A multi-dimensional imaging detection device and method under strong light background

Abstract

This application discloses a multi-dimensional imaging detection device and method under strong light background, relating to the field of target detection. The device includes: a beam splitting unit, an LC-SLM unit, and a visible light polarization CCD unit sequentially arranged on the same optical path; an infrared imaging system arranged on another optical path; the infrared imaging system and the visible light polarization CCD unit are respectively connected to an image processing system; the beam splitting unit is used to split the measured light into a first light ray and a second light ray, and to change the polarization degree of the first light ray to obtain a first modulated light ray; the infrared imaging system is used to convert the second light ray into short-wave infrared images with different polarization degrees; the LC-SLM unit is used to weaken the strong background light in the first modulated light ray to obtain a strong light weakened ray; the visible light polarization CCD unit is used to convert the strong light weakened ray into visible light images with different polarization degrees. This application achieves efficient suppression of strong background light and accurate highlighting of target features.

Description

Technical Field

[0001] This application relates to the field of target detection technology, and in particular to a multi-dimensional imaging detection device and method under strong light background. Background Technology

[0002] In the field of target detection, target detection under strong light background (environment) has always been a challenging problem, especially in fields involving space surveillance, military defense, night vision assistance and infrared photography.

[0003] Traditional detection devices are susceptible to strong light backgrounds. In strong light backgrounds, they often suffer from pixel overexposure, signal being submerged in noise, and difficulty in identifying target features. This greatly limits the effectiveness and reliability of the detection system, resulting in low detection and recognition capabilities in strong light backgrounds. At the same time, the images generated by the detection system also suffer from problems such as loss of detail, limited dynamic range, and complex post-processing. Summary of the Invention

[0004] The purpose of this application is to provide a multi-dimensional imaging detection device and method under strong light background, which improves the detection and recognition capability of target objects under strong light background, and achieves efficient suppression of strong background light and accurate highlighting of target features.

[0005] To achieve the above objectives, this application provides the following solution.

[0006] In a first aspect, this application provides a multi-dimensional imaging detection device under strong light background, the multi-dimensional imaging detection device under strong light background includes: a visible light imaging system, an infrared imaging system, and an image processing system; the visible light imaging system includes: a beam splitting unit, an LC-SLM unit, and a visible light polarization CCD unit; the beam splitting unit, the LC-SLM unit, and the visible light polarization CCD unit are sequentially arranged on the same optical path; the infrared imaging system is arranged on another optical path; the infrared imaging system and the visible light polarization CCD unit are respectively connected to the image processing system.

[0007] The beam splitting unit is used to split the measured light into a first light and a second light, and to change the polarization degree of the first light to obtain a first modulated light; the measured light is the light generated by the target object against a strong light background; the second light is redirected and projected onto the infrared imaging system; the infrared imaging system is used to convert the second light into short-wave infrared images with different polarization degrees; the LC-SLM unit is used to weaken the strong background light in the first modulated light to obtain a strong light weakened light; the visible light polarization CCD unit is used to convert the strong light weakened light into visible light images with different polarization degrees; the image processing system is used to: acquire visible light images and short-wave infrared images with different polarization degrees, and perform the following steps on each visible light image and short-wave infrared image with different polarization degrees: perform multi-scale Retinex decomposition on the original image to obtain the initial illumination component and initial reflection. The initial illumination component is subjected to NSCT transformation, and the initial reflection component is subjected to DT-CWT transformation to obtain high-frequency and low-frequency components under different transformations. Based on U-Net network, ResNet network and region energy weighting method, the high-frequency and low-frequency components under different transformations are fused and enhanced, and the illumination component and reflection component are reconstructed by inverse NSCT transformation and inverse DT-CWT transformation. Based on the reconstructed illumination component and reflection component, multi-scale Retinex is used to reconstruct the image, and the reconstructed visible light image and short-wave infrared image are input into a dual-channel PCNN network for fusion to obtain a fused image. The performance evaluation index of the fused image is determined, and the information value of the fused image is calculated based on the performance evaluation index. The information values ​​of the fused images with different polarization degrees are compared, and the fused image with the information value greater than a set threshold is output.

[0008] Secondly, this application also provides a multi-dimensional imaging detection method under strong light background, the multi-dimensional imaging detection method under strong light background is as follows.

[0009] Acquire visible light and shortwave infrared images with different polarization degrees, and perform the following steps for each polarization degree: Perform multi-scale Retinex decomposition on the original image to obtain initial illumination and initial reflection components; perform NSCT transform on the initial illumination component and DT-CWT transform on the initial reflection component to obtain high-frequency and low-frequency components under different transforms; perform fusion enhancement on the high-frequency and low-frequency components under different transforms based on U-Net, ResNet, and region energy weighting methods, and reconstruct the illumination and reflection components through inverse NSCT and inverse DT-CWT transforms; based on the reconstructed illumination and reflection components, reconstruct the image using multi-scale Retinex, and input the reconstructed visible light and shortwave infrared images into a dual-channel PCNN network for fusion to obtain a fused image; determine the performance evaluation index of the fused image, and calculate the information value of the fused image based on the performance evaluation index; compare the information values ​​of the fused images with different polarization degrees, and output the fused image whose information value is greater than a set threshold.

[0010] Based on the specific embodiments provided in this application, the following technical effects are disclosed.

[0011] This application innovatively constructs a multi-dimensional collaborative detection and intelligent fusion light field detection device. It fully utilizes a beam splitter to effectively divide the measured light into two optical paths, and simultaneously collaborates with other components in an infrared imaging system and a visible light imaging system to construct a multi-spectral collaborative mechanism. This application combines the dynamic control technology of a Liquid Crystal Spatial Light Modulator (LC_SLM) to achieve synchronous capture of light intensity, polarization, and spectral information. Under strong light interference, it can suppress 90% of background noise and improve the target signal-to-noise ratio by 3-5 times. Furthermore, the image processing system of this application effectively decomposes and reconstructs visible light and short-wave infrared images, and then fuses and enhances the reconstructed visible light and short-wave infrared images. The system integrates adaptive light field optimization with nonsubsampled contourlet transform (NSCT), dual-tree complex wavelet transform (DT-CWT), and a two-channel pulse-coupled neural network (PCNN) to perform multi-scale image fusion. Through weighted fusion of high-frequency detail enhancement and low-frequency energy, the target recognition rate is significantly improved. Based on these features, this application improves the detection and recognition capability of target objects against strong light backgrounds, achieving efficient suppression of strong background light and accurate highlighting of target features. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a schematic diagram of the structure of the multi-dimensional imaging detection device under strong light background in the embodiments of this application.

[0014] Figure 2 This is a schematic diagram of the modules of the multi-dimensional imaging detection device under strong light background in the embodiments of this application.

[0015] Figure 3 This is a flowchart illustrating the execution of the multi-dimensional imaging detection method under strong light background in the embodiments of this application.

[0016] Figure reference numerals: Visible light imaging system-1, beam splitting unit-11, dual telecentric lens unit-111, beam splitting prism unit-112, dual polarizer unit-113, LC-SLM unit-12, visible light polarizing CCD unit-13, infrared imaging system-2, short-wave infrared polarizer unit-21, short-wave infrared polarizing CCD unit-22, image processing system-3, motorized rotary wheel unit-4. Detailed Implementation

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0018] Traditional visual imaging devices suffer from problems such as loss of detail, limited dynamic range, and complex post-processing in strong backgrounds. Polarized target detection, on the other hand, can achieve imaging detection in complex environments, significantly improving the working distance and imaging quality of visual imaging devices under strong interference conditions. Liquid crystal spatial light modulators have advantages in strong light background detection, such as precise beam control, high dynamic range, and flexibility. Therefore, polarized target detection and liquid crystal spatial light modulators can be applied to the technical solution of this application.

[0019] The purpose of this application is to provide a multi-dimensional imaging detection device and method under strong light background, which improves the detection and recognition capability of target objects under strong light background, and achieves efficient suppression of strong background light and accurate highlighting of target features.

[0020] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] In one exemplary embodiment, such as Figure 1 and Figure 2 As shown, a multi-dimensional imaging detection device under strong light background is provided. The multi-dimensional imaging detection device under strong light background includes: a visible light imaging system 1, an infrared imaging system 2 and an image processing system 3; the visible light imaging system 1 includes: a beam splitting unit 11, an LC-SLM unit 12 and a visible light polarization charge-coupled device (CCD) unit 13.

[0022] In this system, the beam splitting unit 11, the LC-SLM unit 12, and the visible light polarization CCD unit 13 are sequentially arranged on the same optical path; the infrared imaging system 2 is arranged on another optical path; and the infrared imaging system 2 and the visible light polarization CCD unit 13 are respectively connected to the image processing system 3.

[0023] The beam splitting unit 11 is used to split the measured light into a first light and a second light, and to change the polarization degree of the first light to obtain a first modulated light; the measured light is the light generated by the target object against a background of strong light (strong light generally includes at least one of coherent light and incoherent light); the second light is redirected and projected onto the infrared imaging system 2; the infrared imaging system 2 is used to convert the second light into short-wave infrared images with different polarization degrees; the LC-SLM unit 12 is used to weaken the strong background light in the first modulated light to obtain a strong light weakened light; the visible light polarization CCD unit 13 is used to convert the strong light weakened light into visible light images with different polarization degrees; the image processing system 3 is used to: acquire visible light images and short-wave infrared images with different polarization degrees, and perform the following steps on each visible light image and short-wave infrared image with different polarization degrees: perform multi-scale Retinex decomposition on the original image to obtain an initial The initial illumination component and initial reflection component are identified. The initial illumination component undergoes NSCT transformation, and the initial reflection component undergoes DT-CWT transformation, yielding high-frequency and low-frequency components under different transformations. Based on U-Net, ResNet, and region energy weighting methods, the high-frequency and low-frequency components under different transformations are fused and enhanced. The illumination and reflection components are then reconstructed using inverse NSCT and DT-CWT transformations. Based on the reconstructed illumination and reflection components, a multi-scale Retinex reconstruction is performed. The reconstructed visible light image and short-wave infrared image are then input into a dual-channel PCNN network for fusion, resulting in a fused image. Performance evaluation metrics for the fused image are determined, and the information value of the fused image is calculated based on these metrics. The information values ​​of fused images with different polarization degrees are compared, and fused images with information values ​​greater than a set threshold are output.

[0024] In one preferred embodiment, the angle between the first ray and the second ray is set to 90 degrees.

[0025] Furthermore, the beam splitting unit 11 specifically includes: a dual telecentric lens unit 111, a beam splitting prism unit 112, and a dual polarizer unit 113; the dual telecentric lens unit 111, the beam splitting prism unit 112, the dual polarizer unit 113, the LC-SLM unit 12, and the visible light polarizing CCD unit 13 are sequentially arranged on the same optical path (or the same straight line).

[0026] The dual telecentric lens unit 111 is used for collimation and cross-section adjustment of the measured light, and to ensure that the measured light completely covers the effective light-passing aperture of the beam splitter unit 112; the beam splitter unit 112 is used to split the measured light into a first light and a second light; the first light is projected through the beam splitter unit 112 onto the dual polarizer unit 113; the dual polarizer unit 113 is used to change the polarization degree of the first light to obtain a first modulated light.

[0027] Furthermore, the infrared imaging system 2 includes: a short-wave infrared polarizer unit 21 and a short-wave infrared polarization CCD unit 22; the short-wave infrared polarizer unit 21 and the short-wave infrared polarization CCD unit 22 are sequentially arranged in the optical path where the second light ray is located; the short-wave infrared polarization CCD unit 22 is connected to the image processing system 3.

[0028] The shortwave infrared polarizer unit 21 is used to change the polarization degree of the second light to obtain the second modulated light; the shortwave infrared polarizing CCD unit 22 is used to convert the second modulated light into shortwave infrared images under different polarization degrees.

[0029] Furthermore, the multi-dimensional imaging detection device under strong light background also includes multiple electric rotating wheel units 4; the dual polarizer unit 113 and the short-wave infrared polarizer unit 21 are respectively mounted on different electric rotating wheel units 4; the electric rotating wheel units 4 are used to drive the dual polarizer unit 113 and the short-wave infrared polarizer unit 21 to rotate; the electric rotating wheel units 4 step in a step size of "1°".

[0030] In a preferred embodiment, the dual telecentric lens unit 111 is located in front of the beam splitter unit 112, the beam splitter unit 112 is located in front of the dual polarizer unit 113, the dual polarizer unit 113 is located in front of the LC-SLM unit 12, the dual polarizer unit 113 is mounted on the motorized rotating wheel unit 4, the LC-SLM unit 12 is located in front of the visible light polarization CCD unit 13, the visible light polarization CCD unit 13 is connected to the image processing system 3 so that the target object is imaged at the center of the imaging target surface of the visible light polarization CCD unit 13, the short-wave infrared polarizer unit 21 is located in front of the short-wave infrared polarization CCD unit 22 and is mounted on the motorized rotating wheel unit 4, the short-wave infrared polarization CCD unit 22 is used to receive the light split by the beam splitter unit 112. Among them, the dual telecentric lens unit 111 is XF-PTL15235-F from Canrui Optics, the beam splitter prism unit 112 is BSC0125-4 from Youguang Technology, the motorized rotary wheel unit 4 is K10CR2 / M from THORLABS, the dual polarizer unit 113 is LPVISC050 from THORLABS, the LC-SLM unit 12 is TSLM14U-A from Zhongke Weixing, the visible light polarizing CCD unit 13 is MER2-301-125U3C-HS from Daheng Imaging, the short-wave infrared polarizer unit 21 is GCL-051002 from Daheng Optoelectronics, and the short-wave infrared polarizing CCD unit 22 is MER3-138-136G3M-P-SWIR F02 from Daheng Imaging.

[0031] Furthermore, the specific steps for using the aforementioned multi-dimensional imaging detection device under strong light background are as follows.

[0032] Step 1: Align the large aperture of the dual telecentric lens unit 111 with the target object against a strong light background, and align the small aperture with the beam splitter prism unit 112, so that the measured light rays can pass completely through the beam splitter prism unit 112.

[0033] Step 2: Install the two polarizers in the dual polarizer unit 113 onto the two electric rotating wheel units 4 respectively, and control the rotation angle of the two polarizers (rotating in different directions) through the computer so that different optical signals are transmitted to the LC-SLM unit 12.

[0034] Step 3: The lattice of the liquid crystal layer in the LC-SLM unit 12 can judge the target signal and the strong background signal based on the obtained light signal, and reduce the intensity of the background light signal by changing the lattice transmittance.

[0035] Step 4: The light signal is converted into a visible light image by the visible light polarization CCD unit 13, and a total of 360 visible light images of various polarization degrees are obtained.

[0036] Step 5: Place the short-wave infrared polarizer unit 21 on another beam split by the beam splitter unit 112 and install it on the electric rotating wheel unit 4. Control the rotation angle of the short-wave infrared polarizer unit 21 by computer. The short-wave infrared polarization CCD unit 22 obtains a total of 360 short-wave infrared images of various polarization degrees.

[0037] Step 6: Input all images obtained from the short-wave infrared polarization CCD unit 22 and the visible light polarization CCD unit 13 into the image processing system 3 for processing, and finally output a high-definition image containing the target object.

[0038] In another exemplary embodiment, such as Figure 3 As shown, a multi-dimensional imaging detection method under strong light background is provided. This multi-dimensional imaging detection method under strong light background is applied to the aforementioned multi-dimensional imaging detection device under strong light background. The multi-dimensional imaging detection method under strong light background is as follows.

[0039] First, acquire visible light images and shortwave infrared images with different polarization degrees, and perform the following steps for each visible light image and shortwave infrared image with different polarization degrees.

[0040] In a preferred embodiment, visible light images are named and stored as "rotation angle of the first polarizer in the dual polarizer unit; rotation angle of the second polarizer in the dual polarizer unit; visible light image name", for example, a set of images is named "15°; 30°; Vis"; shortwave infrared images are named and stored as "rotation angle of the shortwave infrared polarizer; shortwave infrared image name", for example, a set of images is named "30°; IR". After storage is completed once, a control signal can be sent to the electric rotary wheel unit via a computer to acquire visible light images and shortwave infrared images with different polarization degrees in "1°" steps. The stored visible light images and shortwave infrared images should correspond one-to-one according to their respective polarization degrees.

[0041] (1) Perform multi-scale Retinex decomposition on the original image to obtain the initial illumination component and the initial reflection component; the original image refers to a set of visible light images and short-wave infrared images under a certain degree of polarization.

[0042] The multi-scale Retinex decomposition method is as follows.

[0043] .

[0044] In the formula, Represents the original image. Indicates the annotation difference as Gaussian filter, Indicates the first n The weights for each scale must satisfy the following conditions. , NThe usual choice is 3.

[0045] The initial illumination component is represented as: .

[0046] The initial reflection component is represented as: .

[0047] (2) Perform NSCT transformation on the initial illumination component and DT-CWT transformation on the initial reflection component to obtain the high-frequency and low-frequency components under different transformations.

[0048] Specifically, the initial illumination components are subjected to NSCT transform to obtain the first initial high-frequency component and the first initial low-frequency component. The NSCT transform consists of non-subsampled pyramid decomposition (NSP) and non-subsampled directional filter bank (NSDFB). The non-subsampled pyramid decomposes the image into multi-scale subbands, with each level generating a low-pass subband and a band-pass subband. Let the image to be decomposed be... I Then the first k The decomposition process is as follows.

[0049] Low-pass filter: .

[0050] Where, For low-pass subband, For the k A low-pass filter of the first stage. This represents the convolution operation. .

[0051] Bandpass sub-band: .

[0052] Each level of bandpass subband Detailed information at the current scale is preserved, and the input directional filter bank will be subjected to directional decomposition subsequently. Each bandpass sub-band Further decomposition via NSDFB Sub-bands in each direction ( (where is the directional decomposition series), the directional filter is constructed as follows.

[0053] Let the prototype directional filter be Multiple directional filters are generated through frequency domain modulation.

[0054] .

[0055] Where, for j Directional filter, For frequency domain coordinates, j For direction index, j=0,1, 2,......, - 1.

[0056] Directional subband generation: .

[0057] In the formula, each directional sub-band Corresponding image at scale k and direction j The coefficient on.

[0058] Then, the initial reflection component is subjected to DT-CWT transformation to obtain the second initial high-frequency component and the second initial low-frequency component.

[0059] (3) Based on U-Net network, ResNet network and regional energy weighting method, the high frequency component and low frequency component under different transformations are fused and enhanced, and the illumination component and reflection component are reconstructed by NSCT inverse transformation and DT-CWT inverse transformation.

[0060] Specifically, the first initial high-frequency component is input into the U-Net network to obtain the first fused enhanced high-frequency component.

[0061] The second initial high-frequency component is input into the ResNet network to obtain the second fused enhanced high-frequency component.

[0062] The first initial low-frequency component and the second initial low-frequency component are fused using a regional energy weighting method to obtain a first fused enhanced low-frequency component and a second fused enhanced low-frequency component. The specific steps of the regional energy weighting method are as follows.

[0063] The first step is to calculate the neighborhood energy of the shortwave infrared image and the visible light image.

[0064] .

[0065] Where, M , N These represent the length and width of the neighboring window, respectively. Indicates the image at position Neighborhood energy, MN Indicates the size of the neighborhood window. Indicates the image at the offset position The pixel value.

[0066] The second step is to divide the obtained neighborhood energy matrix into several non-overlapping sub-regions, and the formula for calculating the regional feature values ​​is as follows.

[0067] .

[0068] Where, Indicates the image number 1 iThe regional feature value of each sub-region is the mean of the sum of squares of all pixel values ​​within the sub-region.

[0069] The third step is to calculate the shortwave infrared image and the visible light image. i The fusion weight of each sub-block.

[0070] .

[0071] .

[0072] Where, and These are the fusion weights for the visible light image and the shortwave infrared image, respectively. and The first part represents the visible light image and the shortwave infrared image. i The energy of each sub-block.

[0073] The fourth step is to calculate the low-frequency coefficients.

[0074] .

[0075] Where, Indicates the low-frequency coefficient. This represents the data input to the visible light image. This represents the data input for the low-frequency infrared image. and This indicates the weight of the corresponding sub-block.

[0076] The first fused enhanced high-frequency component and the first fused enhanced low-frequency component are subjected to NSCT inverse transform to obtain the reconstructed illumination component.

[0077] The reconstructed reflection component is obtained by performing a DT-CWT inverse transform on the second fusion enhanced high-frequency component and the second fusion enhanced low-frequency component.

[0078] (4) Based on the reconstructed illumination and reflection components, the image is reconstructed using multi-scale Retinex, and the reconstructed visible light image and short-wave infrared image are input into a dual-channel PCNN network for fusion to obtain a fused image.

[0079] Specifically, multi-scale Retinex reconstruction is performed on the reconstructed illumination component and the reconstructed reflection component to obtain the reconstructed image; the reconstructed image includes a reconstructed visible light image and a reconstructed shortwave infrared image.

[0080] The reconstructed visible light image and the reconstructed shortwave infrared image are input into a dual-channel PCNN network for image fusion to obtain a fused image. The mathematical model of the dual-channel PCNN network is as follows.

[0081] .

[0082] .

[0083] .

[0084] .

[0085] .

[0086] .

[0087] Where, i and j These represent the coordinates of pixels in the image. n Indicates the number of iterations. Indicates the gain coefficient. This is a weight matrix, representing the nodes ( i , j )and( a , b The connection strength between ) This represents a binary input from another location. Indicates the first k Energy input to the layer, It provides dual-channel input to the neuron. For the connection input of neurons, for Corresponding link strength, This refers to the internal state of the neuron. This is the ignition pulse output of the neuron. and The inputs to the neuron links are respectively and Dynamic attenuation coefficient, This represents an adaptive threshold that decays over time and is adjusted by the input. This indicates the adjustment parameter.

[0088] (5) Determine the performance evaluation index of the fused image and calculate the information value of the fused image based on the performance evaluation index.

[0089] Specifically, performance evaluation metrics include: information entropy. IE Average gradient AG Standard deviation SD and peak signal-to-noise ratio PSNR Among them, information entropy The calculation formula is as follows.

[0090] .

[0091] .

[0092] Where, L Represents the number of gray levels in an image. Represents grayscale value k The probability of appearing in an image.

[0093] Furthermore, the average gradient AG The calculation formula is as follows.

[0094] .

[0095] .

[0096] .

[0097] Where, M and N These represent the number of rows and columns of the image, respectively. and Representing pixels Gradient values ​​in the horizontal and vertical directions.

[0098] Furthermore, standard deviation SD The calculation formula is as follows.

[0099] .

[0100] .

[0101] Where, It is represented as the average value of the image pixels. Indicates the image at position The pixel value at that location.

[0102] Furthermore, peak signal-to-noise ratio PSNR The calculation formula is as follows.

[0103] .

[0104] .

[0105] Where, MAX This represents the maximum value of the image pixels. MSE This represents the mean square error between the original image and the distorted image.

[0106] Finally, the formula for calculating the information value is as follows.

[0107] .

[0108] Where, For information values, , , and These are the weighting coefficients, This represents the maximum peak signal-to-noise ratio. This represents the maximum value of information entropy. The maximum value of the average gradient. This represents the maximum standard deviation.

[0109] (6) Compare the information values ​​of the fused images with different polarization degrees and output the fused image with an information value greater than the set threshold.

[0110] In one preferred implementation, the weight allocation for each performance evaluation index is as follows: =0.4, =0.3, =0.2, =0.1, and when MIQI When the value is greater than 0.75, the current fused image is determined to be the optimal fused image.

[0111] In summary, this application has the following main advantages.

[0112] This application utilizes a multispectral collaborative mechanism between visible light and infrared imaging systems, combined with dynamic control technology of LC-SLM (response time <2ms), to achieve simultaneous capture of light intensity, polarization, and spectral information. Under strong light interference, it can suppress 90% of background noise and improve the target signal-to-noise ratio by 3-5 times. Simultaneously, the image processing system integrates adaptive light field optimization and DT-CWT / NSCT-dual-channel PCNN multi-scale fusion algorithms. Through high-frequency detail enhancement (PSNR improvement of 8-12dB) and low-frequency energy-weighted fusion, it significantly improves the target recognition rate by 65%.

[0113] This application employs a full-link anti-interference design and a modular, scalable architecture, possessing strong environmental adaptability. Through the collaborative efforts of dual telecentric lens units, LC-SLM units, and multi-scale Retinex algorithms, 99% of stray light is eliminated, which can improve the detection range in scenarios such as military reconnaissance and aerospace observation, reduce the false alarm rate, and combine high precision with engineering practicality.

[0114] All actions involving the acquisition of signals, information, or data in this application are carried out in accordance with the relevant data protection laws and policies of the country where the application is located, and with the authorization of the owner of the relevant device.

[0115] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0116] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A multi-dimensional imaging detection device under strong light background, characterized in that, The multi-dimensional imaging detection device under strong light background includes: a visible light imaging system, an infrared imaging system, and an image processing system; the visible light imaging system includes: a beam splitting unit, an LC-SLM unit, and a visible light polarization CCD unit. The beam splitting unit, the LC-SLM unit, and the visible light polarization CCD unit are sequentially arranged on the same optical path; the infrared imaging system is arranged on another optical path; the infrared imaging system and the visible light polarization CCD unit are respectively connected to the image processing system; The beam splitting unit is used to split the measured light into a first light and a second light, and to change the polarization degree of the first light to obtain a first modulated light; the measured light is the light generated by the target object against a strong light background; the second light is redirected and projected onto the infrared imaging system; the infrared imaging system is used to convert the second light into short-wave infrared images with different polarization degrees; the LC-SLM unit is used to attenuate the strong background light in the first modulated light to obtain a strong light attenuated light; the visible light polarization CCD unit is used to convert the strong light attenuated light into visible light images with different polarization degrees; the image processing system is used for: Acquire visible light and shortwave infrared images with different polarization degrees, and perform the following steps for each polarization degree: Perform multi-scale Retinex decomposition on the original image to obtain initial illumination and initial reflection components; perform NSCT transform on the initial illumination component and DT-CWT transform on the initial reflection component to obtain high-frequency and low-frequency components under different transforms; perform fusion enhancement on the high-frequency and low-frequency components under different transforms based on U-Net, ResNet, and region energy weighting methods, and reconstruct the illumination and reflection components through inverse NSCT and inverse DT-CWT transforms; based on the reconstructed illumination and reflection components, reconstruct the image using multi-scale Retinex, and input the reconstructed visible light and shortwave infrared images into a dual-channel PCNN network for fusion to obtain a fused image; determine the performance evaluation index of the fused image, and calculate the information value of the fused image based on the performance evaluation index; compare the information values ​​of the fused images with different polarization degrees, and output the fused image whose information value is greater than a set threshold.

2. The multi-dimensional imaging detection device under strong light background according to claim 1, characterized in that, The beam-splitting unit specifically includes: a dual telecentric lens unit, a beam-splitting prism unit, and a dual polarizer unit; The dual telecentric lens unit, the beam splitter prism unit, the dual polarizer unit, the LC-SLM unit, and the visible light polarizing CCD unit are sequentially arranged on the same optical path; The dual telecentric lens unit is used for collimation and cross-section adjustment of the measured light beam, and to ensure that the measured light beam completely covers the effective light-transmitting aperture of the beam splitter; the beam splitter unit is used to split the measured light beam into a first light beam and a second light beam; the first light beam is projected onto the dual polarizer unit through the beam splitter unit; the dual polarizer unit is used to change the polarization degree of the first light beam to obtain a first modulated light beam.

3. The multi-dimensional imaging detection device under strong light background according to claim 2, characterized in that, The infrared imaging system includes: a short-wave infrared polarizer unit and a short-wave infrared polarizing CCD unit; The short-wave infrared polarizer unit and the short-wave infrared polarizing CCD unit are sequentially arranged in the optical path where the second light ray is located; the short-wave infrared polarizing CCD unit is connected to the image processing system; The shortwave infrared polarizer unit is used to change the polarization degree of the second light to obtain the second modulated light; the shortwave infrared polarizing CCD unit is used to convert the second modulated light into shortwave infrared images under different polarization degrees.

4. The multi-dimensional imaging detection device under strong light background according to claim 3, characterized in that, The multi-dimensional imaging detection device under strong light background also includes multiple electric rotating wheel units; The dual polarizer unit and the short-wave infrared polarizer unit are respectively mounted on different electric rotating wheel units; the electric rotating wheel units are used to drive the dual polarizer unit and the short-wave infrared polarizer unit to rotate.

5. The multi-dimensional imaging detection device under strong light background according to claim 1, characterized in that, The angle between the first ray and the second ray is 90 degrees.

6. A multi-dimensional imaging detection method under strong light background, implemented using the multi-dimensional imaging detection device under strong light background as described in any one of claims 1-5, characterized in that, The multi-dimensional imaging detection method under strong light background includes: Acquire visible light and short-wave infrared images with different polarization degrees, and perform the following steps for each visible light and short-wave infrared image with different polarization degrees: The original image is decomposed into multiple scales using Retinex to obtain the initial illumination component and the initial reflection component. The initial illumination component is subjected to NSCT transformation, and the initial reflection component is subjected to DT-CWT transformation to obtain high-frequency and low-frequency components under different transformations. Based on U-Net network, ResNet network and regional energy weighting method, high-frequency and low-frequency components under different transformations are fused and enhanced, and illumination and reflection components are reconstructed by NSCT inverse transform and DT-CWT inverse transform. Based on the reconstructed illumination and reflection components, a multi-scale Retinex reconstruction is used to reconstruct the image. The reconstructed visible light image and short-wave infrared image are then input into a dual-channel PCNN network for fusion to obtain a fused image. Determine the performance evaluation index of the fused image, and calculate the information value of the fused image based on the performance evaluation index; Compare the information values ​​of fused images with different polarization degrees, and output the fused image whose information value is greater than a set threshold.

7. The multi-dimensional imaging detection method under strong light background according to claim 6, characterized in that, The initial illumination component is subjected to NSCT transformation, and the initial reflection component is subjected to DT-CWT transformation to obtain high-frequency and low-frequency components under different transformations, specifically including: The initial illumination components are subjected to NSCT transformation to obtain the first initial high-frequency component and the first initial low-frequency component; The initial reflection component is subjected to DT-CWT transformation to obtain the second initial high-frequency component and the second initial low-frequency component.

8. The multi-dimensional imaging detection method under strong light background according to claim 7, characterized in that, Based on U-Net, ResNet, and the region energy weighting method, high-frequency and low-frequency components under different transforms are fused and enhanced. Illumination and reflection components are reconstructed through inverse NSCT and DT-CWT transforms. Specifically, this includes: The first initial high-frequency component is input into the U-Net network to obtain the first fused enhanced high-frequency component; The second initial high-frequency component is input into the ResNet network to obtain the second fused enhanced high-frequency component; The first initial low-frequency component and the second initial low-frequency component are fused using the regional energy weighting method to obtain the first fused enhanced low-frequency component and the second fused enhanced low-frequency component. The first fused enhanced high-frequency component and the first fused enhanced low-frequency component are subjected to NSCT inverse transform to obtain the reconstructed illumination component. The second fused enhanced high-frequency component and the second fused enhanced low-frequency component are subjected to DT-CWT inverse transform to obtain the reconstructed reflection component.

9. The multi-dimensional imaging detection method under strong light background according to claim 8, characterized in that, Based on the reconstructed illumination and reflection components, a multi-scale Retinex reconstruction is performed on the image. The reconstructed visible light image and shortwave infrared image are then input into a dual-channel PCNN network for fusion to obtain the fused image, which specifically includes: Multi-scale Retinex reconstruction is performed on the reconstructed illumination component and the reconstructed reflection component to obtain a reconstructed image; the reconstructed image includes a reconstructed visible light image and a reconstructed shortwave infrared image. The reconstructed visible light image and the reconstructed shortwave infrared image are input into a dual-channel PCNN network for image fusion to obtain a fused image.

10. The multi-dimensional imaging detection method under strong light background according to claim 6, characterized in that, The performance evaluation metrics include: information entropy, average gradient, standard deviation, and peak signal-to-noise ratio; the formula for calculating the information value is: ； Where, For information values, , , and These are the weighting coefficients, Peak signal-to-noise ratio, For information entropy, For the average gradient, Standard deviation This represents the maximum peak signal-to-noise ratio. This represents the maximum value of information entropy. The maximum value of the average gradient. This represents the maximum standard deviation.

Images