A label adhesion quality intelligent detection system

Abstract

This invention relates to the field of visual inspection of label bonding quality, and discloses an intelligent inspection system for label bonding quality, comprising: a polarization excitation illumination module for projecting a circularly polarized light field onto a highly reflective curved substrate; a multi-dimensional polarization imaging module for capturing spatially interwoven polarization signals; a geometric contour analysis module for analyzing the geometric contour information of the substrate; a polarization state compensation matrix generation module for determining the local incident angle and generating a compensation matrix according to Fresnel's law of reflection; a polarization distortion correction module for correcting polarization distortion and outputting a full Stokes polarization vector flow; an interface stress field feature extraction module for extracting phase delay features; and a defect location and judgment module for identifying bubble or wrinkle defects. This invention utilizes phase delay features to characterize the surface deformation of the transparent label bonding interface, and combines this with a polarization state compensation mechanism to remove geometric interference from the curved surface, thereby achieving the identification of hidden bonding defects under highly reflective environments.

Description

Technical Field

[0001] This invention relates to an intelligent detection system for label bonding quality, belonging to the field of visual inspection technology for label bonding quality. Background Technology

[0002] Currently, high-brightness stroboscopic light sources combined with area array imaging equipment are commonly used to achieve visual inspection of packaging quality on high-speed production lines. This technology utilizes the intensity distribution characteristics of the light field reflected from the surface of an object and uses a grayscale contrast analysis algorithm to determine the geometric contour or surface defects of the target, maintaining good detection stability when processing regular planar targets.

[0003] With the increasing demands for packaging processes in the pharmaceutical and fast-moving consumer goods industries, the need for quality monitoring of transparent labels on highly reflective curved surfaces is growing. When the bottle under test passes through the assembly line at high speed, its high reflectivity and complex geometric curvature cause specular glare at the apex of the surface. This strong reflected light causes the corresponding pixels of the image sensor to overflow with charge, resulting in saturated highlight areas in the image. Weak diffuse reflection signals from bubbles, wrinkles, or excess adhesive inside the transparent label are completely drowned out by strong background noise, leading to information loss in the detection system. Furthermore, the detection algorithm and control strategy have shortcomings. For example, Chinese invention patent application CN120064298A discloses a method and system for detecting surface defects in sockets based on image processing. This invention is based on scalar analysis of diffuse scattering light spot morphology and is applicable to metal... For non-transparent material surface detection, it is impossible to penetrate the transparent label substrate to capture the internal physical evolution of the interface. Under high-speed production conditions, it lacks the ability to extract the features of stress birefringence effect inside transparent media. Relying solely on frequency domain components or geometric shape clustering leads to physical superposition of bubble and wrinkle stress feature signals with background texture. High curvature edge areas cannot dynamically compensate for geometrically induced polarization distortion, causing imbalance in detection sensitivity and false positives. Conventional improvement methods mostly focus on back-end algorithm enhancement or front-end supplementary lighting angle optimization. However, increasing the brightness of the light source will further aggravate the degree of glare saturation. Moreover, relying solely on software models cannot restore the physical details lost in the initial stage of photoelectric conversion. In addition, under high-speed production cycles of hundreds of pieces per minute, the introduction of complex nonlinear solution models generates huge computational overhead, resulting in detection response delays and difficulty in maintaining detection robustness under fluctuating light environment conditions.

[0004] Therefore, how to utilize the polarization vector properties of the light field to physically remove specular glare interference and achieve accurate demodulation of surface distortion and surface stress defects under low computational load conditions has become the technical problem to be solved by this invention. Summary of the Invention

[0005] To address the problems mentioned in the background art, the technical solution of the present invention is as follows: A label bonding quality intelligent detection system, the system comprising: The polarization excitation illumination module is used to project a circularly polarized light field with a preset rotation direction onto a transparent label on a highly reflective curved substrate under test. The multidimensional polarization imaging module is used to simultaneously acquire spatially interwoven polarization signals reflected by the transparent label in order to obtain light intensity data corresponding to four specific polarization directions in the pixel array of the photosensitive chip. The geometric contour analysis module is used to analyze the geometric contour information of the highly reflective curved surface substrate under test by utilizing the gray-level gradient distribution of light intensity data in the imaging space. The polarization state compensation matrix generation module is used to map the local incident angle of each pixel in the geometric contour information to the reflectivity difference between the s-polarization component and the p-polarization component based on Fresnel's reflection law, and generate a polarization state compensation matrix corresponding to the large incident angle region to compensate for the leakage of pseudo-polarization signals caused by the geometric curvature of the substrate surface. The polarization distortion correction module is connected to the polarization state compensation matrix generation module. It is used to correct the polarization distortion induced by geometric contour information in the spatially interleaved polarization signal using the polarization state compensation matrix, and outputs the corrected full Stokes polarization vector flow. The interface stress field feature extraction module, connected to the polarization distortion correction module, is used to extract the phase delay features generated by the probe light wave penetrating the transparent label in the interface stress field based on the full Stokes polarization vector flow. The phase delay features characterize the surface deformation state of the transparent label substrate and the adhesive layer during the bonding process. The defect location and determination module is used to locate and identify bubble defects or wrinkle defects at the interface of transparent label bonding based on the physical boundary gradient abrupt change characteristics in the phase delay characteristics and in combination with the preset interface stress distribution model.

[0006] Preferably, the multidimensional polarization imaging module includes a micro-polarizer array disposed at the front end of the photosensitive chip pixel array. The micro-polarizer array consists of an imaging macroblock composed of four micro-polarizer pixels arranged in a 2×2 pattern. The transmission directions of the four micro-polarizer pixels are 0°, 45°, 90° and 135°, respectively. The multidimensional polarization imaging module is used to capture the charge signals of four polarization components reflected from the transparent label and located in the same spatial coordinates through the four micro-polarizer pixels in a single exposure cycle, and convert the charge signals into spatially interleaved polarization signals.

[0007] Preferably, the polarization excitation illumination module includes multiple LED beads arranged in a matrix, a linear polarizing film disposed on the light path emitted by the multiple LED beads, and a wide-angle quarter-wave plate; the transmission direction of the linear polarizing film forms a 45° angle with the fast axis direction of the wide-angle quarter-wave plate, which is used to convert the randomly polarized light emitted by the multiple LED beads into a circularly polarized light field with a preset rotation direction.

[0008] Preferably, the polarization state compensation matrix generation module includes a refractive index physical response model. The refractive index physical response model is used to calculate the local incident angle of each pixel in the imaging field of view based on the geometric contour information, and to determine the polarization attenuation factor caused by specular reflection based on the local incident angle, so as to generate a polarization degree correction operator covering the entire imaging field of view as the polarization state compensation matrix.

[0009] Preferably, the interface stress field feature extraction module includes a phase difference demodulation submodule. The phase difference demodulation submodule is used to compare the preset reference phase data of the normal bonding area with the actual dynamic phase modulation amount generated by stress birefringence, and extract the gradient change rate feature that characterizes the defect boundary, so as to separate the specular reflection noise component of the surface of the high reflectivity curved substrate under test at the physical level.

[0010] Preferably, the system also includes an adaptive exposure control module connected to the multidimensional polarization imaging module; the adaptive exposure control module is used to monitor the charge accumulation level of light intensity data in real time, and automatically shorten the exposure integration time of the multidimensional polarization imaging module when the maximum charge value exceeds the preset 80% full-well threshold.

[0011] Preferably, the defect location and determination module includes a signal processing unit. The signal processing unit is used to perform anisotropic filtering on the phase delay characteristics to smooth the discrete noise of the circuit, and to extract the connected regions that meet the stress change characteristics using dual threshold determination logic. Based on the geometric roundness factor of the connected regions, the defects are divided into circular bubbles, strip wrinkles or glue overflow points.

[0012] Preferably, the system further includes a clock synchronization trigger module, which connects the polarization excitation illumination module and the multi-dimensional polarization imaging module; the clock synchronization trigger module is used to send a synchronization pulse signal to align the illumination flicker period with the imaging exposure window on the time axis, with an alignment accuracy deviation of less than 10 ns.

[0013] Preferably, the defect location and determination module is also used to map the spatial coordinates of the identified defects into geometric contour information, generate a three-dimensional distribution cloud map characterizing the bonding quality of the high reflectivity curved substrate surface to be tested, and output a rejection command when the number density of defects exceeds a preset threshold per unit time.

[0014] Compared with the prior art, the beneficial effects of the present invention are: 1. In intelligent detection of label bonding quality, by constructing a physical coupling mechanism at the front end of the optical path to synchronously resolve the circular polarization excitation and the polarization component of the focal plane, and utilizing the physical rotation reversal characteristic of specular reflection at the interface, the glare of highly reflective curved mirror surfaces is suppressed in the initial stage of photoelectric conversion. This avoids the information blind zone formed by the charge trap of the photosensitive chip caused by the strong reflected light field, and allows the underlying defect feature signals that were originally submerged under the high light band to be clearly characterized in a low signal-to-noise ratio environment.

[0015] 2. The low-frequency spatial distribution of the initial light intensity parameter is used to invert the overall geometric contour information of the target under test, and an elliptic leakage compensation baseline corresponding to the large incident angle region is generated according to the Fresnel reflection attenuation law. In this way, the pseudo-polarization distortion caused by the curvature of the surface is removed in the vector decoupling operation, the ring-shaped false positive false alarm generated in the high curvature edge region is eliminated, and the detection sensitivity is smoothly extended from the center of the field of view to the high curvature edge region.

[0016] 3. The surface stress birefringence effect generated during the bonding process between the transparent label substrate and the adhesive layer is transformed into quantifiable polarization angle phase distortion characteristics. By capturing the phase delay difference generated by the transmitted light wave in the stress field, hidden bubbles and tiny wrinkles with small refractive index abrupt changes are mapped into high-contrast physical boundaries in the angular polarization matrix, thus solving the technical bottleneck that traditional scalar imaging systems cannot obtain deformation information inside transparent media. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the logic of polarization analysis and defect identification for label bonding quality in this invention. Figure 2 This is a diagram showing the hardware composition and timing coordination architecture of the intelligent detection system of the present invention.

[0018] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0020] A label bonding quality intelligent detection system, the system includes: The polarization excitation illumination module is used to project a circularly polarized light field with a preset rotation direction onto a transparent label on a highly reflective curved substrate under test. The multidimensional polarization imaging module is used to simultaneously acquire spatially interwoven polarization signals reflected by the transparent label in order to obtain light intensity data corresponding to four specific polarization directions in the pixel array of the photosensitive chip. The geometric contour analysis module is used to analyze the geometric contour information of the highly reflective curved surface substrate under test by utilizing the gray-level gradient distribution of light intensity data in the imaging space. The polarization state compensation matrix generation module is used to map the local incident angle of each pixel in the geometric contour information to the reflectivity difference between the s-polarization component and the p-polarization component based on Fresnel's reflection law, and generate a polarization state compensation matrix corresponding to the large incident angle region to compensate for the leakage of pseudo-polarization signals caused by the geometric curvature of the substrate surface. The polarization distortion correction module is connected to the polarization state compensation matrix generation module. It is used to correct the polarization distortion induced by geometric contour information in the spatially interleaved polarization signal using the polarization state compensation matrix, and outputs the corrected full Stokes polarization vector flow. The interface stress field feature extraction module, connected to the polarization distortion correction module, is used to extract the phase delay features generated by the probe light wave penetrating the transparent label in the interface stress field based on the full Stokes polarization vector flow. The phase delay features characterize the surface deformation state of the transparent label substrate and the adhesive layer during the bonding process. The defect location and determination module is used to locate and identify bubble defects or wrinkle defects at the interface of transparent label bonding based on the physical boundary gradient abrupt change characteristics in the phase delay characteristics and in combination with the preset interface stress distribution model.

[0021] Preferably, the multidimensional polarization imaging module includes a micro-polarizer array disposed at the front end of the photosensitive chip pixel array. The micro-polarizer array consists of an imaging macroblock composed of four micro-polarizer pixels arranged in a 2×2 pattern. The transmission directions of the four micro-polarizer pixels are 0°, 45°, 90° and 135°, respectively. The multidimensional polarization imaging module is used to capture the charge signals of four polarization components reflected from the transparent label and located in the same spatial coordinates through the four micro-polarizer pixels in a single exposure cycle, and convert the charge signals into spatially interleaved polarization signals.

[0022] Preferably, the polarization excitation illumination module includes multiple LED beads arranged in a matrix, a linear polarizing film disposed on the light path emitted by the multiple LED beads, and a wide-angle quarter-wave plate; the transmission direction of the linear polarizing film forms a 45° angle with the fast axis direction of the wide-angle quarter-wave plate, which is used to convert the randomly polarized light emitted by the multiple LED beads into a circularly polarized light field with a preset rotation direction.

[0023] Preferably, the polarization state compensation matrix generation module includes a refractive index physical response model. The refractive index physical response model is used to calculate the local incident angle of each pixel in the imaging field of view based on the geometric contour information, and to determine the polarization attenuation factor caused by specular reflection based on the local incident angle, so as to generate a polarization degree correction operator covering the entire imaging field of view as the polarization state compensation matrix.

[0024] Preferably, the interface stress field feature extraction module includes a phase difference demodulation submodule. The phase difference demodulation submodule is used to compare the preset reference phase data of the normal bonding area with the actual dynamic phase modulation amount generated by stress birefringence, and extract the gradient change rate feature that characterizes the defect boundary, so as to separate the specular reflection noise component of the surface of the high reflectivity curved substrate under test at the physical level.

[0025] Preferably, the system also includes an adaptive exposure control module connected to the multidimensional polarization imaging module; the adaptive exposure control module is used to monitor the charge accumulation level of light intensity data in real time, and automatically shorten the exposure integration time of the multidimensional polarization imaging module when the maximum charge value exceeds the preset 80% full-well threshold.

[0026] Preferably, the defect location and determination module includes a signal processing unit. The signal processing unit is used to perform anisotropic filtering on the phase delay characteristics to smooth the discrete noise of the circuit, and to extract the connected regions that meet the stress change characteristics using dual threshold determination logic. Based on the geometric roundness factor of the connected regions, the defects are divided into circular bubbles, strip wrinkles or glue overflow points.

[0027] Preferably, the system further includes a clock synchronization trigger module, which connects the polarization excitation illumination module and the multi-dimensional polarization imaging module; the clock synchronization trigger module is used to send a synchronization pulse signal to align the illumination flicker period with the imaging exposure window on the time axis, with an alignment accuracy deviation of less than 10 ns.

[0028] Preferably, the defect location and determination module is also used to map the spatial coordinates of the identified defects into geometric contour information, generate a three-dimensional distribution cloud map characterizing the bonding quality of the high reflectivity curved substrate surface to be tested, and output a rejection command when the number density of defects exceeds a preset threshold per unit time.

[0029] Example 1: In a high-speed pharmaceutical production line scenario where hundreds of highly reflective cylindrical glass bottles with transparent or semi-transparent labels are processed per minute, an intelligent label bonding quality detection system faces the problem of physical aliasing of optical signals caused by surface specular reflection glare and the high curvature edges of the substrate. To address this visual inspection condition, an active polarization modulation light source array projects a pre-defined circularly polarized broadband illumination field onto the highly reflective curved substrate under test. When this circularly polarized broadband illumination field penetrates the transparent label interface, it undergoes spatial distortion of polarization state due to stress birefringence caused by internal bubbles or wrinkles, forming a reflected light carrying polarization vector distortion information. The focused plane polarization imaging unit (SPPI) acquires the reflected echo optical path based on a single gated exposure. The photosensitive plane of the SPPI integrates a periodically alternating array of four-directional micro-polarizers. These micro-polarizers are arranged in 2×2 pixel macroblocks, each macroblock containing four micro-polarization units with transmission axes of 0°, 45°, 90°, and 135°. The SPPI synchronously outputs a raw spatially interleaved polarization signal containing discrete components in four polarization directions. The polarization state compensation matrix generation module processes the raw spatially interleaved polarization signal, extracting the intensity components corresponding to the four polarization directions within each macroblock. , , and The initial light intensity parameter was calculated based on the Stokes parametric analytical rule. and initial polarization parameters and ,in The calculation formula is , The calculation formula is , The calculation formula is The polarization state compensation matrix generation module generates the initial light intensity parameters according to the preset window space low-pass filter. The global matrix is ​​used to filter out high-frequency albedo textures and extract the overall geometric light intensity envelope matrix representing the physical curvature contour of the target substrate. Based on the surface physical reflection law under uniform illumination, the low-frequency spatial light intensity envelope gray-level gradient distribution after removing high-frequency albedo textures has a deterministic spatial monotonic mapping relationship with the normal vector of the surface under test. The polarization state compensation matrix generation module extracts the pixel gray-level gradient of the overall geometric light intensity envelope matrix and uses the mapping relationship to calculate the local pixel gray-level gradient difference into the corresponding three-dimensional local incident angle in the field of view coordinates. This mapping relationship is constructed based on Lambert's law of reflection and a preset illumination intensity distribution function. Under a uniform circularly polarized light field, the gray-level values ​​of pixels on the highly reflective surface are... With local normal vector deflection angle A cosine mapping relationship exists, with the gray-level gradient direction corresponding to the principal curvature direction of the surface. By inversely calculating the gray-level change rate between adjacent pixels, this invention obtains the polar angle and azimuth angle of the surface normal in spherical coordinates, thereby determining the three-dimensional local incident angle between the probe ray and the local tangent plane. This ensures that the geometric pose of each pixel has a definite physical quantitative basis, and the local incident angle is calculated based on Fresnel's law of reflection. polarization components and The difference in reflectivity of the polarization components on the dielectric surface generates a baseline matrix characterizing large incident angle distortion ellipsometric leakage compensation, containing a first-direction leakage component S1-base and a second-direction leakage component S2-base. In the specific generation process, this invention extracts the normal vector direction of the local surface based on geometric contour information and uses this to determine the azimuth angle of the local incident surface relative to the reference axis of the photosensitive chip pixel array. The difference in complex amplitude reflectance between the s-component and p-component at the current pixel is calculated based on the Fresnel formula and defined as the pseudo-polarization background value in the local coordinate system. Further, through Jones or Mueller matrix rotation transformation, this reflectance difference in the local coordinate system is projected onto the 0° / 90° and 45° / 135° axes of the global imaging coordinate system, thereby accurately calculating the corresponding first-direction leakage component S1-base and second-direction leakage component S2-base. This ensures that the compensation matrix can cancel the geometric polarization noise caused by the surface tilt angle pixel by pixel. The polarization state compensation matrix generation module extracts the pixel gray-level gradient of the overall geometric intensity envelope matrix and, based on the calibrated Fresnel reflection attenuation mapping function, transforms the pixel gray-level gradient pixel by pixel into an elliptic leakage compensation baseline matrix characterizing large incident angle distortion. The elliptic leakage compensation baseline matrix contains the first-direction leakage component. With the second direction leakage component .

[0030] The polarization distortion correction module constructs a spatial variational physics compensation closed loop based on the elliptic leakage compensation baseline matrix to compensate for the initial polarization parameters. and Obtain the compensation polarization parameters for stripping curvature interference. and The formula for calculating the compensation polarization parameter is as follows: as well as The defect location and judgment module is based on the formula Calculate the degree of linear polarization after spatial variational compensation on a macroblock-by-macroblock basis. Construct a global linear polarization degree topology matrix based on the compensation polarization parameter. and The polarization angle is calculated and a global angular polarization matrix is ​​constructed. The defect location and judgment module sets connected regions with amplitudes below a preset physical threshold in the global linear polarization degree topology matrix as the diffuse reflection contour of the label surface to suppress specular glare and curvature distortion. Pixel clusters with phase abrupt gradients exceeding a preset stress threshold in the global angular polarization matrix are set as internal bubble or wrinkle defect areas in the label adhesive layer. The defect location and judgment module outputs an inspection and judgment signal based on the spatial intersection coordinates of the label surface diffuse reflection contour and the internal bubble or wrinkle defect areas. This overall curvature and surface polarization cross-scale physical separation mechanism dynamically generates a compensation matrix using the geometric topological information contained in the low-frequency scalar light intensity. It cuts off the interference of the inherent curvature-induced ellipsoidal effect of large incident angles on the defect judgment criteria from the bottom layer of physical calculation. The system control unit receives the encoder pulses of the motor driving the chain conveyor belt in real time, based on the bottle diameter. With running speed Calculate the gating pulse period, where The outer diameter of the cross-section of the glass bottle to be measured is... To achieve synchronous acquisition of the active polarization modulation light source array and the focal plane polarization imaging unit, the linear speed of the conveyor belt is synchronously triggered. The pulse width of the active polarization modulation light source array is set to less than 50 μs, and the system control unit adjusts the peak power of the circularly polarized broadband illumination field to 5 to 10 times that of the continuous illumination mode. This maintains the imaging signal-to-noise ratio of the original spatially intertwined polarization signal within a short exposure time and suppresses phase resolution artifacts caused by object motion. When the adaptive exposure control module triggers the exposure integration time reduction action, based on the basic photoelectric model that the charge accumulation of photosensitive pixels is proportional to the exposure time, the system control unit synchronously acquires the actual exposure integration time of the current cycle and divides it by the reference exposure time under the calibration state to generate a dimensionless time normalization coefficient. The multidimensional polarization imaging module divides the output light intensity data by the time normalization coefficient, and maps the acquired signal amplitude back to the standard order of magnitude under the reference exposure time. This maintains the absolute grayscale characteristics of the initial light intensity parameters under dynamic conditions and eliminates the nonlinear drift of pixel grayscale gradient caused by the adaptive suppression action of highlights.

[0031] Example 2: This example constructs a physical verification test platform to determine the accuracy of interface stress field feature extraction. A chain conveyor belt with a running speed of 600 pieces / minute is used as the motion carrier, and a broadband light source array with a light emission wavelength of 450nm to 650nm is erected above it. Gaussian white noise with a signal-to-noise ratio of 20dB and power frequency flicker with a frequency of 50Hz are superimposed as optical interference sources. Cylindrical glass bottles with basic curvature radii of 30mm, 20mm and 10mm are selected as the substrates to be tested for high reflectivity. Transparent labels with microbubbles of 0.5mm in diameter are attached to their surfaces using uniform tension parameters. The test sets up a control group without elliptic leakage compensation baseline matrix and an experimental group with compensation closed loop. When the polarization state compensation matrix generation module generates the overall geometric light intensity envelope matrix characterizing the physical curvature profile of the target substrate, the window size of the spatial low-pass filter is determined. This parameter balances the relationship between the low-frequency component of the overall geometric curvature and the preservation of the high-frequency scattering signal of local bubbles. Based on the spatial frequency sampling theory, when the radius of curvature of the surface under test decreases, leading to an intensified change in pixel grayscale gradient, the system reduces the window size. To avoid background envelope smoothing and absorption of high-frequency defect features; the system increases the window size. By fitting a smooth curved surface, for a cylindrical glass bottle with a radius of curvature of 20 mm, the system calculates the baseline fitting bandwidth and sets the window size. The lower limit is 3×3 pixels, the median is 7×7 pixels, and the upper limit is 15×15 pixels, which constitute the measurement sequence of the verification parameter working window.

[0032] The focal plane polarization imaging unit continuously outputs a raw spatially interlaced polarization signal containing discrete components of four polarization directions under a single gated exposure. The polarization distortion correction module corrects the initial polarization parameters based on the elliptic leakage compensation baseline matrix. In the control group, when the radius of curvature of the glass bottle is reduced from 30 mm to 10 mm, the bubble detection accuracy decreases from 85.4% to 42.5%. The defect location judgment module incorrectly marks the normally fitting area at the edge of the curved surface as a stress defect. In the experimental group, the system uses a median 7×7 pixel window size. Based on the pixel grayscale gradient, an elliptic leakage compensation baseline matrix is ​​generated pixel by pixel. The defect location and judgment module achieves bubble detection accuracy of 98.6%, 98.2%, and 97.5% under curvature radii of 30mm, 20mm, and 10mm, respectively. Extreme value test data shows that when the window size... When the pixel size is 15×15, the background suppression ratio of power frequency flicker and Gaussian noise reaches a saturation plateau, and the processing latency increases to 45.2ms, which does not meet the production line cycle time limit of 600 pieces / minute; when the window size is... When the pixel size is 3×3, the high-frequency feature signals of the microbubbles leak into the elliptic leakage compensation baseline matrix, increasing the local false alarm rate to 31.4%. This test data determines the window size. The working range is from 5×5 pixels to 11×11 pixels; the elliptic leakage compensation baseline matrix extracts the geometric topological information contained in the low-frequency scalar light intensity to establish a surface space mapping. The system uses the computing power of the image processor to reconstruct the basic polarization dimension information. In the detection conditions containing Gaussian white noise and power frequency flicker, this polarization state compensation mechanism separates the curvature-induced ellipticity effect caused by the overall geometric curvature from the stress birefringence effect caused by bubbles or wrinkles, and removes the inherent optical signal aliasing of the large incident angle of the curved surface.

[0033] Example 3: In a high-speed pharmaceutical production line scenario where hundreds of highly reflective cylindrical glass bottles with transparent or semi-transparent labels are processed per minute, an intelligent label bonding quality detection system transforms pixel grayscale gradients based on the Fresnel reflection attenuation mapping function and determines defects by combining a preset stress threshold. Static parameter failures are caused by ambient light attenuation and fluctuations in the optical refractive index of batch substrates. To maintain the operational status of the polarization distortion correction module and the defect location judgment module under continuous detection conditions, the polarization state compensation matrix generation module executes a physical parameter calibration procedure to define the numerical sources of optical model parameters and judgment boundaries. The operator places an unlabeled reference cylindrical glass bottle at the detection station as a physical reference. The polarization excitation illumination module projects a pre-rotating circularly polarized broadband illumination field onto the reference cylindrical glass bottle. The focal plane polarization imaging unit captures the reflected reference spatial interlaced polarization signal. The geometric contour analysis module extracts the discrete pixel grayscale gradient sequence on the surface of the reference cylindrical glass bottle. The polarization state compensation matrix generation module calculates the reference polarization parameters corresponding to each grayscale gradient according to the Stokes parameter analysis rule. and The polarization state compensation matrix generation module constructs a two-dimensional data lookup table, which uses the pixel gray-level gradient as the independent variable and the loss value of the reference polarization parameter deviating from the orthogonal state as the dependent variable.

[0034] The system employs a linear piecewise interpolation algorithm to process discrete data points in a two-dimensional data lookup table, generating a continuous attenuation curve covering the working grayscale gradient range. The polarization state compensation matrix generation module sets this continuous attenuation curve as a Fresnel reflection attenuation mapping function. This process ensures that the generation criterion for the ellipsoidal leakage compensation baseline matrix follows the mapping path of the physical interface reflection characteristics. The system loads 500 standard labeled glass bottles without bubble defects or wrinkle defects as baseline samples. The defect location and judgment module generates a global angular polarization matrix for this baseline sample based on the polarization distortion correction process, extracts the phase abrupt change gradient between adjacent pixels within the global angular polarization matrix, and selects the maximum gradient peak value in this baseline sample group as the extreme value of the substrate stress gradient. The controller sets the preset stress threshold to the extreme value of the base stress gradient. The defect location and judgment module determines the physical region where the phase change gradient exceeds the preset stress threshold in continuous detection tasks as an abnormal fit defect. The parameter determination procedure uses quantitative measurement data to define the physical boundary of the bubble or wrinkle defect by polarization characteristics, and provides a measurement basis for the polarization state compensation mechanism to remove surface geometric interference and identify deformation.

[0035] Example 4: For the condition where fluctuations in the material of the transparent label substrate cause a shift in the background characteristics of stress birefringence, the interface stress field feature extraction module calls the interface stress distribution model. The tension mechanism applies a gradient tensile load from 0.1N to 5.0N to the transparent label sample. The focal plane polarization imaging unit acquires the discrete calibration polarization signal under the corresponding load. The interface stress field feature extraction module processes the discrete calibration polarization signal, extracts the reference phase delay feature matrix, and the controller records the mapping dataset composed of the load scale value and the pixel gradient value of the reference phase delay feature matrix, forming a preset interface stress distribution model. In actual operation... In applications, this interface stress distribution model serves as a calibration bridge from overall mechanical energy to surface optical variables. Specifically, by recording the overall phase fluctuations of the label substrate under different tensile loads, this invention obtains the proportional relationship of the material's elastic-optical coefficients. When the system detects local defects such as bubbles or wrinkles, although the localized stress concentration generated at their edges belongs to the surface scale, the refractive index perturbation induced by them follows the same elastic-optical physical laws. Therefore, through the sensitivity curve stored in the model, this invention can retrospectively derive the extracted phase gradient value into an equivalent local stress intensity, realizing cross-scale quantitative defect determination.

[0036] The defect location and determination module retrieves the interface stress distribution model, the focal plane polarization imaging unit acquires the steady-state baseline polarization degree matrix of the surface under test in the region where no phase change occurs, and the controller calculates the arithmetic mean of the steady-state baseline polarization degree matrix. Spatial standard deviation Determine the preset physical threshold ,in The calculation formula is as follows: ,in, To preset physical thresholds, The mean of steady-state linear polarization. For spatial standard deviation, It is a dimensionless constant; the defect location and determination module compares the global linear polarization degree topology matrix with the preset physical threshold. , with an amplitude smaller than The pixel set is identified as the diffuse reflection contour of the label surface. This process corrects the judgment boundary through optical statistical data, enabling the detection system to maintain recognition consistency under fluctuating light field conditions. During the offline debugging phase, the system determines the dimensionless constants through a traversal search method. That is, while maintaining the running speed of the chain conveyor belt Under constant conditions, a sequence of standard sample images containing preset defects is collected, and different... The receiver operating characteristic curve (ROC) of the defect localization and judgment module is selected based on the given values, maximizing the difference between the defect detection rate and the false alarm rate. Using this value as a benchmark coefficient, the dimensionless constant was experimentally measured under conditions where illuminance fluctuations were less than 10%. The value range is from 2.5 to 3.5, which is used to lock the statistically optimal working point of the decision boundary.

[0037] Example 5: When the system faces the initial deployment conditions of a pharmaceutical production line, the periodically alternating four-directional micro-polarizer array in the multi-dimensional polarization imaging module experiences extinction ratio attenuation and transient field-of-view physical mismatch due to manufacturing tolerances. This physical mismatch causes inter-channel optical crosstalk in the light intensity data output by the photosensitive chip pixel array corresponding to the four specific polarization directions. To address this inter-channel optical crosstalk, the polarization state compensation matrix generation module executes a multi-dimensional spatial polarization response offline calibration procedure. A standard diffuse reflection calibration plate with constant reflectivity is placed at the detection reference position of the chain conveyor belt. The polarization excitation illumination module switches the modulation optical path and projects a uniform flat-field broadband illumination light field without polarization onto the standard diffuse reflection calibration plate. The multi-dimensional polarization imaging module acquires the spatially intertwined polarization signal reflected by the standard diffuse reflection calibration plate from the uniform flat-field broadband illumination light field in a single-gated exposure mode. The polarization state compensation matrix generation module extracts the light intensity response component within each pixel macroblock. , , and ,in, The light intensity response component of the micro-polarization unit at 0 degrees of the transmission axis. The light intensity response component of the micro-polarization unit at 45 degrees of the transmission axis. The light intensity response component of the micro-polarization unit at 90 degrees of the transmission axis. For the light intensity response component of the micro-polarization unit with a transmission axis of 135 degrees, the polarization distortion correction module calculates the system response offset of the actual light intensity response component relative to the theoretical mean based on the physical constraint that the light intensity in the four specific polarization directions is theoretically equal under an unpolarized light field. The polarization distortion correction module uses the system response offset as the input parameter to construct a pixel-level polarization crosstalk decoupling inverse matrix with the same dimension as the pixel array of the photosensitive chip.

[0038] During the continuous detection phase, the multi-dimensional polarization imaging module acquires the original spatially intertwined polarization signal. The polarization distortion correction module multiplies the original spatially intertwined polarization signal with the pixel-level polarization crosstalk decoupling inverse matrix pixel by pixel, outputting independent polarization components stripped of extinction ratio attenuation and field-of-view mismatch errors. The polarization state compensation matrix generation module receives the independent polarization components and calculates the initial light intensity parameters based on them. and initial polarization parameters and This procedure utilizes standard diffuse reflection flat-field calibration to quantify and compensate for the signal background shift caused by the micro-polarization array structure at the front end of the photosensitive chip, maintaining the initial physical consistency of the detection system before generating the elliptic leakage compensation baseline matrix and extracting the interface stress field characteristics.

[0039] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0040] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A label bonding quality intelligent detection system, characterized in that, The system includes: The polarization excitation illumination module is used to project a circularly polarized light field with a preset rotation direction onto a transparent label on a highly reflective curved substrate under test. The multidimensional polarization imaging module is used to simultaneously acquire spatially interwoven polarization signals reflected by the transparent label in order to obtain light intensity data corresponding to four specific polarization directions in the pixel array of the photosensitive chip. The geometric contour analysis module is used to analyze the geometric contour information of the highly reflective curved surface substrate under test by utilizing the gray-level gradient distribution of light intensity data in the imaging space. The polarization state compensation matrix generation module is used to map the local incident angle of each pixel in the geometric contour information to the reflectivity difference between the s-polarization component and the p-polarization component based on Fresnel's reflection law, and generate a polarization state compensation matrix corresponding to the large incident angle region to compensate for the leakage of pseudo-polarization signals caused by the geometric curvature of the substrate surface. The polarization distortion correction module is connected to the polarization state compensation matrix generation module. It is used to correct the polarization distortion induced by geometric contour information in the spatially interleaved polarization signal using the polarization state compensation matrix, and outputs the corrected full Stokes polarization vector flow. The interface stress field feature extraction module, connected to the polarization distortion correction module, is used to extract the phase delay features generated by the probe light wave penetrating the transparent label in the interface stress field based on the full Stokes polarization vector flow. The phase delay features characterize the surface deformation state of the transparent label substrate and the adhesive layer during the bonding process. The defect location and determination module is used to locate and identify bubble defects or wrinkle defects at the interface of transparent label bonding based on the physical boundary gradient abrupt change characteristics in the phase delay characteristics and in combination with the preset interface stress distribution model.

2. The intelligent label bonding quality detection system according to claim 1, characterized in that, The multidimensional polarization imaging module includes a micro-polarizer array disposed at the front end of the photosensitive chip pixel array. The micro-polarizer array consists of an imaging macroblock composed of four micro-polarized pixels arranged in a 2×2 pattern. The transmission directions of the four micro-polarized pixels are 0°, 45°, 90° and 135°, respectively. The multidimensional polarization imaging module is used to capture the charge signals of four polarization components reflected from the transparent label and located in the same spatial coordinates through the four micro-polarized pixels in a single exposure cycle, and convert the charge signals into spatially interleaved polarization signals.

3. The intelligent detection system for label bonding quality according to claim 1, characterized in that, The polarization excitation illumination module includes multiple LED beads arranged in a matrix, a linear polarizing film disposed on the light path emitted by the multiple LED beads, and a wide-angle quarter-wave plate; the transmission direction of the linear polarizing film forms a 45° angle with the fast axis direction of the wide-angle quarter-wave plate, which is used to convert the randomly polarized light emitted by the multiple LED beads into a circularly polarized light field with a preset rotation direction.

4. The intelligent detection system for label bonding quality according to claim 1, characterized in that, The polarization state compensation matrix generation module includes a refractive index physical response model. The refractive index physical response model is used to calculate the local incident angle of each pixel in the imaging field of view based on the geometric contour information, and to determine the polarization attenuation factor caused by specular reflection based on the local incident angle, so as to generate a polarization degree correction operator covering the entire imaging field of view, which serves as the polarization state compensation matrix.

5. The intelligent detection system for label bonding quality according to claim 1, characterized in that, The interface stress field feature extraction module includes a phase difference demodulation submodule. The phase difference demodulation submodule is used to compare the preset reference phase data of the normal bonding area with the actual dynamic phase modulation amount generated by stress birefringence, and extract the gradient change rate feature that characterizes the defect boundary, so as to separate the specular reflection noise component of the surface of the high reflective curved substrate under test at the physical level.

6. The intelligent detection system for label bonding quality according to claim 1, characterized in that, The system also includes an adaptive exposure control module, which is connected to a multidimensional polarization imaging module; The adaptive exposure control module is used to monitor the charge accumulation level of light intensity data in real time, and automatically shortens the exposure integration time of the multidimensional polarization imaging module when the maximum charge value exceeds the preset 80% full-well threshold.

7. The intelligent detection system for label bonding quality according to claim 1, characterized in that, The defect location and determination module includes a signal processing unit. The signal processing unit is used to perform anisotropic filtering on the phase delay characteristics to smooth the discrete noise of the circuit, and uses dual threshold determination logic to extract connected regions that meet the stress change characteristics. Based on the geometric roundness factor of the connected regions, the defects are divided into circular bubbles, strip wrinkles or glue overflow points.

8. The intelligent detection system for label bonding quality according to claim 1, characterized in that, The system also includes a clock synchronization trigger module, which connects the polarization excitation illumination module and the multidimensional polarization imaging module. The clock synchronization trigger module is used to send a synchronization pulse signal to align the illumination flicker period with the imaging exposure window on the time axis, with an alignment accuracy deviation of less than 10 ns.

9. The intelligent detection system for label bonding quality according to claim 1, characterized in that, The defect location and judgment module is also used to map the spatial coordinates of the identified defects to the geometric contour information, generate a three-dimensional distribution cloud map that characterizes the bonding quality of the highly reflective curved substrate surface to be tested, and output a rejection command when the number density of defects exceeds a preset threshold per unit time.

Images