A hyperspectral image compression method based on joint spatial-spectral compressive sensing

Abstract

The application discloses a kind of space spectral domain joint compression sensing method for infrared spectrum image, belong to high resolution spectral imaging and compression imaging field.The method for realizing the application is: generating space spectral domain joint compression imaging encoding matrix, this matrix is used as encoding template and is loaded to DMD, so that it is modulated to target light information in space, using diffraction grating to modulate spectral image after spatial modulation, using low-resolution infrared camera to modulate the image after space and spectrum downsampling, obtain low-resolution two-dimensional aliasing image.Through space compression imaging and spectral compression imaging, the limitation of infrared camera spatial resolution and spectral resolution is broken through, and the low spatial low spectral resolution infrared image collected can restore the original high spatial high spectral resolution infrared image using reconstruction algorithm.The application utilizes the spatial, spectral correlation of spectral image, and compresses data before data acquisition by compression imaging method, which significantly reduces the data amount of acquisition, storage and transmission.

Description

TECHNICAL FIELD

[0001] The present application relates to a compressive sensing method for infrared spectral image reconstruction, in particular to a space-spectrum domain joint compressive sensing image reconstruction method for infrared spectral image, and belongs to the field of high-resolution spectral imaging and compressive imaging. BACKGROUND

[0002] Spectral imaging is known for its good information acquisition capability and good spectral resolution, and has developed rapidly in recent years. However, due to the limitation of the space-spectral resolution of the image acquisition device, the current development of the spectral imager is difficult to meet the pursuit of higher space-spectral resolution of the spectral imager. In addition, the spectral information collected from the target scene can be regarded as a three-dimensional data cube, and the transmission channel is difficult to bear such a huge amount of data, which easily leads to slow data transmission, high packet loss rate and other problems, in addition, the high data acquisition amount brings great pressure to the back-end storage device. Infrared spectral imaging is an important branch of spectral imaging. However, due to the longer wavelength of infrared light and the lower energy of infrared radiation, the size of a single pixel of an infrared detector needs to be designed larger than that of a visible light detector, and under the condition of the same chip size, the number of pixels is less. In addition, due to the limitation of the bandwidth of the spectrometer, the spatial resolution and the spectral resolution are always restricted by each other, and it is difficult to obtain an image with both high spatial resolution and high spectral resolution. Therefore, how to simultaneously obtain an infrared spectral image with high spatial resolution and high spectral resolution has become a difficult problem to be solved.

[0003] In recent years, compressive sensing technology has attracted widespread attention from researchers due to its excellent performance in the field of signal processing. The biggest advantage of this theory is that it can eliminate most of the redundant information and only keep a small amount of measurement values, thereby reducing the pressure on subsequent data calculation. In the field of two-dimensional imaging, compressive sensing technology is mainly applied to space domain compression imaging based on transform domain sparsity or gradient sparsity. Space domain compression imaging can expand the spatial resolution of the camera with the help of high-resolution digital micromirror device (DMD). For three-dimensional spectral images, coded aperture snapshot spectral imaging (CASSI) is a typical example of applying the principle of compressive sensing to spectral reconstruction practical application. The CASSI system maps the target three-dimensional information to a two-dimensional detector array through DMD and grating, and the obtained information is spatial and spectral mixed information, which is reconstructed into complete information through a reconstruction algorithm. Cai et al. proposed a deep unfolding network DAUHST, and the team designed a Half-Shuffle Transformer that can capture local and global dependencies simultaneously. This mechanism is helpful for maintaining the reconstruction effect while reducing the amount of calculation.

[0004] Although the current research on spectral image compressive sensing is very hot, these researches are basically limited in the visible light band, and are only simple spatial domain or spectral domain compressive sensing, and there are few researches on infrared band and compressive sensing in spatial and spectral domains. Therefore, we propose a spatial and spectral domain joint compressive sensing method for infrared spectral images. Based on the compressive sensing theory, the high spatial resolution of the DMD compensates for the low spatial resolution of the infrared detector, and the high spectral resolution of the grating compensates for the low spectral resolution of the infrared detector. The original hyperspectral image is reconstructed by using the deep learning method, so as to break through the limitation of spatial resolution and spectral resolution of the infrared spectral imaging equipment, and obtain the infrared spectral image with high spatial resolution and high spectral resolution. SUMMARY

[0005] In order to solve the problem that the existing infrared spectral image with high spectral resolution and high spatial resolution is difficult to be acquired simultaneously, the purpose of the present application is to provide a spatial and spectral domain joint compressive sensing method and device for infrared spectral images. The method first generates a spatial and spectral domain joint compressive imaging encoding matrix, then loads the matrix as an encoding template to the DMD, so that the target light information is spatially modulated, then uses a diffraction grating to modulate the spatially modulated spectral image in the spectral domain, and finally uses a low-resolution infrared camera to spatially and spectrally downsample the modulated image to obtain a low-resolution two-dimensional mixed image. The low spatial and low spectral resolution infrared image collected can be restored to the original high spatial and high spectral resolution infrared image by using a reconstruction algorithm.

[0006] The purpose of the present application is realized by the following technical solutions.

[0007] A spatial and spectral domain joint compressive sensing method for infrared spectral images, comprising the following steps:

[0008] Step one, the target scene is imaged to the DMD through the imaging system, and the target scene is divided into an original spectral image X of multiple wavebands, X e R H×W×C , wherein R represents a real number set, and the superscript HxWxC represents the data dimension; H and W represent the length and width of the image; C represents the number of spectral bands;

[0009] Step two, an encoding template is generated, and the X in step one is spatially modulated to obtain a spatial modulation image Y' using the encoding template.

[0010] A random binary matrix M e R H×WThe array is used as an encoding template; then the X in step one is encoded by using the encoding template, that is, the encoding template is loaded onto the DMD and the X is encoded, at this time the spatial dimension information of the X is changed, that is, the spatial modulation of the X is completed, and the spatial modulation light Y' ∈ R H×W×C The modulation result of the i-th waveband in Y' is recorded as Y'( :, :, i), and Y'( :, :, i) is represented as follows:

[0011] Y'( :, :, i) = M ⊙ X( :, :, i), i = 1,..., C (1)

[0012] Wherein, X( :, :, i) represents the original image corresponding to the i-th waveband in X; and ⊙ represents matrix point multiplication;

[0013] Step three, the Y' of step two reaches the diffraction grating through the condenser; the position of the diffraction grating is adjusted so that the information light is normally incident on the surface of the diffraction grating, the diffraction grating modulates the spectral dimension of Y', and Y" obtains the spectral modulation light; the modulation result of Y" at (u, v) in the i-th grating image field focal plane coordinate system is recorded as Y"(u, v, i), and the value of Y' at (x, y) in the i-th grating object plane coordinate system is recorded as Y'(x, y, i), and Y"(u, v, i) is represented as follows:

[0014] Y"(u, v, i) = Y'(x, y + n(i - λ c ), i), i = 1,..., C (2)

[0015] Wherein, n represents the grating dispersion rate, λ c represents the central wavelength of the grating, and n(i - λ c ) represents the distance of the light of the i-th waveband being dispersed;

[0016] Step four, the imaging system images Y" obtained in step three to the camera detection array; since the spatial resolution of the DMD is higher than that of the camera, the light reflected by multiple micro-mirrors needs to be focused on one pixel of the camera to realize spatial compression; since the detector plane is a two-dimensional plane, Y" will be compressed into a two-dimensional image on the detector to realize spectral domain compression;

[0017] The low-resolution image resolution of the camera is recorded as Ha×(W + n(C - 1))a, the spatial-spectral domain compression is to simultaneously perform spatial downsampling and spectral downsampling on the modulated image on the detector surface to obtain the compressed image Y; the pixel value of Y in the p-th row and the q-th column of the camera coordinate system is recorded as Y(p, q), and Y(p, q) is represented as follows:

[0018]

[0019] Wherein, a represents the spatial compression ratio; r, s respectively represent the row, column coordinates of the modulated image; Y''(r, s, i) represents the value of the i-th waveband of the spectral modulation result Y'' at (r, s); r, s are represented as:

[0020] r=(p-1) / a+1:p / a (4)

[0021] s=(q-1) / a+1:q / a (5)

[0022] Wherein, the value range of p, q is respectively p=1, 2, …, Ha, q=1, 2, …, (W+n(C-1))a;Colon: represents increasing from the previous number, step length is 1, until the latter number;

[0023] Step five, using the reconstruction algorithm to the compressed image Y obtained in step four is reconstructed, and the infrared spectrum image with high spatial resolution and high spectral resolution is obtained

[0024] The reconstruction algorithm in step five adopts a deep learning method.

[0025] The application further discloses a space-spectrum domain joint compression sensing device for infrared spectrum images, which is used for realizing the space-spectrum domain joint compression sensing method for infrared spectrum images. The space-spectrum domain joint compression sensing device for infrared spectrum images comprises two imaging lenses, a DMD, a relay lens, a diffraction grating, an infrared camera and a computer. A target or a scene reaches the DMD after passing through the imaging lens. The DMD modulates the light information from the target according to a preset loaded coding template. The modulated light information passes through the diffraction grating, and light of different wavelengths is dispersed in the dispersion direction, so that the light information of the target is spectrally modulated. The modulated light signal is imaged onto the infrared camera by the imaging lens. Since only one dispersion element is used, the camera receives a two-dimensional image in which the space and spectrum dimensions are mixed. The obtained image is uploaded to the computer, and a high spatial and spectral resolution image is reconstructed by solving a reconstruction algorithm.

[0026] The imaging lens is used for imaging the target to the digital micromirror array DMD and imaging the DMD and the grating to the detection plane of the detector, respectively.

[0027] The DMD is used for storing a designed coding template and loading the coding template on the DMD, and spatially modulating the image imaged to the DMD.

[0028] The relay lens is used for transmitting the light reflected by the DMD to the diffraction grating.

[0029] The diffraction grating is used for dispersing the light signal of the target in the dispersion direction according to the wavelength, and spectrally modulating the light signal of the target.

[0030] The infrared camera receives the modulated image;

[0031] The computer runs the compressive imaging algorithm to reconstruct the original high spatial and high spectral resolution image from the two-dimensional mixed low resolution image captured by the camera using the measurement matrix.

[0032] Advantages:

[0033] 1. The present application discloses a kind of space spectral domain joint compressive sensing method and device for infrared spectral image, break through the limit of infrared camera spatial resolution and spectral resolution by space domain compressive imaging and spectral domain compressive imaging, so that using low resolution infrared camera, DMD, diffraction grating, imaging light path and control equipment can collect high resolution spectral image.

[0034] 2. The present application discloses a kind of space spectral domain joint compressive sensing method and device for infrared spectral image, using the space, spectral correlation of spectral image, by applying compressive imaging method, compress data before data acquisition, so that only a small amount of data can be recorded to complete high resolution spectral image, significantly reduce the data amount of acquisition, storage, transmission, improve the efficiency of space spectral domain joint compressive sensing. BRIEF DESCRIPTION OF DRAWINGS

[0035] Figure 1 Space spectral domain joint compressive imaging principle schematic diagram;

[0036] Figure 2 The device structure diagram of the system;

[0037] Figure 3 Space spectral domain compressive imaging image modulation and acquisition process schematic diagram;

[0038] Figure 4 A set of original spectral images in the embodiment;

[0039] Figure 5 Reconstructed spectral image in the embodiment;

[0040] 1-target or scene, 2-imaging lens, 3-digital micro-mirror array DMD, 4-relay lens, 5-diffraction grating, 6-imaging lens, 7-infrared camera, 8-computer. DETAILED DESCRIPTION

[0041] In order to better illustrate the purpose and advantages of the present application, the content of the application will be further described below in conjunction with the drawings and examples.

[0042] The spatial resolution is 512x512, the pixel depth is 8bits, the spectral resolution of 10nm wavelength in 610-700nm is 10band images (such asFigure 4 The low-resolution image with a resolution of 256x265 is obtained after down-sampling using the encoding template and the diffraction grating. Finally, a reconstructed image of 10 bands with a spatial resolution of 512x512 and a spectral resolution of 10 nm is recovered from the low-resolution image of 256x265 using a recovery algorithm, as shown in FIG. 3B, so that the spatial resolution is 4 times (2x2) and the spectral resolution is 10 times higher than that of the collected low-resolution image. Figure 5 The low-resolution image with a resolution of 256x265 is obtained after down-sampling using the encoding template and the diffraction grating. Finally, a reconstructed image of 10 bands with a spatial resolution of 512x512 and a spectral resolution of 10 nm is recovered from the low-resolution image of 256x265 using a recovery algorithm, as shown in FIG. 3B, so that the spatial resolution is 4 times (2x2) and the spectral resolution is 10 times higher than that of the collected low-resolution image.

[0043] The compression imaging step in this embodiment is shown in FIG. 2B, and the spectral modulation and image compression process is shown in FIG. 3B. Figure 1 The compression imaging step in this embodiment is shown in FIG. 2B, and the spectral modulation and image compression process is shown in FIG. 3B. Figure 3 The compression imaging step in this embodiment is shown in FIG. 2B, and the spectral modulation and image compression process is shown in FIG. 3B.

[0044] The spectral image compression method disclosed in this embodiment is a joint compression sensing method in the spatial-spectral domain, and the specific implementation steps are as follows:

[0045] Step one: The target scene is imaged onto the DMD by the imaging system, and the target scene is divided into an original spectral image Xe R H×W×C , where R represents the real number set, and the superscript HxWxC represents the data dimension; H and W represent the length and width of the target space; and C represents the number of spectral bands.

[0046] The target used in this embodiment is an original spectral image X containing 10 bands with a spectral resolution of 10 nm from 610 nm to 700 nm, a spatial resolution of 512x512, and a pixel depth of 8 bits. That is, in this example, H=W=512 and C=10.

[0047] Step two: An encoding template is generated, and the X in step one is spatially modulated using the encoding template to obtain a spatial modulation image Y';

[0048] A random binary matrix M e R H×W is used as the encoding template, and in this example, H=W=512; then the X in step one is encoded using the encoding template, that is, the encoding template is loaded onto the DMD and X is encoded, at this time, the spatial dimension information of X is changed, that is, the spatial modulation of X is completed, and a spatial modulation light Y' e R H×W×C is obtained. The modulation result of the i-th band in Y' is denoted as Y'( :, :, i), and is represented as follows:

[0049] Y'( :, :, i) = M o X( :, :, i), i = 1,..., C (6)

[0050] where X( :, :, i) represents the original image corresponding to the i-th band in X; and o represents matrix point multiplication (that is, the corresponding positions in the two matrices are multiplied).

[0051] Step three, Y' of step two reaches the diffraction grating through the condenser; the diffraction grating is a multi-slit diffraction screen with a periodic repeating structure, which has the advantages of high spectral resolution and linear dispersion. Adjust the position of the diffraction grating so that the information light is normally incident on the grating surface. At this time, the grating equation is:

[0052] dsinθ = mλ (7)

[0053] where d is the grating constant, θ is the diffraction angle, m is the diffraction order, and λ is the wavelength. As can be seen from the grating equation, different wavelengths of light correspond to different diffraction angles, so different wavelengths of light can be separated using a diffraction grating, thereby achieving dispersion (assuming dispersion only along the y-axis direction). At this time, the target spectral dimension information has also changed, so the diffraction grating can be regarded as modulating the target light signal in the spectral dimension.

[0054] Adjust the position of the diffraction grating so that the information light is normally incident on the diffraction grating surface. The diffraction grating modulates the Y' in the spectral dimension, and the Y" obtains the spectrum modulated light. The modulation result of Y" at the (u, v) of the i-th grating image plane coordinate system is denoted as Y"(u, v, i), and the value of Y' at the (x, y) of the i-th grating object plane coordinate system is denoted as Y'(x, y, i). The mathematical representation of Y"(u, v, i) is as follows:

[0055] Y"(u, v, i) = Y'(x, y + n(i - λ c ), i), i = 1,..., C (8)

[0056] where n represents the grating dispersion rate (i.e. the distance separated by unit wavelength on the imaging focal plane), λ c represents the center wavelength of the grating, n(i - λ c ) represents the distance dispersed by the light of the i-th waveband; in this example, n = 2 and λ c = 610 nm.

[0057] Step four, the imaging system images Y" obtained in step three to the camera detection array. Since the spatial resolution of the DMD is higher than that of the camera, the light reflected by multiple micro-mirrors needs to be focused on one pixel of the camera to achieve spatial compression. Since the detector plane is a two-dimensional plane, Y" will be compressed into a two-dimensional image on the detector to achieve spectral domain compression.

[0058] The resolution of the low-resolution image captured by the camera is Ha×(W+n(C-1))a, which is 256×265 in this example. The spatial-spectral domain compression is to simultaneously perform spatial down-sampling and spectral down-sampling on the modulated image with a resolution of 512×530×10 on the detector surface, to obtain a compressed image Y with a resolution of 256×265. The value of Y at the camera coordinate system (p, q) is denoted as Y(p, q), and is expressed as follows:

[0059]

[0060] wherein a represents the spatial compression ratio, which is 0.5 in this example; r and s represent the row and column coordinates of the modulated image, respectively; Y"(r, s, i) represents the value of the i-th waveband of the spectral modulation result Y" at (r, s); and r and s are expressed as:

[0061] r = (p-1) / a+1:p / a (10)

[0062] s = (q-1) / a+1:q / a (11)

[0063] wherein p and q take values in the ranges of p = 1, 2, …, Ha and q = 1, 2, …, (W+n(C-1))a, respectively; the colon: represents incrementing from the previous number by a step size of 1 until the latter number is reached;

[0064] Step five: using a reconstruction algorithm to reconstruct the compressed image Y obtained in step four to obtain an infrared spectral image with high spatial resolution and high spectral resolution.

[0065] The image recovery algorithm in step five adopts a deep learning method, and the network model used is a DAUHST model; the compressed image Y, the measurement matrix H, and the trained network model are used to reconstruct the image

[0066] In this example, a high-resolution image of 10 wavebands can be finally recovered. Therefore, the spatial resolution of the image is improved by 4 times (2×2) and the spectral resolution is improved by 10 times using the spatial-spectral domain compression imaging method in this example. The recovery effect is shown in Figure 4 To quantitatively analyze the recovery effect, the peak signal-to-noise ratio (PSNR) of the recovered image is calculated, which is defined as follows:

[0067]

[0068] wherein MSE represents the mean square error, X(:,:,i) represents the original image corresponding to the i-th waveband in X, represents the reconstructed image corresponding to the i-th waveband in , and MAX represents the maximum pixel value of the image.

[0069] Apparatus for implementing the above method, such as Figure 2 As shown, the system includes two imaging lenses 2 and 6, a DMD3, a relay lens 4, a diffraction grating 5, an infrared camera 7, and a computer 8. The target or scene 1 passes through the imaging lens 2 and reaches the DMD3. The DMD3 loads a pre-defined encoding template to spatially modulate the light information from the target. The modulated light information passes through the relay lens 4 and reaches the diffraction grating 5, where light of different wavelengths is dispersed along the dispersion direction, thus spectrally modulating the target's light information. The modulated light signal is then imaged onto the infrared camera 7 by the imaging lens. Since it only passes through one dispersive element, the camera receives a two-dimensional image with both spatial and spectral dimensions superimposed. The obtained image is uploaded to the computer 8, where a high spatial and hyperspectral resolution image is reconstructed using a trained DAUHST network model. The reconstruction result is shown in the figure. Figure 5 As shown.

[0070] Imaging lenses 2 and 6 are used to image the target onto the digital micromirror array (DMD) and to image the DMD and grating onto the detector plane, respectively.

[0071] The function of DMD3 is to store the designed encoding template and load it onto the DMD, and to perform spatial modulation on the image imaged onto the DMD.

[0072] The function of relay lens 4 is to transmit the light reflected by the DMD to the diffraction grating;

[0073] The function of the diffraction grating 5 is to disperse the target light signal along the dispersion direction according to the wavelength, thereby modulating the spectrum of the target light signal;

[0074] The function of infrared camera 7 is to receive modulated images;

[0075] The computer's function is to run the spatial-spectral domain compression imaging restoration algorithm, which uses the measurement matrix to reconstruct the original high spatial-spectral resolution image from the two-dimensional aliased low-resolution image captured by the camera.

[0076] The above detailed description further illustrates the purpose, technical solution, and beneficial effects of the invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A spatial-spectral domain joint compressed sensing method for infrared spectral images, characterized in that: Includes the following steps: Step 1: The target scene is imaged onto the DMD using an imaging system, and the target scene is decomposed into a raw spectral image X∈R with multiple bands. H×W×C Where R represents the set of real numbers, and the superscript H×W×C represents the data dimension; H and W represent the length and width of the image; and C represents the number of spectral bands. Step 2: Generate an encoding template and use the encoding template to spatially modulate X from Step 1 to obtain a spatially modulated image Yπ; Using a random binary matrix M∈R H×W This serves as an encoding template; then, the encoding template is used to encode X, that is, the encoding template is loaded onto the DMD and X is encoded. At this time, the spatial dimension information of X changes, that is, the spatial modulation of X is completed, and the spatially modulated light Y'∈R is obtained. H×W×C Let Y'(:,:,i) be the modulation result of the i-th band in Y'. Then Y'(:,:,i) is represented as follows: Y'(:,:,i)=M⊙X(:,:,i),i=1,...,C (1) Where X(:,:,i) represents the original image corresponding to the i-th band in X; ⊙ represents matrix dot product; Step 3: The Y' from Step 2 reaches the diffraction grating through the condenser lens; adjust the position of the diffraction grating so that the information light is incident orthogonally on the surface of the diffraction grating. The diffraction grating modulates the spectral dimension of Y', resulting in the spectrally modulated light Y'". The modulation result of Y' at (u,v) in the image-side focal plane coordinate system of the i-th grating is denoted as Y'(u,v,i), and the value of Y' at (x,y) in the object-side coordinate system of the i-th grating is denoted as Y'(x,y,i). Then Y'(u,v,i) is represented as follows: Y”(u,v,i)=Y'(x,y+n(i-λ c ),i),i=1,...,C (3) Where n represents the grating dispersion rate, λ c The center wavelength of the grating is represented by n(i-λ). c () represents the distance by which the light in the i-th band is dispersed; Step 4: The imaging system images the Y” obtained in Step 3 onto the camera detector array. Since the spatial resolution of the DMD is higher than that of the camera, it is necessary to focus the light reflected by multiple micromirrors onto one pixel of the camera to achieve spatial compression. Also, since the detector plane is a two-dimensional plane, Y” will be compressed into a two-dimensional image on the detector to achieve spectral compression. Let the resolution of the low-resolution image acquired by the camera be Ha×(W+n(C-1))a. Spatial-spectral domain compression involves simultaneously downsampling the modulated image Y” in the modulated domain on the detector surface to obtain the compressed image Y. The pixel value of Y in the p-th row and q-th column in the camera coordinate system is denoted as Y(p,q). Y(p,q) is represented as follows: Where a represents the spatial compression ratio; r and s represent the row and column coordinates of the modulated image Y”, respectively; Y”(r,s,i) represents the value of the i-th band of the spectral modulation result Y” at (r,s); r and s are expressed as: r=(p-1) / a+1:p / a (5) s=(q-1) / a+1:q / a (6) Where p and q take values ​​in the ranges of p = 1, 2, ..., Ha, and q = 1, 2, ..., (W+n(C-1))a, respectively; the colon indicates that the number increments from the previous number by a step of 1 until the next number is reached. Step 5: Reconstruct the compressed image Y obtained in Step 4 using a reconstruction algorithm to obtain an infrared spectral image with high spatial and spectral resolution.

2. The spatial-spectral domain joint compressed sensing method for infrared spectral images as described in claim 1, characterized in that: The reconstruction algorithm described in step five employs a deep learning method.

3. An apparatus for implementing the spatial-spectral domain joint compressed sensing method for infrared spectral images as described in claim 1, characterized in that: It includes two imaging lenses, a DMD, a relay lens, a diffraction grating, an infrared camera, and a computer; after the target or scene passes through the imaging lens, it reaches the DMD, and the DMD spatially modulates the light information from the target according to a preset encoding template. The modulated light information passes through a diffraction grating, and light of different wavelengths is dispersed along the dispersion direction, thus performing spectral modulation on the light information of the target. The modulated light signal is then imaged onto an infrared camera by an imaging lens. Since it only passes through a dispersive element, the camera receives a two-dimensional image that aliases both spatial and spectral dimensions. The obtained image is uploaded to a computer, and a high spatial and hyperspectral resolution image is reconstructed by solving a reconstruction algorithm.

4. The apparatus as described in claim 3, characterized in that: The imaging lenses function to image the target onto the digital micromirror array (DMD) and to image the DMD and grating onto the detector plane. The function of the DMD is to store the designed encoding template and load it onto the DMD, and to perform spatial modulation on the image image captured onto the DMD. The function of the relay lens is to transmit the light reflected from the DMD to the diffraction grating; The function of a diffraction grating is to disperse the target light signal along the dispersion direction according to the wavelength, thereby modulating the spectrum of the target light signal; The function of an infrared camera is to receive modulated images; The computer's role is to run a spatial-spectral domain compression imaging restoration algorithm, using a measurement matrix to reconstruct the original high spatial-spectral resolution image from a two-dimensional aliased low-resolution image captured by the camera.

Images