Research On The Etalon Effect In Dispersive Hyperspectral Imagers

Hyperspectral imaging is an important tool for remote sensing in the field of earth observation,and its sensitivity and accuracy is the key to supporting data applications.Back-illuminated CCDs will greatly improve the detection sensitivity of spaceborne visible and near-infrared(VNIR)hyperspectral imagers with their superior quantum efficiency.However,dispersive hyperspectral VNIR imagers using back-illuminated CCDs suffer from interference fringes in the near infrared band since the near infrared light will penetrate the photosensitive region and undergo multiple reflections within the CCD,which is known as the Etalon effect.This effect will cause a signal modulation which can become increasingly serious(±25% or more)when the spectral resolution gets higher enough,bringing huge difficulties to subsequent processing and quantitative applications of the hyperspectral data.For the Etalon effect in dispersive hyperspectral VNIR imagers using back-illuminated CCDs,this thesis conducted in-depth study in aspects of the physical mechanism and mathematical modeling of the Etalon effect,aiming at seeking a post-processing approach to remove the impact of the Etalon effect.Similar to the principle of a Fabry-Perot interferometer,the AR coatings,depletion region,insulation,and gate structures of the back-illuminated CCDs can be simplified as parallel planars.When a back-illuminated CCD is illuminated with coherent near-infrared light,the light will be reflected back and forth multiple times before finally being absorbed,and interference fringes occur due to the parallel-plate multiple-beam interference.A multilayer thin-films interference model to describe the fringe pattern was established by taking both the CCD internal structure parameters and the multiple-beam interference of multilayer thin-films into account.This model reveals the contribution of each layer of the CCD to interference fringes,and demonstrates the dominant role of the depletion region in fringe patterns.However,it is difficult to determine and verify the structure parameters of each pixel of the CCD.Thus,the multilayer thin-films interference model was reduced to a single-layer model that only retained the depletion region.The influence of the remainning structrue parameters is expressed by a comprehensive parameter(Fineness factor),which greatly simplifies the model and improves the practicality.According to the model,quantitative analysis of the influence of related factors(cones of illumination,thickness of each layer,fineness factor,spectral resolution,spectral binning,spectral signatures of illumination,and CCD operating temperature)on interference fringes was carried out,revealing the variation of the fringe patterns under various factors.In order to verify the mathematical model of the Etalon effect and related analysis,three VNIR grating-based hyperspectral imagers using thinned,back-illuminated CCDs were used for model validation.The experimental test device was built,and extensive testing in aspects of radiation response characteristics and spectral response characteristics was conducted.The test results show that the experimental test data was in good agreement with the model simulation results.In addition,the spectral response test results show that the Etalon effect would cause the spectral response curve of the hyperspectral imager to deviate from the ideal Gaussian response in the near-infrared bands.This results in a decrease in the spectral calibration accuracy of the near-infrared bands.Therefore,the influence of the Etalon effect on the spectral calibration of hyperspectral imagers was studied,and a spectral correction method was proposed to improve the spectral calibration accuracy.The proposed method was used to correct the spectral calibration data of a hyperspectral imager,and the results demonstrate the effectiveness of the method.The influence of the Etalon effect on the hyperspectral imaging data and the method to remove the interference fringes were studied.Considering the deficiencies of the conventional correction methods,a two-step method to remove fringes for dispersive hyperspectral VNIR imagers was proposed and evaluated.In the first step,to suppress the spectral fringes,a ridge regression based method and a spectral binning based method were proposed and analyzed.In the second step,a spectral-spatial ratio method was presented for the removal of spatial fringes,which made full use of the strong correlation between the adjacent spectral bands and neighboring spatial pixels.The proposed two-step method is highly automated and provides satisfactory results in image quality and spectral fidelity,which makes no assumptions on the instruments and should be transferable to other hyperspectral VNIR imagers.The proposed method shows flexible adaptability to the changing of fringe patterns,which does not rely on any calibration data and computes correction coefficients from the object data itself.We have demonstrated that this method can suppress the peak fringe amplitude by a factor of more than 5,and reduce the RMSE of fringe amplitudes to no more than 2%.