Search

CN-121453187-B - Image real-time radiation correction method for push-broom hyperspectral camera

CN121453187BCN 121453187 BCN121453187 BCN 121453187BCN-121453187-B

Abstract

The invention discloses an image real-time radiation correction method of a push-broom hyperspectral camera, and relates to the technical field of image processing methods. The method comprises the following steps of collecting hyperspectral data of four standard reflectivity plates by using a hyperspectral camera based on a built real-time radiation correction system and defining acquisition parameters, preprocessing an image of a push-broom hyperspectral camera by using a time stamp pair Ji Gao spectral images and spectrometer information and creating a region of interest (ROI) method, and establishing a hyperspectral correction model to predict reflectivity by correcting information acquired by a spectrometer by correcting an azimuth angle of a cosine corrector. The method has the advantages that the data acquisition is accurate, the real-time radiation correction of the hyperspectral image can be realized, and the correction precision is high.

Inventors

  • CEN HAIYAN
  • Ye Angben

Assignees

  • 浙江大学

Dates

Publication Date
20260508
Application Date
20251231

Claims (9)

  1. 1. The image real-time radiation correction method of the push-broom hyperspectral camera is characterized by comprising the following steps of: S1, collecting hyperspectral data of four standard reflectivity plates by using a hyperspectral camera based on a built real-time radiation correction system and defining acquisition parameters; S2, performing image preprocessing of a push-broom hyperspectral camera by using a method of creating a region of interest (ROI) through a time stamp pair Ji Gao of spectral images and spectrometer information; s3, establishing a hyperspectral correction model to predict reflectivity through correcting and correcting information acquired by a spectrometer for correcting and correcting the azimuth angle of the cosine corrector; in the step S2, the radiation measurement deviation is corrected by a reference data optimization method, which includes the following steps: selecting hyperspectral image frames in a continuous time window with the actually measured radiance fluctuation range of less than or equal to 3% of a spectrometer, ensuring the time consistency of data, and selecting a hyperspectral image frame set in a continuous time window T less than or equal to 3 seconds : ; Wherein the method comprises the steps of Is the radiance data measured by the spectrometer at the time point, Dimension 1 XW×C for the kth frame hyperspectral image data cube; Statistical screening of DN values of all pixels in the ROI, removing abnormal values caused by sensor noise, calculating average DN value of the wave band of the rest pixels as standard response value of the wave band under the current illumination condition, and comparing each frame Selecting 100X 60 pixel rectangular ROI in the central region of the image, wherein DN value matrix is as follows Outliers are culled by 3 sigma criterion: ; Wherein the method comprises the steps of A ROI pixel matrix for the kth frame hyperspectral image, For the pixel DN value of the spatial position (i, j), band c in the kth frame, As the average value of all pixels of band c, Standard deviation of pixel DN value for band c.
  2. 2. The method for correcting the image real-time radiation of the push broom hyperspectral camera according to claim 1 is characterized in that the real-time radiation correction system comprises an unmanned aerial vehicle platform (1), a triaxial holder (2) and a computer processing module (3) are arranged on the upper side of the unmanned aerial vehicle platform (1), a spectrometer (4) is arranged on the triaxial holder (2), the spectrometer (4) is used for collecting downlink solar spectrum irradiance in real time and storing data into the computer processing module (3), and a hyperspectral camera (6) is connected to the lower portion of the unmanned aerial vehicle platform (1) through a camera holder (5).
  3. 3. The method for correcting the real-time radiation of the image of the push broom hyperspectral camera as claimed in claim 2, wherein the method comprises the following steps: The hyperspectral camera (6) adopts a push-broom imaging mode, each frame of image corresponds to one scanning line of the ground, the computer processing module (3) is connected with a timestamp of the Ji Guangpu instrument (4), the time of the hyperspectral camera (6) is UTC time, the data acquisition stage is used for carrying out experiments on the whole day to adapt to the change of illumination intensity, the real-time radiation correction system synchronously executes two tasks, namely, firstly, the hyperspectral camera (6) shoots four standard reflectivity plates according to the illumination condition adjusting exposure time, secondly, the spectrometer (4) records solar spectrum irradiance with the independently set integration time, and simultaneously records solar altitude angle, unmanned plane heading angle, solar azimuth angle and corresponding time, and acquires dark current data of the hyperspectral camera and the spectrometer before and after each flight so as to eliminate sensor noise.
  4. 4. The method for correcting the image real-time radiation of the push broom hyperspectral camera according to claim 2 is characterized by setting the flying height of an unmanned aerial vehicle platform (1) in a real-time radiation correction system to be 3 meters, ensuring that imaging of a ground standard reflectivity plate covers the whole field of view, continuously collecting radiation data of a preset number of time points in a single day, eliminating shadow interference of the reflectivity plate through movement of the unmanned aerial vehicle, obtaining reference data without shadow influence, and synchronously recording an original DN value of the hyperspectral camera at each time point, radiation brightness data measured by a spectrometer, a solar incident angle and atmospheric parameters.
  5. 5. The method for correcting real-time radiation of an image of a push broom hyperspectral camera as claimed in claim 1, wherein in S2: Single frame data acquired by a hyperspectral camera (6) is expressed as And associates the corresponding time stamp t with GPS coordinates , ) The time stamp is used for directly correlating the acquisition time of the frame with the position of the unmanned plane platform (1); mapping the time stamp to geographic space coordinates, firstly providing initial geographic coordinates of each frame through GPS ) Determining their spatial arrangement and aligning them in time-stamped order, i.e. Wherein% )= The spliced time dimension directly inherits the original time stamp sequence Each spatial row h=k corresponds to a unique timestamp Thus can pass through Inquiring time information of any pixel row, and finally stacking corrected frames into spliced image according to time sequence And outputs a time stamp mapping table ; In the flight process of the unmanned aerial vehicle platform (1), the hyperspectral camera (6) and the spectrometer (4) realize second-level synchronous acquisition through hardware trigger signals, and the time synchronization relationship is expressed as: ; Wherein the method comprises the steps of And Spec (m) represents the time stamps of the kth frame hyperspectral image and the mth spectrometer sample, respectively, synchronization error 1s。
  6. 6. The method for correcting the image real-time radiation of a push broom hyperspectral camera as claimed in claim 1, wherein in S2, dark current correction and exposure normalization processing are performed to eliminate the influence of sensor noise and different exposure conditions on radiation calibration, specifically comprising the following steps: Under the same ambient temperature, respectively acquiring a full-dark image of a hyperspectral camera (6) and a dark background signal of a spectrometer (4) for dark current correction: For the hyperspectral camera (6), closing the lens cover to collect at least 10 frames of full dark images, and calculating the average value of dark currents in each wave band And from the original image Is subtracted to obtain = ; For the spectrometer (4), the incident light path is closed to collect dark background signals under the same integration time And from the original radiance Is subtracted to obtain = ; Simultaneously, the integration time of the spectrometer (4) is synchronously recorded Exposure time of hyperspectral camera Sum gain value And carrying out exposure normalization processing on the corrected data, wherein the hyperspectral image is obtained by the following formula: normalizing; spectrometer data is calculated by the formula: normalizing; to eliminate the influence of different exposure parameters on the data.
  7. 7. The method for correcting real-time radiation of an image of a push broom hyperspectral camera as claimed in claim 1, wherein in S3: firstly, correcting a downlink solar spectrum collected by a spectrometer (4), and according to a formula The relative azimuth angle of the spectrometer is calculated, wherein, Representing the azimuth angle of the sun, An azimuth angle representing the orientation of the spectrometer; according to the relative azimuth angle of the spectrometer (4) and the solar altitude, according to the formula ( ) Correcting cosine errors of solar spectrums subjected to dark current and exposure normalization; Wherein the method comprises the steps of Representing the solar spectrum after the cosine error correction, Representing the elevation angle of the sun, ( ) Representing the cosine function obtained by fitting the solar altitude and the relative azimuth of the spectrometer.
  8. 8. The method for correcting the real-time radiation of the image of the push broom hyperspectral camera as claimed in claim 1, wherein in the step S3, the accuracy of radiation correction or reflectivity calculation is improved by a band matching method, comprising the following steps: the hyperspectral camera (6) has n discrete wavebands, the wavelength vectors of which are noted as =[ , ,......, The corresponding digital quantized value vector is ; The spectrometer (4) has m discrete wavebands, m > n, the wavelength vectors of which are noted as =[ , ,......, The corresponding DN value vector is ; In order to realize the band matching of the two, firstly, reconstructing spectrometer data in a continuous wavelength space through an interpolation algorithm to establish a continuous spectrum function : ; Linear interpolation or cubic spline interpolation methods are employed, wherein, The linear interpolation formula is: ; the cubic spline interpolation is used for constructing a smooth interpolation function through a piecewise cubic polynomial, and the cubic spline interpolation formula is as follows: ; The output result is the spectrum meter DN value corresponding to the center wavelength of each band of hyperspectral spectrum, forming n x 2 matching matrix , ( )]。
  9. 9. The method for correcting real-time radiation of an image of a push broom hyperspectral camera as claimed in claim 1, wherein in S3: For each band Performing least square linear fitting on hyperspectral DN values corresponding to the four reflectivity plates and DN values of a spectrometer to obtain a response slope N=1, 2,3,4, corresponding to different reflectivity plates, the mathematical expression is: ; where M is the number of sampling points for each band Arithmetic average is carried out on correction factors of the four reflectivity plates to obtain final spectral response coefficient : ; Will be Constructed as a diagonalized correction matrix for raw hyperspectral data Band-by-band correction is performed: = ; The calculation formula of the reflectivity is finally obtained as follows: 。

Description

Image real-time radiation correction method for push-broom hyperspectral camera Technical Field The invention relates to the technical field of image processing methods, in particular to an image real-time radiation correction method of a push-broom hyperspectral camera. Background The current unmanned aerial vehicle hyperspectral imaging system faces two defects of poor dynamic adaptability of a platform and misalignment of a radiation reference. Traditional radiometric calibration depends on satellite remote sensing data, the revisit period exceeds 16 days and is seriously disturbed by the atmosphere, and the band calibration error exceeds 12%. Although the foundation fixed spectrometer has high spectral resolution, the foundation fixed spectrometer cannot adapt to an unmanned aerial vehicle motion platform, so that a standard reflectivity plate is shielded by a body shadow (the coverage area of the actual measurement shadow exceeds 40%). In the prior art, the synchronization of the spectrometer and the camera by adopting a network time protocol still has errors of more than 3 seconds, and a real-time feedback mechanism of the sun angle is lacked, namely, the fluctuation of the radiant flux reaches 35 percent when the sun altitude angle changes by 10 degrees. In addition, CMOS sensors have a weak response in certain bands, and the dark current temperature drift effect (5-10% increase in dark current per 1 degree celsius increase in temperature) in motion further deteriorates radiation uniformity. Push-broom hyperspectral image preprocessing has the core problems of space-time dislocation accumulation and insufficient near-infrared signal-to-noise ratio. The existing method calculates the space coordinates by integrating the GPS speed of the unmanned aerial vehicle, but the speed fluctuation causes the positioning error to exceed 0.8 pixel (equivalent 4 cm ground object offset under the 100 meter altitude). In terms of noise suppression, the near infrared band is degraded to 38 db (over 60 db in the visible band) due to the silicon-based CMOS quantum efficiency of less than 30%. In the traditional static region selection method, under the condition of illumination fluctuation (irradiance change exceeds 15 watts per square meter), single frame data fluctuation reaches +/-18%, and the omission ratio exceeds 15% when the outlier filtering adopts a double standard deviation criterion, so that the radiation correction root mean square error of the low-reflectivity region exceeds 8%. The existing radiation correction model has geometrical error uncompensated and hard injury with incomplete coverage of a calibration range. The azimuth deviation of the spectrometer cosine receiver was uncorrected and when the azimuth deviation from the sun exceeded 30 degrees, the downstream spectral measurement was 18% off the true value. Existing angle compensation schemes only consider that the solar altitude ignores azimuth dynamic changes (15 degrees per hour change resulting in residual errors exceeding 7%). The band matching link adopts nearest neighbor interpolation method, so that the reconstruction error of 400 nm and 1000 nm edge bands exceeds 12%. The calibration model relies on a single 30% reflectance standard, and the reflectance inversion is systematically underestimated (15% target inversion is only 11.3%) due to the detector nonlinear response (determination coefficient lower than 0.92) in a 7% low reflectance scene. Disclosure of Invention The technical problem to be solved by the invention is how to provide the image real-time radiation correction method of the push-broom hyperspectral camera, which has the advantages of accurate data acquisition, realization of the real-time radiation correction of hyperspectral images and high correction precision. In order to solve the technical problems, the technical scheme adopted by the invention is that the image real-time radiation correction method of the push-broom hyperspectral camera comprises the following steps: S1, collecting hyperspectral data of four standard reflectivity plates by using a hyperspectral camera based on a built real-time radiation correction system and defining acquisition parameters; S2, performing image preprocessing of a push-broom hyperspectral camera by using a method of creating a region of interest (ROI) through a time stamp pair Ji Gao of spectral images and spectrometer information; S3, establishing a hyperspectral correction model to predict reflectivity through correcting and correcting information acquired by a spectrometer for correcting and correcting the azimuth angle of the cosine corrector. The real-time radiation correction system comprises an unmanned aerial vehicle platform, wherein a triaxial holder and a computer processing module are installed on the upper side of the unmanned aerial vehicle platform, a spectrometer is arranged on the triaxial holder and used for collecting downlink solar spectrum irradiance in real time