Finished gel product surface defect quality inspection system based on image recognition and restoration

Abstract

The invention relates to the technical field of visual inspection, in particular to a quality inspection system for repairing surface defects of a gel finished product based on image recognition, which comprises a deformation induction imaging module for applying controllable micro-pressure deformation to a gel sample according to the projected area and hardness grade control pressure of the gel sample, capturing a dynamic sequence image of the deformation process, and outputting the dynamic sequence image; surface microcracks are detected; the characteristic analysis module is used for extracting response characteristics of a defect area from the dynamic image sequence, including distinguishing free impurities from fixed impurities according to a relative displacement track, positioning bubble defects by identifying a stress birefringence ring and detecting an opening behavior of surface microcracks in a deformation process; calculating a relaxation time constant ratio and a stress concentration factor of the defect area and the normal area, and evaluating the interface bonding strength; the repair matching module is used for determining a repair feasibility grade according to the interface bonding strength and carrying out self-adaptive correction on the process parameters; and then detecting a verification module, and comparing the variation of the stress concentration factors before and after repair to judge a repair result.

Description

Technical Field

[0001] This invention relates to the field of visual inspection technology, specifically to a quality inspection system for repairing surface defects in finished gel products based on image recognition. Background Technology

[0002] Gel products are widely used in cosmetics, medical care, and industrial materials. During the production of finished gel products, factors such as uneven mixing of raw materials, fluctuations in filling process parameters, and insufficient cleanliness of packaging materials can affect not only the product's appearance but also its performance and safety.

[0003] With the development of machine vision technology, some companies have begun to adopt automated quality inspection systems based on image recognition. However, existing image recognition quality inspection systems face special challenges when applied to gel products: gel materials are usually transparent or semi-transparent, and internal air bubbles are easily confused with surface reflections. Impurities and background contrast are not obvious, resulting in low defect recognition accuracy. In addition, existing systems can usually only detect and mark defects, lacking intelligent analysis of defect type, location, and severity. They cannot automatically determine whether a product can be reworked or should be scrapped, still requiring secondary manual judgment, which affects the overall quality inspection efficiency.

[0004] Therefore, there is a need for a quality inspection system optimized for the characteristics of finished gel products, capable of accurately identifying surface defects and intelligently determining treatment solutions, in order to improve quality inspection efficiency and accuracy and reduce labor costs.

[0005] To address this, a quality inspection system based on image recognition for repairing surface defects in finished gel products is proposed. Summary of the Invention

[0006] The purpose of this invention is to provide a quality inspection system for repairing surface defects of finished gel products based on image recognition. By constructing a quality inspection and repair system for gel surface defects based on image recognition, automatic detection of defects and adaptive adjustment of repair processes can be achieved.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A quality inspection system for repairing surface defects in finished gel products based on image recognition includes: The deformation-induced imaging module applies controllable micro-pressure deformation to the gel sample based on the projected area and hardness level of the gel sample, causing the gel to undergo thickness compression deformation, capturing dynamic sequence images of the deformation process, and detecting surface micro-cracks. The feature analysis module extracts response features of defect areas from dynamic image sequences, distinguishes between free and fixed impurities by tracking the relative displacement trajectory of impurities, locates bubble defects by identifying stress birefringence rings, and detects the opening behavior of surface microcracks during deformation; calculates the ratio of relaxation time constant and stress concentration factor between defect areas and normal areas, and evaluates the interfacial bonding strength between defects and the gel matrix. The repair matching module determines the repair feasibility level based on the interface bonding strength, matches the repair scheme, and adaptively corrects the process parameters based on the interface bonding strength. The re-inspection and verification module repeatedly performs deformation-induced detection on the repaired product to obtain the characteristic parameters after repair, and compares the change in stress concentration factor before and after repair to determine the repair result.

[0008] Preferably, the detection process for surface microcracks includes: Holographic interference fringe patterns are acquired at multiple moments during the deformation process; the interference fringe patterns are converted into surface phase distribution patterns using a phase reconstruction algorithm; the phase difference between the phase distribution patterns before and after deformation is calculated, and phase abrupt change regions are extracted; connectivity analysis and geometric feature extraction are performed on the phase abrupt change regions, and phase abrupt change regions with lengths exceeding a preset threshold and exhibiting linear and / or curved shapes are identified as surface microcracks, and the location coordinates, length, and orientation angle of the microcracks are recorded.

[0009] Preferably, the process of identifying stress birefringence rings to locate bubble defects includes: A sequence of transmitted light images is continuously acquired. The time series standard deviation of grayscale values ​​for each pixel in the image sequence is calculated to generate a grayscale standard deviation distribution map. Pixels with standard deviation values ​​exceeding a preset threshold are extracted from the grayscale standard deviation distribution map as regions of significant change. Connectivity analysis and boundary extraction are performed on the regions of significant change to calculate the geometric shape feature parameters of the boundaries. Regions whose circularity feature values ​​of the boundary shape meet the ring determination criteria are identified as feature rings formed by stress effects, and the center coordinates and radius of the feature rings are obtained. Based on the radius of the feature rings, the applied pressure value, and the photoelastic coefficient of the gel material, the actual radius of the bubble is estimated through a pre-established calibration relationship to determine the location and size of the bubble defect.

[0010] Preferably, the process for detecting the opening behavior of surface microcracks includes: From the acquired deformation process phase data, the phase change curves of the identified surface microcrack region during the pressurization, holding, and depressurization stages are extracted; the phase gradient on both sides of the microcrack is calculated, and the crack opening displacement is estimated by the change in phase gradient; the crack opening displacement is analyzed as a function of applied pressure to obtain the crack opening displacement; based on the crack opening displacement under different pressures, the crack depth and stress state at the crack tip are determined; cracks with opening displacement exceeding a preset threshold are identified as dangerous cracks.

[0011] Preferably, the process for assessing the interfacial bonding strength between defects and the gel matrix includes: An interface bonding strength assessment matrix is ​​established, which divides the relaxation time constant ratio and stress concentration factor into multiple level intervals. For defects where the absolute value of the difference between the relaxation time constant ratio and the unit value is less than a first threshold, and for defects where the difference between the relaxation time constant ratio and the unit value is greater than a second threshold, the interface bonding is determined to be weak. The interface bonding strength level is comprehensively determined by combining the magnitude of the stress concentration factor, and the interface bonding strength is divided into three levels: strong bonding, medium bonding, and weak bonding.

[0012] Preferably, the process for determining the feasibility level of repair includes: A repair feasibility judgment rule base is established based on the interface bonding strength level and stress concentration factor. For defects with weak bonding level, if the stress concentration factor is lower than the first factor threshold, it is judged as excellent repairable level. For defects with medium bonding level, if the stress concentration factor is lower than the second factor threshold, it is judged as generally repairable level. For defects with strong bonding level and / or defects with stress concentration factor exceeding the second factor threshold, it is judged as unrepairable level. At the same time, the proportion of defect size to product size is comprehensively considered. When the proportion is lower than the preset proportion threshold and the stress concentration factor is lower than the first factor threshold, the product is directly judged as qualified. The repair feasibility level is output.

[0013] Preferably, the process of adaptively correcting process parameters includes: An initial repair plan is obtained based on the gel type, defect type, and repair feasibility level, along with an initial combination of process parameters. The repair strength coefficient is adjusted based on the interfacial bonding strength level. A depth correction coefficient is calculated based on the ratio of defect depth to total gel thickness. The initial process parameters are corrected using the repair strength coefficient and the depth correction coefficient, including adjusting at least one of the following parameters: heating temperature, heating time, vibration frequency, vibration acceleration, and ultrasonic power. The corrected combination of process parameters is generated and output as an executable repair instruction.

[0014] Preferably, the process for determining the repair result includes: Perform the same deformation-induced testing process on the repaired product as before repair to obtain the relaxation time constant ratio and stress concentration factor after repair; calculate the change and rate of change of stress concentration factor before and after repair; when the decrease in stress concentration factor exceeds the preset success threshold and the stress concentration factor after repair is lower than the safety threshold, the repair is considered successful; when the stress concentration factor decreases but the decrease is insufficient and / or the value after repair still exceeds the safety threshold, the repair is considered unsuccessful; record the results of successful and unsuccessful repairs, along with the corresponding defect characteristic parameters and process parameters, in the database for optimization and updating of the repair strategy knowledge base.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention utilizes a "deformation-induced imaging" mechanism to transform the thickness compression deformation process of a gel under controlled micro-pressure into observable dynamic sequence images. Furthermore, it leverages holographic interferometric fringe patterns, phase reconstruction algorithms, and phase abrupt change region analysis to achieve highly sensitive detection of surface microcracks. This invention not only identifies the location, length, and direction of cracks but also extracts crack opening behavior at different stages of pressurization, holding, and depressurization. By analyzing the phase gradient change, it infers the crack opening displacement, crack depth, and stress state at the crack tip. Based on a comprehensive analysis model combining dynamic pressure response and phase information, crack detection is elevated from "identifying presence" to "quantifying behavior," significantly improving detection sensitivity and reliability. In practical quality inspection scenarios, this method effectively avoids false detections and missed detections caused by static lighting conditions, surface reflection, and weak textures, thereby comprehensively improving the monitoring accuracy and interpretability of initial micro-damage and evolving cracks.

[0016] 2. This invention not only detects defects but also constructs an interfacial bonding strength evaluation matrix centered on the "relaxation time constant ratio - stress concentration factor." By comparing the difference in relaxation time constants between defective and normal regions, this invention can reflect the local response speed of the gel and changes in the material's microstructure. Combined with the quantitative calculation of the stress concentration factor, this invention further describes the stress accumulation at the defect tip or interface. Furthermore, during microcrack detection, this invention utilizes the crack opening displacement at different pressure stages to estimate crack depth and the stress state at the crack tip, automatically identifying dangerous cracks and forming a complementary evaluation system with the interfacial bonding strength matrix. It can also integrate multi-dimensional parameters such as defect size ratio and stress concentration factor, improving the accuracy of determining whether a product is directly classified as qualified or enters the repair process.

[0017] 3. This invention not only detects and evaluates defects but also establishes an adaptive matching mechanism for repair schemes based on parameters such as interface bonding strength level, defect depth, and stress concentration factor. By using repair strength coefficients and depth correction coefficients to adjust initial process parameters in real time, it ensures sufficient repair of weak bonding defects while avoiding excessive energy application to strong bonding defects, thus significantly improving the accuracy and safety of the repair process. This invention also designs a "deformation-induced detection comparison before and after repair" mechanism. By re-collecting the ratio of relaxation time constants and stress concentration factors after repair and calculating the rate of change, it can automatically determine whether the repair was successful. Through adaptive repair schemes, dynamic adjustment of process parameters, and verifiable results, this invention achieves a closed-loop process from defect detection and repair guidance to repair verification, significantly improving the repair success rate and product consistency. Attached Figure Description

[0018] Figure 1 The flowchart of the quality inspection system for repairing surface defects of finished gel products based on image recognition provided by the present invention; Figure 2 A structural diagram of a quality inspection system for repairing surface defects in finished gel products based on image recognition is provided by the present invention. Figure 3 A flowchart for quality inspection of surface defects in the repair gel product provided by the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention. Example

[0020] Please see Figures 1 to 2 This invention provides a quality inspection system for repairing surface defects of finished gel products based on image recognition. The technical solution is as follows: a deformation-induced imaging module controls the target pressure to apply controllable micro-pressure deformation to the gel sample according to the projected area and hardness level of the gel sample, so that the gel undergoes thickness compression deformation. The system captures dynamic sequence images of the deformation process through intelligent sensors to detect surface micro-cracks. Furthermore, the detection process for surface microcracks includes: Digital holographic interferometry is used to acquire holographic interference fringe patterns at multiple moments during the deformation process. The interference fringe patterns are converted into surface phase distribution patterns using a phase reconstruction algorithm. The phase difference between the phase distribution patterns before and after deformation is calculated, and the phase abrupt change regions are extracted. Connectivity analysis and geometric feature extraction are performed on the phase abrupt change regions. Phase abrupt change regions with a length exceeding a preset threshold and exhibiting a linear or curved shape are identified as surface microcracks. The location coordinates, length, and orientation angle of the microcracks are recorded.

[0021] Specifically, holographic interference fringe patterns are acquired at multiple moments during the deformation process. A phase reconstruction algorithm is used to convert the interference fringe patterns into a surface phase distribution map, which reflects the microscopic changes in the gel surface height. The phase difference between the phase distribution maps before and after deformation is calculated to generate a phase difference map. For the extraction of phase abrupt change regions, the background noise level of the imaging system is pre-determined, and the standard deviation of the phase noise in a sample-free state is obtained. The phase abrupt change threshold is set to three to five times the standard deviation of the background noise to filter out system background noise interference. Pixels with phase difference values ​​exceeding this threshold are extracted as abrupt change regions.

[0022] Connectivity analysis is performed on the phase abrupt change regions, merging adjacent abrupt change pixels into independent objects; the skeleton length of each object is calculated. A preset threshold for microcrack length is set based on the critical crack size of the gel material; in typical applications, this threshold is set to ten times the optical resolution of the imaging system, or twice the average wavelength of the surface roughness of the gel product, to distinguish between real crack defects and normal surface texture undulations. Linear or curved regions with a length exceeding this threshold and an aspect ratio greater than five are identified as surface microcracks, and their geometric parameters are recorded.

[0023] In this embodiment, a quality inspection system for surface defects in finished gel products based on image recognition was constructed, achieving highly sensitive, non-contact detection of microcracks on the gel surface. The system utilizes controllable micro-pressure to induce repeatable deformation in the gel and combines this with digital holographic interferometry to acquire interference fringes at key moments. Phase reconstruction and phase difference analysis are then used to accurately locate phase abrupt change regions. By combining connectivity and geometric feature extraction, automatic identification and quantitative recording of microcracks are achieved. This approach effectively improves the accuracy of microcrack detection and enhances the stability and intelligence of the quality inspection process.

[0024] The feature analysis module extracts response features of defect areas from dynamic image sequences, distinguishes between free and fixed impurities by tracking the relative displacement trajectory of impurities, locates bubble defects by identifying stress birefringence rings, and detects the opening behavior of surface microcracks during deformation; calculates the ratio of relaxation time constant and stress concentration factor between defect areas and normal areas, and evaluates the interfacial bonding strength between defects and the gel matrix. The process of tracking the relative displacement trajectory of impurities to distinguish between free and fixed impurities includes: using an optical flow algorithm to calculate the pixel motion vector field between adjacent frames in a dynamic image sequence to obtain the velocity field distribution across the entire field of view; statistically analyzing the cumulative displacement of each pixel during the entire pressure-relief cycle to calculate the mean and standard deviation of the cumulative displacement in the normal gel matrix area; identifying regions where the cumulative displacement exceeds the mean plus three times the standard deviation as potential impurity regions; for the identified potential impurity regions, extracting the angle between the displacement vector direction and the applied pressure direction during the pressurization phase, as well as the ratio of the displacement to the overall gel thickness compression; when the angle is less than a preset angle threshold and the displacement ratio is close to or greater than the compression ratio, it is determined to be a free impurity, indicating that the impurity is loosely bound to the gel matrix and undergoes relative slippage during deformation; when the angle is large or the displacement ratio is significantly less than the compression ratio, it is determined to be a fixed impurity, indicating that the impurity is tightly bound to the gel matrix and deforms along with the matrix; and recording the position, size, and motion characteristic parameters of free and fixed impurities.

[0025] Preliminary segmentation is performed using the grayscale contrast of static images. The local grayscale gradient of the first frame is calculated, and regions with gradient values ​​significantly higher than the surrounding background are marked as Regions of Interest (ROIs) for impurities. This step is used to locate the pixel set of impurity objects in complex full-field data. Subsequently, an optical flow algorithm is used to calculate the pixel motion vector field between adjacent frames in the dynamic image sequence. For each marked impurity ROI, the average motion vector of all pixels within that region is calculated. Simultaneously, the average background motion vector of the gel matrix pixels in the annular neighborhood surrounding the impurity ROI is calculated. A relative displacement determination logic is defined: the vector difference between the average impurity motion vector and the background motion vector is calculated. If the magnitude of the vector difference is less than the displacement synchronization threshold, and the angle between the impurity movement direction and the background movement direction is less than the direction consistency threshold (it is recommended to take a value within five degrees), it indicates that the impurity moves with the matrix in the same direction with equal amplitude, and is judged as a fixed impurity. The displacement synchronization threshold is set according to the displacement control accuracy of the deformation induction device, which is usually twice the control accuracy. If the magnitude of the vector difference exceeds the displacement synchronization threshold, or the angle between the motion directions is greater than the direction consistency threshold, it indicates that relative slippage or rotation has occurred between the impurity and the matrix, and it is determined to be a free impurity.

[0026] Furthermore, the process of identifying stress birefringence rings to locate bubble defects includes: A sequence of transmitted light images is continuously acquired; the time-series standard deviation of grayscale values ​​is calculated for each pixel in the image sequence to generate a grayscale standard deviation distribution map; pixels with standard deviation values ​​exceeding a preset threshold are extracted from the grayscale standard deviation distribution map as regions of significant change; connected component analysis and boundary extraction are performed on the regions of significant change to calculate the geometric shape feature parameters of the boundaries; regions whose circularity feature values ​​of the boundary shape meet the ring determination criteria are identified as feature rings formed by stress optical effects (i.e., stress birefringence rings), and the center coordinates and radius of the feature rings are obtained; based on the radius of the feature rings, the applied pressure value, and the photoelastic coefficient of the gel material, the actual radius of the bubble is estimated through a pre-established calibration relationship to determine the location and size of the bubble defect.

[0027] The process of obtaining the photoelastic coefficient includes: preparing a standard test sample, which is a pure gel block of the same material and preparation process as the finished gel to be tested, ensuring that the sample is free of internal defects; placing the standard test sample in a photoelastic testing device equipped with a polarization light source and a polarization analyzer, which can apply known and controllable stress to the sample; applying a series of known compressive stresses of different magnitudes to the sample, acquiring transmitted polarized light images at each stress level, and observing stress-induced interference fringes; measuring the number of isotropic series generated at each stress level, and calculating the photoelastic coefficient according to the basic photoelastic equation, which establishes a quantitative relationship between stress difference, photoelastic coefficient, sample thickness, and isotropic series; performing linear regression on the test results of multiple stress levels to obtain the average photoelastic coefficient of the gel material; storing the calibrated photoelastic coefficient in a material parameter database as input parameters for subsequent bubble size estimation and stress analysis; repeating the calibration process for gel materials of different batches or different formulations to obtain the corresponding photoelastic coefficients.

[0028] The specific method for establishing the calibration relationship for the actual radius of the bubble is as follows: Standard gel samples containing bubbles of different known radii (true values ​​measured by microscopy) were pre-prepared. Under a set detection pressure, the system was used to measure the radius of the stress birefringence characteristic ring corresponding to each standard bubble. A calibration curve was plotted with the actual bubble radius as the abscissa and the characteristic ring radius as the ordinate. The least squares method was used to fit the data points to establish a mapping function between the characteristic ring radius and the actual bubble radius. Regarding the setting of the standard deviation threshold: when generating the grayscale standard deviation distribution map, the global average standard deviation of the entire image was calculated. The preset threshold was set to twice the global average standard deviation plus the fluctuation range of the global standard deviation, which served as the statistical boundary for distinguishing stress disturbance areas from background noise.

[0029] In this embodiment, a characteristic analysis module performs in-depth analysis of the mechanical and optical responses of defects on the gel surface and internally, enabling automatic differentiation and quantitative assessment of defect types. The system extracts dynamic features of microcracks, impurities, and bubbles from dynamic images, uses relative displacement to determine impurity properties, and identifies bubbles using a stress birefringence ring to infer their true size. Simultaneously, it assesses the bonding strength of defect interfaces based on the relaxation time constant ratio and stress concentration factor, improving the precision and quantification of defect identification and providing strong data support for the reliability of gel quality.

[0030] Furthermore, the process of detecting the opening behavior of surface microcracks includes: From the phase data of the deformation process acquired by digital holographic interferometry, the phase change curves of the identified surface microcrack region during the pressurization, holding, and depressurization stages are extracted; the phase gradient on both sides of the microcrack is calculated, and the crack opening displacement is estimated by the change in phase gradient; the relationship between the crack opening displacement and the applied pressure is analyzed to obtain the crack opening displacement; based on the crack opening displacement under different pressures, the crack depth and stress state at the crack tip are determined; cracks with opening displacement exceeding a preset threshold are identified as dangerous cracks.

[0031] In this embodiment, the crack opening behavior is accurately analyzed by dynamically tracking the phase changes of surface microcracks during the pressurization, holding, and depressurization stages. The system uses the phase gradient change to calculate the crack opening displacement and combines it with the pressure-displacement relationship to assess the crack depth and tip stress state. Multi-stage phase analysis obtains the crack response characteristics under real mechanical conditions, effectively distinguishing between ordinary cracks and potentially dangerous cracks. When the opening displacement exceeds a threshold, it is automatically marked as a dangerous crack, improving the quality inspection's ability to determine crack risk and significantly enhancing the reliability and safety of gel products.

[0032] For free impurity defects, due to the weak interfacial layer between the impurity and the gel matrix, the relaxation time constant of the defect region is significantly greater than that of the normal region during deformation recovery. The relaxation time constant ratio is usually greater than the preset first ratio threshold, and the stress concentration factor is relatively low due to stress release caused by interfacial slip around the impurity. For fixed impurity defects, the impurity is tightly bonded to the matrix, and the relaxation time constant ratio is close to that of the normal region. However, the stress concentration factor is high due to stress concentration caused by the difference in elastic modulus between the impurity and the matrix. For bubble defects, the recovery speed of the gel matrix around the bubble is slowed down after decompression due to the influence of the bubble cavity. The relaxation time constant ratio is at a moderate level, and the bubble acts as a stress concentration source, so the stress concentration factor is usually at a medium-high level. For surface microcracks, the continuity of the gel matrix at the crack is disrupted. During deformation recovery, the material on both sides of the crack recovers independently. The relaxation time constant ratio may be close to the normal value, but there is significant stress concentration at the crack tip, and the stress concentration factor is the highest. Different types of defects are identified by the combined characteristic pattern of relaxation time constant ratio and stress concentration factor, and the interfacial bonding strength between various defects and the matrix is ​​evaluated.

[0033] Obtaining the relaxation time constant ratio: During the zero-pressure recovery phase of the pressure-relief cycle, side view images of the gel sample were continuously acquired using a high-speed camera. The contour lines of the upper and lower surfaces of the gel were extracted using an edge detection algorithm. The thickness value at each moment was calculated, and the thickness recovery curve over time was plotted. The thickness recovery curves of the defect region and the normal region far from the defect were extracted separately. The viscoelastic recovery was fitted using the nonlinear least squares method to describe the exponential decay process of the thickness recovering from the compressed state to the initial state. The relaxation time constant was extracted from the fitting parameters. The ratio of the relaxation time constant of the defect region to that of the normal region was calculated. Obtaining the stress concentration factor: At the peak deformation moment, digital image correlation technology is used to perform full-field strain measurement on the area surrounding the defect. Regular grids or random speckles are divided on the image, and the components of the strain tensor are calculated by tracking the displacement of each sub-region before and after deformation. Based on the elastic modulus of the gel material, the strain tensor is converted into a stress tensor using Hooke's law under the small strain assumption. The von Mises equivalent stress at each measurement point is calculated, and the maximum value of the stress field around the defect is identified as the maximum equivalent stress. The average stress far from the defect area is calculated as the nominal stress. The stress concentration factor is obtained by dividing the maximum equivalent stress by the nominal stress.

[0034] Furthermore, the process of assessing the interfacial bonding strength between defects and the gel matrix includes: An interface bonding strength assessment matrix is ​​established, which divides the relaxation time constant ratio and stress concentration factor into multiple level intervals. For defects where the absolute value of the difference between the relaxation time constant ratio and the unit value is less than the first threshold, the interface bonding is determined to be strong, indicating that the viscoelastic response of the defect and the matrix is ​​consistent and the interface bonding is tight. For defects where the difference between the relaxation time constant ratio and the unit value is greater than the second threshold, the interface bonding is determined to be weak, indicating that there is a weak interface layer or loose bonding between the defect and the matrix. The interface bonding strength level is comprehensively determined based on the magnitude of the stress concentration factor, and the interface bonding strength is divided into three levels: strong bonding, medium bonding, and weak bonding.

[0035] The process of establishing the interfacial bonding strength assessment matrix includes: firstly, collecting test data from multiple batches and types of gel samples; measuring and recording the relaxation time constant ratio and stress concentration factor in both defective and normal regions of each sample to form a sample dataset containing different defect types and states; statistically analyzing the collected sample datasets to calculate the distribution characteristics of the relaxation time constant ratio and stress concentration factor under different interfacial bonding states, identifying the numerical boundary regions between different interfacial bonding strength levels; determining a first threshold for distinguishing between strong and moderate interfacial bonding and a second threshold for distinguishing between moderate and weak interfacial bonding based on the deviation of the relaxation time constant ratio from the unit value; constructing a two-dimensional judgment space with the relaxation time constant ratio on the horizontal axis and the stress concentration factor on the vertical axis, and dividing this two-dimensional space into multiple level intervals according to the first and second thresholds. Regions where the difference between the ratio of the relaxation time constant and the unit value is less than the first threshold and the stress concentration factor is in a lower range are marked as strong bonding level regions. Regions where the difference between the ratio of the relaxation time constant and the unit value is greater than the second threshold or the stress concentration factor is in a higher range are marked as weak bonding level regions. Regions in between are marked as medium bonding level regions, thus forming a complete interface bonding strength evaluation matrix.

[0036] The process of comprehensively determining the interface bonding strength level by combining the magnitude of the stress concentration factor includes: establishing a two-dimensional determination space, with the relaxation time constant ratio as the horizontal axis and the stress concentration factor as the vertical axis, dividing the determination space into multiple regions; when the relaxation time constant ratio is less than the first ratio threshold and the stress concentration factor is less than the first stress threshold, it is determined to be a weak bonding level, indicating that the mechanical response between the defect and the matrix is ​​highly consistent and the stress concentration degree is low, and the defect has little impact on the overall structure; when the relaxation time constant ratio is between the first ratio threshold and the second ratio threshold, or the stress concentration factor is between the first stress threshold and the second stress threshold, it is determined to be a medium bonding level, indicating that there is a certain interface difference or stress concentration effect between the defect and the matrix, which needs attention but still has the possibility of repair; when the relaxation time constant ratio exceeds the second ratio threshold or the stress concentration factor exceeds the second stress threshold, it is determined to be a strong bonding level, indicating that the mechanical behavior of the defect area is abnormal or there is severe stress concentration, the defect has been deeply embedded in the matrix or caused structural damage; for the case in the boundary region, auxiliary information such as defect type, location and size are comprehensively considered for determination; the determination result is output as the interface bonding strength level, as input for the repair feasibility assessment.

[0037] The first threshold (used to determine the boundary of strong binding) is set to one standard deviation of the normal gel baseline relaxation time constant; the second threshold (used to determine the boundary of weak binding) is set to three standard deviations of the normal gel baseline relaxation time constant.

[0038] Strong bonding level: When the absolute value of the difference between the ratio of the relaxation time constant of the defect region and the unit value is less than the first threshold, and the stress concentration factor is less than the low stress threshold (usually taken as 1.2, which indicates that the stress concentration effect is not significant), it is judged as strong bonding; this indicates that although the defect exists, its viscoelastic response is highly consistent with the matrix. Weak bond level: When the absolute value of the difference is greater than the second threshold, or the stress concentration factor is greater than the high stress threshold (usually 2.0, which represents the local stress multiplication), it is judged as a weak bond; this indicates that there is slip, voids or severe stress singularity at the interface; Medium combination level: The area between the two situations mentioned above.

[0039] In this embodiment, by constructing an interfacial bonding strength assessment matrix and jointly analyzing the relaxation time constant ratio and stress concentration factor, a quantitative classification of the interfacial bonding quality between defects and the gel matrix is ​​achieved. The system uses the relaxation time constant ratio to determine the viscoelastic consistency between the defect and the matrix, distinguishing between strong and weak bonding characteristics. This is then combined with the stress concentration factor reflecting the degree of localized stress anomalies to comprehensively determine the strong, medium, or weak level of interfacial bonding. This method achieves a refined assessment of interfacial bonding strength, providing a basis for defect risk assessment and subsequent process repair.

[0040] The repair matching module determines the repair feasibility level based on the interface bonding strength, matches the repair scheme, and adaptively corrects the process parameters based on the interface bonding strength. Furthermore, the process of determining the feasibility level of remediation includes: A repair feasibility judgment rule base is established based on the interface bonding strength level and stress concentration factor. For defects with weak bonding level, if the stress concentration factor is lower than the first factor threshold, it is judged as excellent repairable. For defects with medium bonding level, if the stress concentration factor is lower than the second factor threshold, it is judged as generally repairable. For defects with strong bonding level or defects with stress concentration factor exceeding the second factor threshold, it is judged as unrepairable. At the same time, the proportion of defect size to product size is comprehensively considered. When the proportion is lower than the preset proportion threshold and the stress concentration factor is lower than the first factor threshold, the product is directly judged as qualified. The repair feasibility level and corresponding processing suggestions are output.

[0041] The method for determining the first factor threshold is as follows: the linear elastic limit stress of the gel material is measured, and the ratio of this stress to the standard stress applied during the deformation induction process is used as the base, and multiplied by a preset safety factor (usually 0.8 to 0.9). When the actual detected stress concentration factor is lower than the first factor threshold, it indicates that the stress level around the defect is still completely within the linear elastic recovery range of the material. The repair process only needs to provide weak thermal or vibrational energy to assist in the rearrangement of molecular chains, thus corresponding to the "high-quality repairable level".

[0042] The second factor threshold is determined by measuring the critical stress at the point where the gel material undergoes micro-yield or microcrack propagation, and using the ratio of this stress to the standard stress as the threshold. When the actual stress concentration factor is between the first factor threshold and the second factor threshold, it indicates that the local area has approached or produced weak plastic deformation, requiring a higher energy repair process, corresponding to the "generally repairable level".

[0043] When the stress concentration factor exceeds the second factor threshold, it indicates that the stress intensity at the defect tip has exceeded the material's bearing limit, and there is a high probability that irreversible structural fracture or deep damage has occurred. At this point, no matter what repair process is applied, the material properties cannot be restored, and it is therefore judged as "unrepairable".

[0044] In this embodiment, the repair matching module comprehensively assesses the interface strength, stress concentration factor, and defect size ratio to establish a repair feasibility level, enabling intelligent decision-making regarding gel defect treatment methods. The system categorizes repair feasibility into high-quality repairable, generally repairable, and unrepairable levels, and automatically outputs corresponding process recommendations. Defects with small size ratios and low stress risk can be directly classified as qualified products, improving inspection efficiency. This module achieves precise matching of repair strategies and adaptive adjustment of process parameters, enhancing product repair resource utilization and ensuring quality consistency.

[0045] Furthermore, the process of adaptively correcting the process parameters includes: The system retrieves an initial repair plan from the repair strategy knowledge base based on gel type, defect type, and repair feasibility level, and obtains an initial combination of process parameters. It then adjusts the repair strength coefficient according to the interfacial bonding strength level, using a lower coefficient for weak bonding and a higher coefficient for moderate bonding. A depth correction coefficient is calculated based on the ratio of defect depth to total gel thickness. The initial process parameters are then corrected using the repair strength coefficient and depth correction coefficient, including adjusting at least one of the following parameters: heating temperature, heating time, vibration frequency, vibration acceleration, or ultrasonic power. Finally, the corrected combination of process parameters is generated and output as an executable repair instruction.

[0046] The specific process of correcting the initial process parameters using the repair strength coefficient and depth correction coefficient includes: For heat repair processes, the initial heating temperature is the characteristic temperature of the gel material. The corrected heating temperature equals the initial heating temperature multiplied by the repair strength coefficient. For weak bonding levels, the repair strength coefficient is set to a smaller value, while for medium bonding levels, it is set to a larger value. Further adjustments to the heating temperature are made based on a depth correction coefficient. This coefficient is calculated according to the ratio of defect depth to total gel thickness. For shallow defects, a lower heating temperature and shorter heating time are used to avoid surface overheating damage. For deep defects, a higher heating temperature or longer heating time is used to ensure sufficient heat conduction to the defect location and avoid overall overheating. The final corrected heating temperature equals the initial heating temperature multiplied by the repair strength coefficient and then multiplied by the depth correction coefficient. For vibration repair... The process involves retrieving the initial vibration frequency and acceleration from the knowledge base based on the defect type. The corrected vibration acceleration equals the initial vibration acceleration multiplied by the repair strength coefficient, and the corrected vibration time equals the initial vibration time multiplied by the depth correction coefficient, allowing deeper defects to receive a longer vibration duration. For ultrasonic repair processes, the corrected ultrasonic power equals the initial ultrasonic power multiplied by the repair strength coefficient, and the focusing depth is directly set based on the defect depth. For composite repair processes, the parameters of each process step are adjusted according to the same correction logic. During the correction process, all parameters undergo safety boundary checks to ensure that the corrected parameters do not exceed the maximum working capacity of the equipment and the tolerance limit of the gel material, avoiding secondary damage. The complete set of corrected process parameters is then output.

[0047] In this embodiment, the precise execution of the gel defect repair scheme is achieved through adaptive correction of process parameters. Based on the gel type, defect type, and repair feasibility level, the system retrieves initial process parameters from a knowledge base. It then adjusts the repair strength coefficient by combining interface strength with the depth of the defect and the gel thickness to calculate a depth correction coefficient. Parameters such as heating temperature, heating time, vibration frequency, vibration acceleration, or ultrasonic power are optimized and corrected. Finally, an executable repair command is generated, achieving intelligent matching and dynamic optimization of process parameters, thus improving repair effectiveness and product quality consistency.

[0048] The re-inspection and verification module repeatedly performs deformation-induced detection on the repaired product to obtain the characteristic parameters after repair, and compares the change in stress concentration factor before and after repair to determine the repair result. Furthermore, the process for determining the repair result includes: performing the same deformation-induced testing procedure on the repaired product as before the repair, obtaining the relaxation time constant ratio and stress concentration factor after repair; calculating the change and rate of change of the stress concentration factor before and after repair; determining the repair as successful when the decrease in stress concentration factor exceeds the preset success threshold and the stress concentration factor after repair is lower than the safety threshold; determining the repair as unsuccessful when the stress concentration factor decreases but the decrease is insufficient or the value after repair still exceeds the safety threshold; and recording the results of successful and unsuccessful repairs, along with the corresponding defect characteristic parameters and process parameters, into the database for optimization and updating of the repair strategy knowledge base.

[0049] The preset success threshold is defined as the relative improvement rate of the stress concentration factor, and it is recommended to set it to 30%. That is, the stress concentration factor after repair should be reduced by at least 30% compared to before repair for the process to be considered effective.

[0050] The safety threshold is an absolute value set based on the fracture mechanics criteria of gel materials. Through material destructive experiments, the critical stress concentration factor that causes macroscopic crack propagation or functional failure of the gel is determined, and 80 percent of this critical value is set as the safety threshold.

[0051] The success of the repair is determined by the decrease in stress concentration factor and the safety threshold. Defect characteristic parameters and process parameters for successful or unsuccessful repairs are recorded and fed back to the repair strategy knowledge base for closed-loop optimization. This module ensures that the repair effect is quantifiable and traceable, improving the reliability of product quality control and the ability to continuously improve repair strategies.

[0052] It also includes a post-repair performance degradation prediction function, used for: measuring not only the instantaneous change of stress concentration factor during the re-inspection of the repaired product, but also the viscoelastic parameters of the repaired area, including elastic modulus and loss factor; comparing the elastic modulus of the repaired area with that of the unrepaired normal area to calculate the modulus deviation rate; comparing the loss factor of the repaired area with that of the unrepaired normal area to assess the energy dissipation characteristics of the repaired area; predicting the performance degradation trend of the repaired area during long-term use based on the differences in mechanical parameters of the repaired area and combined with the aging model of the gel material; calculating the expected service life of the repaired product, and marking the product as substandard and downgrading it or transferring it to the scrap channel when the expected service life meets the requirements; marking the repair information and recommended service life on the product label when the expected service life meets the requirements; and feeding the performance degradation prediction results back to the repair strategy knowledge base to optimize the repair process parameters to improve the long-term stability of the repaired area.

[0053] By using the post-repair performance degradation prediction function, the viscoelastic parameters of the repaired area, such as elastic modulus and loss factor, are measured, the modulus deviation rate is calculated, and the energy dissipation characteristics are evaluated. Combined with the gel material aging model, the long-term performance degradation trend and expected service life are predicted. When the expected service life is lower than the design requirements, the product is marked as substandard and downgraded or scrapped. When the requirements are met, the repair information and recommended service life are marked. The prediction results are fed back to the repair strategy knowledge base to achieve process optimization and improve the long-term stability of the repaired area and the reliability of the product.

[0054] This invention provides an image recognition-based quality inspection and intelligent repair system for surface defects in gel products, achieving closed-loop management of the entire process from defect detection, characteristic analysis, interfacial bonding strength assessment, repair scheme matching and adaptive process adjustment, to repair verification and performance degradation prediction. The system uses deformation-induced imaging and digital holographic interferometry to accurately capture the dynamic sequence of microcracks and quantify crack opening behavior. It combines the relaxation time constant ratio and stress concentration factor to analyze different types of defects and their interfacial bonding strength. The repair matching module determines the repair feasibility level based on the interfacial bonding strength and defect size, obtains an initial process scheme from a knowledge base, and adaptively corrects process parameters such as heating, vibration, and ultrasound using repair strength coefficients and depth correction coefficients to achieve precise and controllable defect repair. The re-inspection and verification module quantifies changes in stress concentration factors and judges the repair effectiveness by repeatedly performing deformation-induced detection on the repaired product, while recording defect characteristics and process parameters and feeding them back to the knowledge base to achieve closed-loop optimization; see details below. Figure 3 The system further combines elastic modulus and loss factor analysis to predict long-term performance degradation and expected lifespan of the repaired area, supporting product quality management and continuous optimization of repair strategies, and significantly improving the reliability and safety of the finished gel product. Example

[0055] This invention provides a quality inspection system for repairing surface defects in finished gel products based on image recognition. The technical solution is as follows: a deformation-induced imaging module, which controls the target pressure to apply controllable micro-pressure deformation to the gel sample according to the projected area and hardness grade of the gel sample, causing thickness compression deformation of the gel, capturing dynamic sequence images of the deformation process, and detecting surface microcracks; a characteristic analysis module, which extracts the response features of the defect area from the dynamic image sequence, distinguishes between free impurities and fixed impurities by tracking the relative displacement trajectory of impurities, locates bubble defects by identifying stress birefringence rings, and detects the opening behavior of surface microcracks during the deformation process; calculates the ratio of the relaxation time constant of the defect area to that of the normal area and the stress concentration factor, and evaluates the interfacial bonding strength between the defect and the gel matrix; a repair matching module, which determines the repair feasibility level based on the interfacial bonding strength, matches repair schemes, and adaptively corrects process parameters based on the interfacial bonding strength; and a re-inspection and verification module, which repeatedly performs deformation-induced detection on the repaired product to obtain the characteristic parameters after repair, and compares the change in stress concentration factor before and after repair to determine the repair result.

[0056] A quality inspection system for surface defects in finished gel products based on image recognition also includes an environmental monitoring and adaptive parameter control module and a multi-defect interaction evaluation module; the specific scheme is as follows: The environmental monitoring and parameter adaptive control module is used for: real-time acquisition of ambient temperature and relative humidity at the testing station; obtaining the temperature-viscosity relationship parameters and glass transition temperature of the material according to the type of gel material; calculating the temperature deviation between the current ambient temperature and the standard testing temperature, and calculating the viscosity correction coefficient of the gel material at the current temperature based on the temperature-viscosity relationship; adjusting the applied target pressure according to the viscosity correction coefficient. When the ambient temperature is higher than the standard temperature, the viscosity decreases, and the target pressure is reduced accordingly to generate sufficient deformation. When the ambient temperature is lower than the standard temperature, the viscosity increases, and the target pressure is increased accordingly to avoid damage caused by excessive stress; simultaneously, adjusting the deformation holding time and recovery observation time after depressurization according to the viscosity correction coefficient to ensure the full development of the viscoelastic response; outputting the adjusted testing parameters to the deformation-induced imaging module to realize the environmentally adaptive testing process; and recording the environmental parameters and adjusted testing parameters for each test for standardized processing of test results and comparative analysis of historical data.

[0057] The process of obtaining temperature-viscosity parameters and glass transition temperature includes: performing temperature scanning tests on the gel material sample using a differential scanning calorimeter, heating from a low-temperature region to a high-temperature region at a preset heating rate, and recording the heat flux versus temperature curve; identifying the glass transition region from the heat flux-temperature curve, and determining the temperature corresponding to the abrupt change in the slope of the heat flux curve as the glass transition temperature; measuring the dynamic viscosity of the gel material at multiple temperature points using a rotational rheometer to obtain discrete temperature-viscosity data points; performing nonlinear fitting on the discrete temperature-viscosity data based on the William Randall-Ferry equation or the Arrhenius equation to extract temperature-viscosity parameters, including the viscosity value at the reference temperature, activation energy, or characteristic constants of the WLF equation; and storing the glass transition temperature and temperature-viscosity parameters in a material parameter database to establish a mapping relationship between gel material types and corresponding parameters.

[0058] The process of obtaining the viscosity correction coefficient includes: real-time acquisition of the ambient temperature at the testing station, calculation of the temperature deviation between the current ambient temperature and the standard testing temperature; calculation of the theoretical viscosity of the gel material at the current ambient temperature and the theoretical viscosity at the standard testing temperature by substituting the temperature-viscosity relationship parameters obtained from the material parameter database into the temperature-viscosity relationship equation; calculation of the viscosity correction coefficient, which is equal to the theoretical viscosity at the current ambient temperature divided by the theoretical viscosity at the standard testing temperature; when the viscosity correction coefficient is greater than the unit value, it indicates that the current ambient temperature is lower than the standard temperature, and the gel viscosity increases; when the viscosity correction coefficient is less than the unit value, it indicates that the current ambient temperature is higher than the standard temperature, and the gel viscosity decreases; and adjustment of the target pressure based on the viscosity correction coefficient, wherein the adjusted target pressure is equal to the initial target pressure multiplied by the viscosity correction coefficient.

[0059] By constructing an environmental monitoring and parameter adaptive control module, the system achieves intelligent response to changes in ambient temperature during the testing process. The system uses a differential scanning calorimeter and a rotational rheometer to pre-calibrate the glass transition temperature and temperature-viscosity relationship parameters of the gel material, and establishes a quantitative model of the material's viscoelastic behavior based on the WLF equation or the Arrhenius equation. In actual testing, the system calculates the viscosity correction coefficient in real time and dynamically adjusts the target pressure, deformation holding time, and recovery observation time to ensure consistent deformation response and sufficient viscoelastic development under different temperature environments. This mechanism effectively eliminates the interference of ambient temperature fluctuations on the test results, improving the stability, repeatability, and data comparability of the quality inspection system in actual production environments.

[0060] The multi-defect interaction assessment module is used to: statistically analyze the number, type, location, and size of all defects identified on a single gel sample; for samples with more than a preset threshold of defects, calculate the center distance between any two defects and identify defect pairs or defect groups with a center distance less than a critical distance; for close defect pairs, assess the spatial superposition effect of the stress fields generated by the two defects based on the superposition principle and calculate the maximum equivalent stress in the superposition area; compare the maximum equivalent stress after superposition with the stress concentration factor in the case of a single defect to calculate the stress amplification factor; when the stress amplification factor exceeds a preset threshold, it is determined that there is a significant defect interaction effect, and the overall risk level of the sample is increased by one level; for samples with defect interactions, when matching repair strategies, priority is given to repairing the critical defects located at the center of the defect group and / or those that contribute the most to stress superposition; for samples with severe defect interactions and whose stress superposition risk cannot be eliminated after repair, they are directly determined to be unrepairable and output to the scrap channel; output a multi-defect interaction assessment report, including a defect distribution map, stress superposition area markings, and repair priority ranking.

[0061] The specific process for obtaining the center distance includes: acquiring the location information of all identified defects on a single gel sample, the location information including the abscissa and ordinate of the geometric center of each defect in the image coordinate system; for three-dimensional detection scenarios, simultaneously acquiring the depth coordinates of the defect centers; for the i-th and j-th defects on the sample, calculating the center distance between them, the center distance is calculated using the Euclidean distance formula, in two dimensions the center distance is equal to the square root of the sum of the squares of the differences in the abscissa and ordinate of the two defects, in three dimensions the center distance is equal to the square root of the sum of the squares of the differences in the abscissa, the squares of the differences in the ordinate and the squares of the differences in the depth coordinate; traversing all defect pair combinations to construct a defect distance matrix, the element in the i-th row and j-th column of the defect distance matrix being the center distance between the i-th and j-th defects; marking defect pairs with a center distance less than a critical distance threshold as adjacent defect pairs, the critical distance threshold being determined based on the equivalent radius of the defect and the stress transfer characteristic length of the gel material.

[0062] The specific process for obtaining the maximum equivalent stress in the superimposed region includes: for each identified pair of adjacent defects, establishing a local stress field model centered on each defect; selecting an analytical or numerical solution for each stress field model based on the defect type, where a spherical cavity stress field model is used for bubble defects, a rigid inclusion stress field model for impurity defects, and a crack tip stress field model for microcrack defects; determining the superimposed region of the stress fields of the two defects, which is the spatial region surrounding the line connecting the centers of the two defects; establishing a computational grid within the superimposed region, and calculating the stress tensor components generated at each node of the grid when the two defects act alone; based on the principle of linear elastic superposition, algebraically superimposing the stress tensor components generated by the two defects at the same node to obtain the superimposed total stress tensor; calculating the von Mises equivalent stress on the superimposed total stress tensor, which is calculated using standard formulas based on the principal stresses or stress components of the stress tensor; traversing all computational nodes within the superimposed region and extracting the maximum value of the von Mises equivalent stress as the maximum equivalent stress in the superimposed region.

[0063] The specific process for obtaining the stress amplification factor includes: obtaining the stress concentration factor of each defect in an adjacent defect pair when it exists alone; the stress concentration factor of a single defect is measured by the characteristic analysis module during deformation-induced detection; calculating the larger value of the two stress concentration factors of a single defect as the reference stress concentration factor in the case of a single defect; dividing the maximum equivalent stress of the superimposed region by the nominal stress in the far field to obtain the equivalent stress concentration factor in the case of defect interaction; calculating the stress amplification factor, which is equal to the equivalent stress concentration factor in the case of defect interaction divided by the reference stress concentration factor in the case of a single defect; when the stress amplification factor is greater than a unit value, it indicates that the defect interaction has produced a stress enhancement effect; when the stress amplification factor is greater than a preset factor threshold, it is determined that there is a significant defect interaction, and the defect pair is marked as a high-risk defect combination.

[0064] By constructing a multi-defect interaction assessment module, a systematic risk analysis of complex defect distributions was achieved. The system uses the Euclidean distance formula to calculate the center-to-center distance between defects and constructs a distance matrix to quickly identify adjacent defect pairs. Corresponding analytical stress field models are established for different defect types, and the von Mises equivalent stress peak value in the superposition region is calculated based on the principle of linear elastic superposition. The enhanced effect of multi-defect interactions is quantified through stress amplification, enabling dynamic adjustment of risk levels. This mechanism effectively identifies defect groups that might be overlooked in individual assessments but pose a serious threat when combined, providing a scientific basis for prioritizing repairs and determining scrap, and significantly improving the quality inspection system's risk identification capability and the accuracy of disposal decisions for multi-defect samples.

[0065] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A quality inspection system for repairing surface defects in finished gel products based on image recognition, characterized in that, include: The deformation-induced imaging module applies controllable micro-pressure deformation to the gel sample based on the projected area and hardness level of the gel sample, causing the gel to undergo thickness compression deformation, capturing dynamic sequence images of the deformation process, and detecting surface micro-cracks. The feature analysis module extracts the response features of the defect area from the dynamic image sequence, distinguishes between free impurities and fixed impurities by tracking the relative displacement trajectory of impurities, locates bubble defects by identifying stress birefringence rings, and detects the opening behavior of surface microcracks during deformation. Calculate the ratio of relaxation time constants and stress concentration factors between defective and normal regions to assess the interfacial bonding strength between defects and the gel matrix; The repair matching module determines the repair feasibility level based on the interface bonding strength, matches the repair scheme, and adaptively corrects the process parameters based on the interface bonding strength. The re-inspection and verification module repeatedly performs deformation-induced detection on the repaired product to obtain the characteristic parameters after repair, and compares the change in stress concentration factor before and after repair to determine the repair result.

2. The quality inspection system for repairing surface defects of finished gel products based on image recognition according to claim 1, characterized in that: The detection process for surface microcracks includes: Holographic interference fringe patterns are acquired at multiple moments during the deformation process; the interference fringe patterns are converted into surface phase distribution patterns using a phase reconstruction algorithm; the phase difference between the phase distribution patterns before and after deformation is calculated, and phase abrupt change regions are extracted; connectivity analysis and geometric feature extraction are performed on the phase abrupt change regions, and phase abrupt change regions with lengths exceeding a preset threshold and exhibiting linear and / or curved shapes are identified as surface microcracks, and the location coordinates, length, and orientation angle of the microcracks are recorded.

3. The quality inspection system for repairing surface defects of finished gel products based on image recognition according to claim 1, characterized in that: The process of identifying stress birefringence rings and locating bubble defects includes: A sequence of transmitted light images is continuously acquired. The time series standard deviation of grayscale values ​​for each pixel in the image sequence is calculated to generate a grayscale standard deviation distribution map. Pixels with standard deviation values ​​exceeding a preset threshold are extracted from the grayscale standard deviation distribution map as regions of significant change. Connectivity analysis and boundary extraction are performed on the regions of significant change to calculate the geometric shape feature parameters of the boundaries. Regions whose circularity feature values ​​of the boundary shape meet the ring determination criteria are identified as feature rings formed by stress effects, and the center coordinates and radius of the feature rings are obtained. Based on the radius of the feature rings, the applied pressure value, and the photoelastic coefficient of the gel material, the actual radius of the bubble is estimated through a pre-established calibration relationship to determine the location and size of the bubble defect.

4. The quality inspection system for repairing surface defects of finished gel products based on image recognition according to claim 1, characterized in that: The process of detecting the opening behavior of surface microcracks includes: From the acquired deformation process phase data, the phase change curves of the identified surface microcrack region during the pressurization, holding, and depressurization stages are extracted; the phase gradient on both sides of the microcrack is calculated, and the crack opening displacement is estimated by the change in phase gradient; the crack opening displacement is analyzed as a function of applied pressure to obtain the crack opening displacement; based on the crack opening displacement under different pressures, the crack depth and stress state at the crack tip are determined; cracks with opening displacement exceeding a preset threshold are identified as dangerous cracks.

5. The quality inspection system for repairing surface defects of finished gel products based on image recognition according to claim 1, characterized in that: The process of assessing the interfacial bonding strength between defects and the gel matrix includes: An interface bonding strength assessment matrix is ​​established, which divides the relaxation time constant ratio and stress concentration factor into multiple level intervals. For defects where the absolute value of the difference between the relaxation time constant ratio and the unit value is less than a first threshold, and for defects where the difference between the relaxation time constant ratio and the unit value is greater than a second threshold, the interface bonding is determined to be weak. The interface bonding strength level is comprehensively determined by combining the magnitude of the stress concentration factor, and the interface bonding strength is divided into three levels: strong bonding, medium bonding, and weak bonding.

6. The quality inspection system for repairing surface defects of finished gel products based on image recognition according to claim 1, characterized in that: The process of determining the feasibility level of repair includes: A repair feasibility judgment rule base is established based on the interface bonding strength level and stress concentration factor. For defects with weak bonding level, if the stress concentration factor is lower than the first factor threshold, it is judged as excellent repairable level. For defects with medium bonding level, if the stress concentration factor is lower than the second factor threshold, it is judged as generally repairable level. For defects with strong bonding level and / or defects with stress concentration factor exceeding the second factor threshold, it is judged as unrepairable level. At the same time, the proportion of defect size to product size is comprehensively considered. When the proportion is lower than the preset proportion threshold and the stress concentration factor is lower than the first factor threshold, the product is directly judged as qualified. The repair feasibility level is output.

7. The quality inspection system for repairing surface defects of finished gel products based on image recognition according to claim 1, characterized in that: The process of adaptively correcting process parameters includes: An initial repair plan is obtained based on the gel type, defect type, and repair feasibility level, along with an initial combination of process parameters. The repair strength coefficient is adjusted based on the interfacial bonding strength level. A depth correction coefficient is calculated based on the ratio of defect depth to total gel thickness. The initial process parameters are corrected using the repair strength coefficient and the depth correction coefficient, including adjusting at least one of the following parameters: heating temperature, heating time, vibration frequency, vibration acceleration, and ultrasonic power. The corrected combination of process parameters is generated and output as an executable repair instruction.

8. The quality inspection system for repairing surface defects of finished gel products based on image recognition according to claim 1, characterized in that: The process of determining the repair results includes: Perform the same deformation-induced testing process on the repaired product as before repair to obtain the relaxation time constant ratio and stress concentration factor after repair; calculate the change and rate of change of stress concentration factor before and after repair; when the decrease in stress concentration factor exceeds the preset success threshold and the stress concentration factor after repair is lower than the safety threshold, the repair is considered successful; when the stress concentration factor decreases but the decrease is insufficient and / or the value after repair still exceeds the safety threshold, the repair is considered unsuccessful; record the results of successful and unsuccessful repairs, along with the corresponding defect characteristic parameters and process parameters, in the database for optimization and updating of the repair strategy knowledge base.

Images