Research Of Fluctuated Skylight Monitoring In Visible To Near-infrared Region And Application In Airborne Hyperspectral Imaging

Airborne hyperspectral imaging is a common method of remote sensing.For aviation operation or hyperspectral imaging in a long time or large area,static radiometric models could hardly fully reflect actual situations as the external environment usually keeps changing.Compared with normal factors,the probabilistic fluctuation of skylight is the major reason contributing to the case that hyperspectral imaging suffers from irregular radiation during the flight,and thus affects the imaging results in a significant way.In order to eliminate the influence of skylight on the airborne hyperspectral imaging in a greatest extent and monitor real-time changes as well,the paper presents a set of skylight monitor established in the aircraft cabin operating in the visible to near-infrared(VNIR)region.The main purpose is to record the skylight data kept synchronization with the VNIR module of airborne hyperspectral imaging along with the positioning device.The skylight monitor comprises with a cosine corrector that receives the hemispherical skylight on the top of cabin,an inside stable optical fiber and a VNIR linear array spectrometer sampling data in a fixed frequency.Besides that,the paper builds a new model of relative radiometric calibration based on the skylight.The correction method focuses on the whole aerial hyperspectral images in the target line by using the irradiance information of skylight matched with image data per frame.After that a detailed comparative evaluation and precision analysis on the calibration result is discussed.In addition,on the basis of actual measurement of reflectance from local surface sampling points,an operation model of reflectance retrieval on the three-field of view(FOV)mosaic hyperspectral images with matching skylight data is also proposed in the paper.At last a comparative result with the inversion curve and the actual measured reflectance spectrum of evaluation reference point on the target line is conveyed along with errors calculation.The main research contents and innovation points in this paper includes the following:1)The paper designs a set of skylight monitor operating in the VNIR region.Relevant system structure and working module are elaborated with responding operating criterions and test standards.In the two flight experiments a sensible and effective platform was established according to actual flight situations,and a matching operational mode with the airborne hyperspectral imaging VNIR module was set as well.Furthermore,a relative method of synchronous radiometric calibration for the stable skylight monitor system is described in detail including the theory introduction,procedure analysis,performance and precision test,which lays a rich foundation for actual flight experiments both on the theory and practice.2)The paper puts forwards a new method of relative radiometric calibration based on the skylight monitor.The solution is to deal with the case that the airborne hyperspectral imaging operating in a long time or large area suffers from troubles as the radiation intensity of skylight may change irregularly or sharply in a short time during a continuous flight with external environment variations.Besides that,the method is also for the problem of inconsistent numerical response in the large-format aerial hyperspectral images.The whole relative radiometric calibration module comprises three main parts,which respectively are the matching hyperspectral image with skylight data,the offset vector of skylight and the calculation of the relative radiometric calibration coefficients.In general,the calibration work is on account of the truncated regional image under skylight with intense fluctuation in the target line.The skylight irradiance data is precisely matched by the means of position data and the satellite remote sensing image.All the relevant data follows the airborne hyperspectral imaging VNIR module.The band matching is based on the regular interpolation method of pchip,while the frame matching is on that of spline.3)The paper presents a new correction method of skylight based on offset reference points in the image.First of all is to select the same surface features widespread in the hyperspectral image.Along with the spectral angle matching(SAM)algorithm and histogram statistic method the skylight offset vector in each band is calculated.Then build the relationship with spectral information of all the offset reference points and the corresponding skylight data.The calibration standard of relative radiometric calibration is based on the specific wavelength(480nm)of skylight at the max value of irradiance during the whole flight line.Finally,the radiometric calibration coefficient matrix in all bands and frames can be determined by skylight data matched with each image frame.The relative radiometric calibrated image would be made at last.4)In the paper,an actual relative radiometric calibration of the airborne hyperspectral image is performed based on the first flight experiment,and the calibrated result is evaluated detailly in the following.The analysis starts from the local area,and the preliminary evaluation is carried out through local contrast images of the truncated regional image in the target line.From specified regions the amplification blocks are selected to set the evaluation reference points and correction reference points.After extracting the spectral characteristic curves of the evaluation reference points,the matching degree between the target reference points and the corresponding templates before and after the calibration is analyzed.Then follows the overall evaluation that the truncated regional image is put under binarization processing to grab the same surface features matched with evaluation reference points.By comparing the response amplitude distributions of digital numbers(DN)for matching pixels before and after radiometric calibration,with median line as benchmark the contrast shows that the method of relative radiometric calibration for hyperspectral imaging based on skylight monitor is effective.Finally,the variation coefficient of matching points in the calibrated image is calculated.Taking green grassland and bare cultivated land as the evaluation reference points,the evaluation results in all the 256 bands are respectively 5%?7% and 4%?6%,with mean values of 5.9% and 5.5%.5)In the paper,a method is proposed to invert the reflectance of the whole airborne hyperspectral image based on skylight data and actual measured reflectance information of the ground sampling points.By measuring the actual reflectance data of all ground sampling points and matching corresponding skylight data,the relationship of the radiance of each point and its DN response in the airborne hyperspectral image is calculated.All the ground sampling points are under outlier detection to avoid the influence of outliers on the DN to radiance(DN-L)coefficient matrix and as well to ensure the effectiveness and accuracy of the final fitting function.The reflectance retrieval coefficient matrix is calculated and the reflectance retrieval image is drawn later in terms of the matching skylight data with corresponding hyperspectral image.The actual related work is based on the second flight experiment and typical selected ground sampling point.Taken cabbage-leaf as the inversion reference point,the whole relative error is less than 10% in the wavelength of 0.45?m?0.98?m,and the local relative error can be lower than 4% in the wavelength of 0.47?m?0.56?m and 0.7?m?0.92?m.