Polarimetric error calibration method for deep learning based polarimetric optical detection system

Abstract

The application discloses a kind of based on deep learning's polarized optical detection system polarization measurement error calibration method, solve the problem that present technique cannot realize the polarization measurement error introduced by the polarization of polarized optical detection system instrument polarization and optical system imaging quality simultaneously calibration, including the following steps:1, obtain the modulation matrix of the polarized optical detection system to be calibrated;2, obtain clean intensity image matrix and degraded intensity image matrix;3, construct polarization image restoration network;4, train polarization image restoration network;5, obtain actual degraded intensity image matrix;6, input actual degraded intensity image matrix into polarization image restoration network, obtain predicted clean intensity image matrix;7, calculate polarization measurement error;8, obtain instrument polarization characteristic measurement error;9, calculate the polarization measurement error of the polarized optical detection system to be calibrated according to polarization measurement error and instrument polarization characteristic measurement error, complete calibration.

Description

Technical Field

[0001] This invention relates to a polarization measurement error calibration method, specifically to a polarization measurement error calibration method for a polarization optical detection system based on deep learning. Background Technology

[0002] Compared to traditional intensity imaging detection technology, polarization optical detection technology carries polarization information of light waves or light waves after their interaction with matter. This allows it to provide more complete light field information, including intensity and polarization, and to obtain higher-dimensional information about the target, such as the target's geometry, composition, and microstructure. It is widely used in fields such as industrial monitoring, medical diagnosis, and astronomical observation.

[0003] According to the principles and methods of polarization optical detection, a polarization detection optical system usually consists of a polarization modulation subsystem and an imaging subsystem. When performing polarization detection on a target, the polarization measurement results will be introduced due to the differences between the polarization characteristics of the instrument and the nominal polarization value of the instrument, and the differences between the imaging quality of the optical system and the ideal imaging quality.

[0004] To improve polarization measurement accuracy, polarization measurement error calibration is required before a polarization detection optical system can perform effective detection. Besides building a polarization calibration system to calibrate the polarization detection optical system, some polarization detection optical systems are equipped with internal polarization calibration units to meet high-precision detection requirements. The polarization calibration units vary depending on the polarization calibration method. For example, in the commonly used four-point polarization calibration method, the polarization calibration unit consists of a linear polarizer and a phase delay element (such as a λ / 4 waveplate). Existing calibration methods mostly focus on the instrument polarization error of the polarization detection optical system, lacking techniques for simultaneously calibrating the imaging quality of the polarization detection optical system. Furthermore, to simultaneously calibrate errors caused by instrument polarization and imaging quality, the corresponding polarization calibration system would have a complex structure and cumbersome operation process, severely hindering the development of high-precision polarization measurement technology, especially high-precision spatial polarization measurement technology. Summary of the Invention

[0005] To address the technical challenges of simultaneously calibrating polarization measurement errors introduced by the polarization of instruments and the imaging quality of optical systems in polarization optical detection systems, and the fact that existing polarization calibration systems are complex in structure and cumbersome in operation, severely restricting further improvements in the polarization measurement accuracy of polarization detection optical systems, this invention provides a deep learning-based method for calibrating polarization measurement errors in polarization optical detection systems.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A deep learning-based method for calibrating polarization measurement errors in a polarization optical detection system, characterized by the following steps:

[0008] Step 1: Obtain the modulation matrix of the polarization optical detection system to be calibrated ;

[0009] Step 2: Obtain the clean intensity image matrix P and degradation intensity image matrix Q ;

[0010] Step 3: Construct a polarization image restoration network based on the UNet model;

[0011] Step 4: Using the degradation intensity image matrix Q The input data is an image matrix that is infinitely close to the clean intensity. P Clean intensity image matrix P t To output the data, train the polarization image restoration network;

[0012] Step 5: Obtain the actual degradation intensity image matrix of the polarization optical detection system to be calibrated. Q r ;

[0013] Step 6: Convert the actual degradation intensity image matrix Q r Input the trained polarization image restoration network to obtain the predicted clean intensity image matrix. P r ;

[0014] Step 7: Based on the modulation matrix Actual degradation intensity image matrix Q r and predict clean intensity image matrix P r Calculate the polarization measurement error introduced by the polarization optical detection system to be calibrated. ;

[0015] Step 8: Obtain the measurement error of the instrument polarization characteristics of the polarization optical detection system to be calibrated. ;

[0016] Step 9: Based on the polarization measurement error Measurement error of instrument polarization characteristics Calculate the polarization measurement error of the polarization optical detection system to be calibrated The calibration was completed.

[0017] Furthermore, step 2 specifically includes:

[0018] 2.1 Generate a clean intensity image matrix with polarization informationP The polarization information is described using Stokes vector parameters.

[0019] 2.2 Generating Zernike polynomials characterizing imaging quality Z ;

[0020] 2.3. The clean intensity image matrix P and Zernike polynomials Z Perform convolution to obtain the corresponding degradation intensity image matrix. Q .

[0021] Further, in step 3, the polarization image restoration network includes a first residual convolutional block to an eighth residual convolutional block connected sequentially from the input end to the output end; simultaneously, another output end of the first residual convolutional block is also connected to another input end of the eighth residual convolutional block; another output end of the second residual convolutional block is also connected to another input end of the seventh residual convolutional block; another output end of the third residual convolutional block is also connected to another input end of the sixth residual convolutional block; wherein the first residual convolutional block to the fourth residual convolutional block are encoders, and the fifth residual convolutional block to the eighth residual convolutional block are decoders.

[0022] Furthermore, in step 3, the output images E1, E2, and E3 from the first residual convolution block to the third residual convolution block have the following dimensions: the width and height of output image E1 are twice that of output image E2, and the width and height of output image E2 are twice that of output image E3.

[0023] Furthermore, step 4 specifically includes:

[0024] 4.1 Construct aberration guidance maps C1, C2, and C3 with the same size as the output images E1, E2, and E3 of the first to third residual convolutional blocks;

[0025] 4.2. Image matrix with degradation intensity Q The input data is then used, and the aberration-guided images C1, C2, and C3 are added to the output images E1, E2, and E3 of the first to third residual convolutional blocks, respectively, to approximate the clean intensity image matrix. P Clean intensity image matrix P t To output the data, the polarization image restoration network is trained using the root mean square function, the multi-scale structural loss function, and / or the total variational function as loss functions until the loss functions converge, thus completing the network training.

[0026] Furthermore, step 4.1 specifically includes:

[0027] Aberration guide maps C1, C2, and C3 with the same size as the output images E1, E2, and E3 of the first to third residual convolutional blocks are constructed using the aberration distribution principle.

[0028] Furthermore, in step 4.1, the principle of aberration distribution is as follows:

[0029]

[0030] in, For aberrations, derived from the Zernike polynomial in step 2.2 Z Characterization, For aberration coefficients, , and for , and Powers of, taking non-negative integers. Normalized image height represents different fields of view. and This indicates the position of the ray at the exit pupil; when using Cartesian coordinates to represent the coordinates of the exit pupil plane, the center of the exit pupil is located at the origin of the corresponding coordinate system. This represents the distance between the position of the corresponding ray at the exit pupil and the center of the exit pupil. For the corresponding light position and The included angle of the axis; , , and corresponding The selection of [aspect] is used to construct different types of aberrations. , The selection of aberrations is used to construct aberrations for different fields of view. .

[0031] Furthermore, step 7 specifically includes:

[0032] 7.1 Calculate the predicted clean intensity image matrix P r Corresponding polarization information :

[0033]

[0034] 7.2 Calculate the actual degradation intensity image matrix Q r Corresponding polarization information :

[0035]

[0036] 7.3. Polarization information With polarization information By performing a difference operation, the polarization measurement error introduced by the polarization optical detection system to be calibrated is obtained. :

[0037] .

[0038] Furthermore, step 8 specifically includes:

[0039] If the polarization optical detection system to be calibrated includes a polarization calibration unit, then the measurement error of the instrument's polarization characteristics can be obtained through this polarization calibration unit. ;

[0040] If the polarization optical detection system to be calibrated does not include a polarization calibration unit, then the measurement error of the instrument's polarization characteristics is obtained through an external polarization calibration system. .

[0041] Furthermore, step 9 specifically includes:

[0042] According to polarization measurement error Measurement error of instrument polarization characteristics Calculate the polarization measurement error of the polarization optical detection system to be calibrated :

[0043]

[0044] Calibration completed.

[0045] The beneficial effects of this invention are:

[0046] 1. The present invention provides a polarization measurement error calibration method for a polarization optical detection system based on deep learning. By constructing a polarization image restoration network with the UNet model as the basic architecture, the polarization image restoration network obtains a predicted clean intensity image matrix without aberrations. Combined with the measurement error of the instrument polarization characteristics of the polarization optical detection system to be calibrated, the calibration accuracy of the polarization detection optical system can be effectively improved, thereby improving the polarization measurement accuracy of the polarization detection optical system.

[0047] 2. This invention provides a deep learning-based method for calibrating polarization measurement errors in a polarization optical detection system. This method utilizes a clean intensity image matrix within a Zernike polynomial. Z Degradation intensity image matrix under the influence of aberrations Q Aberration guidance maps are constructed to train the constructed polarization image restoration network. This enables the trained polarization image restoration network to obtain a clean intensity image matrix with no aberration effects. It can decouple the effect of imaging quality of polarization optical system on polarization measurement accuracy and effectively improve the polarization measurement accuracy of polarization detection optical system.

[0048] 3. The present invention provides a polarization measurement error calibration method for a polarization optical detection system based on deep learning. The method uses the root mean square function, the multi-scale structure loss function and / or the total variational function as loss functions to train the polarization image restoration network. It can adaptively select an appropriate loss function according to the complexity of the polarization image information, thereby improving the prediction accuracy of the polarization image restoration network.

[0049] 4. The present invention provides a polarization measurement error calibration method for a polarization optical detection system based on deep learning. It constructs aberration guide maps C1, C2, and C3 with the same size as the output images E1, E2, and E3 of the first to third residual convolutional blocks using the aberration distribution principle. When training the network, the aberration guide maps C1, C2, and C3 are used as inputs simultaneously, which can further improve the prediction accuracy of the high polarization image restoration network.

[0050] 5. The present invention provides a polarization measurement error calibration method for a polarization optical detection system based on deep learning. Compared with the prior art, it does not require the setting of an additional polarization calibration system, which makes the structure of the polarization optical detection system simpler and more compact. It provides a new method for high-precision on-orbit calibration of space polarization optical detection systems with strict requirements for structural weight and other envelope requirements, and can effectively promote the development of space detection technology based on polarization optics. Attached Figure Description

[0051] Figure 1 This is a flowchart of an embodiment of a polarization measurement error calibration method for a polarization optical detection system based on deep learning according to the present invention;

[0052] Figure 2 This is a schematic diagram of the polarization image restoration network structure constructed in step 3 of the present invention.

[0053] Figure label:

[0054] 1-First residual convolution block, 2-Second residual convolution block, 3-Third residual convolution block, 4-Fourth residual convolution block, 5-Fifth residual convolution block, 6-Sixth residual convolution block, 7-Seventh residual convolution block, 8-Eighth residual convolution block. Detailed Implementation

[0055] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings and embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] This invention provides a deep learning-based method for calibrating polarization measurement errors in a polarization optical detection system, such as... Figure 1As shown, the method includes the following steps:

[0057] Step 1: Obtain the modulation matrix of the polarization optical detection system to be calibrated ;

[0058] The modulation matrix of the uncalibrated polarization optical detection system is determined based on its detection or modulation method. ;

[0059] A polarization optical detection system to be calibrated typically consists of a polarization modulation unit and an imaging optical system. Some polarization optical detection systems also have a polarization calibration unit for calibrating the polarization measurement error of the instrument.

[0060] Step 2: Obtain a clean intensity image matrix with polarization information. P and degradation intensity image matrix Q ;

[0061] 2.1 Generate a clean intensity image matrix with polarization information P This polarization information is described using Stokes vector parameters.

[0062] 2.2 Generating Zernike polynomials characterizing imaging quality Z ;

[0063] 2.3. The clean intensity image matrix P and Zernike polynomials Z Perform convolution to obtain the corresponding degradation intensity image matrix. Q The degradation intensity image matrix Q Contains polarization information;

[0064] Step 3: Construct a polarization image restoration network based on the UNet model;

[0065] like Figure 2 As shown, the polarization image restoration network includes a first residual convolutional block 1 to an eighth residual convolutional block 8 connected sequentially from the input end to the output end; simultaneously, another output end of the first residual convolutional block 1 is also connected to another input end of the eighth residual convolutional block 8; another output end of the second residual convolutional block 2 is also connected to another input end of the seventh residual convolutional block 7; another output end of the third residual convolutional block 3 is also connected to another input end of the sixth residual convolutional block 6; wherein the first residual convolutional block 1 to the fourth residual convolutional block 4 are encoders, and the fifth residual convolutional block 5 to the eighth residual convolutional block 8 are decoders; the output images E1, E2, and E3 of the first residual convolutional block 1 to the third residual convolutional block 3 have the following dimensions: the width and height of the output image E1 are twice that of the output image E2, and the width and height of the output image E2 are twice that of the output image E3;

[0066] Step 4: Train the polarization image restoration network;

[0067] 4.1 Construct aberration guide maps C1, C2, and C3 with the same size as the output images E1, E2, and E3 of the first residual convolutional block 1 to the third residual convolutional block 3 using the aberration distribution principle;

[0068] The principle of aberration distribution is as follows: ;in, For aberrations, derived from the Zernike polynomial in step 2.2 Z Characterization, For aberration coefficients, , and for , and Powers of, taking non-negative integers. Normalized image height represents different fields of view. and This indicates the position of the ray at the exit pupil; when using Cartesian coordinates to represent the coordinates of the exit pupil plane, the center of the exit pupil is located at the origin of the corresponding coordinate system. This represents the distance between the position of the corresponding ray at the exit pupil and the center of the exit pupil. For the corresponding light position and The included angle of the axis;

[0069] By selecting different , , and corresponding Constructing different types of aberrations By selecting different normalized image heights Constructing aberrations in different fields of view Thus, the aberration guidance map is constructed.

[0070] 4.2. Image matrix with degradation intensity Q The input data is then used, and the aberration-guided images C1, C2, and C3 are added to the output images E1, E2, and E3 of the first to third residual convolutional blocks, respectively, to approximate the clean intensity image matrix. P Clean intensity image matrix P t To output the data, the polarization image restoration network is trained using the root mean square function, the multi-scale structure loss function, and / or the total variation function as loss functions until the loss functions converge, thus completing the network training.

[0071] That is: input the degradation intensity image matrix into the polarization image restoration network. Q Then its output is a clean intensity image matrix. P tThe clean polarization image matrix obtained by the evaluation network is evaluated. P t With clean intensity image matrix P The similarity is measured until the loss function converges, thus completing network training.

[0072] The root mean square function, multi-scale structural loss function, and / or total variational function are adaptively selected based on the complexity of the polarization image information in order to improve the prediction accuracy of the polarization image restoration network.

[0073] Step 5: Obtain the actual degradation intensity image matrix of the polarization optical detection system to be calibrated. Q r ;

[0074] Step 6: Convert the actual degradation intensity image matrix Q r Input the trained polarization image restoration network to obtain the predicted clean intensity image matrix. P r ;

[0075] Step 7: Based on the modulation matrix Actual degradation intensity image matrix Q r and predict clean intensity image matrix P r Calculate the polarization measurement error introduced by the polarization optical detection system to be calibrated. ;

[0076] 7.1 Calculate the predicted clean intensity image matrix P r Corresponding polarization information That is, the polarization measurement results :

[0077]

[0078] 7.2 Calculate the actual degradation intensity image matrix Q r Corresponding polarization information That is, the polarization measurement results :

[0079]

[0080] 7.3. Polarization measurement results With polarization measurement results By performing a difference operation, the polarization measurement error introduced by the polarization optical detection system to be calibrated is obtained. :

[0081] ;

[0082] Step 8: Obtain the measurement error of the instrument polarization characteristics of the polarization optical detection system to be calibrated. ;

[0083] If the polarization optical detection system to be calibrated includes a polarization calibration unit, then the measurement error of the instrument's polarization characteristics can be obtained through this polarization calibration unit. ;

[0084] If the polarization optical detection system to be calibrated does not include a polarization calibration unit, then the measurement error of the instrument's polarization characteristics is obtained through an external polarization calibration system. ;

[0085] Step 9: Based on the polarization measurement error Measurement error of instrument polarization characteristics Calculate the polarization measurement error of the polarization optical detection system to be calibrated :

[0086]

[0087] Calibration completed.

[0088] This method can effectively achieve high-precision calibration of system polarization measurement errors caused by optical imaging quality and instrument polarization in polarization optical detection systems. It can provide a new method for high-precision on-orbit calibration of space polarization optical detection systems and provide technical support for further improving the polarization measurement accuracy of polarization detection optical systems.

[0089] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions within the technical scope disclosed in the present invention should be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for calibrating polarization measurement errors in a polarization optical detection system based on deep learning, characterized in that, Includes the following steps: Step 1: Obtain the modulation matrix M of the polarization optical detection system to be calibrated. T ; Step 2: Obtain the clean intensity image matrix P and the degraded intensity image matrix Q; the clean intensity image matrix P contains polarization information, which is described by Stokes vector parameters; Step 3: Construct a polarization image restoration network based on the UNet model; Step 4: Using the degraded intensity image matrix Q as input data, and a clean intensity image matrix P that is infinitely close to the clean intensity image matrix P. t To output data, a polarization image restoration network is trained; specifically, the degraded intensity image matrix Q is input into the polarization image restoration network, and its output is a clean intensity image matrix P. t The clean polarization image matrix P obtained by the polarization image restoration network is evaluated. t The similarity between the image and the clean intensity image matrix P is calculated until the loss function converges, thus completing the training of the polarization image restoration network. Step 5: Obtain the actual degradation intensity image matrix Q of the polarization optical detection system to be calibrated. r ; Step 6: Convert the actual degradation intensity image matrix Q r Inputting the trained polarization image restoration network yields the predicted clean intensity image matrix P. r ; Step 7: Based on the modulation matrix M T Actual degradation intensity image matrix Q r and the predicted clean intensity image matrix P r Calculate the polarization measurement error ΔS introduced by the polarization optical detection system to be calibrated. s ; Step 8: Obtain the instrument polarization characteristic measurement error ΔS of the polarization optical detection system to be calibrated. P ; Step 9: Based on the polarization measurement error ΔS s And the measurement error ΔS of the instrument polarization characteristics P Calculate the polarization measurement error ΔS of the polarization optical detection system to be calibrated, and complete the calibration.

2. The polarization measurement error calibration method for a deep learning-based polarization optical detection system according to claim 1, characterized in that, Step 2 specifically includes: 2.1 Generating a clean intensity image matrix P with polarization information; the polarization information is described by Stokes vector parameters; 2.2 Generate the Zernike polynomial Z that characterizes the imaging quality; 2.3 Convolve the clean intensity image matrix P and the Zernike polynomial Z to obtain the corresponding degraded intensity image matrix Q.

3. The polarization measurement error calibration method for a deep learning-based polarization optical detection system according to claim 2, characterized in that: In step 3, the polarization image restoration network includes a first residual convolutional block to an eighth residual convolutional block connected sequentially from the input end to the output end; simultaneously, another output end of the first residual convolutional block is also connected to another input end of the eighth residual convolutional block; another output end of the second residual convolutional block is also connected to another input end of the seventh residual convolutional block; another output end of the third residual convolutional block is also connected to another input end of the sixth residual convolutional block; wherein the first residual convolutional block to the fourth residual convolutional block are encoders, and the fifth residual convolutional block to the eighth residual convolutional block are decoders.

4. The polarization measurement error calibration method for a deep learning-based polarization optical detection system according to claim 3, characterized in that: In step 3, the output images E1, E2, and E3 from the first residual convolution block to the third residual convolution block have the following dimensions: the width and height of output image E1 are twice that of output image E2, and the width and height of output image E2 are twice that of output image E3.

5. The polarization measurement error calibration method for a deep learning-based polarization optical detection system according to claim 4, characterized in that, Step 4 specifically includes: 4.1 Construct aberration guidance maps C1, C2, and C3 with the same size as the output images E1, E2, and E3 of the first to third residual convolutional blocks; 4.

2. Using the degraded intensity image matrix Q as input data, and adding the aberration-guided images C1, C2, and C3 to the output images E1, E2, and E3 of the first to third residual convolutional blocks respectively, to obtain a clean intensity image matrix P that is infinitely close to the clean intensity image matrix P. t To output the data, the polarization image restoration network is trained using the root mean square function, the multi-scale structural loss function, and / or the total variational function as loss functions until the loss functions converge, thus completing the network training.

6. The polarization measurement error calibration method for a deep learning-based polarization optical detection system according to claim 5, characterized in that, Step 4.1 specifically involves: Aberration guide maps C1, C2, and C3 with the same size as the output images E1, E2, and E3 of the first to third residual convolutional blocks are constructed using the aberration distribution principle.

7. The polarization measurement error calibration method for a deep learning-based polarization optical detection system according to claim 6, characterized in that, In step 4.1, the principle of aberration distribution is as follows: W=∑W i,j,k H i r j cos k f Where W is the aberration, characterized by the Zernike polynomial Z in step 2.2, W i,j,k Let be the aberration coefficients, i, j, and k be powers of H, ρ, and φ, respectively, taking non-negative integers. H is the normalized image height, representing different fields of view. ρ and φ represent the positions of the light rays at the exit pupil. When using Cartesian coordinates to represent the exit pupil plane coordinates, the exit pupil center is located at the origin of the corresponding coordinate system. ρ represents the distance between the corresponding ray position at the system's exit pupil and the exit pupil center, and φ is the angle between the corresponding ray position and the y-axis. i, j, k, and the corresponding W... i,j,k The selection of is used to construct different types of aberrations W, and the selection of H is used to construct different fields of view aberrations W.

8. The polarization measurement error calibration method for a deep learning-based polarization optical detection system according to claim 7, characterized in that, Step 7 specifically includes: 7.1 Calculate the predicted clean intensity image matrix P r Corresponding polarization information S PR : 7.2 Calculate the actual degradation intensity image matrix Q r Corresponding polarization information S QR : 7.

3. Regarding polarization information S PR With polarization information S QR By performing a difference operation, the polarization measurement error ΔS introduced by the polarization optical detection system to be calibrated is obtained. s : ΔS s =S PR -S QR 。 9. The polarization measurement error calibration method for a deep learning-based polarization optical detection system according to claim 8, characterized in that, Step 8 specifically includes: If the polarization optical detection system to be calibrated includes a polarization calibration unit, then the measurement error ΔS of the instrument's polarization characteristics can be obtained through this polarization calibration unit. P ; If the polarization optical detection system to be calibrated does not include a polarization calibration unit, then the instrument polarization characteristic measurement error ΔS is obtained through an external polarization calibration system. P .

10. The polarization measurement error calibration method for a deep learning-based polarization optical detection system according to claim 9, characterized in that, Step 9 specifically includes: According to the polarization measurement error ΔS s And the measurement error ΔS of the instrument polarization characteristics P Calculate the polarization measurement error ΔS of the polarization optical detection system to be calibrated: ΔS=ΔS S -ΔS P Calibration completed.

Images