Polarization method, unit, equipment and medium based on polarization-infrared multi-mode fusion

Abstract

The invention relates to the technical field of optical imaging and sensing, in particular to a polarization method, unit, equipment and medium based on polarization-infrared multi-mode fusion, and the polarization unit is composed of an optical window, a polarization modulation layer, a detection layer and a processing module; all the modules work cooperatively, and the whole process from optical signal collection to data processing and image output is completed. The environmental adaptability is high, multi-scene coverage and stable work are realized, and the existing single-mode sensor is obviously limited by the environment. According to the invention, through dynamic compensation correction (5 * 5 sliding window real-time monitoring residual) and weight adaptive adjustment, a clear image can be stably output in complex environments such as night, foggy days, strong light and the like, and the environmental dependence of a single mode is overcome.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of optical imaging and sensing technology, and in particular to a polarization method, unit, device and medium based on polarization-infrared multi-modal fusion, which is suitable for target detection, identification and enhanced imaging in complex environments, and has important application value in the field of power transmission, etc. The polarization unit integrates the advantages of polarization imaging and infrared imaging, and can efficiently obtain rich information of the target in various complex environments, providing strong support for accurate detection and identification. BACKGROUND

[0002] In the field of optical imaging and sensing, traditional infrared imaging technology is widely used due to its ability to detect targets at night and in harsh environments, but it has obvious defects, is easily disturbed by background radiation, and is difficult to effectively distinguish the material characteristics of the target, which greatly hinders the fine identification of the target. The polarization imaging technology can obtain the polarization characteristics of the target surface, which helps to distinguish different materials, but its dependence on environmental light is relatively strong, and its performance will decrease significantly in poor lighting conditions.

[0003] Due to the limitations of its own principle, the existing single modal sensor cannot balance high sensitivity and high signal-to-noise ratio, and cannot meet the high-precision requirements of target detection in complex environments. The traditional weighted fusion algorithm does not consider the Mie scattering difference between the wire and the fog droplet, so that the SNR (signal-to-noise ratio) after fusion does not increase but decreases; the time synchronization error of the polarization sensor and the infrared detector is greater than 10 ms, causing the image of the moving target (such as a dancing wire) to be misaligned. Poor data synchronization will cause the information after fusion to deviate, affecting the accuracy of target identification; the complex structure makes the device bulky, which is not conducive to application in small and portable devices. Therefore, a compact polarization-infrared multi-modal fusion unit is needed to realize efficient data collaboration and real-time processing, and to meet the needs of target detection in complex environments in various fields. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides a polarization method, unit, device and medium based on polarization-infrared multi-modal fusion, which realizes synchronous acquisition and processing of high-precision multi-modal information by integrating an infrared detector and a polarization-sensitive element, and combining an adaptive fusion algorithm, solves the problems of poor data synchronization, complex structure, and difficulty in balancing high sensitivity and high signal-to-noise ratio in the prior art, and meets the needs of target detection, identification and enhanced imaging in complex environments.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In a first aspect, the present application provides a polarization method based on polarization-infrared multi-modal fusion, comprising:

[0008] Non-uniformity correction is performed on the infrared image and the polarization image to compensate for gain difference and offset between pixels;

[0009] Weight distribution is optimized based on Mie scattering characteristics and dynamic signal-to-noise ratio.

[0010] A morphological filtering algorithm is designed by using an improved closed-open operation of a multi-scale adaptive structure element to eliminate edge artifacts in the fusion image.

[0011] The preferred technical solution of the embodiments of the present application has the following beneficial effects: the inherent defects of a single modality are broken through, the limitations of infrared imaging are solved, traditional infrared imaging is easily disturbed by background thermal radiation, and it is difficult to distinguish materials. The present patent can enhance the material identification capability by fusing polarization information. The deficiencies of polarization imaging are made up, and the polarization imaging relies on ambient light and has poor performance at night. The present application can realize all-weather work. The data quality is significantly improved, and the signal-to-noise ratio (SNR) is optimized by the embodiments of the present application, that is, the Mie scattering correction coefficient (k) and the dynamic weight distribution. In complex scenes such as foggy days, the SNR of the fusion image is improved by more than 30% compared with the traditional method, and the target contrast is improved by 2 times.

[0012] As a preferred scheme of the polarization method based on polarization-infrared multi-modal fusion, wherein the non-uniformity correction of the infrared image comprises:

[0013] The original images I1(x,y) and I2(x,y) of the uniform blackbody radiation surface at temperatures T1 and T2 are collected;

[0014] The gain G(x,y) and the offset O(x,y) of each pixel are calculated:

[0015]

[0016] O(x,y)=I1(x,y)-G(x,y)·T1

[0017] Real-time correction:

[0018]

[0019] wherein I raw (x,y) represents the uncorrected original image output by the infrared detector in real time, and I corr (x,y) represents the infrared image after non-uniformity correction.

[0020] As a preferred scheme of the polarization method based on polarization-infrared multi-modal fusion, the non-uniformity correction on the polarization image comprises:

[0021] Under the condition of no light, a dark image D(x,y) of the polarization detector is collected, and is expressed as:

[0022] S dark (x,y)=S raw (x,y)-D(x,y)

[0023] By using the polarization uniformity of the sky background in the electric power scene, a background region R without conductors in the image is selected, the average response values u0, u 45 , u 90 , u 135 of each polarization direction sub-region in the region are calculated, and the normalization coefficients of each direction are calculated with u0 as the reference, and are expressed as:

[0024]

[0025] The gain correction is performed on each direction polarization image, and is expressed as:

[0026] S' 45 (x,y)=S dark,45 (x,y)·k 45

[0027] S' 90 (x,y)=S dark,90 (x,y)·k 90

[0028] S' 135 (x,y)=S dark,135 (x,y)·k 135

[0029] After each frame of image is collected, the correction residual is monitored in real time through a sliding window, and when the local region standard deviation is greater than 5%, local secondary correction is triggered, so that the correction accuracy of the strong polarization region such as the conductor edge is ensured.

[0030] As a preferred scheme of the polarization method based on polarization-infrared multi-modal fusion, the weight distribution is optimized based on the Mie scattering characteristics and dynamic signal-to-noise ratio, and is expressed as:

[0031]

[0032] Wherein, W IR is the fusion weight of the infrared channel (the value range is 0-1), which determines the contribution proportion of the infrared image in the final fusion result; SNR IR is the signal-to-noise ratio of the infrared channel; SNR PolSNR is the signal-to-noise ratio of the polarization channel; DoP is the degree of polarization; a is a dynamic adjustment factor, which controls the steepness of the weight distribution curve; k is the modified coefficient of Mie scattering:

[0033]

[0034] As a preferred scheme of the polarization method based on polarization-infrared multi-modal fusion, the improved closed-open operation with a multi-scale adaptive structure element is used to design a morphological filtering algorithm, which eliminates edge artifacts in the fusion image through a morphological filtering algorithm, including:

[0035] The gradient amplitude of the fusion image F(x, y) is calculated and expressed as:

[0036]

[0037] The suspected artifact region is extracted, including:

[0038] The gradient direction and the difference between the infrared / polarization original image are greater than 15°;

[0039] The local gradient amplitude mutation rate

[0040] According to the characteristics of the artifact region, a structure element B is dynamically constructed:

[0041]

[0042] The mixed morphological operation is expressed as:

[0043] Fill in the small holes caused by registration errors;

[0044] Detail protection open operation:

[0045]

[0046] Where B' is the erosion version of B;

[0047] Weighted reconstruction:

[0048] F final (x, y) = w·F close +(1-w)·F open

[0049] The weight coefficient w is determined by the artifact confidence:

[0050]

[0051] Where p represents the artifact probability.

[0052] Secondly, the present invention provides a polarization unit based on polarization-infrared multimodal fusion, comprising:

[0053] Optical window, polarization modulation layer, detection layer, processing module.

[0054] The modules work together to complete the entire process from optical signal acquisition to data processing and image output.

[0055] As a preferred embodiment of the polarization unit based on polarization-infrared multimodal fusion described in this invention, wherein:

[0056] The optical window uses a ZnSe substrate and is coated with an anti-reflective film, exhibiting excellent transmittance in the 8-12μm band, with a transmittance >95%. ZnSe material possesses good infrared optical properties, which can effectively reduce the reflection loss of light signals when entering the system, ensuring that the light signal of sufficient intensity enters the subsequent detection module.

[0057] The polarization modulation layer integrates subwavelength aluminum gratings in four directions: 0°, 45°, 90°, and 135°. These aluminum gratings have a period of 200 nm and a height of 300 nm, and are fabricated using precise micro-nano fabrication processes. The subwavelength grating structure allows for selective modulation of light with different polarization directions. When target radiation passes through this layer, light with different polarization directions interacts with the corresponding gratings, thereby enabling precise detection of the target's polarization characteristics.

[0058] The detector layer uses a mercury cadmium telluride (HgCdTe) infrared focal plane array with a pixel size of 640×512 and extremely low noise equivalent temperature difference (NETD < 30mK). Mercury cadmium telluride material has significant advantages in the field of infrared detection, as it is sensitive to infrared radiation and can accurately detect the infrared radiation signal of the target.

[0059] Processing Module: Integrates a GPU-accelerated real-time fusion processor. The GPU (Graphics Processing Unit) possesses powerful parallel computing capabilities, enabling rapid processing of large amounts of data from the polarization modulation and detection layers. The real-time fusion processor employs advanced algorithms to efficiently fuse infrared and polarization data, achieving real-time analysis and processing of multimodal data through optimized data processing flows.

[0060] Thirdly, the present invention provides an electronic device, comprising:

[0061] Memory and processor;

[0062] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of a polarization method based on polarization-infrared multimodal fusion.

[0063] In a fourth aspect, the present application provides a computer readable storage medium storing computer executable instructions, which, when executed by a processor, implement the steps of the polarization method based on polarization-infrared multi-modal fusion.

[0064] Compared with the prior art, the present application has the following advantages: the data synchronization is significantly improved, the motion target misplacement problem is solved, the time synchronization error of the polarization and infrared sensors in the prior art is >10 ms, which leads to misplacement of the image of the dancing conductor and other moving targets, and seriously affects the fusion accuracy. The present application controls the time synchronization error to be <1 ms through the hardware time synchronization and motion compensation algorithm, and the spatial alignment accuracy is <0.5 pixels, which ensures the spatial consistency of the fusion information of the dynamic target and provides a basis for the accurate identification of the moving target.

[0065] The signal-to-noise ratio is greatly optimized, and the target recognition degree is significantly enhanced in a complex environment. The traditional weighted fusion algorithm does not consider the characteristic difference between the conductor and the fog droplet, which leads to a decrease in the SNR after fusion. The present application proposes an adaptive weight distribution algorithm based on Mie scattering characteristics and dynamic signal-to-noise ratio, which dynamically adjusts the weight through the Mie scattering correction factor K and the degree of polarization DoP. In complex environments such as foggy weather, the SNR of the fusion image is improved by >30% compared with the traditional algorithm, and the discrimination between the target and the background is significantly enhanced, solving the inherent defects of single modal imaging, such as "infrared is easily disturbed by background radiation, polarization depends on environmental light".

[0066] The environmental adaptability is strong, and it can work stably in multiple scenes. The single modal sensor in the prior art is obviously limited by the environment. The present application can stably output clear images in complex environments such as night, foggy weather, and strong light through dynamic compensation correction (5x5 sliding window real-time monitoring of residual error) and adaptive adjustment of weight, overcoming the environmental dependence of single modal. BRIEF DESCRIPTION OF DRAWINGS

[0067] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0068] Figure 1 The overall flowchart of a polarization unit based on polarization-infrared multi-modal fusion according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0069] In order to make the above objectives, characteristics and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application are described in detail below with reference to the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor should belong to the protection scope of the present application.

[0070] Embodiment 1, reference Figure 1 For an embodiment of the present application, a polarization unit based on polarization-infrared multi-modal fusion is provided, comprising:

[0071] The polarization unit is composed of an optical window, a polarization modulation layer, a detection layer and a processing module (as shown in Figure 1 Each module works cooperatively to complete the whole process from light signal collection to data processing and image output.

[0072] Optical window: ZnSe substrate is adopted and anti-reflection film is coated, which has excellent transmission performance in the 8-12 μm wave band with a transmission rate > 95%. ZnSe material has good infrared optical characteristics, which can effectively reduce the reflection loss of light signal when entering the system, and ensure that enough strong light signal enters the subsequent detection module. The anti-reflection film coated further optimizes the performance of the optical window, so that more target radiation can smoothly pass through, laying a foundation for accurate detection of the target.

[0073] Polarization modulation layer: this layer integrates 0°, 45°, 90° and 135° four-direction sub-wavelength aluminum gratings. The period of these aluminum gratings is 200 nm, and the height is 300 nm, which is prepared by accurate micro-nano processing technology. The structure design of sub-wavelength grating can selectively modulate light of different polarization directions. When the target radiation passes through this layer, light of different polarization directions will interact with the corresponding grating, so as to realize accurate detection of the polarization characteristics of the target. This multi-directional polarization modulation design enables the unit to comprehensively obtain the information of the target in each polarization direction, providing rich data sources for subsequent multi-modal fusion.

[0074] Detection layer: HgCdTe infrared focal plane array is selected, which has a pixel specification of 640×512 and has an extremely low noise equivalent temperature difference (NETD < 30 mK). HgCdTe material has significant advantages in the field of infrared detection, is sensitive to infrared radiation, and can accurately detect the infrared radiation signal of the target. The high pixel resolution of 640×512 enables the array to clearly capture the detailed information of the target, and the extremely low NETD (noise equivalent temperature difference) ensures that even in complex environments, the target and the background have a small temperature difference, the thermal radiation characteristics of the target can be accurately detected, providing reliable data support for target recognition.

[0075] Processing module: real-time fusion processor integrated with GPU acceleration. GPU (Graphics Processing Unit) has strong parallel computing capability and can quickly process a large amount of data from the polarization modulation layer and the detection layer. The real-time fusion processor uses advanced algorithms to efficiently fuse infrared data and polarization data, and realizes real-time analysis and processing of multi-modal data by optimizing the data processing flow. This real-time fusion can quickly output enhanced images, highlighting the material differences and thermal radiation characteristics of the target, meeting the demand for rapid and accurate detection of targets in complex environments, and has important significance in military reconnaissance, autonomous driving and other applications with high real-time requirements.

[0076] Embodiment 2, which is an embodiment of the present application, based on the above embodiment, provides a polarization method based on polarization-infrared multi-modal fusion. The adaptive fusion algorithm implanted in the processing module includes the following steps:

[0077] Step 1: Non-uniformity correction of infrared image I(x,y) and polarization image S(x,y). Non-uniformity correction aims to eliminate the image deviation caused by the inconsistent response of each pixel in the detector array of the imaging system, and its core is to compensate the gain difference and offset between pixels through algorithm. For the infrared image I(x,y) and the polarization image S(x,y) in the present application.

[0078] It should be noted that the non-uniformity correction of the infrared image in step 1:

[0079] Collect the original images I1(x,y) and I2(x,y) of the uniform blackbody radiation surface at temperatures T1 (such as 20℃) and T2 (such as 40℃);

[0080] Calculate the gain G(x,y) and the bias O(x,y) of each pixel:

[0081]

[0082] O(x,y)=I1(x,y)-G(x,y)·T1

[0083] Real-time correction:

[0084] Further, the polarization channel correction in step 1: According to the characteristics of the sub-wavelength aluminum grating array of the polarization modulation layer, the scene adaptive correction method is adopted:

[0085] Dark current correction: In the absence of light, collect the dark image D(x,y) of the polarization detector to eliminate the dark current offset of each sub-region:

[0086] S dark (x,y)=S raw(x, y)-D(x, y)

[0087] Gain normalization: With the polarization uniformity of sky background in power scene (polarization degree <0.1), select the background area R without wires in the image, calculate the average response value u0, u45, u90, u135 of each polarization direction sub-area in the area, and take u0 as the reference, calculate the normalization coefficient of each direction:

[0088]

[0089] Gain correction is performed on each direction polarization image:

[0090] S' 45 (x, y) = S dark,45 (x, y)·k 45

[0091] S' 90 (x, y) = S dark,90 (x, y)·k 90

[0092] S' 135 (x, y) = S dark,135 (x, y)·k 135

[0093] Dynamic compensation: After each frame of image is collected, the correction residual is monitored in real time through a sliding window (5x5 pixels), and when the local area standard deviation is >5%, local secondary correction is triggered to ensure the correction accuracy of strong polarization areas such as wire edges.

[0094] Step 2: Optimize weight distribution based on Mie scattering characteristics and dynamic signal-to-noise ratio. Based on the target area signal-to-noise ratio (SNR), polarization degree (P) and Mie scattering correction factor (K), the weight is dynamically distributed, and the formula is:

[0095]

[0096] SNR IR is the infrared channel signal-to-noise ratio (dB);

[0097] SNR Pol is the polarization channel signal-to-noise ratio (dB);

[0098] DoP is the polarization degree (0-1 range);

[0099] α is the adjustment factor (2.5 in power scene);

[0100] k is the Mie scattering correction coefficient:

[0101]

[0102] Step 3: Eliminate edge artifacts in the fusion image by morphological filtering. Edge artifacts can affect the clarity and accuracy of the image, and morphological filtering can effectively remove these artifacts to make the edges of the fusion image clearer and more natural.

[0103] The morphological filtering algorithm is designed using an improved closed-open operation of a multi-scale adaptive structure element:

[0104] Artifact detection includes:

[0105] Calculate the gradient amplitude of the fusion image F(x, y):

[0106]

[0107] Extract the suspected artifact area (satisfy any of the following conditions):

[0108] The gradient direction and the difference between the infrared / polarization original image are greater than 15°;

[0109] Local gradient amplitude mutation rate

[0110] According to the characteristics of the artifact area, dynamically construct the structure element B:

[0111]

[0112] Mixed morphological operation, including:

[0113] Artifact suppression closed operation: Fill small holes (diameter < 3 pixels) caused by registration errors;

[0114] Detail protection open operation:

[0115]

[0116] Where B' is the erosion version of B (size reduced by 30%), used to preserve real edges

[0117] Weighted reconstruction:

[0118] F final (x, y) = w·F close +(1-w)·F open

[0119] The weight coefficient w is determined by the artifact confidence:

[0120]

[0121] Where p is the artifact probability, calculated by the SVM classifier.

[0122] It should be noted that the algorithm can accurately locate the artifact area and avoid misprocessing of the real target. The traditional morphological filter often adopts a global processing strategy, which is easy to misjudge the target edge as an artifact or miss subtle artifacts. The artifact detection mechanism of the algorithm is constrained by double features, which significantly improves the accuracy of artifact identification. Based on the gradient direction difference, the "direction contradiction type artifact" caused by the deviation of modality registration can be identified. Combined with the local gradient amplitude mutation rate (> 50%), the "intensity jump type artifact" caused by the response difference of the sensor can be captured.

[0123] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present application, which should be covered in the scope of the claims of the present application.

[0124] Embodiment 3

[0125] The third embodiment of the present application is different from the first two embodiments:

[0126] If the function is realized in the form of a software function unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product, which is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in the embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk, and various media that can store program codes.

[0127] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered a list of executable instructions for implementing logical functions, and can be specifically embodied in any computer-readable medium for use by an instruction execution system, device or apparatus, such as a computer-based system, a system including a processor, or other system that can fetch and execute instructions from the instruction execution system, device or apparatus, or in conjunction with these instructions. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transport programs for use by an instruction execution system, device or apparatus, or in conjunction with these instructions.

[0128] More specific examples (a non-exhaustive list) of the computer-readable medium include the following: an electrical connection (electronic) having one or more wires, a portable computer diskette (magnetic), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can also be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, for example, via an optical scanner, then compiled, interpreted, or otherwise processed, and stored in a computer memory in a form that is then employable by a computer. In some embodiments, the computer-readable medium can be a computer program product that includes a computer program tangibly embodied in a non-transitory medium.

[0129] It should be understood that aspects of the application can be implemented in hardware, software, firmware or combinations thereof. In the embodiments described above, various steps or methods can be implemented, for example in software or firmware, stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, and in another embodiment, any of the following technologies, known in the art, or combinations thereof, can be used: discrete logic circuitry having logic gates for implementing logic functions upon an application of data signals, application specific integrated circuits having appropriate combinational logic gates, programmable gate arrays (PGA), field programmable gate arrays (FPGA), and so forth.

Claims

1. A polarization method based on polarization-infrared multi-modal fusion, characterized in that, The method comprises the following steps: Non-uniformity correction is performed on the infrared image and the polarization image to compensate for the gain difference and offset between pixels; Weight distribution is optimized based on Mie scattering characteristics and dynamic signal-to-noise ratio; A morphological filtering algorithm is designed by using a multi-scale adaptive structure element and improved closed-open operation to eliminate edge artifacts in the fused image.

2. The polarization method based on polarization-infrared multi-modal fusion according to claim 1, characterized in that, Non-uniformity correction is performed on the infrared image, including: Original images I1(x, y) and I2(x, y) of a uniform blackbody radiation surface at temperatures T1 and T2 are collected; The gain G(x, y) and the offset O(x, y) of each pixel are calculated: O(x, y) = I1(x, y) - G(x, y)·T1 Real-time correction: where I raw (x,y) represents the uncorrected raw image output in real time by the infrared detector, I corr (x,y) represents the infrared image after non-uniformity correction.

3. The polarization method based on polarization-infrared multi-modal fusion according to claim 2, characterized in that, Non-uniformity correction is performed on the polarization image, including: In the absence of light, a dark image D(x, y) of the polarization detector is collected, which is represented as: S dark (x,y) = S raw (x,y) - D(x,y) Using the polarization uniformity of the sky background in the electric power scene, a background region R without conductors in the image is selected, the average response values u0, u 45 , u 90 , u 135 of each polarization direction sub-region in the region are calculated, and the normalization coefficients of each direction are calculated with u0 as the reference, which is represented as: Gain correction is performed on the polarization image in each direction, which is represented as: S ' 45 (x,y) = S dark,45 (x,y) · k 45 S ' 90 (x,y) = S dark,90 (x,y) · k 90 S ' 135 (x,y) = S dark,135 (x,y) · k 135 After each frame of image is collected, real-time monitoring of correction residual is performed through a sliding window, and when the local area standard deviation is greater than 5%, local secondary correction is triggered.

4. The polarization method based on polarization-infrared multi-modal fusion according to claim 3, characterized in that, The weight distribution is optimized based on Mie scattering characteristics and dynamic signal-to-noise ratio, which is represented as: wherein W IR The fusion weight of the infrared channel (value range 0~1) determines the contribution proportion of the infrared image in the final fusion result; SNR IR SNR is the signal-to-noise ratio of the infrared channel; SNR Pol SNR is the signal-to-noise ratio of the polarization channel; DoP is the degree of polarization; a is a dynamic adjustment factor, which controls the steepness of the weight distribution curve; k is the Mie scattering correction coefficient:

5. The polarization method based on polarization-infrared multi-modal fusion according to claim 4, characterized in that, The morphological filtering algorithm is designed by using a multi-scale adaptive structure element and improved closed-open operation to eliminate edge artifacts in the fused image, including: The gradient amplitude of the fused image F(x, y) is calculated, which is represented as: The suspected artifact area is extracted, including: The difference between the gradient direction and the infrared / polarization original image is greater than 15° local gradient amplitude jump rate The structure element B is dynamically constructed according to the characteristics of the artifact area: Mixed morphological operation is performed, which is represented as: filling small holes caused by registration errors; Detail protection open operation: Wherein, B' is the erosion version of B; Weighted reconstruction: F final (x,y) = w · F close + (1 - w) · F open The weight coefficient w is determined by the artifact confidence: Wherein, p represents the artifact probability.

6. A polarization unit based on polarization-infrared multi-modal fusion, applying the method of any one of claims 1-5, characterized in that, The method comprises the following steps: An optical window, a polarization modulation layer, a detection layer, and a processing module.

7. The polarization unit based on polarization-infrared multi-modal fusion according to claim 6, characterized in that, The optical window is made of ZnSe substrate and coated with an anti-reflection film; The polarization modulation layer integrates 0°, 45°, 90°, and 135° sub-wavelength aluminum gratings; The detection layer selects a mercury cadmium telluride infrared focal plane array.

8. The polarization unit based on polarization-infrared multi-modal fusion according to claim 6, characterized in that, The processing module integrates a GPU-accelerated real-time fusion processor.

9. An electronic device, comprising: A memory and a processor; The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions, which realize the steps of the polarization method based on polarization-infrared multi-modal fusion of any one of claims 1 to 7 when executed by the processor.

10. A computer readable storage medium storing computer executable instructions, which realize the steps of the polarization method based on polarization-infrared multi-modal fusion of any one of claims 1 to 7 when executed by the processor.

Images