Tobacco moisture meter self-adaptive calibration equipment and calibration method thereof
By using a polarization camera and calibration standard plate in a tobacco moisture meter, analyzing spectral attenuation and polarization images, combined with PLS modeling, the problem of stability and accuracy of spectrometers in tobacco moisture measurement is solved, achieving more efficient moisture measurement.
Patent Information
- Application Number
- CN202410962354.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-18
- Publication Date
- 2024-11-26
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
In tobacco moisture measurement, the stability and accuracy of the measurement results are difficult to maintain due to light source attenuation, equipment aging and sample attribute changes, especially the differences in chemical composition and leaf texture structure of different types of tobacco leaves affect the measurement results.
Adaptive calibration equipment and methods of tobacco moisture meter are adopted, including setting up a polarization camera and calibrating black and white diffuse reflection standard plates, simulating light source attenuation and analyzing polarized images, constructing a relationship mapping table, and combining PLS modeling to achieve moisture measurement.
It improves the stability and accuracy of moisture measurement, suppresses spectral numerical differences, and achieves fast and accurate moisture measurement.
Smart Images

Figure CN119023612A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of spectrometers, and in particular to a tobacco moisture meter self-adaptive calibration device and a calibration method thereof. Background Art
[0002] Spectrometers have been widely used in all aspects of tobacco production. In particular, the use of spectrometers to measure the moisture content in tobacco is a core indicator in production process control. Accurately measuring the moisture content in raw materials such as tobacco and cut tobacco is of great significance for lean control of production parameters. However, during the use of spectrometers, due to the influence of light source attenuation, component aging, environmental factors and changes in sample properties, it is difficult to maintain the accuracy of the numerical value when measuring moisture, and the spectra collected by different spectrometers need to be calibrated.
[0003] During the routine use of moisture meters, in view of the influence of environmental factors, the equipment is usually equipped with special dehumidification and temperature control devices to ensure the reasonable working conditions of the equipment components, so the influence is controllable; for the problem of equipment aging, the functional degradation of components is usually a relatively slow process during the normal use cycle of the equipment, and its influence on accuracy can be ignored. Since the light source (such as halogen tungsten lamp) used by the near-infrared spectrometer will gradually decay over time, the light intensity and spectral characteristics emitted by it may change, thus affecting the stability and accuracy of the measurement results. In addition, different types of tobacco leaves tested also have differences in chemical composition, including differences in the content and proportion of carbohydrates, proteins, volatile substances, etc., which may affect the hygroscopicity and moisture content of tobacco leaves, and thus affect the accuracy of moisture measurement. In addition, the texture structure and color characteristics of different types of tobacco leaves may affect the distribution and diffusion of moisture in the leaves, affect the physical depth of the transmission layer presented by the reflection spectrum and the sensing system, affect the reaction of the sensing system to the sample surface, and then affect the absorption and scattering of near-infrared light by the sample, resulting in spectral variation.
[0004] Therefore, in the actual production process, the accuracy of moisture measurement results based on near-infrared spectrometers needs to give priority to the differences in light source attenuation and the apparent color and material structure characteristics of the target being measured, so as to establish a model to optimize the accuracy and stability of moisture testing. Summary of the invention
[0005] The purpose of the present invention is to solve the defects in the prior art and to propose a tobacco moisture meter self-adaptive calibration device.
[0006] In order to achieve the above object, the present invention adopts the following technical solutions:
[0007] A tobacco moisture meter self-adaptive calibration device comprises a platform, wherein a tobacco leaf body to be tested is arranged in the middle of the platform, and a calibration black and white diffuse reflection standard plate is arranged on a side away from the tobacco leaf body to be tested, wherein the calibration black and white diffuse reflection standard plate is arranged on one side of the platform;
[0008] The industrial computer is set on the top of the platform and has a distance between it and the platform;
[0009] Polarization camera a is electrically connected to the industrial computer, and the direction of polarization camera a is 0°;
[0010] The polarization camera b is electrically connected to the industrial computer, and the direction of the polarization camera b is 45°;
[0011] The polarization camera c is electrically connected to the industrial computer, and the direction of the polarization camera c is 90°;
[0012] The polarization camera d is electrically connected to the industrial computer, and the direction of the polarization camera d is 135°;
[0013] The tungsten-halogen lamp light source is arranged on one side of the polarization camera c and is used to illuminate the platform.
[0014] A tobacco moisture meter adaptive calibration method, using a tobacco moisture meter adaptive calibration device as claimed in claim 1, comprising the following steps:
[0015] S1, using glass slides with different light transmittances to mask the active light source, so as to simulate the intensity attenuation state of the halogen tungsten light source (7), and by selecting glass slides with different light transmittances, a graded simulation of the intensity attenuation degree of the light source is achieved;
[0016] S2. Under the light source attenuation conditions set at each level, the light intensity values of the black and white diffuse reflection standard plates are obtained by the polarization camera under the standard state, and the standard reflection under the standard state is obtained, which is used as the sparse channel reflection reference here;
[0017] S3, respectively obtain the spectral reflectance values of the black block and the white diffuse reflectance standard plate through a near-infrared spectrometer under standard conditions, obtain the standard reflectance of the diffuse reflectance standard plate under standard conditions, and take an average value through N measurements, which is referred to as a dense channel reflectance reference;
[0018] S4. Construct a relationship mapping table from sparse channel reflection reference to dense channel standard reflection reference <F P ,F S >. Using sparse F P The feature query obtains the F that represents the standard reflectance of K spectral channels S ;
[0019] S5. Obtain the black and white squares on the standard plate captured by the polarization camera and calculate the sparse channel standard reference Combined with fast index mapping table <F P ,F S >, use Same as F in the mapping table P The Euclidean distance of the numerical difference is used to retrieve the closest index item and obtain the reflectance reference of the spectrometer standard plate;
[0020] S6. Combined with the spectral reflectance of the actual sample measured by the spectrometer, the spectrometer standard plate reflectance reference is used to directly compare the channel by channel to obtain a more accurate spectral absorption line. The relationship between the spectral absorption line and the water content of the measured target can be modeled by the classic PLS to achieve fast and accurate measurement of moisture.
[0021] Preferably, in S2, a polarization camera is used to obtain N black and white squares on the standard plate, and the black squares White block Number them and count the mean pixel response of each channel in each block and
[0022] Extract the spectral reflection intensity value with statistical significance. Obtain the spectral reflection intensity values of the four channels r, g, b, and ir respectively. FP(r ,g,b,ir), the calculation formula is as follows:
[0023]
[0024] Preferably, the formula for the dense channel reflection reference in S3 is:
[0025]
[0026] in represents the reflection response of the white diffuse reflection standard plate under the i-th measurement condition, where (c1, c2, ..., c K ) is the spectral channel, with a total of K spectral channels.
[0027] Preferably, the polarization image obtained by the polarization color camera is used to measure the surface roughness structure of the tobacco leaf material in the visible area. The combined four-directional polarization camera device is used to analyze the polarization degree and polarization angle of the measured target, so as to measure the texture structure of the surface of the measured target tobacco leaf, and the moisture measurement result is corrected in combination with the texture structure characteristics.
[0028] Combine four sets of polarization cameras to obtain the polarization degree and polarization angle of the target to be measured, and quantitatively analyze the changes in the geometric shape of the object surface (texture characteristics of the target surface) reflected by the polarization state of light after it is reflected from the object surface;
[0029] The Stokes vector and measurement method are used to describe the four-dimensional vector of the polarization state of light. In practice, a polarizer is usually added in front of the detector, and the intensity of light waves with different polarization angles is obtained by rotating the polarizer, and then calculated;
[0030] When measuring, the linear polarizer is rotated by θ, and the phase delay angle β is obtained to obtain the measured image area. I( θ, β). The rotation angle values are: 0°, 45°, 90° and 135°, and the phase delay angle value is: 0, The light intensity values are further obtained to calculate the Stokes parameters.
[0031]
[0032] For the calculation of the linear polarization degree P, the formula is:
[0033]
[0034] The definition of the linear polarization angle φ can be expressed by the Stokes parameter:
[0035]
[0036] Finally, the calculated results of the degree of polarization P and polarization angle φ given different polarizer transmission angles can be obtained by the following equations:
[0037]
[0038] The polarization degree and polarization angle of each pixel in the image area are calculated, the average polarization degree and polarization angle of the entire area are counted, and the variance of the polarization degree distribution and polarization angle distribution is counted.
[0039] Using the above principle, the polarization degree mean P is collected avg , mean polarization angle Polarization degree distribution variance σ avg and the polarization angle distribution variance and the corresponding deviation of the moisture measurement from the mean y Δ , construct a sample of measured data.
[0040] Assuming there are m samples, n polarization degrees and polarization angle distribution variances as input variables, the corresponding moisture measurement error is yΔ As the output variable, the established PLS model can be expressed as:
[0041] Y=TP T +E
[0042] Where Y is the matrix of moisture measurement error (m×1), T is the matrix of distribution variance of polarization degree and polarization angle (m×k), P is the coefficient matrix of PLS model (k×n), and E is the residual matrix (m×1)
[0043] Preferably, the polarized image acquired by the color camera is used to classify and analyze the tobacco leaf material in the visible area, the abundance of each classification is analyzed, and a combined model is used to achieve high-precision moisture determination.
[0044] All tobacco leaves to be tested are manually classified based on color and assumed to be divided into H categories. Tobacco leaf samples are screened based on H category color models, and H moisture prediction models suitable for H category tobacco leaf samples are constructed respectively.
[0045] During actual measurement, a color polarization camera is used to perform superpixel segmentation on the current visible area using the SLIC algorithm, and based on color and spatial distribution, statistics are collected to determine which tobacco leaf types exist in the current sample. The specific steps are as follows:
[0046] (1) Set the number of superpixels K and calculate the superpixel size N / K based on the number of image pixels N.
[0047] (2) Set the initial grid, calculate the gradient of the neighborhood of the central pixel of the grid, and set the pixel with the minimum gradient as the initial clustering seed point.
[0048] (3) Use the formula to convert the regional color from RGB color space to CIELab color space.
[0049]
[0050] (4) In the SLIC algorithm, the color information and position information of the pixel are represented in the five-dimensional space {x, y, l, a, b}. The color and spatial feature distance between pixels in the image is as follows:
[0051]
[0052] Normalize it to get the similarity d s :
[0053]
[0054] (5) Move the cluster center to the mean center of the pixel area with the same label, and repeat step (4) until the error is less than the set threshold or the maximum number of iterations is reached.
[0055] (6) Small scattered areas that are missed in the clustering process are merged into nearby superpixels. Using the superpixel segmentation strategy, the algorithm performs segmentation on the output image after edge enhancement. It can be clearly seen that a smaller superpixel size can obtain a segmentation result with a closer segmentation edge. This solution can effectively ensure the accuracy of the logical segmentation result.
[0056] The mean color value of tobacco leaves in each segmented area is counted, and the KMeans algorithm is used to map each superpixel segmentation block to H tobacco leaf classifications;
[0057] S(cls,area)=KM(SLIC(I))
[0058] Where KM(x) is the KMeans algorithm, SLIC is the segmentation model, I is the tobacco leaf image to be analyzed, S(cls,area) Tobacco leaf types classified based on color and the corresponding pixel areas.
[0059]
[0060] Based on the area proportion of each tobacco leaf classification, the abundance proportion of the material is constructed. Based on the abundance proportion, combined with various moisture measurement models, the fused moisture measurement value is obtained.
[0061] In addition, X is the spectral absorption line of the current case being tested, F i (X) It is a water content prediction model based on spectral absorption lines. T The actual water content.
[0062] T=η i F i (X).
[0063] Compared with the prior art, the beneficial effects of the present invention are: by setting multiple polarization cameras and using the control of an industrial computer, a calibration black and white diffuse reflection standard plate is set on the platform to perform calibration analysis on the tobacco leaf body to be tested, so as to suppress the spectral value difference caused by the difference in tobacco leaf color and material structure characteristics under different light source attenuation conditions, thereby improving the numerical stability and accuracy of moisture measurement;
[0064] Combined with the spectral reflectance of the actual sample measured by the spectrometer, the spectrometer standard plate reflectance reference is used to directly calculate the channel ratio to obtain a more accurate spectral absorption line. The relationship between the spectral absorption line and the water content of the measured target can be modeled using the classic PLS to achieve fast and accurate measurement of moisture;
[0065] The polarization image obtained by the polarization color camera is used to measure the surface roughness structure of the tobacco leaf material in the visible area. The polarization degree and polarization angle of the target are analyzed by the combined four-directional polarization camera device to measure the texture structure of the target tobacco leaf surface, and the moisture measurement result is corrected by combining the texture structure characteristics;
[0066] Polarized images acquired by a color camera are used to classify and analyze tobacco leaf materials in the visible area, the abundance of each classification is analyzed, and a combined model is used to achieve high-precision moisture determination. BRIEF DESCRIPTION OF THE DRAWINGS
[0067] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.
[0068] Figure 1 It is a schematic diagram of the overall structure;
[0069] Figure 2 This is a schematic diagram of the polarization direction structure in the polarization camera;
[0070] Figure 3 is the dotted spectrum;
[0071] Figure 4 It is the absorption spectrum of tobacco leaf material after reference calibration;
[0072] Figure 5 Test spectral response diagrams for reference and standard cases under different light source attenuation conditions;
[0073] Figure 6 It is a mapping table for different absorbances;
[0074] Figure 7 This is the result of super-pixel segmentation based on enhanced edge;
[0075] Figure 8 The following is the overall flow chart of the system.
[0076] In the figure: 1. Tobacco leaf to be tested; 2. Calibrated black and white diffuse reflection standard plates; 3. Polarization camera a; 4. Polarization camera b; 5. Polarization camera c; 6. Polarization camera d; 7. Halogen tungsten light source; 8. Industrial computer. DETAILED DESCRIPTION
[0077] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention;
[0078] Reference Figure 1, a tobacco moisture meter adaptive calibration device, comprising a platform, a tobacco leaf body 1 to be tested is arranged in the middle of the platform, a calibration black and white diffuse reflection standard plate 2 is arranged on a side away from the tobacco leaf body 1 to be tested, and the calibration black and white diffuse reflection standard plate 2 is arranged on one side of the platform;
[0079] The industrial computer 8 is arranged on the top of the platform and has a distance between it and the platform;
[0080] The polarization camera a3 is electrically connected to the industrial computer 8, and the direction of the polarization camera a3 is 0°;
[0081] The polarization camera b4 is electrically connected to the industrial computer 8, and the direction of the polarization camera b4 is 45°;
[0082] The polarization camera c5 is electrically connected to the industrial computer 8, and the direction of the polarization camera c5 is 90°;
[0083] Polarization camera d6 is electrically connected to industrial computer 8, and the direction of polarization camera d6 is 135°'
[0084] The tungsten halogen lamp light source 7 is arranged on one side of the polarization camera c5 and is used to illuminate the platform.
[0085] Reference Figure 2-8 A tobacco moisture meter adaptive calibration method, using a tobacco moisture meter adaptive calibration device as claimed in claim 1, comprising the following steps:
[0086] S1, using glass slides with different light transmittances to mask the active light source, so as to simulate the intensity attenuation state of the halogen tungsten light source (7), and by selecting glass slides with different light transmittances, a graded simulation of the intensity attenuation degree of the light source is achieved;
[0087] S2. Under the light source attenuation conditions set at each level, the light intensity values of the black and white diffuse reflection standard plates are obtained by the polarization camera under the standard state, and the standard reflection under the standard state is obtained, which is used as the sparse channel reflection reference here;
[0088] S3, respectively obtain the spectral reflectance values of the black block and the white diffuse reflectance standard plate through a near-infrared spectrometer under standard conditions, obtain the standard reflectance of the diffuse reflectance standard plate under standard conditions, and take an average value through N measurements, which is referred to as a dense channel reflectance reference;
[0089] S4. Construct a relationship mapping table from sparse channel reflection reference to dense channel standard reflection reference <F P ,F S >. Using sparse F P The feature query obtains the F that represents the standard reflectance of K spectral channels S ;
[0090] S5. Obtain the black and white squares on the standard plate captured by the polarization camera and calculate the sparse channel standard reference Combined with fast index mapping table <F P ,F S >, use Same as F in the mapping table P The Euclidean distance of the numerical difference is used to retrieve the closest index item and obtain the reflectance reference of the spectrometer standard plate;
[0091] S6. Combined with the spectral reflectance of the actual sample measured by the spectrometer, the spectrometer standard plate reflectance reference is used to directly compare the channel by channel to obtain a more accurate spectral absorption line. The relationship between the spectral absorption line and the water content of the measured target can be modeled by the classic PLS to achieve fast and accurate measurement of moisture.
[0092] Preferably, in S2, a polarization camera is used to obtain N black and white squares on the standard plate, and the black squares White block Number them and count the mean pixel response of each channel in each block and
[0093] Extract the spectral reflection intensity value with statistical significance. Obtain the spectral reflection intensity values of the four channels r, g, b, and ir respectively. FP(r ,g,b,ir), the calculation formula is as follows:
[0094]
[0095] Preferably, the formula for the dense channel reflection reference in S3 is:
[0096]
[0097] in represents the reflection response of the white diffuse reflection standard plate under the i-th measurement condition, where (c1, c2, ..., c K ) is the spectral channel, with a total of K spectral channels.
[0098] Preferably, the polarization image obtained by the polarization color camera is used to measure the surface roughness structure of the tobacco leaf material in the visible area. The combined four-directional polarization camera device is used to analyze the polarization degree and polarization angle of the measured target, so as to measure the texture structure of the surface of the measured target tobacco leaf, and the moisture measurement result is corrected in combination with the texture structure characteristics.
[0099] Combine four sets of polarization cameras to obtain the polarization degree and polarization angle of the target to be measured, and quantitatively analyze the changes in the geometric shape of the object surface (texture characteristics of the target surface) reflected by the polarization state of light after it is reflected from the object surface;
[0100] The Stokes vector and measurement method are used to describe the four-dimensional vector of the polarization state of light. In practice, a polarizer is usually added in front of the detector, and the intensity of light waves with different polarization angles is obtained by rotating the polarizer, and then calculated;
[0101] When measuring, the linear polarizer is rotated by θ, and the phase delay angle β is obtained to obtain the measured image area. I( θ, β). The rotation angle values are: 0°, 45°, 90° and 135°, and the phase delay angle value is: 0, The light intensity values are further obtained to calculate the Stokes parameters.
[0102]
[0103] For the calculation of the linear polarization degree P, the formula is:
[0104]
[0105] The definition of the linear polarization angle φ can be expressed by the Stokes parameter:
[0106]
[0107] Finally, the calculated results of the degree of polarization P and polarization angle φ given different polarizer transmission angles can be obtained by the following equations:
[0108]
[0109] The polarization degree and polarization angle of each pixel in the image area are calculated, the average polarization degree and polarization angle of the entire area are counted, and the variance of the polarization degree distribution and polarization angle distribution is counted.
[0110] Using the above principle, the polarization degree mean P is collected avg , mean polarization angle Polarization degree distribution variance σ avg and the polarization angle distribution variance and the corresponding deviation of the moisture measurement from the mean y Δ , construct a sample of measured data.
[0111] Assuming there are m samples, n polarization degrees and polarization angle distribution variances as input variables, the corresponding moisture measurement error is yΔ As the output variable, the established PLS model can be expressed as:
[0112] Y=TP T +E
[0113] Where Y is the matrix of moisture measurement error (m×1), T is the matrix of distribution variance of polarization degree and polarization angle (m×k), P is the coefficient matrix of PLS model (k×n), and E is the residual matrix (m×1)
[0114] Preferably, the polarized image acquired by the color camera is used to classify and analyze the tobacco leaf material in the visible area, the abundance of each classification is analyzed, and a combined model is used to achieve high-precision moisture determination.
[0115] All tobacco leaves to be tested are manually classified based on color and assumed to be divided into H categories. Tobacco leaf samples are screened based on H category color models, and H moisture prediction models suitable for H category tobacco leaf samples are constructed respectively.
[0116] During actual measurement, a color polarization camera is used to perform superpixel segmentation on the current visible area using the SLIC algorithm, and based on color and spatial distribution, statistics are collected to determine which tobacco leaf types exist in the current sample. The specific steps are as follows:
[0117] (1) Set the number of superpixels K and calculate the superpixel size N / K based on the number of image pixels N.
[0118] (2) Set the initial grid, calculate the gradient of the neighborhood of the central pixel of the grid, and set the pixel with the minimum gradient as the initial clustering seed point.
[0119] (3) Use the formula to convert the regional color from RGB color space to CIELab color space.
[0120]
[0121] (4) In the SLIC algorithm, the color information and position information of the pixel are represented in the five-dimensional space {x, y, l, a, b}. The color and spatial feature distance between pixels in the image is as follows:
[0122]
[0123] Normalize it to get the similarity d s :
[0124]
[0125] (5) Move the cluster center to the mean center of the pixel area with the same label, and repeat step (4) until the error is less than the set threshold or the maximum number of iterations is reached.
[0126] (6) Small scattered areas that are missed in the clustering process are merged into nearby superpixels. Using the superpixel segmentation strategy, the algorithm performs segmentation on the output image after edge enhancement. It can be clearly seen that a smaller superpixel size can obtain a segmentation result with a closer segmentation edge. This solution can effectively ensure the accuracy of the logical segmentation result.
[0127] The mean color value of tobacco leaves in each segmented area is counted, and the KMeans algorithm is used to map each superpixel segmentation block to H tobacco leaf classifications;
[0128] S(cls,area)=KM(SLIC(I))
[0129] Where KM(x) is the KMeans algorithm, SLIC is the segmentation model, I is the tobacco leaf image to be analyzed, S(cls ,area) is the tobacco leaf type classified based on color and the corresponding pixel area.
[0130]
[0131] Based on the area proportion of each tobacco leaf classification, the abundance proportion of the material is constructed. Based on the abundance proportion, combined with various moisture measurement models, the fused moisture measurement value is obtained.
[0132] In addition, X is the spectral absorption line of the current case being tested, F i (X) It is a water content prediction model based on spectral absorption lines. T The actual water content.
[0133] T=η i F i (X).
[0134] By setting up multiple polarization cameras and using industrial computers to control them, and setting calibration black and white diffuse reflection standard plates on the platform, the calibration analysis of the tobacco leaves to be tested is carried out to suppress the spectral value differences caused by the differences in tobacco leaf colors and material structure characteristics under different light source attenuation conditions, thereby improving the numerical stability and accuracy of moisture measurement;
[0135] Combined with the spectral reflectance of the actual sample measured by the spectrometer, the spectrometer standard plate reflectance reference is used to directly calculate the channel ratio to obtain a more accurate spectral absorption line. The relationship between the spectral absorption line and the water content of the measured target can be modeled using the classic PLS to achieve fast and accurate measurement of moisture;
[0136] The polarization image obtained by the polarization color camera is used to measure the surface roughness structure of the tobacco leaf material in the visible area. The polarization degree and polarization angle of the target are analyzed by the combined four-directional polarization camera device to measure the texture structure of the target tobacco leaf surface, and the moisture measurement result is corrected by combining the texture structure characteristics;
[0137] Polarized images acquired by a color camera are used to classify and analyze tobacco leaf materials in the visible area, the abundance of each classification is analyzed, and a combined model is used to achieve high-precision moisture determination.
Claims
1. A tobacco moisture meter adaptive calibration device, characterized in that: include: A platform, wherein a tobacco leaf body to be tested (1) is arranged in the middle of the platform, and a calibration black and white diffuse reflection standard plate (2) is arranged on a side away from the tobacco leaf body to be tested (1), and the calibration black and white diffuse reflection standard plate (2) is arranged on one side of the platform; An industrial computer (8) is arranged on the top of the platform and is spaced apart from the platform; A polarization camera a (3) is electrically connected to the industrial computer (8), and the direction of the polarization camera a (3) is 0°; A polarization camera b (4) is electrically connected to the industrial computer (8), and the direction of the polarization camera b (4) is 45°; A polarization camera c (5) is electrically connected to the industrial computer (8), and the direction of the polarization camera c (5) is 90°; A polarization camera d (6) is electrically connected to the industrial computer (8), and the direction of the polarization camera d (6) is 135°; The tungsten-halogen lamp light source (7) is arranged on one side of the polarization camera c (5) and is used to illuminate the platform.
2. A tobacco moisture meter adaptive calibration method, using a tobacco moisture meter adaptive calibration device as claimed in claim 1, characterized in that: The following steps are involved: S1, using glass slides with different light transmittances to mask the active light source, so as to simulate the intensity attenuation state of the halogen tungsten light source (7), and by selecting glass slides with different light transmittances, a graded simulation of the intensity attenuation degree of the light source is achieved; S2. Under the light source attenuation conditions set at each level, the light intensity values of the black and white diffuse reflection standard plates are obtained by the polarization camera under the standard state, and the standard reflection under the standard state is obtained, which is used as the sparse channel reflection reference here; S3, respectively obtain the spectral reflectance values of the black block and the white diffuse reflectance standard plate through a near-infrared spectrometer under standard conditions, obtain the standard reflectance of the diffuse reflectance standard plate under standard conditions, and take an average value through N measurements, which is referred to as a dense channel reflectance reference; S4. Construct a relationship mapping table from sparse channel reflection reference to dense channel standard reflection reference <F P ,F S >. Using sparse F P The feature query obtains the F that represents the standard reflectance of K spectral channels S ; S5. Obtain the black and white squares on the standard plate captured by the polarization camera and calculate the sparse channel standard reference Combined with fast index mapping table <F P ,F S >, use Same as F in the mapping table P The Euclidean distance of the numerical difference is used to retrieve the closest index item and obtain the reflectance reference of the spectrometer standard plate; S6. Combined with the spectral reflectance of the actual sample measured by the spectrometer, the spectrometer standard plate reflectance reference is used to directly compare the channel by channel to obtain a more accurate spectral absorption line. The relationship between the spectral absorption line and the water content of the measured target can be modeled by the classic PLS to achieve fast and accurate measurement of moisture.
3. The method for adaptively calibrating a tobacco moisture meter according to claim 2, characterized in that: In S2, a polarization camera is used to obtain N black and white squares on the standard plate, and the black squares are White block Number them and count the mean pixel response of each channel in each block and Extract the spectral reflection intensity value with statistical significance. Obtain the spectral reflection intensity value F under the four channels of r, g, b, and ir respectively P (r,g,b,ir), the calculation formula is as follows:
4. The method for adaptively calibrating a tobacco moisture meter according to claim 2, characterized in that: The formula for the dense channel reflection reference in S3 is: in represents the reflection response of the white diffuse reflection standard plate under the i-th measurement condition, where (c1, c2, ..., c K ) is the spectral channel, with a total of K spectral channels.
5. The method for self-adapting calibration of a tobacco moisture meter according to claim 2, characterized in that: The polarization image obtained by the polarization color camera is used to measure the surface roughness structure of the tobacco leaf material in the visible area. The polarization degree and polarization angle of the target are analyzed by the combined four-directional polarization camera device to measure the texture structure of the target tobacco leaf surface, and the moisture measurement result is corrected by combining the texture structure characteristics.
6. The method for self-adapting calibration of a tobacco moisture meter according to claim 2, characterized in that: Polarized images acquired by a color camera are used to classify and analyze tobacco leaf materials in the visible area, the abundance of each classification is analyzed, and a combined model is used to achieve high-precision moisture determination.
