Passive temperature and distance measurement method and system based on multispectral infrared

Abstract

The invention relates to a passive temperature and distance measurement method and system based on multispectral infrared, designs a multispectral long-wave infrared detection system, constructs a multispectral infrared temperature and distance measurement model, takes the difference between a minimum system observation value and a model calculation value as a target function, and preliminarily solves model parameters by adopting a KKT and gradient descent method. Furthermore, a temperature and distance correction model is established, a loss function fusing a data error item and a physical information item is designed, and the corrected temperature and distance are obtained. According to the method, the infrared thermal image with the specific wavelength is obtained through the multispectral long-wave infrared detection system, synchronous obtaining of the temperature and the distance of the target object is achieved through the established multispectral infrared detection model, and high accuracy and reliability are achieved.

Description

Technical Field

[0001] The invention relates to the technical field of passive temperature measurement and ranging, in particular to a passive temperature measurement and ranging method and system based on multi-spectrum infrared. Background Art

[0002] Infrared thermal imaging is a widely used non-contact temperature measurement technology, which has many applications in the fields of steel, agriculture, power electronics, and medical auxiliary diagnosis. Existing infrared thermal imagers need to set parameters such as target distance and emissivity in advance. When there are multiple targets of different distances and materials in the scene, the infrared thermal imager can only correct the temperature measurement results according to the fixed parameters set in advance. Obviously, there are errors in the temperature measurement results. In the field of depth measurement, the existing depth sensing technology usually requires clear and visible textures, and is easily affected by light, making it difficult to achieve passive depth measurement under low light conditions. Using thermal signals to measure distance has significant advantages, and it still has reliable ranging capabilities even in dark and harsh environments. In the above case, the multi-spectral infrared temperature measurement and ranging method can achieve the simultaneous acquisition of the target object's emissivity, temperature and distance information by receiving radiation information of multiple wavelengths and combining spectral radiation and attenuation laws. Due to its accuracy and robustness in temperature and depth measurement, it is gradually becoming a key technical path in the field of temperature measurement and ranging.

[0003] Existing multispectral infrared temperature and ranging methods can be divided into analytical optimization methods and data-driven methods. Analytical optimization methods usually include two key components: objective function and constraints. The objective function defines the target to be optimized, while the constraints limit the feasible solution space of the problem. In this framework, the optimal solution is the solution that makes the objective function reach the maximum or minimum value under given constraints. However, this type of method is sensitive to initial values ​​and has strong model dependence. Especially under nonlinear conditions, its solution accuracy may be limited, or even fall into a local optimal solution, resulting in deviations in the inversion results. Feature extraction and parameter inversion in data-driven methods are usually based on a deep learning framework, which captures the mapping relationship between input multispectral data and target physical parameters by building a complex network architecture. This type of method has high requirements for data quality and network structure design. In the case of insufficient labeled data or high complexity of feature space, it may face problems such as insufficient generalization ability or feature learning bias, which in turn affects the final parameter inversion accuracy. At the same time, most of the existing methods involving mechanism modeling do not fully consider the source of the radiation energy received by the infrared thermal imager. Most methods ignore the radiation energy from the atmosphere and surrounding objects, which leads to systematic deviations in the estimation of the radiation characteristics of the target object, thereby affecting the accuracy and reliability of the measurement.

[0004] The present invention combines the advantages of rapid decision-making and physical constraints of analytical optimization methods with the powerful nonlinear fitting and mapping capabilities of data-driven methods, fully considers the source of radiation energy, and innovatively proposes a passive temperature measurement and ranging method and system based on multi-spectral infrared. The present invention designs a multi-spectral long wavelength infrared (LWIR) detection system, constructs a multi-spectral infrared temperature measurement and ranging model, designs a minimization objective function between system observation values ​​and model calculation values, constructs a fitting equation of emissivity to wavelength, reduces the number of unknown parameters, constrains the solution of the objective function, simplifies the preliminary solution of the temperature, distance and emissivity of the target object, establishes a temperature and distance correction model based on a fully connected neural network (Fully Connected Neural Network, FCNN), and designs a model loss function that integrates data error terms and physical information error terms, overcomes the problem that the traditional temperature measurement and ranging method does not fully consider the source of radiation energy and has low solution accuracy, and realizes accurate detection of the temperature and distance of the target object.

[0005] Compared with the prior art, the differences are as follows:

[0006] Application (patent) number: CN202410174012.X Application date: 2024-02-07

[0007] Application publication number: CN118067250A Public announcement date: 2024-05-24

[0008] Invention Title: Infrared temperature measurement compensation method, device, equipment and medium based on distance measurement

[0009] The invention proposes a method, device, equipment and medium for infrared temperature measurement compensation based on distance measurement, which obtains the first distance between the infrared sensor and the target object through an ultrasonic sensor, and compensates the infrared temperature measurement result of the target object based on the first distance and the ambient temperature. However, the invention requires the use of an ultrasonic sensor for distance measurement, and the ultrasonic distance measurement must be performed before or after the infrared temperature measurement. The temperature measurement and distance measurement cannot be performed synchronously, making it difficult to meet real-time requirements.

[0010] Application (patent) number: CN202010688429.X Application date: 2020-07-16

[0011] Application publication number: CN112013964A Publication date: 2020-12-01

[0012] Invention Title: A method, system, device and storage medium for automatic compensation of infrared temperature measurement distance

[0013] The invention proposes an automatic compensation method, system, device and storage medium for infrared temperature measurement distance, which obtains the target distance through visible light images and compensates the infrared temperature measurement results based on the distance. However, the measurement system of the invention is complex, requiring the use of dual-channel visible light cameras and infrared thermal imagers at the same time, and there are difficulties in multi-sensor calibration and time synchronization data collection, and the measurement effect is limited in low-light environments. Summary of the invention

[0014] In view of the above problems, the present invention proposes a passive temperature measurement and ranging method and system based on multi-spectral infrared. The method proposed in the present invention can accurately obtain the temperature and distance of the target object, providing a new idea for temperature measurement and ranging in industrial processes.

[0015] To achieve the above object, the technical solution adopted by the present invention is:

[0016] The invention provides a passive temperature and distance measurement system based on multi-spectral infrared, comprising a signal line, a thermometer, a lifting handle, a protective shell, a focusing lens, a motor, an infrared thermal imager, a rotating rod, a filter wheel buckle and a filter wheel, wherein the motor, the infrared thermal imager and the filter wheel are installed inside the protective shell, the motor and the infrared thermal imager are connected to a host computer through a signal line, a rotating rod is installed at an output end of the motor, one end of the rotating rod is connected to the filter wheel, the filter wheel is fixed inside the protective shell through the filter wheel buckle, the filter wheel is provided with four narrow-band filters, a full-transparent sheet and a light-shielding sheet at equal angles, the motor outputs a control signal to rotate the rotating rod and thus drive the filter wheel to rotate, so as to switch between the narrow-band filter and the light-shielding sheet, a lifting handle is installed on the upper end surface of the protective shell, a thermometer and a focusing lens are installed outside the protective shell, the thermometer is also connected to the host computer through a signal line, and the focusing lens is installed in front of the protective shell to converge external radiation energy and enter the inside of the protective shell.

[0017] The present invention provides a passive temperature measurement and ranging method based on a multi-spectral infrared passive temperature measurement and ranging system, comprising the following steps:

[0018] (1) Design a multispectral LWIR detection system. The detection system consists of an infrared thermal imager, a filter wheel with four narrow-band filters, a fully transparent filter and a light shield, a thermometer, and a focusing lens. The filter wheel is rotated by a motor to switch between the narrow-band filters and the light shield. The atmospheric temperature is measured by a thermometer, and the infrared radiation is focused onto the system surface by a focusing lens.

[0019] (2) Based on the principle of infrared radiation temperature measurement and Lambert-Beer law, a multi-spectral infrared temperature and ranging model is constructed by fully considering the radiation energy of the target object itself, the radiation energy reflected by the target object from the surrounding objects, the atmospheric radiation energy and radiation attenuation;

[0020] (3) Construct a minimization objective function between the system observation value and the model calculation value, and construct a fitting equation of the emissivity to wavelength based on the variation law of the spectral emissivity with wavelength. Use the fitting equation as the constraint condition of the objective function, use KKT and gradient descent method to iteratively optimize the objective function, and finally solve for temperature, distance and emissivity;

[0021] (4) Construct a temperature and distance correction model based on FCNN, and use the emissivity information obtained to correct the temperature and distance. In addition, design a loss function that integrates the data error term and the physical information error term to provide feedback guidance for the update of FCNN parameters.

[0022] As a further improvement of the method of the present invention, the multi-spectral LWIR detection system is constructed in step (1), specifically as follows;

[0023] The system consists of an infrared thermal imager, a filter wheel with four narrow-band filters, a full-transmitter and a light-shielding plate, a thermometer and a focusing lens. According to the specific spectral response curve of the infrared thermal imager in the long-wave infrared band, four long-wave infrared narrow-band filters at the peak of the response are selected. The spectral band allowed to pass by the narrow-band filter is very narrow, that is, for the spectral channel i = 1, 2, ..., K, it can be approximately regarded as all the incident radiation is only at the discrete wavelength λ i The detection system controls the filter wheel to rotate through the motor to switch the narrow-band filter and the light shield. The atmospheric temperature is measured through the thermometer. The infrared radiation is focused to the surface of the system through the focusing lens.

[0024] The shield is used to reduce the two special noises existing in the LWIR imaging process: the "daffodil effect" and 1 / f noise. The shield is introduced into the system design. By quickly switching the narrow-band filter and the shield, the camera captures a pair of images with / without scene radiation, namely, the scene and the reference. The two images are subtracted, as shown in formula (1), to obtain an image containing only the scene radiation, which effectively eliminates the "daffodil effect":

[0025] I(λ i )=I s (λi)-I r (λ i ) (1)

[0026] Among them, I s (λ i ) and I r (λ i ) are the scene image affected by the “daffodil effect” and the reference image, respectively. i ) is an image containing only scene radiation;

[0027] By quickly switching narrowband filters and light shields, multiple pairs of images are acquired in a short period of time, and the images are averaged after subtraction, which can significantly suppress 1 / f noise:

[0028]

[0029] Where N is the number of image pairs captured, For N I(λ i ) is the image obtained after averaging.

[0030] As a further improvement of the method of the present invention, the step (2) establishes a multi-spectral infrared temperature and distance measurement model, which is specifically as follows:

[0031] Step 1: Starting from the object emission term, an ideal black body with a temperature of T emits light with a wavelength of λ. According to Planck's black body radiation law:

[0032]

[0033] Where λ is the wavelength of light, c is the speed of light, h is Planck's constant, and k is the Boltzmann constant;

[0034] The emissivity ε(λ) of an object is defined as the ratio of the radiation emitted by the object to the radiation emitted by a blackbody source at that temperature. Therefore, the radiant energy emitted by an object with emissivity ε(λ) and temperature T is expressed as:

[0035] E(λ;T)=ε(λ)M e (λ; T) (4)

[0036] As light propagates through the atmosphere to the sensor, some of it is absorbed. The ratio between the radiation reaching the sensor and the radiation emitted by the object is defined as the transmittance T. air (λ), then the object’s contribution to the observed radiation is given by:

[0037] E obj (λ; T) = τ air (λ)ε(λ)M e (λ; T) (5)

[0038] According to Kirchhoff's law of thermal radiation, under thermal equilibrium, the absorptivity is equal to the emissivity, and the emissivity of air is related to the transmittance, that is:

[0039] ε air (λ)=1-τ air (λ) (6)

[0040] According to (6), air at temperature T air The emissions are:

[0041] E air (λ; Tair )=(1-τ air (λ))M e (λ; T air ) (7)

[0042] Among them, T air is the air temperature, obtained by a thermometer;

[0043] The target will reflect the radiation energy of the surrounding objects, which can be quantified as follows:

[0044] E ref (λ) = τ air (λ)(1-ε(λ))X λ (8)

[0045] Among them, X λ Indicates the radiation energy of surrounding objects at wavelength λ;

[0046] Taking into account the contribution of the target object, atmosphere and surrounding environment, the spectrum observed by the measurement system is expressed as:

[0047] I(λ)=E obj (λ;T)+E air (λ; T air )+E ref (λ) (9)

[0048] Using the Lambert-Beer law, the transmittance of a medium can be expressed as:

[0049] i out (λ)=exp(-σ air (λ)d) in (λ) (10)

[0050] where i in (λ) and i out (λ) are the input and transmitted intensities of the spectrum, σ air (λ) is the extinction coefficient of the medium, d is the thickness of the medium, τ air (λ)=exp(-σ air (λ)d);

[0051] Based on formulas (3)-(10), the radiation intensity I(λ) at wavelength λ can be expressed as:

[0052]

[0053] Among them, R v (λ) is the sensitivity relating energy density to observed intensity;

[0054] To simplify the calculation, it is assumed that the atmospheric pressure and water vapor content are known and the extinction coefficient is fixed for all λ. Then the parameter σ air (λ) can be obtained by using the high-resolution spectral modeling software SpectralCalc based on the readings of air temperature, humidity and pressure at the weather station; a black body furnace ε(λ) = 1 with a known temperature is placed 0 m in front of the thermal imager, and formula (11) can be simplified to:

[0055] I(λ)=R v (λ)M e (λ; T) (12)

[0056] Where I(λ) is the observed radiation intensity, M e (λ; T) can be calculated according to formula (3), and they are all known values. Therefore, R can be calculated according to formula (12): v (λ);

[0057] Step 2: The radiation energy X of the surrounding objects at wavelength λ λ It is expressed as follows:

[0058]

[0059] Among them, α represents the target object, β represents all other objects in the surrounding environment, and V αβ is the view factor, which indicates the ratio of radiation leaving β to α, E βλ represents the infrared radiation emitted by object β, V αβ satisfy:

[0060]

[0061] Consider the radiation emitted by a finite number of dominant objects, X λ E βλ Part is observed as I(λ), and I(λ) is sampled into m spectra to approximately describe the m most important environmental objects. Assume that the size of each thermal image collected is H×W×C, where H is the image height, W is the image width, and C is the number of spectral channels. Considering m environmental objects, the image H×W is spatially divided into m×1 quadrants, and the dimension of each quadrant is

[0062] Each quadrant is then spatially averaged into a spectrum of length C, i.e., for each of the C channels, a pair of size The sub-images are averaged to obtain their average values. Considering m = 2 environmental objects, the two spectra obtained according to the above method are expressed as E 1λ and E 2λ , is the equivalent object of the environment, now X λ Given by:

[0063] X λ =V α1 E 1λ +V α2 E 2λ (15)

[0064] By substituting (15) into (11), the set of unknown parameters to be estimated becomes {d, T, ε(λ i ), V α1 , V α2}, where i=1, 2,…,K.

[0065] As a further improvement of the method of the present invention, the construction and solution of the objective function in step (3) are as follows:

[0066] In formula (11), for the measurement of K spectral channels, there are K+4 unknowns, namely, K unknown values ​​of object emissivity, 1 unknown value of temperature, 1 unknown value of distance, and 2 unknown values ​​of view factor. This is an underdetermined problem. The objective function is used to solve the problem. The minimization objective function between the system observation value and the model calculation value is constructed. The fitting equation of emissivity to wavelength is constructed. The fitting equation is used as the constraint condition of the objective function, which reduces the number of unknown parameters and the search range of the solution. The specific steps include:

[0067] Step 1: Construct the objective function. The underdetermined problem has insufficient constraints because the number of equations is less than the number of unknowns. The set of solutions to the underdetermined equations has a high degree of freedom and there may be infinite solutions. Therefore, the objective function is introduced for optimization to change the solution selection process from disordered to ordered. The objective function is as follows:

[0068]

[0069] Among them, argmin is the input point that makes the function output value reach the minimum value, ||·|| 2 represents the 2-norm of the matrix, I(λ i ) is the observed radiation intensity, I(λ i ; d, ε(λ i ), T, V α1 , V α2 ) is the radiation intensity calculated according to formula (11);

[0070] Step 2: Add constraints. A commonly accepted assumption is that the spectral emissivity changes with the wavelength. The spectral emissivity is fitted using a wavelength function containing M, M < K adjustable parameters. The equation is assumed to be:

[0071]

[0072] lnε(λ)=a+bλ (18)

[0073] ε(λ)=a 0 +a 1 λ (19)

[0074]

[0075] ε(λ)=exp[-(a 0 +a 1 λ) 2 ] (twenty one)

[0076] Where K represents the number of spectral channels;

[0077] Using equation (20) as the constraint condition of emissivity, equation (16) becomes:

[0078]

[0079] Step 3: Solving the objective function. First, to avoid the model falling into a local optimal solution, initialize the model parameters to make them as close to the true value as possible. The initialization parameters are as follows:

[0080]

[0081]

[0082] Among them, M e -1 is the inverse function of Planck's blackbody radiation law, K is the number of spectral channels, and U(0,10) represents random sampling from a uniform distribution between 0 and 10 m;

[0083] Then, construct the Lagrangian function:

[0084]

[0085] Among them, η i , i is the value of ε(λ i ) of the inequality constraints, v, ξ, γ α1 , γ α2 is the Lagrange multiplier for other inequality constraints, μ is the equality constraint V α1 +V α2 = 1, are all non-negative values, and for each inequality constraint, they satisfy the complementary relaxation condition:

[0086]

[0087] Taking partial derivatives for each variable, we get the KKT gradient condition:

[0088]

[0089]

[0090] Finally, combined with the above conditions, the gradient descent method is used to solve the model parameters.

[0091] As a further improvement of the method of the present invention, the step (4) constructs a temperature and distance correction model, which is specifically as follows;

[0092] The temperature and distance are corrected using the emissivity information obtained in step (3). First, a correction model network architecture is built. The network used is FCNN. The network input features are the emissivity, temperature and distance information obtained by solving the objective function. The main body of the network consists of multiple fully connected layers to extract the deep nonlinear relationship of the input features. The activation function ReLU is introduced in each layer to enhance the model's ability to fit complex distributions. Finally, the corrected temperature and distance values ​​are generated through the output layer. In order to ensure physical consistency, a loss function that integrates data error terms and physical information error terms is designed to provide feedback guidance for the update of FCNN parameters. Specifically, the following steps are included:

[0093] Step 1: Verify the rationality of using emissivity to correct temperature and distance. First, calculate the derivative of equation (11) with respect to distance, emissivity and temperature:

[0094]

[0095] Eliminate the common term R v (λ)·exp(-σ air (λ)d), the core comparison of the three derivatives is simplified to:

[0096]

[0097] Based on equations (22), (34)-(39), the derivatives of equation (22) with respect to each parameter are calculated and simplified. It is found that the order of magnitude of the derivative of equation (22) with respect to emissivity is much greater than the order of magnitude of the derivative of equation (22) with respect to distance and temperature, indicating that the accuracy of emissivity is much higher than that of temperature and distance. Therefore, it is reasonable to use emissivity to correct temperature and distance.

[0098] Step 2: In order to solve the problem that traditional regression prediction only uses the accuracy of the predicted value as the evaluation indicator, a loss function that integrates the data error term and the physical information error term is designed:

[0099] Loss = ω 1 MSE data +ω 2 MSE physics (40)

[0100] Among them, MSE data is the data error term, which measures the error between the network prediction value and the true value, MSE physics is the physical information error term, which is the constraint residual based on the physical model, ω 1 ,ω 2 are the weight hyperparameters of the data error term and the physical information error term respectively;

[0101] The data error term is the parameter value d predicted by the network pred and T pred and the true value d true and T true The specific calculation of the error is as follows:

[0102]

[0103] Where N is the number of samples;

[0104] The physical information error term is the parameter value d predicted by the neural network pred 、T pred Substituting into the mechanism model (11) we obtain I model (λ; d pred , T pred ) and the error of the observed value I(λ), which is calculated as follows:

[0105]

[0106] Finally, the total loss is minimized using the gradient descent method and the FCNN parameters are updated.

[0107] Beneficial effects:

[0108] (1) The present invention designs a multi-spectral LWIR detection system, which captures the spectral information of specific wavelengths through an infrared thermal imager and a narrow-band filter, fully considering the specific contributions of different wavelengths, effectively overcoming the problem that the existing infrared thermal imager only considers the cumulative radiation in a specific band interval, and is conducive to improving the measurement accuracy of temperature and distance.

[0109] (2) Based on the principle of infrared radiation temperature measurement and Lambert-Beer law, the present invention fully considers the radiation energy of the target object itself, the radiation energy of the surrounding objects reflected by the target object, the atmospheric radiation energy and radiation attenuation, and constructs a multi-spectral infrared temperature measurement and ranging model. It effectively overcomes the limitation of traditional methods that only focus on the radiation of the target object itself, significantly improves the accuracy and stability of multi-spectral infrared temperature measurement and ranging, and provides a reliable technical solution for high-precision measurement in complex environments.

[0110] (3) The present invention constructs a minimization objective function between the system observation value and the model calculation value, and constructs a fitting equation of the emissivity to wavelength based on the law of spectral emissivity changing with wavelength. The fitting equation is used as a constraint condition of the objective function, and the KKT and gradient descent methods are used to iteratively optimize the objective function, ultimately achieving the solution of parameters such as temperature, distance and emissivity.

[0111] (4) The present invention constructs a temperature and distance correction model based on FCNN, and uses the emissivity information obtained to correct the temperature and distance. In addition, a loss function that integrates the data error term and the physical information error term is designed to optimize the two target information at the same time, providing feedback guidance for the update of FCNN parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0112] Figure 1 A multispectral LWIR detection system is shown;

[0113] Figure 2 A schematic diagram of the internal structure of the multi-spectral LWIR detection system housing is shown;

[0114] Figure 3 A cross-sectional view of a multi-spectral LWIR detection system is shown;

[0115] Figure 4 A method flow chart is shown;

[0116] Figure 5 A distance regression prediction result graph is shown;

[0117] Figure 6 A graph of temperature regression prediction results is shown.

[0118] The reference numerals are as follows:

[0119] 1. Signal line; 2. Thermometer; 3. Carrying handle; 4. Protective shell; 5. Focusing lens; 6. Motor; 7. Infrared thermal imager; 8. Rotating rod; 9. Filter wheel buckle; 10. Filter wheel; 11. Narrowband filter; 12. Shading film; 13. Fully transparent film. DETAILED DESCRIPTION

[0120] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. The following embodiments are used to illustrate the present invention but are not used to limit the scope of the present invention.

[0121] The present invention provides a passive temperature and distance measurement method and system based on multi-spectrum infrared. Figure 1-3 is a system diagram. Figure 4 It is an implementation step diagram, including the following steps:

[0122] (3) Design a multi-spectral LWIR detection system, which consists of an infrared thermal imager, a filter wheel with four narrow-band filters, a fully transparent filter and a light shield, a thermometer and a focusing lens. The filter wheel is rotated by a motor to switch between the narrow-band filters and the light shield. The atmospheric temperature is measured by a thermometer, and the infrared radiation is focused onto the system surface by a focusing lens.

[0123] (4) Based on the principle of infrared radiation temperature measurement and Lambert-Beer law, a multi-spectral infrared temperature and ranging model is constructed by fully considering the radiation energy of the target object itself, the radiation energy reflected by the target object from the surrounding objects, the atmospheric radiation energy and radiation attenuation;

[0124] (5) Construct a minimization objective function between the system observation value and the model calculation value, and construct a fitting equation of the emissivity to wavelength based on the variation law of the spectral emissivity with wavelength. Use the fitting equation as the constraint condition of the objective function, use KKT and gradient descent method to iteratively optimize the objective function, and finally solve the parameters such as temperature, distance and emissivity;

[0125] (6) Construct a temperature and distance correction model based on FCNN, and use the emissivity information obtained to correct the temperature and distance. In addition, design a loss function that integrates the data error term and the physical information error term to provide feedback guidance for the update of FCNN parameters.

[0126] The specific implementation plan is as follows:

[0127] (1) Construction of multi-spectral LWIR detection system;

[0128] In view of the problem that the existing infrared thermal imager only considers the cumulative radiation within a specific band range, the present invention obtains radiation information of multiple specific wavelengths and designs a multi-spectral LWIR detection system such as Figure 1-3 As shown. The system consists of an infrared thermal imager 7, a filter wheel 10 with four narrow-band filters 11, a fully transparent plate 13 and a light shielding plate 12, a thermometer 2 and a focusing lens 5. According to the specific spectral response curve of the infrared thermal imager in the long-wave infrared band, the present invention selects four long-wave infrared narrow-band filters at the peak of the response. The spectral band allowed to pass by the narrow-band filter is very narrow, that is, for the spectral channel i = 1, 2, ..., K, it can be approximately regarded as all incident radiation only at the discrete wavelength λ i The detection system controls the rotation of the filter wheel through a motor to switch between narrow-band filters and light shields; measures the atmospheric temperature through a thermometer; and focuses the infrared radiation onto the system surface through a focusing lens.

[0129] The present invention discloses a passive temperature and distance measurement system based on multi-spectral infrared. A motor 6, an infrared thermal imager 7 and a filter wheel 10 are installed inside the protective shell 4. The motor 6 and the infrared thermal imager 7 are connected to a host computer through a signal line 1. A rotating rod 8 is installed at the output end of the motor 6. One end of the rotating rod 8 is connected to the filter wheel 10. The filter wheel 10 is fixed inside the protective shell 4 through a filter wheel buckle 9. The filter wheel 10 is provided with four narrow-band filters 11, a full-transparent plate 13 and a light shielding plate 12 at equal angles. The motor 6 outputs a control signal to rotate the rotating rod 8 to drive the filter wheel 10 to rotate, so as to switch the narrow-band filter 11 and the light shielding plate 12. A lifting handle is installed on the upper end surface of the protective shell 4. A thermometer 2 and a focusing lens 5 are installed outside the protective shell 4. The thermometer 2 is also connected to the host computer through the signal line 1. The focusing lens 5 is installed in front of the protective shell 4 to converge external radiation energy and enter the protective shell 4. The present invention uses a shading sheet to reduce two special noises existing in the LWIR imaging process: the "daffodil effect" and 1 / f noise. In LWIR observation, radiation from the camera body is reflected by the surface of the optical element (including the narrowband filter and the shading sheet), and is imaged again at the plane where the sensor is located, which has a great impact on the quality of the system imaging. This is the "daffodil effect". In order to eliminate the "daffodil effect", the present invention introduces a shading sheet into the system design. By quickly switching the narrowband filter and the shading sheet, the camera captures a pair of images with / without scene radiation, namely the scene and the reference. By subtracting the two images, an image containing only the scene radiation can be obtained:

[0130] I(λ i )=I s (λ i )-I r (λ i ) (1)

[0131] Among them, I s (λ i ) and I r (λ i ) are the scene image affected by the “daffodil effect” and the reference image, respectively. i ) is an image containing only scene radiation.

[0132] Another unique noise characteristic of the LWIR imaging process is 1 / f noise. 1 / f noise is low-frequency fluctuations that appear strongly when processing images taken at long intervals. By rapidly switching narrowband filters and masking sheets, acquiring multiple pairs of images in a short period of time, and averaging them after subtraction, 1 / f noise can be significantly suppressed:

[0133]

[0134] Where N is the number of image pairs captured, For N I(λ i ) is the image obtained after taking the average.

[0135] In summary, the present invention designs a multispectral LWIR detection system, which suppresses typical noise that may occur in the imaging process and creates the prerequisite for subsequent model establishment.

[0136] (2) Establish a multi-spectral infrared temperature and ranging model;

[0137] In order to solve the problem that the existing infrared temperature measurement and ranging methods focus on the radiation of the target object itself and ignore other radiation factors, the present invention is based on the infrared radiation temperature measurement principle and Lambert-Beer law, fully considers all radiation energy sources that the infrared thermal imager may receive, and constructs a multi-spectral infrared temperature measurement and ranging model. Specifically, it includes the following steps:

[0138] Step 1: Starting from the object emission term, an ideal black body with a temperature of T emits light with a wavelength of λ. According to Planck's black body radiation law:

[0139]

[0140] Where λ is the wavelength of light, c is the speed of light, h is Planck's constant, and k is the Boltzmann constant.

[0141] Real-world materials emit less thermal radiation than a true black body. The emissivity of an object, ε(λ), is defined as the ratio of the radiation emitted by the object to the radiation emitted by a black body source at that temperature. Therefore, the radiated energy emitted by an object with emissivity ε(λ) and temperature T can be expressed as:

[0142] E(λ;T)=ε(λ)M e (λ; T) (4)

[0143] As light propagates through the atmosphere to the sensor, some of it is absorbed. The ratio between the radiation reaching the sensor and the radiation emitted by the object is defined as the transmittance τ air (λ). The object's contribution to the observed radiation is then given by:

[0144] E obj (λ; T) = τ air (λ)ε(λ)M e (λ; T) (5)

[0145] According to Kirchhoff's law of thermal radiation, under thermal equilibrium, the absorptivity is equal to the emissivity, and the emissivity of air is related to the transmittance, that is:

[0146] ε air (λ)=1-τ air(λ) (6)

[0147] According to (6), air at temperature T air The emissions are:

[0148] E air (λ; T air )=(1-τ air (λ))M e (λ; T air ) (7)

[0149] Among them, T air is the air temperature, obtained by a thermometer.

[0150] The target will reflect the radiation energy of the surrounding objects, which can be quantified as follows:

[0151] E ref (λ) = τ air (λ)(1-ε(λ))X λ (8)

[0152] Among them, X λ It represents the radiation energy of surrounding objects at wavelength λ.

[0153] Taking into account the contributions of the target object, atmosphere and surrounding environment, the spectrum observed by the measurement system can be expressed as:

[0154] I(λ)=E obj (λ;T)+E air (λ; T air )+E ref (λ) (9)

[0155] Using the Lambert-Beer law, the transmittance of a medium can be expressed as:

[0156] i out (λ)=exp(-σ air (λ)d)i in (λ) (10)

[0157] where i in (λ) and i out (λ) are the input and transmitted intensities of the spectrum, σ air (λ) is the extinction coefficient of the medium, d is the thickness of the medium, τ air (λ)=exp(-σ air (λ)d). Thermal imagers either eliminate this attenuation by allowing the user to manually enter the depth, or simply ignore it. In contrast, the present invention will take this into account in the subsequent solution process to estimate the depth of the target.

[0158] Based on formulas (3)-(10), the radiation intensity I(λ) at wavelength λ can be expressed as:

[0159]

[0160] Among them, R v (λ) is the sensitivity relating the energy density to the observed intensity.

[0161] To simplify the calculation, it is assumed that the atmospheric pressure and water vapor content are known and the extinction coefficient is fixed for all λ. Then the parameter σ air (λ) can be obtained using the high-resolution spectral modeling software SpectralCalc based on the readings of air temperature, humidity and pressure at the weather station; a blackbody furnace (ε(λ) = 1) with known temperature is placed 0 m in front of the thermal imager, and formula (11) can be simplified to:

[0162] I(λ)=R v (λ)M e (λ; T) (12)

[0163] Where I(λ) is the observed radiation intensity, M e (λ; T) can be calculated according to formula (3), and they are all known values. Therefore, R can be calculated according to formula (12): v (λ).

[0164] Step 2: The radiation energy X of the surrounding objects at wavelength λ λ It is expressed as follows:

[0165]

[0166] Among them, α represents the target object, β represents all other objects in the surrounding environment, and V αβ is the view factor, which indicates the ratio of radiation leaving β to α, E βλ Represents the infrared radiation emitted by object β. αβ satisfy:

[0167]

[0168] Although it is difficult to fully consider the radiation emitted by all objects in the target object's surrounding environment, it is worth noting that there are usually several objects that dominate the scattering contribution, such as the sky, the ground, and buildings. Now only consider the radiation emitted by a limited number of dominant objects, X λ E βλThe part is observed as I(λ), and I(λ) is sampled into m spectra to approximately describe the m most important environmental objects. The present invention assumes that the size of each thermal image collected is H×W×C, where H is the image height, W is the image width, and C is the number of spectral channels. Considering m environmental objects, the image H×W is spatially divided into m×1 quadrants, and the dimension of each quadrant is

[0169] Each quadrant is then spatially averaged into a spectrum of length C, i.e., for each of the C channels, a pair of size The sub-images are averaged to obtain their average values. Considering m = 2 environmental objects, the two spectra obtained according to the above method are expressed as E 1λ and E 2λ , is the equivalent object of the environment, now X λ Given by:

[0170] X λ =V α1 E 1λ +V α2 E 2λ (15)

[0171] By substituting (15) into (11), the set of unknown parameters to be estimated becomes {d, T, ε(λ i ), V α1 , V α2}, where i=1, 2,…,K.

[0172] In summary, based on the infrared radiation temperature measurement principle and the Lamhert-Beer law, the present invention constructs a multi-spectral infrared temperature measurement and ranging model, fully considers all the radiation that the infrared thermal imager may receive, accurately reflects the source of radiation energy, improves the reliability and interpretability of the model, and at the same time performs approximate processing on the environmental radiation, reducing the complexity of the model.

[0173] (3) Construction and solution of objective function;

[0174] The inversion of the forward model in formula (11) is underdetermined. For the measurement of K spectral channels, there are K+4 unknowns (K unknown values ​​of object emissivity, 1 unknown value of temperature, 1 unknown value of distance, and 2 unknown values ​​of viewing angle factor), which is an underdetermined problem. In order to solve the inverse problem, the present invention constructs a minimization objective function between the system observation value and the model calculation value. In view of the redundant calculation or meaningless solution exploration problem caused by the excessively broad solution space of the objective function, the present invention constructs a fitting equation of the emissivity to the wavelength based on the law of variation of the spectral emissivity with the wavelength, and uses the fitting equation as the constraint condition of the objective function, thereby reducing the number of unknown parameters and the search range of the solution. Specifically, the following steps are included:

[0175] Step 1: Construct the objective function. In an underdetermined problem, the number of equations is less than the number of unknowns, resulting in insufficient constraints. The set of solutions to the underdetermined equations has a high degree of freedom and there may be an infinite number of solutions. Therefore, an objective function is introduced for optimization to transform the solution selection process from disordered to ordered. The objective function is as follows:

[0176]

[0177] Among them, argmin is the input point that makes the function output value reach the minimum value, ||·|| 2 represents the 2-norm of the matrix, I(λ i ) is the observed radiation intensity, I(λ i ; d, ε(λ i ), T, V α1 , V α2 ) is the radiation intensity calculated according to formula (11).

[0178] Step 2: Add constraints. In order to further simplify the problem solving, the present invention uses a wavelength function containing M (M < K) adjustable parameters to fit the spectral emissivity, reducing the number of unknown parameters and the search range of the solution. At present, a generally accepted assumption in the field of multi-spectral radiation thermometry is that the spectral emissivity changes with the wavelength. Some well-known assumption equations are:

[0179]

[0180] lnε(λ)=a+bλ (18)

[0181] ε(λ)=a 0 +a 1 λ (19)

[0182]

[0183] ε(λ)=exp[-(a 0 +a 1 λ) 2 ] (twenty one)

[0184] Where K represents the number of spectral channels.

[0185] The present invention adopts formula (20) as the constraint condition of emissivity. Formula (16) becomes:

[0186]

[0187] Step 3: Solving the objective function. First, to avoid the model falling into a local optimal solution, initialize the model parameters to make them as close to the true value as possible. The initialization parameters are as follows:

[0188]

[0189] Among them, M e -1 is the inverse function of Planck's blackbody radiation law, K is the number of spectral channels, and U(0, 10) represents random sampling from a uniform distribution between 0 and 10 m.

[0190] Then, construct the Lagrangian function:

[0191]

[0192] Among them, η i , i is the value of ε(λ i ) of the inequality constraints, v, ξ, γ α1 , γ α2 is the Lagrange multiplier for other inequality constraints, μ is the equality constraint V α1 +V α2 = 1, all are non-negative. For each inequality constraint, it satisfies the complementary relaxation condition:

[0193]

[0194] Taking partial derivatives for each variable, we get the KKT gradient condition:

[0195]

[0196]

[0197] Finally, combined with the above conditions, the gradient descent method is used to solve the model parameters.

[0198] In summary, the present invention constructs a minimization objective function between the system observation value and the model calculation value, and constructs a fitting equation of the emissivity to the wavelength based on the law of variation of the spectral emissivity with the wavelength. The fitting equation is used as the constraint condition of the objective function, which reduces the number of unknown parameters and the search range of the solution, effectively simplifies the model solution, and at the same time, the KKT and gradient descent methods are used to solve the model parameters, which provides a basis for the subsequent realization of the temperature and distance correction model.

[0199] (4) Construct temperature and distance correction model;

[0200] In order to further improve the accuracy of temperature and distance, the present invention uses the emissivity information obtained by solution to correct the temperature and distance. First, a correction model network architecture is built. The network used in the present invention is FCNN. The network input features are the emissivity, temperature and distance information obtained by solving the objective function. The main body of the network is composed of multiple fully connected layers to extract the deep nonlinear relationship of the input features, and introduce an activation function (ReLU) in each layer to enhance the model's ability to fit complex distributions. Finally, the corrected temperature and distance values ​​are generated through the output layer. To ensure physical consistency, the present invention designs a loss function that fuses data error terms and physical information error terms to provide feedback guidance for the update of FCNN parameters. Specifically, the following steps are included:

[0201] Step 1: Verify the rationality of using emissivity to correct temperature and distance. First, calculate the derivative of equation (11) with respect to distance, emissivity and temperature:

[0202]

[0203]

[0204] Eliminate the common term R v (λ)·exp(-σ air (λ)d), the core comparison of the three derivatives simplifies to:

[0205]

[0206] Based on equations (22), (34)-(39), the derivatives of equation (22) with respect to various parameters are calculated and simplified. It is found that the order of magnitude of the derivative of equation (22) with respect to emissivity is much larger than the order of magnitude of the derivative of equation (22) with respect to distance and temperature, indicating that the accuracy of emissivity is much higher than the accuracy of temperature and distance. Therefore, it is reasonable to use emissivity to correct temperature and distance.

[0207] Step 2: In order to solve the problem that traditional regression prediction only uses the accuracy of the predicted value as the evaluation indicator, the present invention designs a loss function that integrates the data error term and the physical information error term:

[0208] Loss = ω 1 MSE data +ω 2 MSE physics (40)

[0209] Among them, MSE data is the data error term, which measures the error between the network prediction value and the true value, MSE physics is the physical information error term, which is the constraint residual based on the physical model, ω 1 ,ω 2are the weight hyperparameters of the data error term and the physical information error term respectively.

[0210] The data error term is the parameter value predicted by the network (d pred and T pred ) and the true value (d true and T true The specific calculation is as follows:

[0211]

[0212] Where N is the number of samples.

[0213] The physical information error term is the parameter value (d pred 、T pred ) is substituted into the mechanism model (11) to obtain I model (λ; d pred ,T pred ) and the error of the observed value I(λ), which is calculated as follows:

[0214]

[0215] Finally, the total loss is minimized using the gradient descent method and the FCNN parameters are updated.

[0216] In summary, the present invention constructs a temperature and distance correction model based on FCNN, and uses the emissivity information obtained by solving to correct the temperature and distance. In addition, the present invention designs a loss function that integrates the data loss term and the physical information error term to provide feedback guidance for the update of FCNN parameters, so that the model can optimize the two target information at the same time during the training process, ensuring physical consistency.

[0217] Example 1

[0218] This embodiment takes a black body furnace as the object, and is carried out in a laboratory environment (temperature: 24°C, relative humidity: 53%), and applies the invented multi-spectral infrared temperature and distance measurement method and system to the temperature and distance measurement of the black body furnace. The temperature of the black body furnace is set to 50°C, the distance range is 0-10m, and the step length is 1m. Figure 5 , 6 The absolute error and relative error of the distance measurement and temperature measurement results of the present invention are shown respectively. Figure 3 It can be seen that the root mean square error of the distance measurement result of the present invention is 0.210, and the mean absolute error is 0.191. Figure 4 It can be seen that the root mean square error of the temperature measurement result of the present invention is 0.689, and the mean absolute error is 0.611. The experimental results fully prove that the method proposed in the present invention performs well in temperature measurement and distance measurement, and can meet the measurement accuracy requirements.

[0219] The above description is only a preferred embodiment of the present invention and does not constitute any other form of limitation to the present invention. Any modification or equivalent change made based on the technical essence of the present invention still falls within the scope of protection required by the present invention.

Claims

1. A passive temperature and distance measurement system based on multi-spectral infrared, comprising a signal line (1), a thermometer (2), a carrying handle (3), a protective housing (4), a focusing lens (5), a motor (6), an infrared thermal imager (7), a rotating rod (8), a filter wheel buckle (9) and a filter wheel (10), characterized in that: A motor (6), an infrared thermal imager (7) and a filter wheel (10) are installed inside the protective housing (4); the motor (6) and the infrared thermal imager (7) are connected to a host computer via a signal line (1); a rotating rod (8) is installed at the output end of the motor (6); one end of the rotating rod (8) is connected to the filter wheel (10); the filter wheel (10) is fixed inside the protective housing (4) via a filter wheel buckle (9); and the filter wheel (10) is provided with four narrow-band filters (11), a fully transparent plate (13) and a light shielding plate (14) at equal angles. The motor (6) outputs a control signal to rotate the rotating rod (8) so as to drive the filter wheel (10) to rotate, so as to switch between the narrow-band filter (11) and the light shielding plate (12). A lifting handle is installed on the upper end surface of the protective housing (4). A thermometer (2) and a focusing lens (5) are installed outside the protective housing (4). The thermometer (2) is also connected to the host computer through a signal line (1). The focusing lens (5) is installed in front of the protective housing (4) so ​​that external radiation energy is gathered and enters the interior of the protective housing (4).

2. The passive temperature measurement and ranging method based on the multi-spectrum infrared passive temperature measurement and ranging system of claim 1 is characterized by: The following steps are involved: (1) Design a multispectral LWIR detection system. The detection system consists of an infrared thermal imager, a filter wheel with four narrow-band filters, a fully transparent filter and a light shield, a thermometer, and a focusing lens. The filter wheel is rotated by a motor to switch between the narrow-band filters and the light shield. The atmospheric temperature is measured by a thermometer, and the infrared radiation is focused onto the system surface by a focusing lens. (2) Based on the principle of infrared radiation temperature measurement and Lambert-Beer law, a multi-spectral infrared temperature and ranging model is constructed by fully considering the radiation energy of the target object itself, the radiation energy reflected by the target object from the surrounding objects, the atmospheric radiation energy and radiation attenuation; (3) Construct a minimization objective function between the system observation value and the model calculation value, and construct a fitting equation of the emissivity to wavelength based on the variation law of the spectral emissivity with wavelength. Use the fitting equation as the constraint condition of the objective function, use KKT and gradient descent method to iteratively optimize the objective function, and finally solve for temperature, distance and emissivity; (4) Construct a temperature and distance correction model based on FCNN, and use the emissivity information obtained to correct the temperature and distance. In addition, design a loss function that integrates the data error term and the physical information error term to provide feedback guidance for the update of FCNN parameters.

3. The passive temperature measurement and ranging method based on the multi-spectrum infrared passive temperature measurement and ranging system according to claim 2 is characterized in that: The step (1) constructs a multi-spectral LWIR detection system, as follows: The system consists of an infrared thermal imager, a filter wheel with four narrow-band filters, a full-transmitter and a light-shielding plate, a thermometer and a focusing lens. According to the specific spectral response curve of the infrared thermal imager in the long-wave infrared band, four long-wave infrared narrow-band filters at the peak of the response are selected. The spectral band allowed to pass by the narrow-band filter is very narrow. For the spectral channel i = 1, 2, ..., K, it can be approximately regarded as all the incident radiation is only at the discrete wavelength λ i The detection system controls the filter wheel to rotate through the motor to switch the narrow-band filter and the light shield. The atmospheric temperature is measured through the thermometer. The infrared radiation is focused to the surface of the system through the focusing lens. The shield is used to reduce two special noises in the LWIR imaging process: the "daffodil effect" and 1 / f noise. The shield is introduced into the system design. By quickly switching the narrow-band filter and the shield, the camera captures a pair of images with / without scene radiation, namely the scene and the reference. The two images are subtracted, as shown in formula (1), to obtain an image containing only scene radiation, effectively eliminating the "daffodil effect": I(λ i )=I s (l i )-I r (l i ) (1) Among them, I s (λ i ) and I r (λ i ) are the scene image affected by the "daffodil effect" and the reference image, I(λ i ) is an image containing only scene radiation; By quickly switching narrowband filters and light shields, multiple pairs of images are acquired in a short period of time, and the images are averaged after subtraction, which can significantly suppress 1 / f noise: Where N is the number of image pairs captured, For N I(λ i ) is the image obtained after averaging.

4. The passive temperature measurement and ranging method based on the multi-spectrum infrared passive temperature measurement and ranging system according to claim 2 is characterized in that: The step (2) establishes a multi-spectral infrared temperature and distance measurement model, which is specifically as follows; Step 1: Starting from the object emission term, an ideal black body with a temperature of T emits light with a wavelength of λ. According to Planck's black body radiation law: Where λ is the wavelength of light, c is the speed of light, h is Planck's constant, and k is the Boltzmann constant; The emissivity ε(λ) of an object is defined as the ratio of the radiation emitted by the object to the radiation emitted by a blackbody source at that temperature. Therefore, the radiant energy emitted by an object with emissivity ε(λ) and temperature T is expressed as: E(λ;T)=ε(λ)M e (λ;T) (4) As light propagates through the atmosphere to the sensor, some of it is absorbed. The ratio between the radiation reaching the sensor and the radiation emitted by the object is defined as the transmittance τ air (λ), then the object’s contribution to the observed radiation is given by: E obj (λ;T)=τ air (l)e(l)M e (λ;T) (5) According to Kirchhoff's law of thermal radiation, under thermal equilibrium, the absorptivity is equal to the emissivity, and the emissivity of air is related to the transmittance, that is: e air (λ)=1-τ air (l) (6) According to (6), air at temperature T air The emissions are: E air (λ;T air )=(1-τ air (l))M e (λ;T air ) (7) Among them, T air is the air temperature, obtained by a thermometer; The target will reflect the radiation energy of the surrounding objects, which can be quantified as follows: E ref (λ)=τ air (λ)(1-e(λ))X λ (8) Among them, X λ Indicates the radiation energy of surrounding objects at wavelength λ; Taking into account the contributions of the target object, atmosphere and surrounding environment, the spectrum observed by the measurement system is expressed as: I(λ)=E obj (λ;T)+E air (λ;T air )+E ref (l) (9) Using the Lambert-Beer law, the transmittance of a medium can be expressed as: I out (λ)=exp(-σ air (λ)d)i in (l) (10) where λ in (λ) and i out (λ) are the input and transmitted intensities of the spectrum, σ air (λ) is the extinction coefficient of the medium, d is the thickness of the medium, τ air (λ)=exp(-σ air (λ)d); Based on formulas (3)-(10), the radiation intensity I(λ) at wavelength λ can be expressed as: Among them, R v (λ) is the sensitivity relating energy density to observed intensity; To simplify the calculation, it is assumed that the atmospheric pressure and water vapor content are known and the extinction coefficient is fixed for all λ. Then the parameter σ air (λ) can be obtained by using the high-resolution spectral modeling software SpectralCalc based on the readings of air temperature, humidity and pressure at the weather station; a black body furnace ε(λ) = 1 with a known temperature is placed 0 m in front of the thermal imager, and formula (11) can be simplified to: I(λ)=R v (λ)M e (λ;T) (12) Where I(λ) is the observed radiation intensity, M e (λ;T) can be calculated according to formula (3), and they are all known values. Therefore, R can be calculated according to formula (12): v (λ); Step 2: The radiation energy X of the surrounding objects at wavelength λ λ It is expressed as follows: Among them, α represents the target object, β represents all other objects in the surrounding environment, and V αβ is the view factor, which indicates the ratio of radiation leaving β to α, E βλ represents the infrared radiation emitted by object β, V αβ satisfy: Consider the radiation emitted by a finite number of dominant objects, X λ E in βλ Part is observed as I(λ), and I(λ) is sampled into m spectra to approximately describe the m most important environmental objects. Assume that the size of each thermal image collected is H×W×C, where H is the image height, W is the image width, and C is the number of spectral channels. Considering m environmental objects, the image H×W is spatially divided into m×1 quadrants, and the dimension of each quadrant is Each quadrant is then spatially averaged into a spectrum of length C, i.e., for each of the C channels, a pair of size The sub-images are averaged to obtain their average values. Considering m = 2 environmental objects, the two spectra obtained according to the above method are expressed as E 1λ and E 2λ , is the equivalent object of the environment, now X λ Given by: X λ =V α1 E 1λ +V α2 E 2λ (15) By substituting (15) into (11), the set of unknown parameters to be estimated becomes {d, T, ε(λ i ), V α1 , V α2 }, where i=1, 2,…,K.

5. The passive temperature measurement and ranging method based on the multi-spectrum infrared passive temperature measurement and ranging system according to claim 2 is characterized in that: The construction and solution of the objective function in step (3) are as follows: In formula (11), for the measurement of K spectral channels, there are K+4 unknowns, namely, K unknown values ​​of object emissivity, 1 unknown value of temperature, 1 unknown value of distance, and 2 unknown values ​​of view factor. This is an underdetermined problem. The objective function is used to solve the problem. The minimization objective function between the system observation value and the model calculation value is constructed. The fitting equation of emissivity to wavelength is constructed. The fitting equation is used as the constraint condition of the objective function, which reduces the number of unknown parameters and the search range of the solution. The specific steps include: Step 1: Construct the objective function. The underdetermined problem has insufficient constraints because the number of equations is less than the number of unknowns. The set of solutions to the underdetermined equations has a high degree of freedom and there may be infinite solutions. Therefore, the objective function is introduced for optimization to change the solution selection process from disordered to ordered. The objective function is as follows: Among them, argmin is the input point that makes the function output value reach the minimum value, ||||2 represents the 2-norm of the matrix, and I(λ i ) is the observed radiation intensity, I(λ i ; d, ε(λ i ), T, V α1 , V α2 ) is the radiation intensity calculated according to formula (11); Step 2: Add constraints. A commonly accepted assumption is that the spectral emissivity changes with the wavelength. The spectral emissivity is fitted using a wavelength function containing M, M < K adjustable parameters. The equation is assumed to be: lnε(λ)=a+bλ (18) ε(λ)=a0+a1λ (19) Where K represents the number of spectral channels; Using equation (20) as the constraint condition of emissivity, equation (16) becomes: Step 3: Solving the objective function. First, to avoid the model falling into a local optimal solution, initialize the model parameters to make them as close to the true value as possible. The initialization parameters are as follows: Among them, M e -1 is the inverse function of Planck's blackbody radiation law, K is the number of spectral channels, and U(0,10) represents random sampling from a uniform distribution between 0 and 10 m; Then, construct the Lagrangian function: Among them, η i , i is the value of ε(λ i ) of the inequality constraints, v, ξ, γ α1 , γ α2 is the Lagrange multiplier for other inequality constraints, μ is the equality constraint V α1 +V α2 = 1, are all non-negative values, and for each inequality constraint, they satisfy the complementary relaxation condition: Taking partial derivatives for each variable, we get the KKT gradient condition: Finally, combined with the above conditions, the gradient descent method is used to solve the model parameters.

6. The passive temperature measurement and ranging method based on the multi-spectrum infrared passive temperature measurement and ranging system according to claim 2, characterized in that: The step (4) constructs a temperature and distance correction model, which is as follows; The temperature and distance are corrected using the emissivity information obtained in step (3). First, a correction model network architecture is built. The network used is FCNN. The network input features are the emissivity, temperature and distance information obtained by solving the objective function. The main body of the network consists of multiple fully connected layers to extract the deep nonlinear relationship of the input features. The activation function ReLU is introduced in each layer to enhance the model's ability to fit complex distributions. Finally, the corrected temperature and distance values ​​are generated through the output layer. In order to ensure physical consistency, a loss function that integrates data error terms and physical information error terms is designed to provide feedback guidance for the update of FCNN parameters. Specifically, the following steps are included: Step 1: Verify the rationality of using emissivity to correct temperature and distance. First, calculate the derivative of equation (11) with respect to distance, emissivity and temperature: Eliminate the common term R v (λ)·exp(-σ air (λ)d), the core comparison of the three derivatives simplifies to: Based on equations (22), (34)-(39), the derivatives of equation (22) with respect to each parameter are calculated and simplified. It is found that the order of magnitude of the derivative of equation (22) with respect to emissivity is much greater than the order of magnitude of the derivative of equation (22) with respect to distance and temperature, indicating that the accuracy of emissivity is much higher than that of temperature and distance. Therefore, it is reasonable to use emissivity to correct temperature and distance. Step 2: In order to solve the problem that traditional regression prediction only uses the accuracy of the predicted value as the evaluation indicator, a loss function that integrates the data error term and the physical information error term is designed: Loss=ω1·MSE data +ω2·MSE physics (40) Among them, MSE data is the data error term, which measures the error between the network prediction value and the true value, MSE physics is the physical information error term, which is the constraint residual based on the physical model. ω1 and ω2 are the weight hyperparameters of the data error term and the physical information error term respectively. The data error term is the parameter value d predicted by the network pred and T pred and the true value d ture and T true The specific calculation of the error is as follows: Where N is the number of samples; The physical information error term is the parameter value d predicted by the neural network pred 、T pred Substituting into the mechanism model (11) we obtain I model (λ; d pred , T pred ) and the error of the observed value I(λ), which is calculated as follows: Finally, the total loss is minimized using the gradient descent method and the FCNN parameters are updated.

Images