Camouflage metal material classification method and device fusing heat and polarization parameters

Abstract

The invention relates to a classification method and device for camouflage metal materials fusing heat and polarization parameters, and the method comprises the steps: heating a metal material of which the surface is a camouflage medium through point laser, and obtaining an infrared image sequence of a heat diffusion process of a heating point region and a polarization image of the material surface through an infrared polarization camera; an image sequence of a thermal diffusion and absorption process is simulated through a finite difference method, a real thermal diffusion process sequence diagram shot by an infrared polarization camera is used as a constraint to perform multiple iterations so as to acquire diffusion and absorption thermal parameters, and a convolutional neural network is used to perform characteristic value extraction on a polarization degree image so as to acquire a polarization degree parameter. And fusing the parameters obtained in the two modes, and inputting the fused parameters into a convolutional neural network for training and classification. Through verification, the method can be used for camouflage metal material classification.

Description

Technical Field

[0001] The present invention belongs to the technical field of computer vision, and relates to a classification method and device for camouflaged metal materials integrating thermal and polarization parameters. Background Art

[0002] The identification and classification of metal materials with camouflage coatings on their surfaces are of great significance in the military, security and industrial fields. Camouflage coatings can be used to conceal, deceive the enemy and reduce the visibility of the target, so identifying and classifying metal materials under the coating can help ensure security and ensure the correct handling and utilization of materials. In the field of national defense and security, understanding the types of metals under camouflage coatings can help identify military technologies and equipment that may be used by the enemy, identify potential threats in advance and formulate corresponding national defense strategies. In the field of intelligence collection, it can infer the nature, purpose and weaknesses of the target, provide key information for intelligence collection, and provide important reference for military tactical planning. In the field of anti-counterfeiting and security, it can be used to identify product authenticity, ensure product authenticity and prevent the circulation of counterfeit goods. In the field of border security, understanding the nature and types of potential threat substances can strengthen border security measures and prevent the spread of illegal items. In the field of environmental protection, it can effectively classify and treat waste in an environmentally friendly manner and improve the recycling rate of resources. In the manufacturing industry, it can ensure that products meet standards and specifications, while promoting material engineering research and promoting the development and innovation of new materials.

[0003] The methods for identifying and classifying metal under camouflage coating mainly involve ultrasonic technology, X-ray technology, electromagnetic detection technology, optical imaging technology, etc. The ultrasonic detection method is to propagate mechanical waves through a medium. When the mechanical wave encounters an interface of different density or medium, part of the energy will be reflected back. By analyzing the reflected ultrasonic wave, the interface between metal and non-metal can be detected, thereby identifying and classifying the metal under the camouflage coating. Ultrasonic waves are affected by the interface during the detection process. When there is a complex multi-layer structure or camouflage coating, it may cause signal attenuation and scattering, reducing the detection accuracy. The X-ray detection method uses the property that X-rays can penetrate materials, and metals have a strong absorption capacity for X-rays, so metals can be detected by measuring the transmitted X-rays. Using X-ray imaging technology, images can be obtained, and metals under camouflage coatings can be identified and classified by analyzing the images. However, X-rays are a kind of ionizing radiation, and long-term exposure may cause harm to human health. Therefore, protective measures need to be taken in the application scenario, which increases the complexity of operation. Electromagnetic detection technology introduces an external electromagnetic field. When the metal is in the electromagnetic field, a current is induced, and the metal can be identified by detecting the change of the induced current. Detect and classify metals under camouflage coatings by measuring changes in magnetic fields using the magnetic differences of metals. Electromagnetic interference in the external environment may interfere with the detection results, especially in complex electromagnetic environments, where detection accuracy is challenged. Optical imaging technology uses technologies such as infrared imaging and lidar to classify by analyzing the reflection, absorption, and scattering properties of metals. Optical technology may be interfered with in the presence of camouflage layers because camouflage materials can change the propagation path of light and affect recognition accuracy. Factors such as ambient lighting and atmospheric conditions may have a negative impact on optical imaging, resulting in unstable recognition results.

[0004] In summary, current methods have made some progress in identifying and classifying metal types under camouflage coatings, but there are still some challenges and limitations. In response to the problems faced by the above methods, this patent proposes a new camouflage metal material classification method and device that integrates thermal and polarization parameters. It does not require direct contact with the metal surface and can monitor and classify in real time. Since different metals have different thermal and polarization properties, this method can effectively distinguish similar metals by simultaneously utilizing two modal information on the material surface. In addition, since this method uses a stable external heat source to heat the material surface, the detection signal will not attenuate, and it is simple to operate and is not affected by ambient light or shadows. Summary of the invention

[0005] Technical issues to be solved

[0006] In order to avoid the shortcomings of the prior art, the present invention proposes a classification method and device for camouflaged metal materials that integrates thermal and polarization parameters, which overcomes signal attenuation, complex operation, and the influence of ambient light and shadow, and solves the problem of identifying and classifying metals under camouflage coatings. The method and device of this patent can make full use of the heat diffusion information and polarization degree information on the surface of the material, thereby obtaining thermal parameters and polarization degree parameters and using parameter fusion to classify metal materials under camouflage coatings.

[0007] Technical Solution

[0008] A classification method for camouflaged metal materials by combining heat and polarization parameters, characterized in that the fusion heat and polarization parameters are combined, and the classification steps are as follows:

[0009] Step 1: Paste a camouflage thin layer medium 4 on the surface of a metal material plate 5, use a point laser to heat the metal surface to obtain a point heating point 6, and collect an infrared time series corresponding to the heat diffusion at the point heating point and a polarization image of the material surface;

[0010] Step 2: Process the polarization image sequence to obtain a polarization degree image, and extract the polarization degree parameters of the polarization degree image;

[0011] Step 3: Use the finite difference method forward propagation model to simulate the infrared image sequence of the heat diffusion process, and obtain the estimated values ​​of the thermal parameters of diffusivity and absorptivity through multiple iterations;

[0012] Step 4: For each metal, steps 1 to 3 are repeated multiple times to obtain multiple sets of thermal parameters and polarization parameters, and the two parameters are fused and input into the convolutional neural network for training;

[0013] During each training, a set of parameters for each material was set aside as a validation set, and the remaining data was used for training;

[0014] Repeat this process until every sample has been used as a validation set;

[0015] After the convolutional neural network is trained, it is used to classify various metal materials. The metal materials that need to be classified use steps 1 to 2 to obtain the thermal parameters and polarization degree parameters of various metal materials input into the convolutional neural network.

[0016] The infrared polarization camera collects the time series corresponding to the heat diffusion at the heating point and the polarization image of the material surface.

[0017] In step 1, polarization images corresponding to polarizers at different angles on the material surface are first collected, and then a point on the material surface is heated using a point-shaped laser lamp, and a thermal diffusion sequence of the heated point area is collected using an infrared polarization camera.

[0018] The process of simulating the thermal diffusion process image sequence using the finite difference method forward propagation model and obtaining the estimated values ​​of the thermal parameters of the diffusivity and absorptivity through multiple iterations:

[0019] The finite difference method forward propagation model is used to calculate the temperature at any position on the material surface at each time point based on the known initial conditions;

[0020] The simulated image generated at each time point is compared with the thermal diffusion sequence infrared images taken by a real infrared polarization camera, and the root mean square error of the corresponding pixel values ​​of all images in the simulated image sequence and the real image sequence is used as the loss function.

[0021] Use an optimization algorithm to update the estimated values ​​of the relevant parameters to minimize the loss function;

[0022] The optimization process of the optimization algorithm is continuously iterated, and each iteration makes the simulation results gradually approach the real data until the preset number of iterations is reached to obtain the estimated values ​​of the diffusion rate and absorption rate parameters.

[0023] The process of processing a polarization image sequence to obtain a polarization degree image and extracting a polarization degree parameter of the polarization degree image includes: processing polarization images corresponding to linear polarizers at different angles to generate a polarization degree image, cropping the polarization degree image to remove edge features, using a convolution layer and a pooling layer to perform feature selection on the part with the same central texture of the polarization degree image to generate a feature map; expanding the feature map into a one-dimensional vector containing multiple eigenvalues, and using the multiple eigenvalues ​​as polarization degree parameters.

[0024] When solving the polarization degree image and converting it into polarization degree parameters, polarizers with four angles of 0°, 45°, 90°, and 135° are selected to collect polarization images corresponding to I(0°), I(45°), I(90°), and I(135°), respectively.

[0025] When calculating the polarization degree parameters at four angles, the polarization state of light is Stokes vector S=[S0, S1, S2, S3], where S0 represents the total intensity of light, S1 represents the linear polarization energy in the horizontal or vertical direction, S2 represents the linear polarization energy in the +45° or -45° direction, and S3 represents the left-handed or right-handed circular polarization component; the intensity of the polarizer in different directions:

[0026]

[0027] S1=I(0°)-I(90°)

[0028] S2=I(45°)-I(135°)

[0029] The obtained Stokes parameters are further used to calculate the degree of polarization DoLP:

[0030]

[0031] An infrared thermal image and polarization image acquisition device, characterized in that: it is used for infrared thermal image and polarization image acquisition; it includes a data processor 1, a laser light 2 and an infrared polarization camera 3; the laser light 2 and the infrared polarization camera 3 are located directly above the metal material plate 5 to be inspected, and the data processor 1 is connected to the infrared polarization camera 3 via a data line to obtain image information taken by the infrared polarization camera 3.

[0032] A computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, the classification method of camouflaged metal materials combining thermal and polarization parameters as described in any one of claims 1 to 7 is implemented.

[0033] A computer program product, characterized by comprising computer executable instructions, which, when executed, are used to implement the classification method of camouflaged metal materials integrating thermal and polarization parameters as described in any one of claims 1 to 7.

[0034] Beneficial Effects

[0035] The present invention proposes a classification method and device for camouflaged metal materials that integrates thermal and polarization parameters. The method uses a point laser to heat a metal material whose surface is a camouflaged medium, and uses an infrared polarization camera to obtain an infrared image sequence of the thermal diffusion process in the heating point area and a polarization image of the material surface. The image sequence of the thermal diffusion and absorption process is simulated by the finite difference method, and the actual thermal diffusion process sequence diagram taken by the infrared polarization camera is used as a constraint to perform multiple iterations to obtain the diffusion and absorption thermal parameters. The convolutional neural network is used to extract the feature values ​​of the polarization image to obtain the polarization parameters. The parameters obtained in the two modes are fused and input into the convolutional neural network for training and classification. It has been verified that this method can be used for the classification of camouflaged metal materials.

[0036] The beneficial effects of the present invention are as follows: the existing methods for identifying and classifying metals under camouflage coatings face problems such as attenuation and scattering of detection signals, complex experimental operations, and interference caused by complex electromagnetic environments and lighting conditions. As a result, it is difficult for existing methods to stably obtain the characteristics of metals, making the classification results unstable. The present method uses a stable external heat source to heat the surface of the material, so the detection signal will not attenuate, and the operation is simple and is not affected by the environmental electromagnetic field and light. The surface of the camouflaged metal absorbs laser energy and converts it into heat, and the local temperature rises to form an obvious gradient. Different metals have different thermal properties and polarization properties, so the speed and range of the thermal gradient change on the surface of different camouflaged metal materials and the degree of polarization of the material surface will be different, so that the simulated diffusion rate, absorptivity thermal parameters and polarization degree parameters are different, and the classification of camouflaged metal materials is achieved by analyzing the number of thermal parameters and polarization degree parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 A flow chart of the method of the present invention;

[0038] Figure 2 Different types and specifications of metals;

[0039] Figure 3 The original metal surface (left) and the surface with a camouflage thin layer of dielectric pasted on it (right);

[0040] Figure 4 Infrared thermography and polarization image acquisition device;

[0041] Figure 5 The thermal diffusion comparison chart of the original image taken by the thermal imager and the thermal diffusion comparison chart obtained by simulation;

[0042] Figure 6 Accuracy of leave-one-out-validation training and classification for thermal and polarization parameters. DETAILED DESCRIPTION

[0043] The present invention will now be further described with reference to the embodiments and the accompanying drawings:

[0044] The present invention adopts the following technical scheme: a metal material with a camouflage medium on its surface is heated by a point laser, and an infrared polarization camera is used to obtain an infrared image sequence of the heat diffusion process in the heating point area and a polarization image of the material surface. The image sequence of the heat diffusion and absorption process is simulated by the finite difference method, and the real heat diffusion process sequence diagram taken by the infrared polarization camera is used as a constraint to iterate multiple times to obtain the diffusion and absorption heat parameters, and a convolutional neural network is used to extract the feature value of the polarization image to obtain the polarization parameter, and the parameters obtained in the two modes are fused and input into the convolutional neural network for training and classification.

[0045] The present invention provides a method and device for classifying camouflaged metal materials by integrating thermal and polarization parameters. Figure 1 As shown, the overall roadmap of the method of the present invention; Figure 2 As shown, different types and specifications of metals; Figure 3 As shown in the figure, the comparison between the original metal surface and the camouflage thin layer of medium pasted on the surface; Figure 4 As shown, infrared thermal image and polarization image acquisition device; as Figure 5 As shown in Figure 2, the thermal diffusion of the original image taken by the thermal imager and the thermal diffusion obtained by simulation are compared in different frames; Figure 6 As shown, the accuracy of the leave-one-out validation training and classification for thermal and polarization parameters.

[0046] The method is specifically implemented according to the following steps:

[0047] Step 1: Figure 2 , Figure 3 , Figure 4 As shown, the same camouflage thin layer medium is pasted on the metal surface, and polarization images corresponding to polarizers at different angles on the material surface are collected. A point laser is used to heat a point on the camouflage metal surface, and an infrared polarization camera is used to collect the time series corresponding to the heating point area and the polarization image of the material surface;

[0048] Step 2: Figure 5 As shown, choosing the finite difference method as the forward propagation model can generate the thermal diffusion process within a given time range and realize the simulation of the infrared intensity image sequence of the thermal diffusion and absorption processes; the mean square error loss function is used to compare the thermal diffusion process generated by the simulation with the thermal diffusion process sequence diagram taken by the infrared polarization camera, and multiple iterations are performed to optimize the estimated values ​​of the diffusion and absorption related parameters.

[0049] The finite difference method forward propagation model is used to calculate the temperature at each time point at any position on the surface of the material based on the known initial conditions. The simulated image generated at each time point is compared with the thermal diffusion sequence image taken by the real infrared polarization camera, and the root mean square error of the corresponding pixel values ​​of all pixels in the simulated image and the real image is used as the loss function. The estimated values ​​of the relevant parameters are updated using the optimization algorithm to minimize the loss function. This optimization process is continuously iterated, and each iteration will gradually make the simulation results approach the real data until the preset number of iterations is reached to obtain the estimated values ​​of the diffusion rate and absorption rate parameters.

[0050] Step 3: Process the polarization images corresponding to polarizers at different angles obtained in each experiment to generate polarization images, retain the part with the same central texture of the polarization image for feature extraction, generate feature images through convolutional layers and pooling layers, and expand the feature images to obtain a one-dimensional vector containing multiple eigenvalues. These eigenvalues ​​can reflect the statistical data related to the polarization effect on the surface of the material, and use these eigenvalues ​​as polarization parameters.

[0051] The polarization images corresponding to the linear polarizers at different angles (0°, 45°, 90°, 135°) obtained in each experiment are processed to generate polarization images. The polarization images are cropped to remove edge features, and the part with the same texture in the center of the material is retained for feature extraction. The image is processed multiple times using convolutional layers and pooling layers to perform feature selection to reduce the feature dimension and remove redundant information. Finally, a feature map is generated, and the image is expanded into a one-dimensional vector containing multiple eigenvalues. These eigenvalues ​​can reflect the statistical data related to the polarization effect on the surface of the material, and these eigenvalues ​​are used as polarization parameters.

[0052] Step 4: Figure 6 As shown, each material can obtain a set of thermal parameters and polarization parameters in each experiment. Multiple experiments are performed on each material to obtain multiple sets of parameters. A convolutional neural network is used as a classifier to classify the relevant parameters of different materials using the leave-one-out validation method.

[0053] For each metal, multiple sets of thermal parameters and polarization parameters are obtained by repeating the cycle multiple times. The two parameters are fused and input into the convolutional neural network. During each training, a set of parameters for each material is set aside as a validation set, and training is performed on the remaining data. This process is repeated until each sample has been used as a validation set.

[0054] In the technical solution of the present invention, it is first necessary to solve the polarization image and convert it into polarization parameters. The polarization images collected by polarizers at four angles of 0°, 45°, 90°, and 135° correspond to I(0°), I(45°), I(90°), and I(135°), respectively. These four angles can maximize the measurement signal-to-noise ratio and minimize the system error. The polarization state of light is usually represented by the Stokes vector, that is, S=[S0, S1, S2, S3], where S0 represents the total intensity of light, S1 represents the linear polarization energy in the horizontal or vertical direction, S2 represents the linear polarization energy in the +45° or -45° direction, and S3 represents the left-handed or right-handed circular polarization component. The intensity measurement results of the polarizer in different directions are:

[0055] The relationship between Stoke's parameter and polarization images collected by polarizers at four angles is as follows:

[0056]

[0057] S1=I(0°)-I(90°)

[0058] S2=I(45°)-I(135°)

[0059] The obtained Stokes parameters can be further used to calculate the degree of polarization (DoLP):

[0060]

[0061] Through the above solution, the polarization image of the material surface can be obtained from the polarization images corresponding to the polarizers taken at four angles. The part with the same texture in the center of the polarization image is retained for feature extraction. The feature image is generated through the convolution layer and the pooling layer. After the feature image is expanded, a one-dimensional vector containing multiple eigenvalues ​​is obtained. These eigenvalues ​​can reflect the statistical data related to the polarization effect on the material surface. These eigenvalues ​​are used as polarization parameters.

[0062] Secondly, the collected heat diffusion sequence diagram needs to be converted into thermal parameters. Thermal imagers can only measure radiance, not temperature directly. To solve this problem, first assume that the object's absorption rate is ε = 1, that is, the radiation of an ideal black body, and the thermal imager temperature measurement value υ C Satisfy C =υ black . Real blackbody temperature υ and ideal blackbody temperature υ black The relationship is Where σ is the Boltzmann constant, then:

[0063]

[0064] Where x represents the position of the material surface, t represents time, k = σ0 / cρ and β = 1 / cρ, σ0 represents the conductivity of the material, c is the specific heat of the material, ρ is the density of the material. γ is the external heat energy injected per unit time per unit volume, ε hs is the absorption rate of the material to the heat source.

[0065] Now there is a measurement value υ about the thermal imager C , input the light source γ and the expressions for the material properties ε and k at each spatial position, and use the measured values ​​to estimate the thermal diffusion parameters ε and k. Use the following symbols instead:

[0066]

[0067] where υ t Indicates the measurement value of the thermal imager υ t The partial derivative with respect to time, υ xx Indicates t Second-order partial derivatives with respect to the coordinates.

[0068] Expand the above one-dimensional case to three dimensions, and xyz =υ xx +υ yy +υ zz ,have to:

[0069] υ t = k·Δυ xyz +ε′·γ

[0070] where υ xx ,υ yy ,υ zz Respectively represent υ t The second-order partial derivative with respect to the x, y, and z coordinates, Δυ xyz Indicates t The second-order partial derivative of the three-dimensional coordinates. This formula is the core of material classification. k represents the thermal diffusion coefficient of the material, and ε′ is a factor that depends on the absorptivity of the heat source. The parameter tuple k, ε′ is the feature value used for the classification task.

[0071] The temperature distribution in the material is derived using the forward difference method as follows:

[0072]

[0073] where γ t (x) represents the state of the external heat source at time t, k(x,y,z) represents the diffusion rate at point (x,y,z), and ε′(x,y,z) represents the absorption rate at point (x,y,z).

[0074] This is the time step equation. Given a time step of t, we get the temperature value at time step t + 1. This experiment starts these calculations with the initial conditions at time t = 0. This method can be used to obtain the temperature value at any point in space and time through time stepping.

[0075] Therefore, k and ε′ are optimized given the initial conditions, and the optimization process is constrained using a real two-dimensional image of the surface temperature of the object over time. The above optimization process converts the thermal diffusion function into thermal parameters and uses them for classification. Nine metal materials of different types and specifications were used in this experiment.

[0076] Multiple experiments are conducted on each material, and the thermal parameters of the material are fused with the polarization parameters as a parameter set for material classification. This set can be used to classify the thermal parameters and polarization parameters recovered from unknown materials to determine the corresponding material type. This experiment uses a convolutional neural network as a classifier to classify materials based on the created thermal parameter and polarization parameter set. The classifier is simple, fast, and suitable for small data sets.

[0077] Example:

[0078] The camera used in this embodiment is an uncooled focal plane infrared polarization camera. The overall roadmap of the method of this patent is as follows: Figure 1 Different types and specifications of metals, such as Figure 2 As shown. A thin layer of camouflage medium is pasted on the metal surface to remove the reflection effect. Figure 3 As shown. Infrared thermal image and polarization image acquisition device, such as Figure 4 As shown. In order to make the surface temperature more obvious, a 200mW laser lamp is used for fixed point heating. The infrared polarization camera captures different frames of the thermal diffusion process on the material surface, collects the time series of heating and cooling to obtain the temperature diffusion map, and the original image thermal diffusion captured by the thermal imager and the thermal diffusion comparison map obtained by simulation, as shown in Figure 5 The accuracy of the training and classification of thermal parameters and polarization parameters by leave-one-out validation method is shown in Figure 6 shown.

Claims

1. A classification method for camouflaged metal materials combining thermal and polarization parameters, characterized in that: Combining the fusion heat with the polarization parameters, the classification steps are as follows: Step 1: Paste a camouflage thin layer medium (4) on the surface of a metal material plate (5), use a point laser to heat the metal surface to obtain a point heating point (6), and collect an infrared time series corresponding to the heat diffusion at the point heating point and a polarization image of the material surface; Step 2: Process the polarization image sequence to obtain a polarization degree image, and extract the polarization degree parameters of the polarization degree image; Step 3: Use the finite difference method forward propagation model to simulate the infrared image sequence of the heat diffusion process, and obtain the estimated values ​​of the thermal parameters of diffusivity and absorptivity through multiple iterations; Step 4: For each metal, steps 1 to 3 are repeated multiple times to obtain multiple sets of thermal parameters and polarization parameters, and the two parameters are fused and input into the convolutional neural network for training; During each training, a set of parameters for each material was set aside as a validation set, and the remaining data was used for training; Repeat this process until every sample has been used as a validation set; After the convolutional neural network is trained, it is used to classify various metal materials. The metal materials that need to be classified use steps 1 to 2 to obtain the thermal parameters and polarization degree parameters of various metal materials input into the convolutional neural network.

2. The classification method of camouflaged metal materials combining thermal and polarization parameters according to claim 1, characterized in that: The infrared polarization camera collects the time series corresponding to the heat diffusion at the heating point and the polarization image of the material surface.

3. The classification method of camouflaged metal materials combining thermal and polarization parameters according to claim 1, characterized in that: In step 1, polarization images corresponding to polarizers at different angles on the material surface are first collected, and then a point on the material surface is heated using a point-shaped laser lamp, and a thermal diffusion sequence of the heated point area is collected using an infrared polarization camera.

4. The classification method of camouflaged metal materials combining thermal and polarization parameters according to claim 1, characterized in that: The process of simulating the thermal diffusion process image sequence using the finite difference method forward propagation model and obtaining the estimated values ​​of the thermal parameters of the diffusivity and absorptivity through multiple iterations: The finite difference method forward propagation model is used to calculate the temperature at any position on the material surface at each time point based on the known initial conditions; The simulated image generated at each time point is compared with the thermal diffusion sequence infrared images taken by a real infrared polarization camera, and the root mean square error of the corresponding pixel values ​​of all images in the simulated image sequence and the real image sequence is used as the loss function. Use an optimization algorithm to update the estimated values ​​of the relevant parameters to minimize the loss function; The optimization process of the optimization algorithm is continuously iterated, and each iteration makes the simulation results gradually approach the real data until the preset number of iterations is reached to obtain the estimated values ​​of the diffusion rate and absorption rate parameters.

5. The classification method of camouflaged metal materials combining thermal and polarization parameters according to claim 1, characterized in that: The process of processing the polarization image sequence to obtain the polarization degree image and extracting the polarization degree parameters of the polarization degree image includes: processing the polarization images corresponding to the linear polarizers at different angles to generate the polarization degree image, cropping the polarization degree image to remove edge features, and using the convolution layer and the pooling layer to perform feature selection on the part with the same central texture of the polarization degree image to generate a feature map; The characteristic map is expanded into a one-dimensional vector containing multiple eigenvalues, and the multiple eigenvalues ​​are used as polarization degree parameters.

6. The classification method of camouflaged metal materials combining thermal and polarization parameters according to claim 1, characterized in that: When solving the polarization degree image and converting it into polarization degree parameters, polarizers with four angles of 0°, 45°, 90°, and 135° are selected to collect polarization images corresponding to I(0°), I(45°), I(90°), and I(135°), respectively.

7. The classification method of camouflaged metal materials combining thermal and polarization parameters according to claim 6, characterized in that: When calculating the polarization degree parameters at four angles, the polarization state of light is Stokes vector S=[S0, S1, S2, S3], where S0 represents the total intensity of light, S1 represents the linear polarization energy in the horizontal or vertical direction, S2 represents the linear polarization energy in the +45° or -45° direction, and S3 represents the left-handed or right-handed circular polarization component; the intensity of the polarizer in different directions: S1=I(0°)-I(90°) S2=I(45°)-I(135°) The obtained Stokes parameters are further used to calculate the degree of polarization DoLP:

8. An infrared thermal image and polarization image acquisition device, characterized in that: Used for collecting infrared thermal images and polarization images in claim 1; comprising a data processor (1), a laser light (2) and an infrared polarization camera (3); the laser light (2) and the infrared polarization camera (3) are located directly above the metal material plate (5) to be inspected, and the data processor (1) is connected to the infrared polarization camera (3) via a data line to obtain image information captured by the infrared polarization camera (3).

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for classifying camouflaged metal materials by integrating thermal and polarization parameters as claimed in any one of claims 1 to 7 is implemented.

10. A computer program product, characterized in that The invention comprises computer executable instructions, which are used to implement the classification method of camouflaged metal materials integrating thermal and polarization parameters as claimed in any one of claims 1 to 7 when being executed.

Images