiation calibration.This paper focuses on edge radiation calibration of hyperspectral remote sensing with optical mechanical scanner. First, the reasons and principles of edge radiation distortion are intensively studied. Four algorithms of simplified radiant transfer model, statistical law of hyperspectral image, moment matching and histogram matching are adopted to realize the calibration of edge radiation distortion. The space dimension validity of the four algorithms is verified through subjective qualitative standards and objective quantitative standards. Meanwhile, the spectrum dimension validity is verified through tests on the recovery capacity of spectral feature in marginal area and the preservation capacity in middle area using spectral angle evaluation criteria. Then, according to the characteristic of huge amount of data and the miniaturization of remote sensing platforms, parallel optimization of the four algorithms is designed and realized based on multi-core processor and GPU/CUDA. Basic laws and conclusions concerning the optimization are made through the speed-up ratio experiments and time elapse proportion experiments, which contribute to the solutions of applying high performance parallel computing system to the question of edge radiation calibration. Finally, an edge radiation calibration system of hyperspectral data is designed and realized on the basis of related theories and technologies.