Packaging paper humidity detection method based on infrared imaging temperature distribution analysis

Abstract

The invention discloses a packaging paper humidity detection method based on infrared imaging temperature distribution analysis, and relates to the technical field of humidity detection.The method comprises the steps that according to the type of packaging paper, a programmable metasurface optical filter array is triggered to conduct dynamic waveband selection, and multi-waveband infrared radiation data of the surface of the packaging paper is collected; generating a temperature and humidity mapping data set by adopting a double-generative adversarial network; a Poincare disk model is used for processing a packaging paper center gentle area, a Lorentz model is used for processing an edge mutation area, and a mixed feature tensor is extracted in a hyperbolic European-style mixed space; humidity distribution reasoning is carried out through a hyperbolic neural network, a packaging paper surface humidity distribution matrix is output, and the hyperbolic neural network comprises a dynamic curvature convolution layer and a hyperbolic attention mechanism. According to the method, the humidity diffusion path is simulated and the anti-noise temperature field is generated by using the double generative adversarial networks at the same time, so that the adaptability to different packaging paper materials and microstructures is enhanced, and the packaging paper humidity detection precision is improved.

Description

Technical Field

[0001] The invention relates to the technical field of humidity detection, in particular to a packaging paper humidity detection method based on infrared imaging temperature distribution analysis. Background Art

[0002] With the growing demand for logistics, storage and industrial packaging, packaging paper plays a key role in protecting all kinds of goods. The moisture state of packaging paper directly affects its mechanical properties and moisture-proof effect. Due to differences in material and structure, the moisture distribution of packaging paper itself is uneven and time-varying. Traditional humidity detection methods mostly use contact measurement methods, which have problems such as low detection accuracy, response delay and environmental interference, and are difficult to meet the requirements of modern production and quality control.

[0003] In addition, non-contact detection technology has gradually attracted attention in recent years. It indirectly reflects humidity information by monitoring the temperature distribution on the surface of packaging paper, but it still faces many challenges in practical applications. Different packaging papers have significant differences in manufacturing process, surface texture and hygroscopicity, which makes the acquisition and analysis of signals during the detection process complicated. How to improve the accuracy and stability of data processing while ensuring detection efficiency has become an important issue that needs to be solved urgently in current scientific research and industry. Summary of the invention

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

[0005] Therefore, the present invention provides a packaging paper humidity detection method based on infrared imaging temperature distribution analysis to solve the problems of low detection accuracy and poor adaptability to packaging paper materials in traditional packaging paper humidity detection.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, the present invention provides a packaging paper moisture detection method based on infrared imaging temperature distribution analysis, which comprises triggering a programmable metasurface filter array to perform dynamic band selection according to the type of packaging paper, and collecting multi-band infrared radiation data on the packaging paper surface; A dual generative adversarial network is used to generate a temperature and humidity mapping dataset, where the first generator simulates the humidity diffusion path and the second generator generates a noise-resistant temperature field through magnetorheological thermal excitation; The Poincare disk model is used to process the flat area in the center of the wrapping paper, and the Lorentz model is used to process the edge mutation area, and the mixed characteristic tensor is extracted in the hyperbolic Euclidean mixing space. The humidity distribution is inferred by a hyperbolic neural network, and a surface humidity distribution matrix of the packaging paper is output. The hyperbolic neural network includes a dynamic curvature convolution layer and a hyperbolic attention mechanism.

[0007] As a preferred solution of the method for detecting the humidity of wrapping paper based on infrared imaging temperature distribution analysis according to the present invention, wherein: the first generator combines the properties of the wrapping paper material and the constraints of the humidity diffusion partial differential equation to generate a humidity diffusion path matching the microstructure of the wrapping paper.

[0008] As a preferred solution of the method for detecting the humidity of wrapping paper based on infrared imaging temperature distribution analysis according to the present invention, wherein: the second generator generates a non-uniform thermal excitation field by regulating the magnetic field distribution and suppresses the temperature field noise in combination with the heat conduction equation.

[0009] As a preferred solution of the method for detecting the humidity of wrapping paper based on infrared imaging temperature distribution analysis according to the present invention, wherein: the dynamic band selection adjusts the transmission band combination of the filter array in real time by analyzing the spectral absorption characteristics of the wrapping paper and compensates the band weights based on the ambient temperature fluctuation.

[0010] As a preferred solution of the method for detecting the humidity of wrapping paper based on infrared imaging temperature distribution analysis according to the present invention, wherein: the steps for extracting the hybrid feature tensor in the hyperbolic Euclidean hybrid space are as follows. Map the infrared radiation data in the gentle area at the center of the wrapping paper to the geodesic coordinate system of the Poincaré disk model, and extract the curvature features in the hyperbolic space. Input the infrared radiation data in the edge mutation area into the Lorentz model for gradient field decomposition, and extract the mutation features in the Euclidean space. Fuse the hyperbolic and Euclidean features through tensor splicing and normalization operations to generate a hybrid feature tensor.

[0011] As a preferred solution of the method for detecting the humidity of wrapping paper based on infrared imaging temperature distribution analysis according to the present invention, wherein: the dynamic curvature convolution layer adaptively adjusts the geometric shape of the convolution kernel according to the curvature change of the local feature manifold, and dynamically constrains the curvature parameters by using the hyperbolic tangent function.

[0012] As a preferred solution of the method for detecting the humidity of wrapping paper based on infrared imaging temperature distribution analysis according to the present invention, wherein: the hyperbolic attention mechanism generates attention weights by calculating the geodesic distance of the hybrid feature tensor in the Poincaré disk, and performs non-linear weighted fusion on the feature channels based on the hyperbolic space projection.

[0013] As a preferred solution of the method for detecting the humidity of wrapping paper based on infrared imaging temperature distribution analysis according to the present invention, wherein: the humidity distribution matrix on the surface of the wrapping paper is used to generate a visualized humidity distribution map and perform humidity detection.

[0014] In a second aspect, the present invention provides a computer device, including a memory and a processor, where the memory stores a computer program, and: when the computer program is executed by the processor, any step of the method for detecting the humidity of wrapping paper based on infrared imaging temperature distribution analysis as described in the first aspect of the present invention is implemented.

[0015] In a third aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and: when the computer program is executed by the processor, any step of the method for detecting the humidity of wrapping paper based on infrared imaging temperature distribution analysis as described in the first aspect of the present invention is implemented.

[0016] The beneficial effects of the present invention are as follows: By using a programmable metasurface filter array to achieve dynamic band selection, accurately capturing multi-band infrared radiation data on the surface of the wrapping paper, the accuracy of humidity detection is improved; at the same time, a dual generative adversarial network is used to simulate the humidity diffusion path and generate a noise-resistant temperature field, enhancing the adaptability to different wrapping paper materials and microstructures; by applying the Poincaré disk model and the Lorentz model to the central flat area and the edge mutation area of the wrapping paper respectively, and extracting the hybrid feature tensor in the hyperbolic Euclidean hybrid space, a fine modeling of complex temperature and humidity distributions is realized; combined with the efficient non-linear feature processing of the dynamic curvature convolution layer and the hyperbolic attention mechanism in the hyperbolic neural network, not only the computational efficiency is optimized, but also the anti-interference ability is improved; the finally output humidity distribution matrix can be directly used to generate an intuitive visualization graph, providing reliable technical support for quality control and detection decision-making, and having a certain industrial integration and application prospect. Description of the Drawings

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for description in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.

[0018] Figure 1 It is the overall framework diagram of the humidity detection of the wrapping paper by infrared imaging temperature distribution analysis in Embodiment 1.

[0019] Figure 2 It is the schematic diagram of the humidity detection process of the wrapping paper by infrared imaging temperature distribution analysis in Embodiment 1.

[0020] Figure 3 It is the schematic diagram of the infrared imaging and humidity mapping relationship in Embodiment 1.

[0021] Figure 4 It is the schematic diagram of the hyperbolic neural network structure in Embodiment 1. Detailed Embodiments

[0022] To make the above objects, features, and advantages of the present invention more apparent and understandable, the following provides a detailed description of the specific embodiments of the present invention with reference to the accompanying drawings of the specification.

[0023] In the following description, numerous specific details are set forth to facilitate a thorough understanding of the present invention. However, the present invention may be implemented in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the spirit of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0024] Secondly, as used herein, "an embodiment" or "embodiments" refers to specific features, structures, or characteristics that may be included in at least one implementation of the present invention. The appearances of "in an embodiment" in different places in this specification do not all refer to the same embodiment, nor are they separate or alternative embodiments that exclude each other.

[0025] Example 1, referring to Figures 1 to 4 , this embodiment provides a method for detecting the humidity of wrapping paper based on infrared imaging temperature distribution analysis, including the following steps: S1: Trigger the programmable metasurface filter array for dynamic band selection according to the wrapping paper category, and collect multi-band infrared radiation data on the surface of the wrapping paper.

[0026] Specifically, it includes the following steps: S1.1: Perform wrapping paper category identification and spectral characteristic matching.

[0027] Specifically, set the wrapping paper category code , where is the total number of preset wrapping paper types. Store the absorption spectra of each type of wrapping paper in the infrared band (3 - 14 μm) . According to the input wrapping paper category code retrieve the corresponding absorption spectrum , and extract its characteristic absorption peak band set , where is the number of effective bands , is the absorption spectrum, is the wrapping paper category code, is the wavelength, is the characteristic absorption peak band set.

[0028] It should be noted that the absorption spectrum It refers to the infrared absorption spectral characteristics of the wrapping paper material in the infrared band (3 - 14 μm), specifically manifested as the absorption rate of the wrapping paper material to infrared radiation at different wavelengths. It is generated by measuring the absorption characteristics of the wrapping paper material in the infrared band through an infrared spectrometer (such as a Fourier transform infrared spectrometer) and matching the characteristic peaks with a standard database (such as the Sadtler spectral library).

[0029] S1.2: Achieve dynamic band selection.

[0030] Specifically, dynamically select the optimal band combination from the absorption peak band set to maximize humidity sensitivity and environmental robustness.

[0031] For each candidate band , calculate its humidity sensitivity factor and environmental interference factor, and the expressions are as follows:

[0032] Among them, is the humidity sensitivity factor, is the environmental interference factor, is the band number index, is the wrapping paper category The infrared radiation absorption rate at the characteristic absorption peak band , is the humidity value of the sample wrapping paper, is the band The full width at half maximum, is the width suppression coefficient , is the radiation variance caused by the environmental temperature in the band , is the environmental temperature, is the characteristic absorption peak band The reciprocal of the signal-to-noise ratio, is the radiation variance weight (can be taken as 0.6), is the signal-to-noise ratio weight (can be taken as 0.4, adjusted through experiments).

[0033] It should be noted that the humidity value of the sample wrapping paper refers to the moisture content of the sample wrapping paper calculated by the oven drying method, and the corresponding sample is selected according to the wrapping paper category.

[0034] Combining the humidity sensitivity factor and the environmental interference factor for dynamic band selection, the band optimization function expression is:

[0035] Among them, is the band optimization value, is the natural exponential function, is the reference baseline wavelength (taking 8 - 10μm), is the wavelength deviation penalty coefficient (which can be taken as 2.5).

[0036] Select the highest 3 bands as the activation bands.

[0037] S1.3: Control the programmable metasurface filter array.

[0038] It should be understood that the metasurface filter array is composed of multiple metasurface units, and each unit can dynamically adjust the transmission wavelength by applying a voltage.

[0039] Specifically, the control strategy is as follows: Divide the metasurface filter array into multiple sub - regions equally according to fixed rows and columns, and assign an activation band to each sub - region; Measure the transmission wavelengths of the metasurface units at different voltages through experiments, establish a look - up table. For the target band corresponding to each sub - region, reverse - query the look - up table to obtain the corresponding voltage, perform voltage mapping on the activation band, and obtain the driving voltage of each sub - region; When the number of activated bands is greater than 1, cycle - switch the bands according to time slices for time - sequence polling.

[0040] S1.4: Perform environmental temperature compensation.

[0041] Specifically, after collecting the original infrared radiation intensity on the surface of the wrapping paper, dynamically compensate for the influence of the environmental temperature through the principle of black - body radiation. Collect the current environmental temperature through a sensor. In the no - wrapping - paper no - load state, pre - record the background radiation baseline values of each band at different environmental temperatures to form a temperature - radiation mapping database. When actually detecting, retrieve the environmental radiation baseline value of the corresponding band from the database according to the current environmental temperature, regard it as the interference component generated by the environmental heat source (such as the equipment's own heat generation, air heat convection), and remove it. Finally, output the infrared radiation data after environmental compensation.

[0042] S1.5: Collect multi - band infrared radiation data on the surface of the wrapping paper.

[0043] It should be understood that the infrared radiation data after environmental temperature compensation and the coordinates of the infrared image are stored in a three - dimensional tensor to obtain the multi - band infrared radiation data on the surface of the wrapping paper.

[0044] Preferably, the corresponding infrared absorption spectrum is coded and matched according to the wrapping paper category, the characteristic absorption peak band is accurately extracted, and the optimal band combination is dynamically screened in combination with humidity sensitivity and environmental robustness to improve the humidity detection sensitivity and reduce the influence of environmental interference. A programmable metasurface filter array is used to dynamically adjust the transmission wavelength by voltage control, and the target band is efficiently activated according to the sub-region allocation and time-slot polling strategy, thereby improving the adaptability and real-time performance of multi-band information acquisition. During the data acquisition process, environmental temperature compensation is performed in combination with the blackbody radiation formula to correct the influence of temperature fluctuations on the infrared radiation data and ensure the stability and accuracy of the measurement results. The compensated infrared radiation data is integrated with the infrared image coordinates, and the multi-band radiation data is stored in matrix form to provide high-quality input for subsequent humidity distribution inference based on deep learning.

[0045] S2: Use a dual generative adversarial network to generate a temperature and humidity mapping dataset.

[0046] Specifically, it includes the following steps: S2.1: Construct a dual generator adversarial network.

[0047] Specifically, the dual generator adversarial network includes a first generator, a second generator, a first discriminator, a second discriminator, and a feature interaction unit.

[0048] The first generator generates a humidity distribution field based on the wrapping paper microstructure and humidity diffusion physical model.

[0049] The second generator generates a noise-resistant temperature field based on the magnetorheological heat excitation and heat conduction model.

[0050] The first discriminator discriminates the authenticity of the humidity distribution field.

[0051] The second discriminator discriminates the authenticity of the noise-resistant temperature field.

[0052] The feature interaction unit realizes the feature coupling between the first generator and the second generator through cross-modal attention.

[0053] Furthermore, the input of the dual generator adversarial network is the multi-band infrared radiation data on the wrapping paper surface, the magnetic field distribution on the wrapping paper surface, and the wrapping paper material parameters.

[0054] The magnetic field distribution on the wrapping paper surface is obtained by using the finite element simulation method. First, a computational domain is established, and parameters are set according to the wrapping paper material properties. At the same time, the calculation accuracy is optimized by using a non-uniform adaptive grid. During the simulation process, an external alternating magnetic field is applied as the excitation source, and boundary conditions such as zero magnetic scalar potential boundary and absorption boundary are set. The finite element method is used to calculate the steady-state or alternating magnetic field distribution at different positions on the wrapping paper surface, and the magnetic field data is interpolated and smoothed. Finally, the magnetic field distribution matrix on the wrapping paper surface is obtained.

[0055] The wrapping paper material parameters include porosity, fiber orientation, and thickness.

[0056] The outputs are a normalized humidity distribution matrix and a temperature field matrix.

[0057] S2.2: Modeling the humidity diffusion path of the first generator.

[0058] Specifically, the encoder uses 5-layer 3D convolution to extract the geometric features of the wrapping paper material parameters.

[0059] The PDE is discretized using the implicit finite difference method, a residual constraint term is constructed, and a physical residual is added to the loss function. The expression is:

[0060] where is the physical residual loss, is the total number of samples for calculating the physical residual, is the time variable, is the humidity diffusion term based on Fick's law, is the gradient, is the diffusion coefficient, are the wrapping paper material parameters, is the spatial abscissa, is the spatial ordinate, is the humidity source term generated by the first generator.

[0061] The decoder uses transposed convolution to generate the humidity distribution matrix, and the output layer is Sigmoid activation.

[0062] S2.3: Generating the noise-resistant temperature field of the second generator.

[0063] Specifically, the input multi-band infrared radiation data on the wrapping paper surface is used to extract multi-band features by the Inception unit, and the magnetic field distribution on the wrapping paper surface is encoded as a spatial modulation vector through a fully connected layer.

[0064] Based on the magnetic field distribution on the wrapping paper surface, the thermal excitation power density is regulated, and the heat conduction equation constraint is embedded in the temperature field generation. The expression is:

[0065] where is the temperature on the wrapping paper surface, is the thermal diffusivity, is the thermal excitation power density, is the material density, is the specific heat capacity, is the noise suppression factor (optimized through adversarial training), is the Gaussian noise field, is the coefficient of heat conversion efficiency , is the vacuum permeability, is the magnetic susceptibility.

[0066] The U-Net structure is used, and the skip connection injects the magnetic field modulation information. The output layer uses the Tanh activation and maps to the temperature range .

[0067] S2.4: Set the adversarial training strategy.

[0068] It should be understood that the discriminant loss is generated according to the surface temperature of the wrapping paper generated by the second generator and the surface humidity of the wrapping paper generated by the first generator, and the weighted sum is performed to establish the total loss function for joint optimization.

[0069] The total loss function includes the discriminant loss of the first discriminator, the discriminant loss of the second discriminator, the physical residual loss, and the temperature field reconstruction loss, and the weight coefficients are set to 1, 1, 0.5, and 0.2 respectively.

[0070] S2.5: Generate the temperature and humidity mapping dataset.

[0071] Specifically, the multi-band infrared radiation data on the surface of the wrapping paper is standardized, and the magnetic field distribution on the surface of the wrapping paper is obtained through finite element simulation.

[0072] The first generator, the second generator, the first discriminator, and the second discriminator are alternately updated, and the Adam optimizer is used. After each round of training, the sample pairs generated by the first generator and the second generator are frozen.

[0073] Random affine transformation is applied to the generated surface temperature of the wrapping paper, and random pulse noise is injected into the generated surface temperature of the wrapping paper.

[0074] The temperature and humidity mapping dataset includes the multi-band infrared radiation data on the surface of the wrapping paper, the magnetic field distribution on the surface of the wrapping paper, the generated surface temperature of the wrapping paper, and the generated surface humidity of the wrapping paper. The temperature and humidity mapping dataset is divided into a training set, a validation set, and a test set in a ratio of 70:15:15.

[0075] Preferably, the dual-generator architecture combines the humidity diffusion physical model and the magnetorheological heat conduction model, making the generation of the humidity distribution field and the noise-resistant temperature field more in line with the physical characteristics of the wrapping paper material, and enhancing the physical consistency and generalization ability of the architecture. The first generator models based on the microstructure of the wrapping paper and the humidity diffusion path, introducing physical residual constraints to ensure that the generated humidity distribution conforms to Fick's law. The second generator combines the thermal excitation mechanism regulated by the magnetic field and the heat conduction equation, making the temperature field generation more noise-resistant. The feature interaction unit enables the temperature and humidity data to interact with each other during the generation process through the cross-modal attention mechanism, enhancing the authenticity and robustness of the generation. Through the adversarial training strategy, the adversarial loss, physical residual loss, and temperature reconstruction loss are comprehensively optimized to make the generation of higher quality. Combining finite element simulation and random affine transformation and impulse noise injection ensures that the generated temperature and humidity mapping dataset has higher diversity and robustness, and reasonably divides the training set, validation set, and test set to ensure the rationality of training.

[0076] S3: Use the Poincaré disk model to process the flat central area of the wrapping paper and the Lorentz model to process the edge mutation area, and extract the mixed feature tensor in the hyperbolic-Euclidean hybrid space.

[0077] Specifically, it includes the following steps: S3.1: Perform region division and preprocessing.

[0078] Specifically, based on the temperature and humidity mapping dataset, for each band , use the Sobel operator to calculate the radiation intensity gradient amplitude , and the expression is:

[0079] Among them, is the infrared radiation data of band , is the radiation intensity gradient amplitude, is the band index.

[0080] Set the gradient threshold , where is the mean value of the gradient amplitude, is the standard deviation of the gradient amplitude.

[0081] According to the gradient threshold, divide the central flat area and the edge mutation area, and the expression is:

[0082] Among them, is the central flat area, is the edge mutation area, coordinate at the radiation intensity gradient amplitude, is the gradient threshold.

[0083] S3.2: Process the central flat area through the Poincaré disk model.

[0084] It should be understood that the goal of this step is to map the flat area data to the hyperbolic space and capture the hierarchical humidity diffusion pattern.

[0085] Specifically, for the central flat area of the infrared radiation data perform normalization and map it to the Poincaré disk , and the expression is:

[0086] where is the hyperbolic space mapping point, is 's mean value, is 's standard deviation, is the natural base, is the imaginary unit, satisfying , is the local gradient direction angle.

[0087] Ensure through the scaling factor.

[0088] Furthermore, construct a geodesic coordinate system, define the hyperbolic metric tensor and calculate the local curvature feature, and the expression is:

[0089] where is the hyperbolic metric tensor, is the local curvature feature.

[0090] Extract the curvature tensor , where is the curvature dimension, is the set of real numbers.

[0091] S3.3: Process the edge mutation area through the Lorentz model.

[0092] It should be understood that the goal of this step is to decompose the gradient field in the Euclidean space and capture the edge humidity mutation feature.

[0093] Specifically, for the infrared radiation data of the edge mutation area , construct the tangent space projection on the Lorentz manifold , define the pseudo-Riemannian metric and decompose the gradient into the normal vector and the tangent vector, and the expression is:

[0094] Among them, is the normal vector, is the tangent vector, is the random perturbation vector.

[0095] Furthermore, according to the normal vector and the tangent vector, the covariant derivative is calculated, and the expression is:

[0096] Among them, is the component of the tangent vector in the direction, is the component of the tangent vector in the direction. Extract the mutation feature tensor .

[0097] S3.4: Generate the hybrid feature tensor.

[0098] It should be understood that the goal of this step is to fuse hyperbolic and Euclidean features and maintain geometric consistency.

[0099] Specifically, perform a logarithmic mapping on and apply a hyperbolic to Euclidean projection. Normalize .

[0100] Concatenate along the spatial dimension and perform channel normalization to extract the hybrid feature tensor .

[0101] Preferably, based on gradient magnitude analysis, the central smooth region and the edge mutation region are divided to adapt to the geometric characteristics of different regions. For the central smooth region, the Poincaré disk model is used to map the infrared radiation data to the hyperbolic space, define the hyperbolic metric tensor and calculate the local curvature features, so as to capture the hierarchical humidity diffusion pattern. For the edge mutation region, the Lorentz manifold is used to decompose the gradient field, extract the normal vector and the tangent vector, and calculate the covariant derivative to analyze the mutation characteristics of the humidity change. By applying the hyperbolic to Euclidean projection, a unified hybrid feature tensor is formed. It effectively combines the hierarchical information expression ability of the hyperbolic space with the local change analysis ability of the Euclidean space, enhances the modeling ability for complex humidity distribution patterns, and provides accurate and efficient support for the humidity monitoring and quality control of wrapping paper.

[0102] S4: Perform humidity distribution inference through a hyperbolic neural network and output the humidity distribution matrix on the surface of the wrapping paper.

[0103] Specifically, it includes the following steps: S4.1: Construct a hyperbolic neural network.

[0104] Specifically, the input of the hyperbolic neural network is the hybrid feature tensor, and the output is the humidity distribution matrix on the surface of the wrapping paper.

[0105] The structure of the hyperbolic neural network includes: a dynamic curvature convolution layer (3 layers) for adaptively adjusting the geometric shape of the convolution kernel; a hyperbolic attention unit (2 layers) for cross-channel feature weighted fusion; a hyperbolic pooling layer for downsampling and maintaining the hyperbolic manifold structure; and a hyperbolic regression head for mapping to the Euclidean space to output the humidity matrix.

[0106] S4.2: Adjust the dynamic curvature convolution layer.

[0107] Specifically, perform a hyperbolic tangent constraint on the input mixed feature tensor and calculate the local curvature. The expression is:

[0108] Among them, is the local curvature, is the mixed feature tensor, is the Poincaré disk logarithmic mapping.

[0109] Deform the base kernel according to the local curvature to obtain a curvature adaptive convolution kernel. The expression is:

[0110] Among them, is the curvature adaptive convolution kernel, is the base kernel, is the learnable scaling coefficient, , is the element-wise product.

[0111] Furthermore, based on the curvature adaptive convolution kernel, perform a convolution operation in the Poincaré disk for hyperbolic convolution.

[0112] S4.3: Calculate the cross-channel attention weight through the hyperbolic attention mechanism.

[0113] Specifically, calculate the geodesic distance. The expression is:

[0114] Among them, is the geodesic distance, is the feature channel, is another feature channel, is the inverse hyperbolic cosine function.

[0115] Calculate the similarity weight between feature channels based on the geodesic distance and normalize it. The expression is:

[0116] Among them, is the similarity weight between feature channels, is the number of channels, is the channel number index, and is the feature dimension.

[0117] Based on the similarity between channels, weighted aggregation of features is performed for hyperbolic feature fusion.

[0118] S4.4: Map the hyperbolic features to the Euclidean space and output the humidity distribution matrix on the surface of the wrapping paper.

[0119] Specifically, perform a geometric projection on the output feature tensor. Through the Poincaré logarithmic mapping based on the origin, project each hyperbolic feature point radially onto the tangent space at the origin. This process uses the inverse hyperbolic tangent function to calculate the scaling factor and normalize the original features to ensure that the projected features are completely in the Euclidean space.

[0120] Adjust the spatial dimension, and expand the projected feature map through bilinear interpolation. After aligning the feature dimensions, perform a non-linear transformation on the feature vector of each pixel point. Compress the multi-channel information into a single-channel humidity prediction value. Apply the Sigmoid activation function to the output of each pixel to constrain the value between 0 and 1, corresponding to the dry to wet state of the surface of the wrapping paper.

[0121] The finally output humidity distribution matrix, where each matrix element corresponds to the normalized humidity value at a specific position on the wrapping paper.

[0122] S4.5: Set the loss function and training strategy.

[0123] Specifically, the expression of the humidity distribution loss (main loss) is:

[0124] where, is the humidity distribution loss function, is the width of the humidity distribution matrix, is the depth of the humidity distribution matrix, is the predicted humidity value at the coordinate , is the true humidity value at the coordinate .

[0125] The expression of the hyperbolic manifold constraint loss (regular term) is:

[0126] where, is the hyperbolic manifold constraint loss function, is the number of layers of the neural network, is the index of the number of layers of the neural network, is the hidden representation of the -th layer neural network in the hyperbolic feature space, is the curvature constant of the hyperbolic space, is the characteristic radius of the

[0127] The Riemannian Adam optimizer is adopted, and the learning rate is set to , and the momentum is set to 0.9 and 0.999.

[0128] Preferably, the state curvature convolution layer adaptively adjusts the convolution kernel morphology according to the local curvature to more accurately capture the humidity characteristics in the non-Euclidean space. The hyperbolic attention mechanism calculates the similarity between channels through the geodesic distance and weighted fuses the information of different feature channels to improve the feature expression ability. The hyperbolic features are projected into the Euclidean space through logarithmic mapping, and the fully connected layer is used to generate the humidity distribution matrix, so as to ensure that the output meets the actual application requirements. During the training process, the humidity distribution loss function constrains the error between the predicted value and the true value, while the hyperbolic manifold constraint loss further maintains the manifold structure of the network in the hyperbolic space and improves the generalization ability. Combining with the efficient training strategy of the Riemannian Adam optimizer can improve the convergence speed and stability of the hyperbolic neural network while maintaining the hyperbolic geometry consistency.

[0129] This embodiment also provides a computer device, applicable to the case of the packaging paper humidity detection method based on infrared imaging temperature distribution analysis, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to implement the packaging paper humidity detection method based on infrared imaging temperature distribution analysis as proposed in the above embodiment.

[0130] This computer device can be a terminal. This computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of this computer device is used to provide computing and control capabilities. The memory of this computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of this computer device is used to communicate with external terminals in a wired or wireless manner, and the wireless manner can be implemented through WIFI, operator network, NFC (Near Field Communication) or other technologies. The display screen of this computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of this computer device can be a touch layer covered on the display screen, or a button, a trackball or a touchpad set on the computer device shell, or an external keyboard, touchpad or mouse, etc.

[0131] This embodiment also provides a storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the method for detecting the humidity of wrapping paper based on infrared imaging temperature distribution analysis proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM for short), Electrically Erasable Programmable Read-Only Memory (EEPROM for short), Erasable Programmable Read Only Memory (EPROM for short), Programmable Read-Only Memory (PROM for short), Read-Only Memory (ROM for short), magnetic memory, flash memory, magnetic disk or optical disc.

[0132] In summary, the present invention: realizes dynamic band selection by adopting a programmable metasurface filter array, accurately captures multi-band infrared radiation data on the surface of the wrapping paper, and improves the accuracy of humidity detection; at the same time, uses a dual generative adversarial network to simulate the humidity diffusion path and generate a noise-resistant temperature field, enhancing the adaptability to different wrapping paper materials and microstructures; by applying the Poincaré disk model and the Lorentz model to the flat area in the center and the mutation area at the edge of the wrapping paper respectively, and extracting the mixed feature tensor in the hyperbolic Euclidean hybrid space, realizes the fine modeling of complex temperature and humidity distributions; combines the efficient non-linear feature processing of the dynamic curvature convolutional layer and the hyperbolic attention mechanism in the hyperbolic neural network, not only optimizes the calculation efficiency, but also improves the anti-interference ability; the finally output humidity distribution matrix can be directly used to generate an intuitive visualization graph, providing reliable technical support for quality control and detection decision-making, and having a certain industrial integration and application prospect.

[0133] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A method for detecting the humidity of wrapping paper based on infrared imaging temperature distribution analysis, characterized in that: Including, Triggering a programmable metasurface filter array for dynamic band selection according to the wrapper category, and collecting multi-band infrared radiation data on the surface of the wrapper; Using a dual generative adversarial network to generate a temperature and humidity mapping dataset, where the first generator simulates the humidity diffusion path, and the second generator generates a noise-resistant temperature field through magnetorheological thermal excitation; Using the Poincaré disk model to process the flat area at the center of the wrapper, and using the Lorentz model to process the edge mutation area, and extracting a mixed feature tensor in the hyperbolic-Euclidean hybrid space; Performing humidity distribution inference through a hyperbolic neural network, and outputting a humidity distribution matrix on the surface of the wrapper, where the hyperbolic neural network includes a dynamic curvature convolutional layer and a hyperbolic attention mechanism.

2. The humidity detection method of wrapping paper based on infrared imaging temperature distribution analysis according to claim 1, characterized in that: The first generator combines the material properties of the wrapper and the constraints of the humidity diffusion partial differential equation to generate a humidity diffusion path matching the microstructure of the wrapper.

3. The humidity detection method for wrapping paper based on infrared imaging temperature distribution analysis according to claim 1, wherein: The second generator generates a non-uniform thermal excitation field by regulating the magnetic field distribution, and suppresses the temperature field noise in combination with the heat conduction equation.

4. The humidity detection method of wrapping paper based on infrared imaging temperature distribution analysis according to claim 1, wherein: The dynamic band selection adjusts the transmission band combination of the filter array in real time by analyzing the spectral absorption characteristics of the wrapper, and compensates the band weights based on the ambient temperature fluctuation.

5. The humidity detection method of wrapping paper based on infrared imaging temperature distribution analysis according to claim 1, characterized in that: The step of extracting the mixed feature tensor in the hyperbolic-Euclidean hybrid space is specifically as follows: Mapping the infrared radiation data of the flat area at the center of the wrapper to the geodesic coordinate system of the Poincaré disk model, and extracting the curvature feature in the hyperbolic space; Inputting the infrared radiation data of the edge mutation area into the Lorentz model for gradient field decomposition, and extracting the mutation feature in the Euclidean space; Fusing the hyperbolic and Euclidean features through tensor splicing and normalization operations to generate a mixed feature tensor.

6. The humidity detection method for wrapping paper based on infrared imaging temperature distribution analysis according to claim 1, wherein: The dynamic curvature convolutional layer adaptively adjusts the geometric shape of the convolutional kernel through the curvature change of the local feature manifold, and dynamically constrains the curvature parameter using the hyperbolic tangent function.

7. The humidity detection method of wrapping paper based on infrared imaging temperature distribution analysis according to claim 1, characterized in that: The hyperbolic attention mechanism generates attention weights by calculating the geodesic distance of the mixed feature tensor in the Poincaré disk, and performs non-linear weighted fusion on the feature channels based on the hyperbolic space projection.

8. The humidity detection method of wrapping paper based on infrared imaging temperature distribution analysis according to claim 1, wherein: The humidity distribution matrix on the surface of the wrapper is used to generate a visualized humidity distribution map and perform humidity detection.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the wrapper humidity detection method based on infrared imaging temperature distribution analysis according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the wrapper humidity detection method based on infrared imaging temperature distribution analysis according to any one of claims 1 to 8.

Images