A full polarization spectral imaging device and detection method

Abstract

The application discloses a kind of full polarization spectral imaging device and detection method, contain sequentially arranged polarization modulator, imaging spectrometer, data acquisition processing display system to incident light;Polarization modulator is single fixed retarder and single fixed linear polarizer sequentially arranged along incident light, can carry out spectral modulation to full stokes parameter, it has the superiority of simple, ultra-compact, miniaturization, anti-system error in structure.By constructing complete imaging physical model or constructing imaging physical model based on compressed sensing, then using untrained artificial neural network realizes the full stokes parameter image reconstruction under each narrow waveband, and the deep neural network of "black box operation" is endowed with physical meaning.The reconstructed polarization spectral resolution can be consistent with the resolution of the imaging spectrometer used;And it has the characteristics of strong anti-noise ability, so as to ensure high-quality reconstruction.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of optical remote sensing detection, and particularly relates to a full-polarization spectral imaging device and a detection method. BACKGROUND

[0002] The electromagnetic wave radiated by an object contains spectral polarization information varying with spatial position, which can be used to invert the morphology and physical and chemical characteristics of the target. Polarization spectral imaging technology is a cutting-edge remote sensing technology that simultaneously acquires two-dimensional spatial target polarization spectral information, and has certain potential to improve the efficiency and accuracy of target detection, identification and classification. It will have important application value and prospect in many fields such as earth resource survey, environmental health monitoring, natural disaster prediction, atmospheric detection, astronomical observation, machine vision bionics, biomedical diagnosis and the like.

[0003] Polarization spectral imaging technology is a combination of polarization modulation technology and spectral imaging technology. Polarization modulation technology includes rotating element method, split-aperture method, split-amplitude method and spectral modulation method; among them, the spectral modulation method has the advantages of real-time modulation and single-channel due to the use of only one optical path to realize polarization modulation coding, and is suitable for fusion with any spectral imaging system. However, the existing polarization module based on spectral modulation includes two fixed retarders and one fixed polarizer, and the thickness ratio and azimuth angle of the two retarders determine the amount of coded information and the subsequent recovery effect. So far, there is no spectral modulation technology based on a single fixed retarder and a fixed polarizer, mainly because there is no algorithm available for subsequent decoding in intuition.

[0004] In addition, in the decoding algorithm of the existing polarization spectral technology based on spectral modulation, the commonly used is the Fourier transform decoding algorithm, which mainly performs channel interception operation in the frequency domain, and has problems such as channel crosstalk, narrow bandwidth, low signal-to-noise ratio, etc. Moreover, the spectral resolution of the recovered Stokes parameters is lower than that of the used spectrometer, which is not conducive to subsequent processing. Although the newly emerging advanced iterative method based on least squares and the optimization method based on compressed sensing can solve the problem of low spectral resolution, their noise suppression capability is limited, especially the introduction of polarization elements greatly reduces the detection intensity, and low signal-to-noise ratio will lead to great distortion of the recovered spectrum. The untrained network method is a new type of iterative method that does not need a large amount of data set for training. It relies on the nonlinear fitting advantage of the network itself, regards its structure as a depth image prior, can be equivalent to a regularization process, and brings hope to the reconstruction problem in this direction, and has successfully solved many imaging problems and image denoising, repair and super-resolution and many other image processing problems. However, the untrained network has not been applied to polarization spectral imaging recovery, mainly because the imaging model has not been established.

[0005] It can be seen that the prior art has not yet realized the hardware of spectral modulation by using a single fixed retarder and a single polarizer, and the existing demodulation method is not suitable for the above hardware, and it is urgent to develop corresponding hardware and demodulation algorithm to realize a polarization spectral imaging software and hardware technology with the advantages of simple and compact structure, rapid and convenient, economical and practical, high polarization spectral resolution, strong anti-noise capability and the like. SUMMARY

[0006] In order to overcome the above-mentioned defects of the prior art, the purpose of the present application is to provide a full polarization spectral imaging device and a detection method to solve one or more of the above-mentioned technical problems. The present application has the advantages of simple and compact structure, rapid and convenient, economical and practical, high resolution, strong anti-noise capability, simultaneous performance improvement and the like.

[0007] In order to achieve the above-mentioned purpose, the technical scheme adopted by the present application is:

[0008] The full polarization spectral imaging device comprises a polarization modulator, an imaging spectrometer and a data acquisition, processing and display system arranged in sequence along the incident light; the polarization modulator is composed of a fixed retarder and a fixed linear polarizer arranged in sequence along the incident light.

[0009] For convenience of description, the present application sets an x-y-z orthogonal coordinate system conforming to the right-hand rule, wherein the positive direction of the z-axis is the incident direction; the positive direction of the y-axis is vertically upward; the x-axis is perpendicular to the y-axis and the z-axis; the optical axis of the full polarization spectral imaging device is horizontally arranged, and the z-axis is parallel to the optical axis.

[0010] In one embodiment, the fast axis of the fixed retarder is parallel to the x-y plane, and the included angle θ between the x-z plane and the fast axis is any value between 0° and 180°, and θ cannot be 0°, 45°, 90°, 135° or 180°; the polarization direction of the fixed linear polarizer is parallel to the x-axis or the y-axis.

[0011] In one embodiment, a first fixed retarder can be arranged before the fixed retarder, wherein the fast axis direction of the first fixed retarder is parallel to the polarization direction of the fixed linear polarizer, the fast axis direction of the fixed retarder is parallel to the x-y plane, and the included angle between the x-z plane and the fast axis is 45°, and the polarization direction of the fixed linear polarizer is parallel to the x-axis or the y-axis.

[0012] In one embodiment, the first fixed retarder is a quarter-wave achromatic retarder.

[0013] In one embodiment, the imaging spectrometer is a snapshot imaging spectrometer or a scanning imaging spectrometer; the data acquisition, processing and display system comprises data acquisition, preprocessing, restoration and display functional modules.

[0014] In one embodiment, the polarization modulator is placed in the collimated optical path of the imaging spectrometer.

[0015] The present invention also provides a detection method based on the aforementioned full polarization spectral imaging device, comprising the following steps:

[0016] The beam of light carrying spatial polarization spectral information emitted by the target is polarized by a polarization modulator and then imaged by an imaging spectrometer. The data acquisition, processing, and display system controls the imaging spectrometer to acquire polarization-modulated images under all resolvable narrow bands; the polarization-modulated image under the k-th resolvable narrow band is then vectorized, i.e. Simultaneously, the Stokes parameters of the incident scene under the k-th resolvable narrow band are vectorized, i.e. Where N x N y N λ and N s Let represent the number of pixels along the x-axis, the number of pixels along the y-axis, the number of resolvable narrow bands, and the number of Stokes parameters to be measured, respectively; then the linear imaging model of the fully polarized spectral imaging device in the k-th resolvable narrow band is expressed as:

[0017] I k =M k S k +ε k (1)

[0018] in and These represent the measurement matrix and measurement noise matrix of the full polarization spectral imaging device, respectively; the specific representations of each parameter are as follows:

[0019]

[0020]

[0021]

[0022]

[0023]

[0024]

[0025]

[0026]

[0027]

[0028]

[0029]

[0030] Where A, B, and C are the modulation variables formed after modulation by the polarization modulator, and the superscripts are [1,1], [2,1], ..., [N]. x N y ] represents the spatial coordinates of an image pixel;

[0031] When the polarization modulator consists of a fixed delay unit and a fixed linear polarizer, then:

[0032] A = cos 2 2θ+sin 2 2θcosδ B=sin 2 (δ / 2)sin4θ C=-sin2θsinδ

[0033] When a first fixed delay unit is placed before the fixed delay unit and the fixed linear polarizer, then...

[0034] A=cosδ B=sinδsinδ1 C=-sinδcosδ1

[0035] Where θ is the angle between the fast axis of the fixed delay and the xz plane, δ is the delay of the fixed delay, and δ1 is the delay of the first fixed delay.

[0036] In one embodiment, a nonlinear mapping function of an untrained artificial neural network (which can be any network model) is utilized. As a priori knowledge of the depth image, we solve the following ill-conditioned inverse problem:

[0037]

[0038] Where Θ k For parameters of an untrained artificial neural network, For Θ k The estimated value; the input is the polarization modulation image I under the k-th narrow band. k After multiple iterations and optimizations, the optimal result was obtained. pass By performing mapping, we obtain the estimated Stokes parameter image for the k-th narrowband:

[0039]

[0040] Therefore, the Stokes parameter images for all narrow bands can be obtained.

[0041] In one embodiment, when the polarization information of the incident scene varies slowly with wavelength, it is assumed that the polarization information under a band unit composed of several adjacent narrow bands is consistent. In this case, the underdetermined problem of equation (5) is transformed into the well-posed problem as follows:

[0042]

[0043] Where u = 0, 1, ..., N s -1 indicates the order of narrowbands within the band cell. To ensure that the narrowbands at both ends of the entire narrowband array can form similar band cell combinations, a mirror-filling method is used at both ends to supplement the required number of narrowbands. In this case, the input data is N. s Polarization modulation image under a nearby narrow band After multiple iterations and optimizations, the optimal solution was obtained. Through optimal Perform a mapping to obtain these N s Estimated Stokes parameter images under neighboring narrow bands:

[0044]

[0045] By misaligning a single narrow band, the adjacent N bands are recombined s A narrow band, to obtain another set of N s The estimated Stokes parameter images under each narrow band are obtained, and so on. The estimated Stokes image under each narrow band is solved by N. s Then, by taking the average value of equation (9), the final Stokes parameter image for each narrow band is obtained:

[0046]

[0047] In one embodiment, based on the principle of compressed sensing, the underdetermined problem of equation (5) is transformed into the following optimization problem:

[0048]

[0049] Among them W k For an appropriate basis (e.g., Legendre polynomial basis, or discrete cosine transform basis, or a combination of both, or other applicable basis), the superscripts (1), (2), ..., (N) are used. s ) represents N s W k The order; after multiple iterations of optimization, the optimal order is obtained. Through optimal By performing mapping, we obtain the estimated sparse image under the k-th narrowband:

[0050]

[0051] The estimated Stokes parameter image for the corresponding k-th narrowband is as follows:

[0052]

[0053] All narrow bands are sequentially processed through the above solution process, ultimately yielding estimated Stokes parameter images for all narrow bands.

[0054] Compared with the prior art, the beneficial effects of the present invention are:

[0055] This invention provides a fully polarization spectral imaging device that features a simple and compact structure, affordability, high timeliness, high resolution, strong noise resistance, and simultaneous performance improvement. Specifically, this invention integrates the multi-complex spectral modulation characteristics of a single fixed delayer and a single fixed polarizer, along with the powerful fitting ability of an untrained network. This simplifies the system structure, compresses the system size, preserves spatial dimension information, and performs multi-complex polarization modulation encoding and sampling only in the spectral dimension. Subsequent reconstruction models only need to enhance spectral demodulation, thereby reducing model complexity requirements, improving reconstruction quality and noise resistance, and maximizing spectral range and resolution performance.

[0056] In this invention, the polarization modulation device employs a combination of a single fixed delay unit and a single fixed polarizer. Compared to a device using two fixed delay units and a single fixed polarizer, it offers advantages in structure such as simplicity, ultra-compactness, miniaturization, and resistance to systematic errors. However, the detection method of this invention is also applicable to devices using two fixed delay units and a single fixed polarizer.

[0057] In this invention, by establishing a physical imaging model and using an untrained artificial neural network to reconstruct the final data, a suitable imaging hardware system and reconstruction software system can be obtained, giving physical meaning to the "black box operation" of deep neural networks. The reconstructed polarization spectral resolution can be kept consistent with the resolution of the imaging spectrometer used; and it has strong noise resistance, thus ensuring high-quality reconstruction.

[0058] The full polarization spectral imaging device and detection method of the present invention have potential application value in fields such as astronomical observation, space exploration, Earth remote sensing, machine vision and biomedical diagnosis. Attached Figure Description

[0059] Figure 1 This is a schematic diagram of the structure of the full polarization spectral imaging device of the present invention.

[0060] Figure 2 This is a schematic diagram of another structure of the polarization modulator of the present invention.

[0061] Figure 3This is a schematic diagram of the detection method of the full polarization spectral imaging device of the present invention.

[0062] Figure 4 This is a schematic diagram of the detection method of the full polarization spectral imaging device of the present invention.

[0063] Figure 5 This is a specific implementation example of the full polarization spectral imaging device and detection method of the present invention.

[0064] In the figure, 11 is a polarization modulator, 111 is a fixed delay unit, 112 is a fixed linear polarizer, 113 is the first fixed delay unit, 12 is an imaging spectrometer, and 13 is a data acquisition, processing, and display system. Detailed Implementation

[0065] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings and examples.

[0066] As mentioned earlier, given the limitations of current decoding algorithms, it is currently believed that a polarization modulation device consisting of a single fixed delay unit and a single fixed linear polarizer is not feasible.

[0067] like Figure 1 As shown, in Embodiment 1 of the present invention, a fully polarized spectral imaging device is provided, mainly comprising a polarization modulator 11, an imaging spectrometer 12, and a data acquisition, processing, and display system 13 arranged sequentially along the incident light direction. The polarization modulator 11 is preferably placed in the collimated optical path of the imaging spectrometer 12. The polarization modulator 11 consists only of a fixed retarder 111 and a fixed linear polarizer 112 arranged sequentially along the incident light direction, enabling spectral modulation of all Stokes parameters. Structurally, it possesses advantages such as simplicity, ultra-compactness, miniaturization, and resistance to systematic errors.

[0068] Among them, the imaging spectrometer 12 can be a snapshot imaging spectrometer or a scanning imaging spectrometer, etc. The data acquisition, processing and display system 13 mainly includes functional modules such as data acquisition, preprocessing, restoration and display. Both the imaging spectrometer 12 and the data acquisition, processing and display system 13 adopt existing conventional equipment or modules and systems.

[0069] refer to Figure 1 For ease of description, this invention establishes an orthogonal xyz coordinate system that conforms to the right-hand rule, that is, the positive z-axis is the incident direction of the light beam; the positive y-axis is vertically upward; the x-axis is perpendicular to the y-axis and the z-axis; the optical axis of the fully polarized spectral imaging device is set horizontally, and the z-axis is parallel to the optical axis.

[0070] In this coordinate system, the polarization direction of the fixed linear polarizer 112 is parallel to the x-axis, the fast axis direction of the fixed delayer 111 is parallel to the xy plane, and the angle θ between it and the xz plane can be arbitrarily taken between 0° and 180°. In particular, θ cannot be 0°, 45°, 90°, 135° or 180°, and the delay amount of the fixed delayer 111 is δ.

[0071] Based on this structure, the detection method of the present invention is as follows:

[0072] The beam of light carrying spatial polarization spectral information emitted by the target is polarized by polarization modulator 11, and then spectrally imaged by imaging spectrometer 12. Data acquisition, processing and display system 13 controls imaging spectrometer 12 to acquire polarization modulation images under all resolvable narrow bands; the polarization modulation image under the k-th resolvable narrow band is vectorized, i.e. Simultaneously, the Stokes parameters of the incident scene under the k-th resolvable narrow band are vectorized, i.e. The linear imaging model of the fully polarized spectral imaging device in the k-th resolvable narrow band is expressed as:

[0073] I k =M k S k +ε k (1)

[0074] in and Let N represent the measurement matrix and measurement noise matrix of the total polarization spectral imaging device, respectively. x N y N λ and N s These represent the number of pixels in the x-axis direction, the number of pixels in the y-axis direction, the number of resolvable narrowbands, and the number of Stokes parameters to be measured, respectively.

[0075] Therefore, each parameter can be specifically represented as follows:

[0076]

[0077]

[0078]

[0079]

[0080]

[0081]

[0082]

[0083]

[0084]

[0085]

[0086]

[0087] Among them, the superscripts [1,1], [2,1], ..., [N] x N y ] represents the spatial coordinates of an image pixel;

[0088] For example, I [2,1] This represents the measured intensity at spatial coordinates [2,1] in a narrow band. The first Stokes parameter at image spatial coordinates [2,1] in a certain narrow band. This represents the second Stokes parameter at image spatial coordinates [2,1] in a certain narrow band. The third Stokes parameter at image spatial coordinates [2,1] in a narrow band. The fourth Stokes parameter at image spatial coordinates [2,1] in a narrow band represents the fourth Stokes parameter. [2,1] This indicates that the image spatial coordinates [2,1] in a certain narrow band represent the modulation constants 1 and A. [2,1] This represents the first modulation variable, B, at image spatial coordinates [2,1] in a certain narrow band. [2,1] The second modulation variable, C, is located at image spatial coordinates [2,1] in a narrow band. [2,1] This represents the third modulation variable at the image spatial coordinates [2,1] in a certain narrow band.

[0089] s0 is a vector composed of the first Stokes parameter S0 of all spatial coordinate positions of the image, s1 is a vector composed of the second Stokes parameter S1 of all spatial coordinate positions of the image, s2 is a vector composed of the second Stokes parameter S2 of all spatial coordinate positions of the image, and s3 is a vector composed of the second Stokes parameter S3 of all spatial coordinate positions of the image.

[0090] m0 is a diagonal matrix composed of the modulation constant 1 at all spatial coordinate positions of the image; m1 is a diagonal matrix composed of the first modulation variable A at all spatial coordinate positions of the image; m2 is a diagonal matrix composed of the second modulation variable B at all spatial coordinate positions of the image; and m3 is a diagonal matrix composed of the third modulation variable C at all spatial coordinate positions of the image.

[0091] In Embodiment 1 of the present invention, it is obvious that:

[0092] A = cos 2 2θ+sin 2 2θcosδ B=sin 2 (δ / 2)sin4θ C=-sin2θsinδ

[0093] To solve for the Stokes parameter images in all narrow bands, this invention can utilize the nonlinear mapping function of an untrained artificial neural network. As a priori knowledge of the depth image, we solve the following ill-conditioned inverse problem:

[0094]

[0095] The artificial neural network of this invention can adopt any network model. In equation (5), Θ k For parameters of an untrained artificial neural network, Represents Θ k The estimated value; the input is the polarization modulation image I under the k-th narrow band. k After multiple iterations and optimizations, the optimal result was obtained. pass By performing mapping, we can obtain the estimated Stokes parameter image for the k-th narrowband:

[0096]

[0097] Therefore, the Stokes parameter images for all narrow bands can be obtained.

[0098] Based on this, the present invention provides two specific solutions to equation (5).

[0099] Method 1, see reference Figure 3 When the polarization information of the incident scene changes slowly with wavelength, the polarization information of a band unit composed of several (at least three or four) neighboring narrow bands can be considered to be consistent. In this case, the underdetermined problem of equation (5) is transformed into the well-posed problem as follows:

[0100]

[0101] Where u = 0, 1, ..., N s -1 indicates the order of narrowbands within the band cell. To ensure that the narrowbands at both ends of the entire narrowband array can form similar band cell combinations, a mirror-filling method is used at both ends to supplement the required number of narrowbands. In this case, the input data is N. s Polarization modulation image under a nearby narrow band After multiple iterations and optimizations, the optimal solution can be obtained. Then through the optimal By mapping, we can obtain these N. s Estimated Stokes parameter images under neighboring narrow bands:

[0102]

[0103] Using this method, neighboring N bands can be recombined by misaligning a single narrow band. s From a narrow band, another set of N can be obtained. s The estimated Stokes parameter images under each narrow band are obtained, and so on. The estimated Stokes image under each narrow band is solved by N. s Then, by taking the average value of equation (9), the final Stokes parameter image for each narrow band is obtained:

[0104]

[0105] Taking four adjacent wavebands as an example, the form of formula (7) is as follows:

[0106]

[0107] At this point, the input data consists of four polarization modulation images I from adjacent narrow bands: (k-3, k-2, k-1, and k). k-3 I k-2 I k-1 I k Correspondingly, the Stokes parameter images for these four adjacent narrow bands can be obtained as follows:

[0108]

[0109] Using this method, by misaligning a single narrow band and recombining four adjacent narrow bands, another set of four narrow band Stokes parameter images can be obtained. This process is repeated, with the Stokes image for each narrow band being solved four times. The average value is then taken to obtain the final Stokes parameter image for each narrow band.

[0110] Method 2, see reference Figure 4 Based on the principle of compressed sensing, the underdetermined problem of formula (5) is transformed into the following optimization problem:

[0111]

[0112] Among them W k For an appropriate basis (e.g., Legendre polynomial basis, or discrete cosine transform basis, or a combination of both, or other applicable basis), the superscripts (1), (2), ..., (N) are used. s ) represents N s W kThe order; after multiple iterations of optimization, the optimal order is obtained. Through optimal By performing mapping, we obtain the estimated sparse image under the k-th narrowband:

[0113]

[0114] The estimated Stokes parameter image for the corresponding k-th narrowband is as follows:

[0115]

[0116] All narrow bands are sequentially processed through the above solution process, ultimately yielding estimated Stokes parameter images for all narrow bands.

[0117] The detection method described above is also applicable to polarization modulation devices consisting of two fixed delayers and one fixed linear polarizer. For example... Figure 2 As shown, in Embodiment 2 of the present invention, a first fixed delay unit 113 is provided before the fixed delay unit 111.

[0118] Also using the xyz orthogonal coordinate system of Embodiment 1, in this embodiment, the fast axis direction of the first fixed delay unit 113 is parallel to the polarization direction of the fixed linear polarizer 112, the fast axis direction of the fixed delay unit 111 is parallel to the xy plane, and the angle between it and the xz plane is 45°, the polarization direction of the fixed linear polarizer 112 is parallel to the x-axis or y-axis; the delay amount δ1 of the first fixed delay unit 113.

[0119] In Embodiment 2 of the present invention, it is obvious that:

[0120] A=cosδ B=sinδsinδ1 C=-sinδcosδ1

[0121] For example, the first fixed delay unit 113 can be a quarter-wave achromatic delay unit with a delay of δ1 = π / 2. In this case, we have:

[0122] A = cosδ B = sinδ C = 0

[0123] This means that at this point, the measurement matrix M in equation (4) k The fourth element m3 does not exist; only the first three elements [m0, m1, m2] exist.

[0124] A specific embodiment of the total polarization spectral imaging device and detection method of the present invention is as follows: Figure 5As shown, the fixed delay unit 111 in the polarization modulator is made of a 1.4 mm thick quartz crystal, with its fast axis parallel to the xy plane and an angle of 22.5° with the xz plane. The fixed polarizer 112 is a polymer polarizer with its polarization direction parallel to the x-axis. The imaging spectrometer 113 can be the snapshot imaging spectrometer (ORRIS) based on a staircase lens array and a focal plane gradient filter proposed in the invention patent "ZL201710571025.0, Compact Miniature Snapshot Spectral Imaging Detection Device and Detection Method". The system can ultimately acquire polarization modulation images in 72 narrow bands within the operating wavelength range of the spectrometer (400-850nm), and import them into the data acquisition, processing, and display system. In one solution method, an untrained network is formed using a five-layer symmetrical U-net network. In the first solution method, polarization modulation images from four adjacent narrow bands are grouped into a single unit and input into the untrained network for solving. In the second solution method, a basis combining Legendre polynomial and discrete cosine transform bases is used to sparsely represent the Stokes parameters, and polarization modulation images from individual narrow bands are successively input into the untrained network for solving. Using the theoretical value of the measurement matrix formed by polarization modulator 11 as the starting value, after thousands of iterations and optimizations, the spectral images in 72 narrow bands corresponding to each of the four Stokes parameters (S0, S1, S2, S3) can be reconstructed. Figure 5 The results are the restoration results under a narrow band of 550nm. Both solution methods yielded high peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) values ​​for each Stokes parameter image under the 550nm narrow band.

[0125] Therefore, this invention, by constructing a complete imaging physical model or an imaging physical model based on compressed sensing, and then utilizing an untrained artificial neural network to reconstruct the full Stokes parameter image under each narrow band, gives physical meaning to the "black box operation" of deep neural networks. The reconstructed polarization spectral resolution can be kept consistent with the resolution of the imaging spectrometer used; and it has strong noise resistance, thus ensuring high-quality reconstruction.

[0126] The above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art can still make modifications or equivalent substitutions to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention are within the protection scope of the claims of the present invention pending approval.

Claims

1. A detection method based on a fully polarized spectral imaging device, wherein the fully polarized spectral imaging device comprises a polarization modulator (11), an imaging spectrometer (12), and a data acquisition, processing, and display system (13) arranged sequentially along the incident light direction; the polarization modulator (11) consists of a fixed delay unit (111) and a fixed linear polarizer (112) arranged sequentially along the incident light direction; Its features are, Includes the following steps: The beam of light carrying spatial polarization spectral information emitted by the target is polarized by the polarization modulator (11) and then spectrally imaged by the imaging spectrometer (12). The data acquisition, processing and display system (13) controls the imaging spectrometer (12) to acquire polarization-modulated images under all resolvable narrow bands; the polarization-modulated image under the k-th resolvable narrow band is then vectorized, i.e. Simultaneously, the Stokes parameters of the incident scene under the k-th resolvable narrow band are vectorized, i.e. Where N x N y N λ and N s Let represent the number of pixels along the x-axis, the number of pixels along the y-axis, the number of resolvable narrow bands, and the number of Stokes parameters to be measured, respectively; then the linear imaging model of the fully polarized spectral imaging device in the k-th resolvable narrow band is expressed as: I k =M k S k +ε k (1) in and These represent the measurement matrix and measurement noise matrix of the full polarization spectral imaging device, respectively; the specific representations of each parameter are as follows: Where A, B, and C are the modulation variables formed after modulation by polarization modulator (11), and the superscripts are [1,1], [2,1], ..., [N]. x N y ] represents the spatial coordinates of an image pixel; A=cos 2 2θ+sin 2 2θcosδ B=sin 2 (δ / 2)sin 4θ C=-sin2θsinδ Where θ is the angle between the fast axis of the fixed delay unit (111) and the xz plane, and δ is the delay amount of the fixed delay unit (111); Using nonlinear mapping functions of untrained artificial neural networks As a priori knowledge of the depth image, we solve the following ill-conditioned inverse problem: Where Θ k For parameters of an untrained artificial neural network, For Θ k The estimated value; the input is the polarization modulation image I under the k-th narrow band. k After multiple iterations and optimizations, the optimal result was obtained. Through optimal By performing mapping, we obtain the estimated Stokes parameter image for the k-th narrowband: Therefore, the Stokes parameter images for all narrow bands can be obtained.

2. The detection method according to claim 1, characterized in that, When the polarization information of the incident scene changes slowly with wavelength, it is assumed that the polarization information under a band unit composed of several adjacent narrow bands is consistent. In this case, the underdetermined problem of equation (5) is transformed into the well-posed problem as follows: Where u = 0, 1, ..., N s -1 indicates the order of narrowbands within the band cell. To ensure that the narrowbands at both ends of the entire narrowband array can form similar band cell combinations, a mirror-filling method is used at both ends to supplement the required number of narrowbands. In this case, the input data is N. s Polarization modulation image under a nearby narrow band After multiple iterations and optimizations, the optimal solution was obtained. Through optimal Perform a mapping to obtain these N s Estimated Stokes parameter images under neighboring narrow bands: By misaligning a single narrow band, the adjacent N bands are recombined s A narrow band, to obtain another set of N s The estimated Stokes parameter images under each narrow band are obtained, and so on. The estimated Stokes image under each narrow band is solved by N. s Then, by taking the average value of equation (9), the final Stokes parameter image for each narrow band is obtained:

3. The detection method according to claim 1, characterized in that, Based on the principle of compressed sensing, the underdetermined problem of equation (5) is transformed into the following optimization problem: Among them W k For an appropriate basis, superscripts (1), (2), ..., (N) are used. s ) represents N s W k The order; after multiple iterations of optimization, the optimal order is obtained. Through optimal By performing mapping, we obtain the estimated sparse image under the k-th narrowband: The estimated Stokes parameter image for the corresponding k-th narrowband is as follows: All narrow bands are sequentially processed through the above solution process, ultimately yielding estimated Stokes parameter images for all narrow bands.

4. The detection method according to claim 1, characterized in that, An orthogonal xyz coordinate system conforming to the right-hand rule is established, wherein the positive z-axis is the incident direction; the positive y-axis is vertically upward; the x-axis is perpendicular to the y-axis and z-axis; the optical axis of the fully polarized spectral imaging device is set horizontally, and the z-axis is parallel to the optical axis; the fast axis of the fixed retarder (111) is parallel to the xy plane, and the angle θ between it and the xz plane is arbitrarily taken between 0° and 180°, and θ cannot be 0°, 45°, 90°, 135° or 180°; the polarization direction of the fixed linear polarizer (112) is parallel to the x-axis or y-axis.

5. The detection method according to claim 1, characterized in that, A first fixed delay unit (113) is provided before the fixed delay unit (111) and the fixed linear polarizer (112), then: A=cosδ B=sinδsinδ1 C=-sinδcosδ1 Wherein, δ1 is the delay amount of the first fixed delay unit (113).

6. The detection method according to claim 5, characterized in that, An orthogonal xyz coordinate system conforming to the right-hand rule is set, wherein the positive z-axis is the incident direction; the positive y-axis is vertically upward; the x-axis is perpendicular to the y-axis and z-axis; the optical axis of the fully polarized spectral imaging device is set horizontally, and the z-axis is parallel to the optical axis; wherein the fast axis direction of the first fixed retarder (113) is parallel to the polarization direction of the fixed linear polarizer (112), the fast axis direction of the fixed retarder (111) is parallel to the xy plane, and the angle between it and the xz plane is 45°, and the polarization direction of the fixed linear polarizer (112) is parallel to the x-axis or y-axis.

7. The detection method according to claim 6, characterized in that, The first fixed delay unit (113) is a quarter-wave achromatic delay unit.

8. The detection method according to claim 1, 5, 6, or 7, characterized in that, The imaging spectrometer (12) is a snapshot imaging spectrometer or a scanning imaging spectrometer; the data acquisition, processing and display system (13) includes data acquisition, preprocessing, restoration and display function modules.

9. The detection method according to claim 1, 5, 6, or 7, characterized in that, The polarization modulator (11) is placed in the collimated optical path of the imaging spectrometer (12).

Images