Interferometric hyperspectral stripe removal method and apparatus

Abstract

The embodiment of the present application discloses a kind of interference hyperspectral stripe removal method and device.The method comprises: according to the data characteristics of interferometric spectrometer, the original interference data cube of each scene image is arranged in sequence according to row direction;MNF transformation is carried out to the original interference data after arrangement, and transformed interference data cube is obtained;From the transformed interference data cube, interference data cube without ghost image is screened out;Spectral restoration processing is carried out to the interference data cube, and spectral image is obtained.The embodiment of the present application removes the stripe noise caused by parasitic ghost image by combining minimum noise separation, well meets the spectral quality demand of interferometric imager data processing, improves the signal-to-noise ratio and spectral accuracy of satellite interferometric imager data processing.

Description

Technical Field

[0001] This invention relates to the field of aerospace technology, and in particular to a method and apparatus for removing interference hyperspectral fringes. Background Technology

[0002] The hyperspectral imager carried on the Environment and Disaster Reduction Satellite-2 is a spatiotemporally modulated interferometric imaging spectrometer. This spectrometer generates a three-dimensional image containing both spatial and interferometric information by pushing and sweeping the entire field of view. Fourier transforming the interferogram yields the spectral distribution of the light source target, and after processing by the data image system, a hyperspectral image can be synthesized. Spatiotemporally modulated interferometric spectrometers possess the advantages of potentially high throughput, multiple channels, and high spectral resolution, demonstrating significant development potential and application prospects.

[0003] The interferometric spectrometer aboard the Environment and Disaster Reduction Satellite-2 faces inherent challenges and imperfections during imaging due to limitations in current devices and manufacturing processes. Among these, suppressing and processing parasitic ghost images is a crucial aspect of data processing. Parasitic ghost images are caused by insufficient light transmission enhancement at the interface between the detector's silicon surface and air. Light reflected from the detector passes through a Fourier mirror and re-enters the interferometer, where it is then dispersed before passing back through the Fourier mirror and entering the detector a second time. This results in new interferometric data superimposed on the true interferometric data. The presence of parasitic ghost images negatively impacts the signal-to-noise ratio and spectral accuracy of the hyperspectral data. Summary of the Invention

[0004] The technical problem solved by this invention is to overcome the shortcomings of the prior art and provide a method and apparatus for removing interference hyperspectral fringes.

[0005] The technical solution of this invention is:

[0006] In a first aspect, embodiments of the present invention provide a method for removing interference hyperspectral fringes, the method comprising:

[0007] Based on the characteristics of interferometric spectrometer data, the original interferometric data cubes for each image are arranged in row order;

[0008] The original interferometric data after arrangement is subjected to MNF transformation to obtain the transformed interferometric data cube;

[0009] Filter out the interferometric data cubes without ghost images from the transformed interferometric data cubes;

[0010] The interferometric data cube is subjected to spectral restoration processing to obtain a spectral image.

[0011] Optionally, after arranging the original interferometric data cubes of each image in row-oriented order according to the characteristics of the interferometric spectrometer data, the method further includes:

[0012] The original interferometric data after arrangement are subjected to relative radiometric correction.

[0013] Optionally, after performing MNF transformation on the arranged original interferometric data to obtain the transformed interferometric data cube, the method further includes:

[0014] The transformed interferometric data cube is subjected to noise reduction and signal separation processing to obtain the processed interferometric data cube.

[0015] Optionally, the step of filtering out interferometric data cubes without ghost images from the transformed interferometric data cubes includes:

[0016] Perform an inverse MNF transform on the processed interferometric data cube to obtain an inverse transformed interferometric data cube;

[0017] The inverse transform interferometric data cubes are sorted according to the signal-to-noise ratio of the bands to obtain sorted inverse transform interferometric data cubes.

[0018] Select the top N interferometric data cubes without ghost images from the sorted inverse transform interferometric data cubes, where N is a positive integer.

[0019] Optionally, the third dimension of the original interferometric data cube is the interferometric dimension, and each band of the original interferometric data cube is a quick view of the optical path difference at the corresponding position.

[0020] Secondly, embodiments of this application provide an interference hyperspectral fringe removal device, the device comprising:

[0021] The interferometric data arrangement module is used to arrange the original interferometric data cubes of each image in row order according to the characteristics of the interferometric spectrometer data.

[0022] The interferometric data transformation module is used to perform MNF transformation on the arranged original interferometric data to obtain the transformed interferometric data cube;

[0023] An interferometric data filtering module is used to filter out interferometric data cubes without ghost images from the transformed interferometric data cubes;

[0024] The spectral image acquisition module is used to perform spectral restoration processing on the interference data cube to obtain a spectral image.

[0025] Optionally, the device further includes:

[0026] The interferometric data correction module is used to perform relative radiometric correction processing on the original interferometric data after it has been arranged.

[0027] Optionally, the device further includes:

[0028] The interference data acquisition module is used to perform noise reduction and signal separation processing on the transformed interference data cube to obtain the processed interference data cube.

[0029] Optionally, the interference data filtering module includes:

[0030] The inverse transform interferometric data acquisition unit is used to perform an inverse MNF transform on the processed interferometric data cube to obtain an inverse transform interferometric data cube.

[0031] The inverse transform interferometric data sorting unit is used to sort the inverse transform interferometric data cube according to the signal-to-noise ratio of the band, so as to obtain the sorted inverse transform interferometric data cube.

[0032] The interferometric data cube filtering unit is used to filter out the top N interferometric data cubes without ghost images from the sorted inverse transformation interferometric data cubes, where N is a positive integer.

[0033] Optionally, the third dimension of the original interferometric data cube is the interferometric dimension, and each band of the original interferometric data cube is a quick view of the optical path difference at the corresponding position.

[0034] The advantages of this invention compared to existing technologies are as follows: This invention denoises the interferometric dimensions of interferometric imaging data and separates signal and noise through Mini-Nutrition (MNF) transformation. The signal residue caused by parasitic ghost images remains after the principal components. By performing an inverse MNF transformation on the first five principal components, the parasitic ghost images are eliminated. This method combines minimum noise separation to remove the fringe noise caused by parasitic ghost images, effectively meeting the spectral quality requirements of interferometric imager data processing and improving the signal-to-noise ratio and spectral accuracy of satellite interferometric imager data processing. Furthermore, the successful removal of fringe noise caused by parasitic ghost images from the spectral data of the Environment and Disaster Reduction Satellite-2 hyperspectral imager is of great significance for improving the spectral accuracy and image signal-to-noise ratio of the restored interferometric imaging spectrometer. Attached Figure Description

[0035] Figure 1 A flowchart illustrating the steps of an interference hyperspectral fringe removal method provided in an embodiment of the present invention;

[0036] Figure 2 A schematic diagram illustrating a process for removing interference hyperspectral fringes according to an embodiment of the present invention;

[0037] Figure 3 This is a schematic diagram of an interference hyperspectral stripe removal device provided in an embodiment of the present invention. Detailed Implementation

[0038] During the spectral reconstruction of interferometric images using an interferometric imaging spectrometer, parasitic ghost images are generated, with fringes superimposed on the interferometric image, resulting in a wavy appearance across the entire optical field of view in the reconstructed image. This ghost image noise can be removed by performing a Multi-Negative-Negative-Negative (MNF) transform on the interferometric data. First, a forward transform is performed to identify which spectral bands contain relevant images. Then, an inverse MNF transform is performed, using spectral subsets or methods to smooth the noise before the inverse transform to eliminate the noise. Finally, the image obtained after spectral reconstruction is free of the noise caused by the ghost images.

[0039] The technical solutions of the embodiments of the present invention will be described in detail below with reference to specific examples.

[0040] Example 1

[0041] Reference Figure 1 The flowchart illustrates the steps of an interference hyperspectral fringe removal method provided by an embodiment of the present invention, as follows: Figure 1 As shown, the method may include the following steps:

[0042] Step 101: Based on the characteristics of the interferometric spectrometer data, arrange the original interferometric data cubes of each image in row order.

[0043] In this embodiment, during the spectral reconstruction of the interferometric image by the interferometric imaging spectrometer, parasitic ghost images are generated and superimposed on the interferometric image.

[0044] After acquiring the interferometric images, the raw interferometric data cubes for each image can be arranged in row order according to the characteristics of the interferometric spectrometer data. In specific implementation, the LAMIS can be transposed into a (space * pushbroom * interferometric) image based on the characteristics of the interferometric spectrometer data. The specific operation is as follows: The raw interferometric data LAMIS for each image is arranged in row order. The third dimension of this interferometric data cube is the interferometric dimension, where each band is a quick view of the optical path difference at the corresponding position. Taking the visible and near-infrared spectrum of the HAVE2 satellite as an example, the V dimension of the transposed interferometric data cube is 1990 (rows) * 2048 (detectors) * 256 (interferometric dimensions).

[0045] Simultaneously, relative radiometric correction can be performed on the original interferometric data after arrangement. Radiometric correction refers to the correction of systematic and random radiometric distortions or aberrations caused by external factors and data acquisition and transmission systems, eliminating or correcting image distortions caused by radiometric errors. The main factors contributing to radiometric errors (i.e., grayscale distortion) in remote sensing images include: ① atmospheric scattering and absorption of electromagnetic radiation; ② changes in solar altitude and sensor observation angle; ③ changes in radiation intensity caused by terrain undulations; ④ differences in sensor detection system performance, such as differences in sensitivity, spectral response, and light transmission performance of optical systems or different detectors; ⑤ image processing, such as photographic processing. Image grayscale distortion is related to the image's spatial frequency. The higher the spatial frequency, i.e., the smaller the target, the greater the radiometric error. Radiometric correction is actually a component of image restoration (or reconstruction).

[0046] After arranging the raw interferometric data cubes of each image in row order according to the characteristics of the interferometric spectrometer data, step 102 is executed.

[0047] Step 102: Perform MNF transformation on the original arranged interferometric data to obtain the transformed interferometric data cube.

[0048] After arranging the original interferometric data cubes of each image in row order according to the characteristics of the interferometric spectrometer data, the arranged original interferometric data can be subjected to MNF transformation to obtain the transformed interferometric data cubes.

[0049] Minimum Noise Fraction Rotation (MNF) can be used to determine the inherent dimensionality (i.e., number of bands) of image data, separate noise from the data, and reduce the computational requirements in subsequent processing.

[0050] MNF is essentially a two-tiered principal component transformation. The first transformation (based on the estimated noise covariance matrix) separates and readjusts the noise in the data, resulting in transformed noisy data with minimal variance and no inter-band correlation. The second step is a standard principal component transformation on the noise-whitened data. For further spectral processing, the intrinsic dimension of the data is determined by examining the final eigenvalues ​​and related images. The data space can be divided into two parts: one part correlated with larger eigenvalues ​​and their corresponding eigenimages, and the other part correlated with approximately the same eigenvalues ​​and noise-dominated images.

[0051] After performing MNF transformation on the original arranged interferometric data to obtain the transformed interferometric data cube, step 103 is executed.

[0052] Step 103: Select the interferometric data cubes without ghost images from the transformed interferometric data cubes.

[0053] After performing a Multi-Nutrition Fault Transform (MNF) on the original arranged interferometric data to obtain transformed interferometric data cubes, interferometric data cubes without ghost images can be selected from them. In specific implementations, after obtaining the transformed interferometric data cubes, denoising and signal separation processes are performed on them to obtain processed interferometric data cubes. Then, an inverse MNF transform is performed on the processed interferometric data cubes to obtain inverse transformed interferometric data cubes. The inverse transformed interferometric data cubes are then sorted according to the signal-to-noise ratio of the bands to obtain sorted inverse transformed interferometric data cubes. The first N interferometric data cubes without ghost images, where N is a positive integer, are selected from the sorted inverse transformed interferometric data cubes. In this example, N = 5.

[0054] After filtering out the interferometric data cubes without ghost images from the transformed interferometric data cubes, step 104 is performed.

[0055] Step 104: Perform spectral restoration processing on the interferometric data cube to obtain a spectral image.

[0056] After filtering out interferometric data cubes without ghost images from the transformed interferometric data cubes, spectral restoration processing can be performed on the interferometric data cubes to obtain spectral images.

[0057] In the above process, the interference dimension of the interferometric imaging data is denoised. Signal-noise separation is achieved through Minimal Noise Reduction (MNF) transformation. The signal residue caused by parasitic ghost images remains after the principal components. The parasitic ghost images are eliminated by performing an inverse MNF transformation on the first five principal components. This method combines minimum noise separation to remove the fringe noise caused by parasitic ghost images, effectively meeting the spectral quality requirements of interferometric imager data processing.

[0058] The above implementation process can be combined with Figure 2 The following description is provided.

[0059] Reference Figure 2 The diagram illustrates a process for removing interference hyperspectral fringes according to an embodiment of this application. Figure 2As shown, after obtaining the original LAMIS interferogram, relative radiometric correction and LAMIS transpose can be performed on it to obtain a (spatial * pushbroom * interferometry) image. Then, MNF transformation and inverse MNF transformation are performed, and the parasitic ghost image LAMSIS is selected. Finally, the parasitic ghost image-removed LAMSIS is subjected to spectral restoration processing to obtain fringe-free spectral data. This process incorporates minimum noise separation to remove the fringe noise caused by parasitic ghost images, effectively meeting the spectral quality requirements of interferometric imager data processing.

[0060] Example 2

[0061] Reference Figure 3 The diagram shows a schematic representation of an interference hyperspectral fringe removal device provided in an embodiment of the present invention. Figure 3 As shown, the device may include the following modules:

[0062] Interference data arrangement module 310 is used to arrange the original interference data cubes of each image in row order according to the characteristics of the interferometric spectrometer data.

[0063] The interferometric data transformation module 320 is used to perform MNF transformation on the arranged original interferometric data to obtain the transformed interferometric data cube;

[0064] Interference data filtering module 330 is used to filter out interference data cubes without ghost images from the transformed interference data cubes;

[0065] The spectral image acquisition module 340 is used to perform spectral restoration processing on the interference data cube to obtain a spectral image.

[0066] Optionally, the device further includes:

[0067] The interferometric data correction module is used to perform relative radiometric correction processing on the original interferometric data after it has been arranged.

[0068] Optionally, the device further includes:

[0069] The interference data acquisition module is used to perform noise reduction and signal separation processing on the transformed interference data cube to obtain the processed interference data cube.

[0070] Optionally, the interference data filtering module includes:

[0071] The inverse transform interferometric data acquisition unit is used to perform an inverse MNF transform on the processed interferometric data cube to obtain an inverse transform interferometric data cube.

[0072] The inverse transform interferometric data sorting unit is used to sort the inverse transform interferometric data cube according to the signal-to-noise ratio of the band, so as to obtain the sorted inverse transform interferometric data cube.

[0073] The interferometric data cube filtering unit is used to filter out the top N interferometric data cubes without ghost images from the sorted inverse transformation interferometric data cubes, where N is a positive integer.

[0074] Optionally, the third dimension of the original interferometric data cube is the interferometric dimension, and each band of the original interferometric data cube is a quick view of the optical path difference at the corresponding position.

[0075] The specific embodiments described in this application are intended to enable those skilled in the art to gain a more comprehensive understanding of this application, but do not limit this application in any way. Therefore, those skilled in the art should understand that modifications or equivalent substitutions can still be made to this application; and all technical solutions and improvements that do not depart from the spirit and technical essence of this application should be covered within the scope of protection of this patent application.

[0076] The contents not described in detail in this specification are common knowledge to those skilled in the art.

Claims

1. An interferometric hyperspectral striping removal method, characterized in that, The method comprises: According to the characteristics of the interferometric spectrometer data, the original interference data cube of each scene image is arranged in sequence in the row direction; The MNF transformation is performed on the arranged original interference data to obtain a transformed interference data cube; The processed interference data cube is obtained by performing denoising and signal separation processing on the transformed interference data cube; The inverse MNF transformation is performed on the processed interference data cube to obtain an inverse transformed interference data cube; The inverse transformed interference data cube is sorted according to the signal-to-noise ratio of the wave band to obtain a sorted inverse transformed interference data cube; The first N interference data cubes without ghost images are selected from the sorted inverse transformed interference data cube, and N is a positive integer; The spectral image is obtained by performing spectral restoration processing on the interference data cube.

2. The method of claim 1, wherein, After the original interference data cube of each scene image is arranged in sequence in the row direction according to the characteristics of the interferometric spectrometer data, the arranged original interference data is further subjected to relative radiation correction processing.

3. The method of claim 1, wherein, The third dimension of the original interference data cube is the interference dimension, and each wave band of the original interference data cube is a fast view of the corresponding position optical path difference.

4. An apparatus for interferometric hyperspectral stripe removal, characterized in that, The device comprises: An interference data arrangement module is configured to arrange the original interference data cube of each scene image in sequence in the row direction according to the characteristics of the interferometric spectrometer data; An interference data transformation module is configured to perform MNF transformation on the arranged original interference data to obtain a transformed interference data cube; A processed interference data acquisition module is configured to obtain a processed interference data cube by performing denoising and signal separation processing on the transformed interference data cube; An interference data screening module is configured to screen interference data cubes without ghost images from the transformed interference data cube; A spectral image acquisition module is configured to obtain a spectral image by performing spectral restoration processing on the interference data cube; The interference data screening module specifically comprises: An inverse transformed interference data acquisition unit is configured to perform MNF inverse transformation on the processed interference data cube to obtain an inverse transformed interference data cube; An inverse transformed interference data sorting unit is configured to sort the inverse transformed interference data cube according to the signal-to-noise ratio of the wave band to obtain a sorted inverse transformed interference data cube; An interference data cube screening unit is configured to screen the first N interference data cubes without ghost images from the sorted inverse transformed interference data cube, and N is a positive integer.

5. The apparatus of claim 4, wherein, The device further comprises: An interference data correction module is configured to perform relative radiation correction processing on the arranged original interference data.

6. The apparatus of claim 4, wherein, The third dimension of the original interference data cube is the interference dimension, and each wave band of the original interference data cube is a fast view of the corresponding position optical path difference.

Images