Spectral calibration system for a computational spectral imager and method for extracting an observation matrix

Abstract

The present invention discloses a spectral calibration system and an observation matrix extraction method for a computational spectral imager. An adjustable monochromatic light source transmits monochromatic light, which enters the computational spectral imager to be calibrated through a collimator. A light source control system controls the adjustable monochromatic light source to adjust the wavelength of the monochromatic light according to a step value, and an image acquisition system acquires the monochromatic light to obtain a set of image sequences. Based on the image sequences, the image similarity is calculated. A curve of the relationship between the image similarity and the wavelength is plotted. In the curve of the relationship between the image similarity and the wavelength, the wavelength corresponding to the peak value of the relationship curve is determined. The sequence images corresponding to the peak wavelength are extracted. According to the sequence images, the observation matrix is determined. The invention has universality for the spectral calibration system and the observation matrix extraction of the computational spectral imager, and does not need to introduce additional identifiers on the coding plate, thereby improving the space utilization rate of the coding plate.

Description

Technical Field

[0001] The present invention relates to a spectral calibration system for a computational spectral imager and a method for extracting an observation matrix, belonging to the field of spectral calibration. Background Art

[0002] Spectral calibration is used to determine the relative spectral response of a camera, and obtain performance parameters such as the central wavelength, spectral range, effective spectral bandwidth, and out-of-band response of each spectral band of the spectral imager. These performance parameters will be used as the basis for the camera radiometric calibration test and on-orbit calibration data processing and applications.

[0003] Different from traditional spectral imaging systems, the coded aperture computational spectral imaging system is a new type of snapshot spectral imaging system. It does not directly obtain the spectral bands of the scene, but completes aliased sampling through a coded mask and a dispersive prism, and then uses spectral reconstruction technology to complete spectral division, with the advantages of snapshot and high light utilization rate.

[0004] For a coded aperture computational spectral imager, the position of the primary image is no longer a slit, but a two-dimensional coded plate. The image obtained on the detector is no longer the spectral dispersion of a one-dimensional scene, but a coded two-dimensional spectral aliased image of the scene. For the calibration of a coded aperture computational spectral imager, due to the introduction of the two-dimensional coded mask, spectral calibration is no longer the determination of the central wavelength and the spectral response function, but the calibration of the coded images of the coded mask at different wavelengths, which will be used to construct an observation matrix to reconstruct the spectral map information. The extraction of the observation matrix is crucial for spectral reconstruction.

[0005] Patent CN109186762 describes a spectral calibration system and method for a compressive sensing coded super-resolution spectral imager. The pixel size of the applicable instrument is larger than the unit size of the coded plate. The feature of its calibration method is to use a special coded plate with crosshair identifiers at designated positions on the coded plate, irradiate the spectrometer to be calibrated with an adjustable monochromatic light source, and calculate the pixel position corresponding to each wavelength using the standard gray centroid method. This method requires engraving identifiers on the coded plate, reducing the space utilization rate of the coded plate, and is also not applicable to the calibration of a coded aperture computational spectral system without identifiers. Summary of the Invention

[0006] The technical problem to be solved by the present invention is: overcoming the deficiencies of the prior art, providing a spectral calibration system for a computational spectral imager and a method for extracting an observation matrix, which is used to solve the calibration problem of a coded aperture system without identifiers and is not restricted by the relationship between the pixel size of the instrument and the unit size of the coded plate.

[0007] The technical solution of the present invention is:

[0008] The present invention discloses a method for extracting a spectral calibration observation matrix of a computational spectral imager, including:

[0009] An adjustable monochromatic light source transmits monochromatic light, which enters the computational spectral imager to be calibrated through a collimator;

[0010] The light source control system controls the adjustable monochromatic light source to adjust the wavelength of the monochromatic light according to a step value, and the image acquisition system acquires a set of image sequences of the monochromatic light;

[0011] Based on the image sequences, calculate the image similarity;

[0012] Draw a curve of the corresponding relationship between the image similarity and the wavelength;

[0013] In the curve of the corresponding relationship between the image similarity and the wavelength, determine the wavelength corresponding to the peak of the relationship curve;

[0014] Extract the sequence images corresponding to the peak wavelength;

[0015] Determine the observation matrix according to the sequence images.

[0016] In the above method for extracting the spectral calibration observation matrix, the light source control system controls the adjustable monochromatic light source to gradually adjust the wavelength of the monochromatic light according to a step value, and the image acquisition system acquires a set of image sequences. The specific method is as follows:

[0017] The light source control system controls the adjustable monochromatic light source to adjust from the wavelength λ0 to λ with a step value of Δλ n , when adjusting to the wavelength λ i , the image acquisition system acquires a two-dimensional image, and obtains an image sequence I λi (x, y) composed of n + 1 two-dimensional images, and the number of image pixels is X × Y;

[0018] Among them, λ0 is the starting wavelength of the working wavelength range of the computational spectral imager to be calibrated; λ i is the intermediate wavelength of the working wavelength range of the computational spectral imager to be calibrated, and λ i = λ0 + i × Δλ (i = 1, 2,..., n); Δλ is the wavelength adjustment step value; λ n is the cut-off wavelength of the working wavelength range of the computational spectral imager to be calibrated; I λi (x, y) is the gray value of the monochromatic image of the encoding plate on the detector at the monochromatic wavelength λ i , (x, y) is the pixel position on the detector; X is the number of pixels in the row direction of the image sensor, and x is the pixel serial number in the row direction of the image sensor; Y is the number of pixels in the column direction of the image sensor, and y is the pixel serial number in the column direction of the detector.

[0019] In the above method for extracting the spectral calibration observation matrix, the specific method for calculating the image similarity is as follows:

[0020] Calculate each image and the first two-dimensional image of the image difference The formula is:

[0021]

[0022] Calculate the standard deviation of the image difference for each wavelength The formula is: The formula is:

[0023]

[0024] where is The average value, and the calculation formula is:

[0025]

[0026] Calculate the similarity between each wavelength image and the first two-dimensional image. The calculation formula is:

[0027]

[0028] where is the standard deviation of the image difference for each wavelength The standard deviation.

[0029] In the above method for extracting the spectral calibration observation matrix, in the curve of the correspondence between the image similarity and the wavelength, to determine the wavelength corresponding to the peak of the relationship curve, the specific method is as follows:

[0030] (1) Visually determine the peak position λ on the curve of the image similarity i and the wavelength λ Pj (j = 1, 2... M);

[0031] (2) Near the peak position, compare the peak position of λ Pj with the Pj±△λ corresponding to the nearby wavelength λ value. Take the wavelength corresponding to the maximum as the peak wavelength, where △λ is the theoretical spectral resolution at the wavelength λ Pj ;

[0032] (3) Traverse each peak, and use the above steps to obtain and λ i on the curve at the M peak positions and the corresponding wavelengths λ Pk (k = 1, 2... M).

[0033] In the above method for extracting the spectral calibration observation matrix, the method for extracting the sequence image corresponding to the peak is specifically as follows:

[0034] Select from the image sequence the wavelength λ i and the sequence image corresponding to the corresponding peak wavelength λ Pk (k = 1, 2... M).

[0035] In the above method for extracting the spectral calibration observation matrix, the method for determining the observation matrix is specifically as follows:

[0036] Form an observation matrix from the sequence images in ascending order according to the corresponding wavelength λ Pk .

[0037] The present invention discloses a spectral calibration system for a computational spectral imager, including: a tunable monochromatic light source, a collimator, an image acquisition system, and a light source control system;

[0038] The light source control system is connected to the tunable monochromatic light source to control the illumination intensity and wavelength of the tunable monochromatic light source;

[0039] The image acquisition system is connected to the computational spectral imager to be calibrated to perform image acquisition;

[0040] One end of the light source outlet of the tunable monochromatic light source is aligned with the incident port of the collimator;

[0041] The outlet end of the collimator is connected to the computational spectral imager to be calibrated, and the outgoing light perpendicularly enters the lens of the computational spectral imager to be calibrated, and all positions of the encoding plate of the computational imager to be calibrated are evenly illuminated.

[0042] In the above spectral calibration system, the tunable monochromatic light source consists of a broadband light source and a monochromator. The light source outlet of the broadband light source is aligned with the light inlet of the monochromator. The broadband light source provides a broadband light source covering the working range of the spectral imager, and the broadband light source forms a narrowband light source for calibration after passing through the monochromator.

[0043] In the above spectral calibration system, the working ranges of the broadband light source and the monochromator of the tunable monochromatic light source cover the working spectral band of the computational spectral imager to be calibrated.

[0044] In the above spectral calibration system, the spectral bandwidth of the monochromator is not greater than 1 / 12 to 1 / 10 of the spectral resolution of the computational spectral imager to be calibrated, and monochromatic light is continuously output at a wavelength interval not greater than 1 / 12 to 1 / 10 of the spectral resolution of the computational spectral imager to be calibrated.

[0045] The beneficial effects of the present invention compared with the prior art are as follows:

[0046] (1) The present invention relates to a spectral calibration system and an observation matrix extraction method for a general-purpose computational spectral imager, and has no special requirements for the size relationship between the detector pixels and the coding plate units of a compressed coding imaging spectrometer;

[0047] (2) The spectral calibration system and the observation matrix extraction method adopted by the present invention can solve the calibration problem of a coded aperture system without identifiers;

[0048] (3) The present invention does not require the introduction of additional identifiers on the coding plate, releases the constraint of calibration on the coding plate identifiers, and at the same time improves the space utilization rate of the coding plate and reduces the size of the coding plate. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 is a calibration system diagram of the computational spectral imager of the present invention;

[0050] Figure 2 is a schematic diagram of a coded aperture computational spectral imager of the present invention;

[0051] Figure 3 is a schematic diagram of a two-dimensional image of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0052] The following further describes the present invention in detail with reference to the drawings and specific embodiments.

[0053] The following further describes the present invention in detail with reference to the drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the present invention, rather than limiting the present invention. In addition, it should be noted that for the sake of description, only the parts related to the present invention rather than all the structures are shown in the drawings.

[0054] As Figure 1 shown, the present invention discloses a method for extracting an observation matrix for spectral calibration of a computational spectral imager, including the following steps:

[0055] Step (1), a tunable monochromatic light source 1 sends monochromatic light through a collimator 2 into a computational spectral imager to be calibrated;

[0056] Step (2), a light source control system 3 controls the tunable monochromatic light source 1 to adjust the wavelength of the monochromatic light according to a step value, and an image acquisition system 4 acquires a set of image sequences of the monochromatic light. The specific method is as follows:

[0057] The light source control system 3 controls the tunable monochromatic light source 1 to adjust from a wavelength λ0 to λ with a step value of Δλ n , when adjusting to the wavelength λ i , the image acquisition system 4 acquires a two-dimensional image, and an image sequence composed of n + 1 two-dimensional images is obtained The number of image pixels is X×Y;

[0058] Where λ0 is the starting wavelength of the working wavelength range of the spectral imager to be calibrated; in this example, it is 450 nm; λ i is the middle wavelength of the working wavelength range of the spectrometer to be determined, λ i =λ0+i×Δλ, i=1,2,...,n, in this example n=25; Δλ is the wavelength adjustment step value, in this example Δλ is 1nm; λ n is the cutoff wavelength of the working wavelength range of the spectral imager to be calibrated. In this example, λ n 700nm; is the monochromatic wavelength λ i The grayscale value of the monochrome image of the lower encoding plate on the detector, (x, y) is the pixel position on the detector; X is the number of pixels in the row direction of the image sensor, x is the pixel serial number in the row direction of the image sensor, and it is also the dispersion direction. In this example, X is 512; Y is the number of pixels in the column direction of the image sensor, y is the pixel serial number in the column direction of the detector, which is perpendicular to the dispersion direction.

[0059] Step (3): Calculate the image similarity based on the image sequence. The specific method is:

[0060] Calculate each image With the first 2D image Image difference The formula is:

[0061]

[0062] Calculate the image difference for each wavelength Standard deviation The formula is:

[0063]

[0064] in, for The average value is calculated as:

[0065]

[0066] Calculate the similarity between each wavelength image and the first two-dimensional image. The calculation formula is:

[0067]

[0068] in, Image difference for each wavelength The standard deviation of .

[0069] In this example, an image sequence consisting of 26 two-dimensional images is obtained. The 26 image sequences correspond to the images of the coding plate on the detector under 26 monochromatic wavelengths. As the wavelength increases, the image sequence shifts in sequence.

[0070] Step (4), drawing a curve of the relationship between image similarity and wavelength;

[0071] Step (5): In the image similarity and wavelength correspondence curve, determine the wavelength corresponding to the peak of the relationship curve. The specific method is:

[0072] In image similarity With wavelength λ i Visually determine the peak position λ on the curve Pj j=1,2…M;

[0073] Near the peak position, compared with the λ peak position Pj With nearby wavelength λ Pj±△λ Corresponding Value size, maximum The corresponding wavelength is the peak wavelength, where △λ is the wavelength λ Pj Theoretical spectral resolution at ;

[0074] Traverse each peak and use the above steps to obtain and λ i The peak position at point M on the curve and the corresponding wavelength λ Pk k=1,2…M.

[0075] Step (6) extracts the sequence image of the wavelength corresponding to the peak value, and the specific method is as follows:

[0076] From the image sequence Select wavelength λ i and the corresponding peak wavelength λ Pk Sequence images corresponding to k=1,2…M

[0077] Step (7): Determine the measurement matrix based on the sequence image. The specific method is:

[0078] The sequence of images According to the corresponding wavelength λ Pk The observation matrix is constructed in order from small to large.

[0079] The present invention discloses a spectrum calibration system for a computational spectrum imager, comprising: an adjustable monochromatic light source 1, a collimator 2, a light source control system 3 and an image acquisition system 4;

[0080] The light source control system 3 is connected to the adjustable monochromatic light source 1 to control the light intensity and wavelength of the adjustable monochromatic light source 1;

[0081] The image acquisition system 4 is connected to the to-be-calibrated computational spectral imager for image acquisition;

[0082] One end of the light source outlet of the tunable monochromatic light source 1 is aligned with the incident port of the collimator 2;

[0083] The outlet end of the collimator 2 is connected to the to-be-calibrated computational spectral imager, and the outgoing light perpendicularly enters the lens of the to-be-calibrated computational spectral imager, and all positions of the encoding plate of the to-be-calibrated computational imager are evenly illuminated.

[0084] The tunable monochromatic light source 1 consists of a broadband light source 5 and a monochromator 6. The light source outlet of the broadband light source 5 is aligned with the light inlet of the monochromator 6. The broadband light source 5 provides a broadband light source covering the working range of the spectral imager. The broadband light source forms a narrowband light source for calibration after passing through the monochromator 6

[0085] The working ranges of the broadband light source and the monochromator of the tunable monochromatic light source cover the working spectral band of the to-be-calibrated computational spectral imager. The spectral bandwidth of the monochromator is not greater than 1 / 12 - 1 / 10 of the spectral resolution of the to-be-calibrated computational spectral imager, and monochromatic light is continuously output at a wavelength interval not greater than 1 / 12 - 1 / 10 of the spectral resolution of the to-be-calibrated computational spectral imager.

[0086] Taking the working spectral band of 450nm - 700nm of the to-be-calibrated computational spectral imager as an example, the working ranges of the broadband light source 5 and the monochromator 6 of the tunable monochromatic light source need to cover 450nm - 700nm. The spectral bandwidth of the monochromator is not greater than 1 / 10 of the spectral resolution of the to-be-calibrated computational spectral imager, and monochromatic light is continuously output at a wavelength interval not greater than 1 / 10 of the spectral resolution of the to-be-calibrated computational spectral imager. Taking the spectral resolution of the to-be-calibrated computational spectral imager as 10nm as an example, the spectral bandwidth of the monochromator is 1nm, and monochromatic light is continuously output at a 1nm wavelength interval.

[0087] As described above, it is only the best specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any change or replacement that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

[0088] The content not detailedly described in the specification of the present invention belongs to the well-known technology of those skilled in the art.

Claims

1. A method for extracting a spectrum calibration observation matrix of a computational spectrum imager, characterized in that: include: The adjustable monochromatic light source (1) sends monochromatic light through a collimator (2) and enters a computational spectrum imager to be calibrated; The light source control system (3) controls the adjustable monochromatic light source (1) to adjust the wavelength of the monochromatic light according to a step value, and the image acquisition system (4) acquires the monochromatic light to obtain a set of image sequences; Calculate image similarity based on image sequence; Draw a curve of the relationship between image similarity and wavelength; In the image similarity and wavelength correspondence curve, determining the wavelength corresponding to the peak of the relationship curve; Extract the sequence images of the wavelength corresponding to the peak; According to the sequence of images, the observation matrix is determined.

2. The method for extracting a spectrum calibration observation matrix of a computational spectral imager according to claim 1, wherein: The light source control system (3) controls the adjustable monochromatic light source (1) to gradually adjust the wavelength of the monochromatic light according to the step value, and the image acquisition system (4) acquires a set of image sequences. The specific method is: The light source control system (3) controls the adjustable monochromatic light source (1) to adjust the wavelength from λ0 to λ with a step value Δλ n , adjusted to wavelength λ i When the image acquisition system (4) acquires two-dimensional images, an image sequence consisting of n+1 two-dimensional images is obtained. The number of image pixels is X×Y; Wherein, λ0 is the starting wavelength of the working wavelength range of the spectral imager to be calibrated; λ i is the middle wavelength of the working wavelength range of the spectrometer to be determined, λ i =λ0+i×Δλ(i=1,2,......,n); Δλ is the wavelength adjustment step value; λ n is the cutoff wavelength of the working wavelength range of the computational spectrometer to be calibrated; is the monochromatic wavelength λ i The grayscale value of the monochrome image of the lower encoding plate on the detector, (x, y) is the pixel position on the detector; X is the number of pixels in the row direction of the image sensor, x is the pixel serial number in the row direction of the image sensor; Y is the number of pixels in the column direction of the image sensor, y is the pixel serial number in the column direction of the detector.

3. The method for extracting a spectrum calibration observation matrix of a computational spectral imager according to claim 1, wherein: The specific method for calculating image similarity is as follows: Calculate each image With the first 2D image Image difference The formula is: Calculate the image difference for each wavelength Standard deviation The formula is: in, for The average value is calculated as: Calculate the similarity between each wavelength image and the first two-dimensional image. The calculation formula is: in, Image difference for each wavelength The standard deviation of .

4. The method for extracting a spectrum calibration observation matrix for a computational spectral imager according to claim 1, wherein: In the image similarity and wavelength correspondence curve, the wavelength corresponding to the peak of the relationship curve is determined by: (1) In image similarity With wavelength λ i Visually determine the peak position λ on the curve Pj (j=1,2…M); (2) Near the peak position, compare the λ peak position Pj With nearby wavelength λ Pj±△λ Corresponding Value size, maximum The corresponding wavelength is the peak wavelength, where △λ is the wavelength λ Pj Theoretical spectral resolution at ; (3) Traverse each peak and use step (2) to obtain and λ i The peak position at point M on the curve and the corresponding wavelength λ Pk (k=1,2…M).

5. The method for extracting a spectrum calibration observation matrix of a computational spectral imager according to claim 1, wherein: The specific method for extracting the sequence image of the wavelength corresponding to the peak is as follows: From the image sequence Select wavelength λ i and the corresponding peak wavelength λ Pk (k=1,2…M) corresponding sequence images 6. The method for extracting a spectrum calibration observation matrix for a computational spectral imaging instrument according to claim 1, wherein: The measurement matrix is determined by: The sequence of images (k=1,2…M) according to the corresponding wavelength λ Pk The observation matrix is constructed in order from small to large.

Images