An automotive parts coating quality monitoring system and method

Abstract

This invention discloses a coating quality monitoring system and method for automotive parts. It enhances the optical image of the coating surface to generate a scattering spectrum and extracts defect features. Based on defect severity grading and identification priority calculation, it achieves intelligent defect identification and priority management. By constructing the association path between major and minor defects and detecting aggregation intensity, it locates key problem areas and predicts potential risk areas based on multi-dimensional similarity analysis. It constructs a coupling spectrum by performing a three-dimensional correlation mapping of gloss distribution features, thickness gradient, and color difference distribution and determines the judgment criteria. It performs diffusion tracking on key problem areas and generates influence radius parameters through dynamic boundary processing, constructing a zonal detection strategy based on detection sensitivity. Through abnormal coupling area identification and imbalance compensation parameter generation, it performs dynamic grading adjustments to achieve accurate quality level determination.

Description

[[IDID=0]]Technical Field

[0001] The present invention relates to the technical field of coating quality detection, and particularly to a coating quality monitoring system and method for automotive parts. Background Art

[0002] The coating quality of automotive parts directly affects the anti-corrosion performance, appearance quality and service life of products, and is a key quality control link in the automotive manufacturing process. Common defects on the coating surface include various types such as cracks, bubbles, particles, sagging, orange peel, color difference, etc. These defects have significant differences in formation mechanisms, severity levels, and impacts on product performance. Currently, coating quality detection methods in industrial production mainly rely on manual visual inspection and conventional instrument measurements. Manual visual inspection relies on the experience judgment of inspectors, which has problems such as strong subjectivity, low efficiency, and easy fatigue, and it is difficult to detect minor defects. Conventional instrument measurement methods such as gloss meters, thickness gauges, color difference meters, etc. can obtain single-index data such as the gloss, thickness, and color difference of the coating, but various instruments work independently, and the obtained data is isolated from each other, lacking the comprehensive analysis ability of the correlation relationships between multiple quality indicators.

[0003] Existing detection technologies also have the following limitations: the identification of defects stays at the single-point detection level, and fails to analyze the distribution law and aggregation characteristics of defects as a whole; the quality judgment criteria are fixed and unified, without considering the differences in the impacts of different regions and different defect types on product quality; the detection process lacks pertinence, using the same detection method and detection density for all regions, which not only causes waste of detection resources but may also miss key problem areas. Therefore, there is an urgent need for a method to solve at least one of the above problems. Summary of the Invention

[0004] The present invention discloses a coating quality monitoring system and method for automotive parts. By enhancing the optical image of the coating surface to generate a scattering pattern and extracting defect features, intelligent identification and priority management of defects are achieved based on defect severity grading and recognition priority calculation; by constructing the association path between major defects and minor defects and detecting the aggregation intensity to locate key problem areas, and predicting potential risk areas based on multi-dimensional similarity analysis; mapping the glossiness distribution characteristics, thickness gradient, and color difference distribution in three dimensions to construct a coupling map and determine the judgment criteria; implementing diffusion tracking on key problem areas and generating influence radius parameters through dynamic boundary processing, and constructing a partition detection strategy in combination with detection sensitivity; through abnormal coupling region identification and unbalance compensation parameter generation for dynamic grading adjustment, accurate determination of quality grades is achieved, providing strong support for intelligent quality control and process optimization.

[0005] In the first aspect of the present invention, a coating quality monitoring method for automotive parts is proposed, including the following steps:

[0006] Obtain optical images and process parameter information of the coating surface, perform dark field enhancement processing on the optical images of the coating surface to generate a scattering enhancement spectrum, and generate defect identification weights based on the scattering enhancement spectrum;

[0007] The key defect cluster area is located by the defect identification weight, the shape similarity feature of the key defect cluster area is extracted, and the shape similarity feature is analyzed in multiple dimensions to determine the potential defect risk area.

[0008] The gloss distribution characteristics and color difference spatial distribution of the potential defect risk area are obtained. Thickness fluctuation characteristics are identified from the process parameter information to generate thickness gradient parameters. The gloss distribution characteristics, the thickness gradient parameters and the color difference spatial distribution are three-dimensionally correlated and mapped to generate a gloss-thickness-color difference coupling spectrum. The defect judgment boundary is determined based on the gloss-thickness-color difference coupling spectrum.

[0009] The detection sensitivity is determined by combining the defect judgment boundary and the scattering enhancement spectrum. An influence range analysis is performed on the key defect cluster area to generate an influence radius parameter. A zonal detection strategy is constructed based on the influence radius parameter and the detection sensitivity.

[0010] For the gloss-thickness-color difference coupled spectrum, abnormal coupling regions are identified and detected. Based on the abnormal coupling regions and the morphological similarity features, similarity matching is performed to generate imbalance compensation parameters. Based on the imbalance compensation parameters and the partition detection strategy, a quality level judgment result is generated.

[0011] A second aspect of this invention provides a coating quality monitoring system for automotive parts, comprising:

[0012] The data acquisition module is used to acquire optical images and process parameter information of the coating surface, perform dark field enhancement processing on the optical images of the coating surface to generate a scattering enhancement spectrum, and generate defect identification weights based on the scattering enhancement spectrum;

[0013] The defect localization module is used to locate key defect clusters through the defect identification weights, extract the shape similarity features of the key defect clusters, and perform multi-dimensional similarity analysis on the shape similarity features to determine potential defect risk areas;

[0014] The coupling mapping module is used to obtain the gloss distribution characteristics and color difference spatial distribution of the potential defect risk area, identify thickness fluctuation characteristics from the process parameter information to generate thickness gradient parameters, perform three-dimensional correlation mapping between the gloss distribution characteristics, the thickness gradient parameters and the color difference spatial distribution to generate a gloss-thickness-color difference coupling map, and determine the defect judgment boundary based on the gloss-thickness-color difference coupling map.

[0015] The detection optimization module is used to determine the detection sensitivity by combining the defect judgment boundary and the scattering enhancement spectrum, perform influence range analysis on the key defect cluster area to generate influence radius parameters, and construct a zonal detection strategy based on the influence radius parameters and the detection sensitivity.

[0016] The determination output module is used to identify abnormal coupling regions in the gloss-thickness-color difference coupled spectrum, perform similarity matching between the abnormal coupling regions and the morphological similarity features to generate imbalance compensation parameters, and generate quality level determination results based on the imbalance compensation parameters and the partition detection strategy.

[0017] The beneficial effects of this invention are reflected in the following points: First, by performing dark-field enhancement processing on the coating surface image to highlight the scattered light signal of defects, the geometric, morphological, and optical features of defects are extracted to establish a multi-dimensional evaluation system. Combined with priority calculation, differentiated allocation of detection resources is achieved. Furthermore, by constructing a network of associated paths between major and minor defects and calculating the aggregation intensity index to locate key problem areas, and by constructing multi-dimensional feature vectors based on morphological similarity, scattering intensity, size, and process parameters for similarity analysis, areas with highly similar feature vectors, even if no obvious defects are currently detected, can be identified as potential risk areas. This achieves intelligent monitoring throughout the entire process from defect identification and hierarchical management to risk prediction. Second, by constructing a coupling spectrum through three-dimensional correlation mapping of gloss distribution characteristics, thickness gradient parameters, and color difference spatial distribution, the limitations of traditional methods that only focus on a single quality parameter are overcome. This allows for the identification of parameter coupling imbalances such as mismatch between gloss and thickness, and incoordination between color difference and gloss. Based on the data distribution characteristics of normal quality areas and defective quality areas in the coupling spectrum, the defect judgment boundary is determined, making quality evaluation more comprehensive and accurate. Third, by tracking and identifying diffusion segments in key problem areas to accelerate diffusion and generating diffusion enhancement factors, combined with dynamic boundary processing to determine the influence radius parameters, the coating surface is divided into core area, influence area, and peripheral area, and a differentiated detection strategy is constructed based on detection sensitivity. Furthermore, by identifying abnormal coupling areas and generating imbalance compensation parameters, the quality judgment criteria are dynamically adjusted. By combining the differences in detection response to establish a constraint response distribution field and dynamically adjusting the classification based on the imbalance utilization factor, the optimal allocation of detection resources and the accurate determination of quality level are achieved. Attached Figure Description

[0018] The accompanying drawings illustrate specific examples of the technical solutions described in this invention and, together with the detailed embodiments, form part of the specification, serving to explain the technical solutions, principles, and effects of this invention.

[0019] Unless otherwise specified or otherwise, the same reference numerals in different figures represent the same or similar technical features, and different reference numerals may be used to represent the same or similar technical features.

[0020] Figure 1 This is a flowchart of a method for monitoring the coating quality of automotive parts according to the present invention.

[0021] Figure 2 This is a structural block diagram of an automotive parts coating quality monitoring system according to the present invention. Detailed Implementation

[0022] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0023] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0024] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0025] The technical solutions of the embodiments of this application are described below.

[0026] like Figure 1 As shown, this embodiment of the invention provides a method for monitoring the coating quality of automotive parts, including the following steps S110-S150:

[0027] Step S110: Obtain optical images and process parameter information of the coating surface, perform dark field enhancement processing on the optical images of the coating surface to generate scattering enhancement spectrum, and generate defect identification weights based on the scattering enhancement spectrum.

[0028] Specifically, the process involves acquiring optical images and process parameters of the coating surface. Complete image data of the coating surface, including surface morphology, color distribution, and reflected light intensity, is collected using an optical imaging system. A high-resolution industrial camera, with a resolution of at least 20 megapixels, is deployed to scan and photograph the coating surface to ensure the capture of minute surface defects. Uniform illumination is achieved using a ring light source or coaxial light source, with the illumination angle controlled between 45 and 60 degrees to avoid strong reflections interfering with image quality. Grayscale and color images of the coating surface are acquired; grayscale images are used to identify surface unevenness and roughness variations, while color images are used to identify color differences and contamination. Illumination parameters during image acquisition are recorded, including light source type, light intensity, illumination angle, and exposure time. Process parameter information is collected to obtain coating thickness measurement data. The thickness of each area of ​​the coating is measured using an ultrasonic or electromagnetic thickness gauge, and the uniformity of the thickness distribution is recorded. Multiple measurement points are set in each measurement area, with a spacing of 10 to 20 millimeters to ensure coverage of thickness variation characteristics. The thickness value at each measurement point is recorded, with the unit of thickness being micrometers. In the automotive parts painting workshop, the coating thickness varies in different areas. The thickness is relatively uniform in the flat area of ​​the door, while the thickness fluctuates greatly in complex structural areas such as corners and reinforcing ribs.

[0029] Dark-field enhancement processing is performed on optical images of the coating surface to generate a scattering enhancement map. Coating surface images are extracted from an optical image database and subjected to dark-field enhancement to highlight surface defect features. Dark-field enhancement enhances the scattered light signal in the image, making minute defects, particles, and uneven areas more apparent. Background separation processing is performed on the image to identify background and coating surface areas. The light intensity distribution of the coating surface area is analyzed to identify the location of light intensity anomalies. High-frequency information in the image is extracted; this corresponds to detailed features such as surface edges, cracks, and particles. A high-pass filter is used to process the image, filtering out low-frequency background information while retaining high-frequency defect information. Image contrast is enhanced, widening the grayscale difference between defective and normal areas. In automotive engine hood coating inspection, minute scratches on the coating surface are difficult to observe under normal lighting. After dark-field enhancement, the scattered light at the scratch edges is magnified, making the scratch outline clearly discernible. Pseudo-color mapping is applied to the enhanced image, mapping different grayscale values ​​to different colors to form a scattering enhancement map. In the scattering enhancement map, high-scattering areas are displayed as warm tones, and low-scattering areas are displayed as cool tones.

[0030] In some embodiments, generating defect identification weights based on the scattering enhancement map includes: extracting defect features from the scattering enhancement map to generate a defect feature set; classifying the defect feature set according to severity to form a defect level identifier; calculating identification priority based on the defect level identifier; and generating defect identification weights based on the identification priority.

[0031] Defect feature sets are generated by extracting defect features from the scattering enhancement map. Abnormal regions are identified from the scattering enhancement map, corresponding to the locations of defects on the coating surface. The scattering enhancement map is divided into normal and abnormal regions using an image segmentation algorithm. For each abnormal region, geometric, morphological, and optical features are extracted. Geometric features include the defect's area, perimeter, major axis length, minor axis length, and aspect ratio. Morphological features include the defect's roundness, convexity, extension direction, and edge roughness. Optical features include the average scattering intensity, maximum scattering intensity, scattering intensity variance, and scattering intensity gradient of the defect region. In automotive bumper coating inspection, dust particle defects appear as small, high-scattering-intensity circular areas; sagging defects appear as strip-shaped, medium-scattering-intensity areas; and orange peel defects appear as large, wavy, low-to-medium scattering-intensity areas. Spatial location information for each defect is extracted, and its coordinates in the image are recorded. All extracted defect features are integrated to form a defect feature set, which records the defect's number, geometric features, morphological features, optical features, and spatial location.

[0032] The defect feature set is graded according to severity to form a defect level label. The characteristic parameters of each defect in the feature set are analyzed to assess the severity of the defect. Defect severity is comprehensively considered based on defect size, morphological complexity, scattering intensity, and location sensitivity. An area threshold is set: defects exceeding 10 square millimeters are classified as large-size defects, 1 to 10 square millimeters as medium-size defects, and less than 1 square millimeter as small-size defects. Morphological complexity is determined using roundness and convexity parameters: roundness less than 0.5 or convexity less than 0.7 is classified as complex-shaped defects. A scattering intensity threshold is set: average scattering intensity exceeding 1.5 times the threshold is classified as high-scattering defects. Location sensitivity refers to the degree to which the defect's location affects the product's appearance or performance. The severity of each defect is scored based on these four dimensions, using a weighted summation method: S = w1 × size score + w2 × morphological score + w3 × intensity score + w4 × location score, where w1, w2, w3, and w4 are weighting coefficients. Defects are categorized into three levels based on severity scores: Level A (Severe) defects (scores above 80), Level B (Medium) defects (scores between 50 and 80), and Level C (Minor) defects (scores below 50). Each defect is assigned a level label: Level A indicates high risk requiring immediate action; Level B indicates medium risk requiring monitoring; and Level C indicates low risk, acceptable, or subject to periodic inspection. The resulting defect level labels are recorded according to defect number and level category.

[0033] Identification priority is calculated based on defect level identifiers. The priority order of each defect in detection and processing is determined based on the defect level identifier. Level A severe defects have the highest identification priority, Level B medium defects have a medium identification priority, and Level C minor defects have a low identification priority. The identification priority value P is obtained as follows: P = 100 - 10 × (Level No. - 1) × (1 + D_density), where P is the identification priority, level no. 1 is level A, 2 is level B, and 3 is level C, and D_density is the defect density coefficient of the area surrounding the defect. The identification priority value ranges from 0 to 100, with a higher value indicating a higher priority. The defect density coefficient reflects whether other defects cluster around the defect; the priority of a single defect in a densely populated defect area will be appropriately increased. For example, a Level A crack defect located in a densely populated defect area near a car door weld, surrounded by multiple bubble and particle defects, has an identification priority of 95, higher than the priority of 90 for an isolated Level A defect.

[0034] Defect identification weights are generated based on identification priorities. These priorities are then normalized to distribute the weight values ​​between 0 and 1. The normalized weight W = P / P_max, where W is the defect identification weight, P is the identification priority of the defect, and P_max is the maximum identification priority among all defects. Defects with the highest identification priority have weights close to 1, while those with the lowest priority have weights close to 0. The distribution characteristics of the defect identification weights are analyzed to identify high-weight and low-weight defects. High-weight defects are the primary focus of detection, allocating more computational resources and detection time in the automatic detection algorithm. Low-weight defects are secondary targets and can be detected using rapid screening or manual sampling. During quality assessment, the presence of high-weight defects directly leads to product rejection, requiring rework or scrapping. The quantity and distribution of medium-weight defects affect the product's quality level; a small number of medium-weight defects may be deemed acceptable but downgraded. Low-weight defects, within acceptable limits, do not affect product acceptance. A correlation is established between defect identification weights and detection resource allocation. Defects with a weight greater than 0.8 are allocated 30% of the detection resources, defects with a weight between 0.5 and 0.8 are allocated 50% of the detection resources, and defects with a weight less than 0.5 are allocated 20% of the detection resources. This weight-driven resource allocation prioritizes ensuring the detection accuracy and efficiency of critical defects.

[0035] Step S120: Locate the key defect cluster area by defect identification weight, extract the shape similarity features of the key defect cluster area, and perform multi-dimensional similarity analysis on the shape similarity features to determine the potential defect risk area.

[0036] In some embodiments, locating the critical defect cluster area by means of the defect identification weight includes: decomposing the defect identification weight into primary defects and secondary defects; tracing the secondary defects to form a defect association path based on the primary defects; detecting the clustering intensity of the defect association path to obtain a clustering location set; and determining the critical defect cluster area based on the clustering location set.

[0037] The defect identification weights are decomposed into primary and secondary defects. Defects are classified based on the distribution characteristics of the weight values ​​from the defect identification weight data. A weight threshold is used to divide defects into primary and secondary categories. Primary defects are those with high weight values ​​and significant impact on product quality, while secondary defects are those with low weight values ​​and relatively minor impact on product quality. The primary / secondary threshold is set as the median of the weight values ​​or a fixed value of 0.6. Defects with weights exceeding the threshold are classified as primary defects, and those with weights below the threshold are classified as secondary defects. The quantity and distribution characteristics of primary and secondary defects are statistically analyzed. The number of primary defects reflects the overall quality level of the coating; a higher number of primary defects indicates a more serious quality problem. The composition of primary defect types is analyzed to identify which defect types dominate among primary defects. In automotive component coating inspection, primary defects typically include cracks, peeling, and foreign object embedding, which significantly affect the coating's protective performance and appearance quality. Secondary defects typically include small particles, minor scratches, and color differences, which mainly affect appearance details. Extract detailed feature information of major defects, including defect number, weight value, spatial location, defect type, and geometric dimensions. Extract basic feature information of minor defects, and record defect number, weight value, and spatial location.

[0038] Defect association paths are formed by tracing primary defects to secondary defects. For each primary defect in the primary defect dataset, the distribution of secondary defects in its surrounding area is analyzed. The influence range of each primary defect is obtained; the influence range is typically the area extending outward from the boundary of the primary defect by a certain distance. The radius of the influence range is determined based on the size and type of the primary defect. For example, a primary crack defect with a length of 15 mm has an influence range radius of 30 mm, while a particle defect with a diameter of 2 mm has an influence range radius of only 5 mm. Secondary defects within the influence range of each primary defect are searched, and the association between primary and secondary defects is established. Associations include spatial proximity, morphological similarity, and causal correlation. In the inspection of automotive engine hood coatings, eight bubble defects and five pinhole defects were found within a 30 mm radius of a primary crack defect. These secondary defects are distributed along the same stress direction as the primary crack, forming a clear association. The established associations are represented in the form of paths, with the starting point of the path being the primary defect and the ending point being the associated secondary defect. A defect association path network is formed, where nodes represent defects and edges represent the associations between defects.

[0039] The clustering intensity of defects is detected by analyzing defect association paths to obtain a set of cluster locations. Based on the defect association path network, the clustering intensity of defects in each region of the coating surface is detected. The coating surface is divided into regular grid cells, typically 50 mm × 50 mm in size. The number of defect association paths within each grid cell is counted; a higher number of paths indicates a higher clustering intensity. The clustering intensity index I_cluster = (N_paths × W_total weight) / A_cell, where I_cluster is the clustering intensity index, N_paths is the number of defect association paths within the grid cell, W_total weight is the sum of the weights of all defects within the cell, and A_cell is the area of ​​the grid cell. For example, in the inspection of the trunk lid coating of an SUV, a certain grid cell in the tailgate weld area contains 12 defect association paths, involving 3 major defects and 15 minor defects, with a total defect weight of 8.5 and a grid cell area of ​​0.0025 square meters. The obtained clustering intensity index is 40800, significantly higher than other areas. A clustering intensity threshold is set; grid cells with a clustering intensity index exceeding the threshold are marked as cluster locations. All marked cluster locations are integrated to form a cluster location set.

[0040] Critical defect clusters are identified based on the cluster location set. The locations with the highest cluster intensity index are extracted from the cluster location set; these locations represent the areas with the most severe coating quality problems. Spatial connectivity analysis is performed on the cluster locations to identify adjacent or nearby cluster locations and merge them into continuous cluster regions. The spatial connectivity criterion is that the distance between cluster locations is less than twice the grid cell size. The merged continuous cluster regions are then identified as critical defect clusters. Critical defect clusters represent the areas with the highest concentration of coating surface defects and the highest quality risk, requiring in-depth root cause analysis and targeted quality improvement. The boundary coordinates, area size, cluster intensity, and defect composition of each critical defect cluster are obtained. Boundary coordinates are used to pinpoint the specific location of the cluster, area size reflects the influence range of the cluster, cluster intensity reflects the severity of the cluster, and defect composition reflects the main defect types within the cluster. The causes of critical defect cluster formation are analyzed, and process factors leading to defect clustering are identified using process parameter datasets. Common causes include sagging due to improper coating ratios, particle aggregation due to unstable spraying pressure, and pinhole aggregation due to low curing temperature.

[0041] Extracting morphological similarity features from key defect clusters. For each identified key defect cluster, extract the morphological features of the defects within the cluster, including shape, texture, direction, and distribution pattern. Analyze the shape features of each defect within the cluster, and statistically analyze the proportion of defects with different shapes such as circles, ellipses, stripes, and irregular shapes. Extract the texture features of the defects within the cluster, and obtain texture direction, texture period, and texture contrast parameters through texture analysis algorithms. Analyze the directional features of the defects to identify whether the defects have a unified extension direction. Obtain the morphological similarity feature S_morphology = α × shape similarity + β × texture similarity + γ × direction consistency, where S_morphology is the morphological similarity feature value, and α, β, and γ are weighting coefficients. Shape similarity is obtained by comparing the standard deviation of the defect shape parameters; the smaller the standard deviation, the higher the similarity. Texture similarity is obtained through the correlation coefficient of texture feature vectors. Direction consistency is obtained through the angular variance of the defect principal axis direction.

[0042] Multi-dimensional similarity analysis is performed on morphological similarity features to identify potential defect risk areas. Morphological feature parameters such as shape similarity, texture similarity, and orientation consistency are extracted from the morphological similarity feature data. Morphological similarity features are integrated with scattering intensity features, size features, and process-related features to construct a multi-dimensional feature vector E=[S_morphology, I_scattering, D_size, C_process], where E is the multi-dimensional feature vector, S_morphology is the morphological similarity feature value, I_scattering is the scattering intensity feature, D_size is the size feature, and C_process is the process-related feature. Similarity calculation is performed between feature vectors to obtain the similarity coefficient R_similarity=1 / (1+||E1-E2||), where ||E1-E2|| is the Euclidean distance between the vectors. A similarity matrix is ​​constructed to identify highly similar regions with a similarity coefficient greater than 0.7, indicating that these regions have similar defect characteristics and causes. By comparing the multi-dimensional feature vectors of various regions on the coating surface with the feature vectors of known defect regions, areas with highly similar feature vectors, even if no obvious defects are currently detected, are identified as potential defect risk areas. For example, in the painting of car doors, a defect appears in the corner area of ​​the rear door. Its multi-dimensional feature vector shows that the coating gap is small and the spraying angle is limited. Through similarity analysis, it is found that the corner of the front door and the corner of the side skirt have similar feature vector combinations and are identified as potential defect risk areas.

[0043] Step S130: Obtain the gloss distribution characteristics and color difference spatial distribution of the potential defect risk area; identify the thickness fluctuation characteristics from the process parameter information to generate the thickness gradient parameter; perform a three-dimensional correlation mapping between the gloss distribution characteristics, the thickness gradient parameter and the color difference spatial distribution to generate a gloss-thickness-color difference coupling map; and determine the defect judgment boundary based on the gloss-thickness-color difference coupling map.

[0044] Obtain the gloss distribution characteristics and color difference spatial distribution of potential defect risk areas. For the identified potential defect risk areas, a gloss meter is used to measure the surface gloss value of each area. Multiple measurement points are set within each area, with a spacing of 10 to 20 mm. The gloss value of each measurement point is recorded in gloss units (GU). The spatial distribution pattern of gloss values ​​is analyzed to identify high-gloss, medium-gloss, and low-gloss areas. High-gloss areas typically have a gloss value exceeding 80 GU, medium-gloss areas are between 50 and 80 GU, and low-gloss areas are below 50 GU. Gloss distribution characteristics are extracted, including the area's average gloss, standard deviation of gloss, maximum gloss value, minimum gloss value, and gloss gradient distribution. A colorimeter is used to measure the surface color difference values ​​of each area. The color difference measurement adopts the CIELab color space system, recording three color parameters: L, a, and b. L represents lightness (brightness), a represents red-green hue (positive values ​​lean towards red, negative values ​​lean towards green), and b represents yellow-blue hue (positive values ​​lean towards yellow, negative values ​​lean towards blue). The total color difference value ΔE is calculated as √(ΔL² + Δa² + Δb²), where ΔE is the total color difference, ΔL is the lightness difference compared to the standard color chart, Δa is the red-green hue difference, and Δb is the yellow-blue hue difference. Measurement point arrays are set up within each area, and the color difference values ​​at each point are recorded. The color difference values ​​at each measurement point are organized and arranged according to their spatial location to form color difference spatial distribution data. This spatial distribution uses the spatial coordinates of the measurement points as an index and the total color difference value ΔE and the color difference components (L, a, b) as attribute values, comprehensively recording the spatial distribution of color differences in each potential defect risk area.

[0045] Thickness gradient parameters are generated by identifying thickness fluctuation characteristics from process parameter information. Coating thickness measurement data corresponding to each potential defect risk area are extracted from the process parameter dataset. Thickness values ​​at multiple measurement points within each area are analyzed to identify thickness fluctuation characteristics. Thickness fluctuations include random fluctuations and systematic fluctuations. Random fluctuations manifest as irregular changes in thickness values, while systematic fluctuations manifest as gradual or periodic changes in thickness along a certain direction. The average and standard deviation of the thickness are calculated; the standard deviation reflects the severity of the thickness fluctuation. The thickness difference between adjacent measurement points is obtained; the thickness difference reflects the steepness of the thickness change. The thickness gradient parameter G_thickness = ΔH / Δd is generated, where G_thickness is the thickness gradient parameter, ΔH is the thickness difference between adjacent measurement points, and Δd is the spatial distance between measurement points. For example, in the coating inspection of a car door outer panel, the coating thickness in the door reinforcing rib area shows significant fluctuations due to surface unevenness and limited spraying angle, with a thickness gradient parameter reaching 15 micrometers / cm, while the thickness gradient parameter in the door planar area is only 3 micrometers / cm.

[0046] In some embodiments, the step of generating a gloss-thickness-color difference coupled map by performing a three-dimensional correlation mapping of the gloss distribution features, the thickness gradient parameters, and the color difference spatial distribution includes: performing hierarchical analysis on the gloss distribution features to generate a gloss hierarchical structure; projecting the color difference spatial distribution onto the gloss hierarchical structure to form a color difference mapping layer; pairing and fusing the color difference mapping layer with the thickness gradient parameters to form a three-dimensional correlation structure; and constructing a gloss-thickness-color difference coupled map based on the three-dimensional correlation structure.

[0047] A hierarchical analysis was performed on the gloss distribution characteristics to generate a gloss hierarchy structure. Gloss values ​​for each potential defect risk area were extracted from the gloss distribution characteristic data and stratified according to gloss level. Based on the distribution range of gloss values, the gloss was divided into three levels: high-gloss, medium-gloss, and low-gloss. The high-gloss layer corresponds to areas with gloss values ​​exceeding 75 GU; these areas have smooth and flat coating surfaces with good specular reflection. The medium-gloss layer corresponds to areas with gloss values ​​between 45 and 75 GU; the coating surface quality is moderate. The low-gloss layer corresponds to areas with gloss values ​​below 45 GU; the coating surface is rough or has defects. The spatial distribution characteristics of each gloss level were analyzed to identify the area proportion and spatial continuity of each level. The boundary positions of each gloss level were extracted; the boundary is the area where adjacent gloss levels meet. The transition characteristics between gloss levels were analyzed to identify whether the transition is gradual or abrupt. A gloss hierarchy structure data was constructed, recording the gloss range, spatial location, area size, and boundary coordinates of each level.

[0048] For example, the step of projecting the color difference spatial distribution onto the gloss hierarchy to form a color difference mapping layer includes: calculating the gloss difference between the highlight layer and the low gloss layer of the gloss hierarchy to identify the maximum gradient position; locating the color difference inversion region of the color difference spatial distribution at the maximum gradient position; obtaining the color difference attenuation rate from the color difference inversion region to generate an attenuation utilization value; and projecting the attenuation utilization value to form a color difference mapping layer.

[0049] The gloss difference between the highlight and low-gloss layers in the gloss hierarchy is calculated to identify the location of the maximum gradient. Boundary information between the highlight and low-gloss layers is extracted from the gloss hierarchy data. A significant difference in gloss exists between the highlight and low-gloss layers, forming the gloss difference. The gloss difference value between the highlight and low-gloss layer boundary points is obtained as D_gloss = G_highlight - G_lowlight, where G_highlight is the gloss level at the highlight layer boundary point, and G_lowlight is the gloss level at the low-gloss layer boundary point. The gloss difference is scanned along the boundary between the highlight and low-gloss layers, and the gloss difference value is obtained at different locations on the boundary. The gloss gradient is obtained as G_gradient = D_gloss / L_distance, where G_gradient is the gloss gradient, and L_distance is the spatial distance from the highlight layer boundary point to the low-gloss layer boundary point. The locations with the maximum gloss gradient are identified; these locations are key areas where gloss changes are most drastic and quality differences are most pronounced. The maximum gradient location typically corresponds to the edge of a coating defect, abrupt changes in process parameters, or a boundary point of material properties. For example, in the inspection of car bumper coatings, the gloss gradient in the transition area between the bumper and the body color is extremely large, with the black coating of the bumper showing low gloss and the body color coating showing high gloss.

[0050] Locate the color difference reversal region in the spatial distribution of color difference at the location of maximum gradient. Based on the identified location of maximum gradient, analyze the spatial distribution characteristics of color difference at this location and its surrounding area. Due to the drastic change in gloss, the location of maximum gradient is often accompanied by abnormal color difference performance. Extract the color difference values ​​in the area surrounding the location of maximum gradient and plot the color difference variation curve along the spatial distribution. The color difference variation curve shows the trend of color difference change with location. Analyze the color difference variation curve to identify the color difference reversal region. The color difference reversal region is the area where the direction of color difference value change reverses, that is, the inflection point where the color difference first increases and then decreases or first decreases and then increases. The color difference reversal phenomenon reflects the complexity and discontinuity of color distribution on the coating surface. In the color difference reversal region, the optical properties and color of the coating exhibit non-monotonic changes, which may be related to abrupt changes in coating thickness, changes in surface roughness, or differences in material composition. Locate the specific location and spatial range of the color difference reversal region. The color difference reversal region is usually located near the location of maximum gradient, but not necessarily completely coincident. Record the boundary coordinates, color difference variation amplitude, and reversal point location of the color difference reversal region. The color difference reversal point is the specific location where the trend of color difference change reverses. Analyze the changes in color difference components within the color difference reversal region to identify whether it is a reversal of lightness (L), red-green hue (a), or yellow-blue hue (b). The reversal of different color difference components reflects different physical causes.

[0051] The color difference attenuation rate is obtained from the color difference inversion region to generate an attenuation utilization value. For the identified color difference inversion region, the attenuation characteristics of the color difference value along spatial location are analyzed. Color difference attenuation refers to the phenomenon that the color difference value gradually decreases from the inversion point outwards. The color difference attenuation rate reflects the speed of color difference change. Color difference values ​​at multiple points on the boundary of the color difference inversion region are extracted and sorted according to their distance from the inversion point. The color difference change and spatial distance between adjacent points are obtained to generate the color difference attenuation rate R_attenuation = ΔE_color difference / Δx_distance, where R_attenuation is the color difference attenuation rate, ΔE_color difference is the color difference change at adjacent points, and Δx_distance is the spatial distance between adjacent points. A large color difference attenuation rate value indicates rapid color difference attenuation, while a small value indicates slow color difference attenuation. The spatial distribution pattern of the color difference attenuation rate is analyzed to identify rapid attenuation regions and slow attenuation regions. In rapid attenuation regions, the color difference decreases rapidly over short distances, while in slow attenuation regions, the color difference decreases gradually over longer distances. The color difference attenuation rate is related to the coating's hiding power, thickness variation, and light scattering characteristics. Coatings with strong hiding power exhibit rapid color difference decay, while coatings with weak hiding power experience slow color difference decay. The decay utilization value U_decay = R_decay × L_influence, where U_decay is the decay utilization value, R_decay is the color difference decay rate, and L_influence is the length of the influence range of the color difference reversal area. The decay utilization value comprehensively reflects the speed and range of color difference decay; a larger value indicates a stronger color difference decay effect. Areas with high decay utilization values ​​have strong color difference recovery capabilities, while areas with low decay utilization values ​​have a large range of sustained color difference influence.

[0052] A color difference mapping layer is formed by projecting the attenuation utilization values. Based on the attenuation utilization values ​​of each color difference inversion region, the spatial distribution data of color difference is mapped onto the gloss hierarchy structure to form a color difference mapping layer. The color difference mapping layer superimposes color difference information on the spatial framework of the gloss hierarchy structure, realizing the spatial correspondence and correlation between gloss and color difference. The attenuation utilization values ​​are used as mapping weights to weight the spatial distribution data of color difference. Color difference data in areas with high attenuation utilization values ​​have a larger weight in the mapping, while color difference data in areas with low attenuation utilization values ​​have a smaller weight. Through weighted mapping, key areas with significant color difference changes and strong attenuation effects are highlighted. For each level in the gloss hierarchy structure, the color difference distribution characteristics within that level are calculated. The color difference in the highlight layer is usually small and uniform, the color difference in the mid-gloss layer is moderate, and the color difference in the low-gloss layer is large and uneven. At the boundaries of the gloss levels, the color difference often exhibits abrupt changes or inversion characteristics. The color difference values ​​are displayed in partitions according to the gloss levels to form a layered color difference distribution map. The color difference mapping layer uses gloss levels as its framework, color difference values ​​as fillers, and attenuation utilization values ​​as weights to form a three-dimensional color difference distribution pattern. It analyzes abnormal color difference regions within the color difference mapping layer to identify high color difference areas and regions with abrupt color difference changes.

[0053] A three-dimensional correlation structure is formed by pairing and fusing the color difference mapping layer and thickness gradient parameters. Spatial matching is performed between the color difference mapping layer data and the thickness gradient parameter data. For each potential defect risk area, the color difference mapping layer data and thickness gradient parameter data for that area are extracted. The measurement points of the two types of data are spatially aligned to ensure that color difference and thickness gradient at the same spatial location can be correlated. The correlation between color difference and thickness gradient is analyzed to identify whether a location with a large color difference corresponds to a location with a large thickness gradient. A three-dimensional correlation structure is constructed, with gloss, color difference, and thickness gradient as its three dimensions. Each data point in the three-dimensional correlation structure corresponds to a combination of gloss, color difference, and thickness gradient at a certain location on the coating surface. The clustering patterns and outliers of the data are identified through the distribution of data points in three-dimensional space. In the inspection of automotive wheel hub coatings, data points in normal areas cluster in regions with high gloss, low color difference, and low thickness gradient, while data points in defective areas deviate from the cluster center, exhibiting a combination of low gloss, high color difference, or high thickness gradient.

[0054] A gloss-thickness-color difference coupling map is constructed based on a three-dimensional correlation structure. The coupling map displays the relationship between gloss, thickness gradient, and color difference in a three-dimensional coordinate system. The X-axis represents gloss, the Y-axis represents thickness gradient, and the Z-axis represents color difference. Measurement data points for each potential defect risk area are plotted in the three-dimensional coordinate system; the position of each data point is determined by its gloss, thickness gradient, and color difference values. The spatial distribution of the data points visually demonstrates the coupling relationship between the three parameters. Data clustering regions in the coupling map are identified, representing the main distribution pattern of coating quality status. Data from normal quality areas clusters in the high gloss, low thickness gradient, and low color difference quadrants. Data from defective quality areas is scattered across various combinations of low gloss, high thickness gradient, and high color difference. Outliers in the coupling map are analyzed, corresponding to abnormal locations on the coating surface. The three-dimensional coordinates of outliers reveal the specific quality problems at that location. For example, the coordinates of an outlier are (gloss 30 GU, thickness gradient 18 μm / cm, color difference ΔE=4.5), indicating that there is a composite defect of low gloss, drastic thickness fluctuation and severe color difference at that location.

[0055] Defect judgment boundaries are determined based on the gloss-thickness-color difference coupled spectrum. Data distribution characteristics of normal and defective quality regions are extracted from the gloss-thickness-color difference coupled spectrum. The gloss range, thickness gradient range, and color difference range of the normal quality region are statistically analyzed, and the mean and standard deviation of each parameter are obtained. The parameter range of the normal quality region defines the benchmark standard for coating acceptance. The parameter distribution of the defective quality region is analyzed to identify the key parameter thresholds leading to defect judgment. A lower limit threshold for gloss is set, typically the average gloss of the normal region minus twice the standard deviation. Upper limits thresholds for thickness gradient and color difference are set, with the color difference threshold typically set to ΔE = 2.0 or 3.0. Considering gloss, thickness gradient, and color difference comprehensively, defect judgment boundaries are established. The defect judgment boundaries divide the three-dimensional space of the coupled spectrum into acceptable and unacceptable regions. The judgment boundaries adopt a three-dimensional boundary surface form, and the boundary surface equation integrates the threshold conditions of gloss, thickness gradient, and color difference. Different judgment boundaries are set for different quality requirements. The determined defect judgment limits are clearly recorded according to parameter thresholds and boundary equations, serving as a quantitative standard for coating quality inspection and judgment.

[0056] Step S140: Combine the defect judgment boundary and the scattering enhancement spectrum to determine the detection sensitivity, perform influence range analysis on the key defect cluster area to generate influence radius parameters, and construct a zonal detection strategy based on the influence radius parameters and the detection sensitivity.

[0057] Specifically, the detection sensitivity is determined by combining the defect judgment boundary with the scattering enhancement spectrum. The defect judgment boundary and the scattering enhancement spectrum are overlaid and analyzed to identify the correspondence between changes in scattering intensity and the defect judgment boundary. The gradient of scattering intensity near the judgment boundary is analyzed; areas with larger gradients have higher sensitivity to defect identification, while areas with smaller gradients have lower sensitivity. The detection sensitivity is calculated as S_sensitivity = ΔI_scattering / ΔP_parameter, where S_sensitivity is the detection sensitivity, ΔI_scattering is the change in scattering intensity, and ΔP_parameter is the change in defect judgment parameters (including changes in gloss, thickness gradient, or color difference). Detection sensitivity reflects the responsiveness of scattering intensity to changes in defect parameters. Areas with high sensitivity can accurately identify minute defects through the scattering spectrum, while areas with low sensitivity have weaker ability to identify minute defects. Detection sensitivity varies for different types of defects. Surface cracks and flaking have high detection sensitivity due to significant scattering effects. Fine particles and color differences have relatively low detection sensitivity due to weak scattering effects. In automotive component coating inspection practice, conventional inspection parameters are sufficient to effectively identify defects in high-sensitivity areas, while low-sensitivity areas require increased image resolution or adjusted lighting conditions to enhance defect visibility. The detection sensitivity distribution covers the entire coating surface. Analyzing the detection sensitivity distribution in potential defect risk areas and other coating regions helps identify high-sensitivity and low-sensitivity areas. High-sensitivity areas can be inspected using rapid methods, while low-sensitivity areas require more precise inspection methods.

[0058] In some embodiments, performing influence range analysis on the critical defect cluster area to generate influence radius parameters includes: performing diffusion tracking on the critical defect cluster area to obtain a diffusion trajectory curve; identifying accelerated diffusion segments from the diffusion trajectory curve to generate a diffusion enhancement factor; using the diffusion enhancement factor in conjunction with the diffusion trajectory curve to perform dynamic boundary processing to obtain a boundary parameter set; and determining the influence range based on the boundary parameter set to generate influence radius parameters.

[0059] Diffusion tracing was performed on key defect clusters to obtain diffusion trajectory curves. For each key defect cluster, a layer-by-layer scan was performed from the cluster boundary outwards to track the spatial attenuation changes of defect characteristic parameters. Defect characteristic parameters include defect density, scattering intensity, gloss deviation, and color difference deviation. Multiple concentric circular or rectangular rings were set outside the cluster boundary, with a ring spacing of 5 to 10 mm. The defect density, average scattering intensity, and quality parameter deviation within each ring were statistically analyzed. Diffusion trajectory curves were plotted, with the horizontal axis representing the radial distance from the cluster boundary and the vertical axis representing the numerical values ​​of the defect characteristic parameters. The morphological characteristics of the diffusion trajectory curves were analyzed to identify rapid attenuation zones and slow attenuation zones. In the inspection of automotive front fender coatings, the defect density in the critical defect cluster area outside the boundary of a wheel arch edge welding area decreased rapidly from 12 defects / 100 cm² to 3 defects / 100 cm² within 0 to 20 mm, exhibiting a rapid attenuation characteristic, while the defect density in the range of 20 to 50 mm decreased slowly from 3 defects / 100 cm² to 0.5 defects / 100 cm², exhibiting a slow attenuation characteristic. Extract key feature points on the diffusion trajectory curve, including the starting point, inflection point, and ending point.

[0060] For example, the step of identifying accelerated diffusion segments from the diffusion trajectory curve to generate diffusion enhancement factors includes: performing curvature analysis on the diffusion trajectory curve to determine curvature jump points; locating gradient change rates at the curvature jump points to generate change rate benchmarks; comparing the change rate benchmarks with the diffusion trajectory curve to divide it into accelerated diffusion segments; and extracting diffusion enhancement factors based on the accelerated diffusion segments.

[0061] Curvature analysis is performed on the diffusion trajectory curve to determine curvature abrupt change points. Curvature analysis reflects the degree of curvature bending. Curvature values ​​κ are obtained for each point on the curve; curvature is determined by the relationship between the first and second derivatives of the curve. Larger curvature values ​​indicate more pronounced curvature bending, while smaller values ​​indicate a smoother curve. A curvature distribution curve is plotted to identify abrupt changes in curvature. These abrupt changes are key points where the curve shape undergoes a significant change, corresponding to inflection points or accelerations in the diffusion process. Curvature abrupt change points are determined when the curvature change between adjacent points exceeds a threshold. The magnitude of the curvature abrupt change is determined by the ratio of the curvature difference between adjacent points to the average curvature; a ratio exceeding 0.5 is considered a curvature abrupt change. Curvature abrupt change points typically correspond to the transition from slow to rapid decay or vice versa in the diffusion trajectory curve. The physical significance of curvature abrupt change points is analyzed; they may be related to spatial differences in coating material properties, abrupt changes in process conditions, or discontinuities in stress distribution. For example, in the inspection of the coating of the inner panel of a car door, the curvature jump point near the door reinforcing rib corresponds to the transition position of the coating thickness from the thin area to the thick area.

[0062] Gradient change rate is located at curvature abrupt change points to generate a rate of change benchmark. For each identified curvature abrupt change point, the gradient change characteristics of the diffusion trajectory curve near that point are analyzed. The gradient is the slope of the curve at a point, reflecting the rate of change of defect characteristic parameters with distance. Gradient values ​​are obtained for the segments before and after the curvature abrupt change point; the gradient of the segment before the abrupt change point reflects the diffusion rate before the abrupt change, and the gradient of the segment after the abrupt change point reflects the diffusion rate after the abrupt change point. The gradient change rate γ = (G_after - G_before) / G_before is generated, where γ is the gradient change rate, G_after is the average gradient of the segment after the abrupt change point, and G_before is the average gradient of the segment before the abrupt change point. A positive gradient change rate indicates that the diffusion rate accelerates at the abrupt change point, while a negative value indicates that the diffusion rate slows down. The larger the absolute value of the gradient change rate, the more significant the change in diffusion rate. The gradient change rate is used as a rate of change benchmark to identify acceleration or deceleration phenomena in the diffusion process. Curvature abrupt change points with a gradient change rate greater than 0.3 are marked as acceleration abrupt change points, indicating that diffusion is significantly accelerated at that location. Curvature abrupt change points with a gradient change rate less than -0.3 are marked as deceleration abrupt change points, indicating that diffusion significantly slows down at these locations. Analyze the distribution pattern of the rate of change baseline to identify the number of acceleration and deceleration abrupt change points on the diffusion trajectory curve.

[0063] Accelerated diffusion segments are formed by comparing the rate of change benchmark with the diffusion trajectory curve. Based on the rate of change benchmark data, different diffusion segments are divided on the diffusion trajectory curve. These segments are categorized into accelerated diffusion segments, uniform diffusion segments, and decelerating diffusion segments based on gradient change characteristics. Locations marked as acceleration jump points are extracted; these locations are the starting points or key feature points of accelerated diffusion segments. The boundaries of accelerated diffusion segments are determined by extending forward and backward from each acceleration jump point. The boundary determination criterion for accelerated diffusion segments is that the gradient value reverts to the benchmark gradient range or encounters the next jump point. Accelerated diffusion segments are sections on the diffusion trajectory curve where the slope continuously increases or remains high. The diffusion trajectory curve is compared point-by-point with the rate of change benchmark to identify all segments where the curve gradient exceeds the benchmark threshold. These segments are integrated to form a complete set of accelerated diffusion segments. The length and gradient characteristics of accelerated diffusion segments are analyzed; longer accelerated diffusion segments indicate that the defect impact propagates rapidly over a larger area, while accelerated diffusion segments with larger gradients indicate that the defect impact propagates extremely quickly. In the inspection of automotive sunroof frame coatings, the accelerated diffusion section often corresponds to areas with weak coating material performance or areas with continuous distribution of process defects such as drainage grooves. These areas require key inspection and quality improvement.

[0064] The diffusion enhancement factor is extracted from the accelerated diffusion segment. The diffusion enhancement factor comprehensively reflects the propagation intensity and influence range of the defect impact within the accelerated diffusion segment. The total length L_acceleration of the accelerated diffusion segment is calculated, reflecting the spatial range covered by the accelerated diffusion phenomenon. The average gradient G_acceleration within the accelerated diffusion segment is obtained, reflecting the rate level of accelerated diffusion. The diffusion enhancement factor F_enhancement = (L_acceleration / L_total) × (G_acceleration / G_average) is generated, where F_enhancement is the diffusion enhancement factor, L_acceleration is the total length of the accelerated diffusion segment, L_total is the total length of the diffusion trajectory curve, G_acceleration is the average gradient of the accelerated diffusion segment, and G_average is the average gradient of the entire curve. The larger the diffusion enhancement factor value, the more significant the accelerated diffusion phenomenon and the stronger the propagation ability of the defect impact. Clusters with a diffusion enhancement factor greater than 1.5 have strong diffusion characteristics, requiring a larger influence range and stricter monitoring measures. Clusters with a diffusion enhancement factor less than 0.8 have limited diffusion ability and a relatively concentrated influence range. The correlation between the diffusion enhancement factor and the characteristics of key defect clusters is analyzed to identify which cluster characteristics lead to a high diffusion enhancement factor. Typically, areas with high defect density, complex defect types, or severe process abnormalities have a larger diffusion enhancement factor.

[0065] A boundary parameter set is obtained by combining a diffusion enhancement factor with the diffusion trajectory curve for dynamic boundary processing. The diffusion enhancement factor is used as a weighting coefficient to dynamically adjust the diffusion trajectory curve. Dynamic boundary processing adjusts the criteria for determining the boundary of the influence range based on the magnitude of the diffusion enhancement factor. For clusters with large diffusion enhancement factors, a stricter boundary judgment criterion is adopted, defining the boundary as the position where the defect feature parameters on the diffusion trajectory curve decay to a lower level. For clusters with small diffusion enhancement factors, a relatively lenient boundary judgment criterion is adopted, defining the boundary as the position where the defect feature parameters decay to the point of near disappearance. The boundary judgment threshold T_boundary = T_benchmark × (1 - α × F_enhancement) is set, where T_boundary is the dynamically adjusted boundary threshold, T_benchmark is the benchmark boundary threshold, α is an adjustment coefficient (usually between 0.2 and 0.4), and F_enhancement is the diffusion enhancement factor. The larger the diffusion enhancement factor, the lower the boundary threshold, the outerier the boundary position, and the larger the influence range. The position where the defect feature parameters decrease below the boundary threshold on the diffusion trajectory curve is identified, and this position is determined as the dynamic boundary position. Extract the spatial coordinates of the dynamic boundary location, the radial distance from the boundary of the cluster area, and the numerical values ​​of the defect characteristic parameters at that location. Integrate the dynamic boundary location information of each key defect cluster area to form a boundary parameter set. The boundary parameter set contains key information such as the boundary coordinates, boundary radius, defect density at the boundary, and quality parameters at the boundary for each cluster area.

[0066] The influence radius parameter is generated based on the boundary parameter set to determine the influence range. Boundary radius data for each critical defect cluster area is extracted from the boundary parameter set. The boundary radius is the radial distance from the center of the cluster area to the dynamic boundary position. The boundary radius values ​​of each cluster area in different radial directions are statistically analyzed to obtain the average boundary radius R_average and the maximum boundary radius R_maximum. The influence radius parameter R_influence = w_average × R_average + w_maximum × R_maximum, where R_influence is the influence radius parameter, w_average and w_maximum are weighting coefficients, and w_average + w_maximum = 1. Typically, w_average is set to 0.7 and w_maximum to 0.3. The influence radius parameter determines the spatial range within which critical defect cluster areas require focused monitoring and detection. In the inspection of car rearview mirror coatings, the average boundary radius of a certain critical defect cluster area is 45 mm, and the maximum is 68 mm. The generated influence radius parameter is 51 mm. A circular area with a radius of 51 mm, centered on the center of the cluster area, is determined as the influence range, and intensive inspection is required within this range.

[0067] A zonal detection strategy is constructed based on the influence radius parameter and detection sensitivity. The coating surface is divided into different detection zones by combining the influence radius parameter and detection sensitivity data. Based on the influence radius parameter of the critical defect cluster area, the coating surface is divided into three regions: core zone, influence zone, and peripheral zone. The core zone corresponds to the critical defect cluster area itself, the influence zone corresponds to the area within the influence radius, and the peripheral zone corresponds to the area outside the influence radius. The detection parameters and detection frequency for each region are determined by combining the detection sensitivity characteristics of each region. For the core zone, due to the high and severe defect density, a high-precision detection method is adopted, with the detection resolution set to the highest level, achieving 100% detection coverage. For the influence zone, a medium-precision detection method is adopted, with a detection coverage of 80% to 90%. For the peripheral zone, a standard-precision detection method is adopted, with a detection coverage of 60% to 70%. In areas with low detection sensitivity, detection parameters are improved to compensate for insufficient sensitivity. Improvement measures include increasing light intensity, adjusting the light angle, increasing image resolution, or increasing the number of detections. A zonal detection strategy matrix is ​​constructed, where the rows represent the region type, the columns represent the detection parameter type, and the matrix elements are the specific settings of each parameter for each region.

[0068] Step S150: For the gloss-thickness-color difference coupled spectrum, abnormal coupling regions are identified and detected. Imbalance compensation parameters are generated by similarity matching based on the abnormal coupling regions and morphological similarity features. Based on the imbalance compensation parameters and the partition detection strategy, a quality level judgment result is generated.

[0069] Specifically, anomaly identification is performed on the gloss-thickness-color difference coupling spectrum to detect abnormal coupling regions. The distribution pattern of data points in the coupling spectrum is analyzed to identify normal data clusters and abnormal data deviation areas. Data points in normal data clusters are concentrated in specific areas of three-dimensional space, representing the normal combination of coating quality states. Data points in abnormal data deviation areas are far from normal clusters, indicating abnormal parameter coupling relationships in the coating. A clustering analysis algorithm is used to identify the center and boundary range of normal clusters. The center of the cluster represents the standard state of coating quality, and the boundary range represents the acceptable range of parameter fluctuations. The Euclidean distance from each measurement point to the center of the normal cluster is calculated as D_deviation = √[(G-G0)²+(H-H0)²+(C-C0)²], where D_deviation is the deviation distance, G, H, and C are the gloss, thickness gradient, and color difference at that point, respectively, and G0, H0, and C0 are the corresponding parameter values ​​at the center of the normal cluster. Measurement points with large deviation distances have abnormal parameter combinations, while measurement points with small deviation distances have normal parameter combinations. An anomaly detection threshold is set, and measurement points deviating from the normal cluster area by more than twice its radius are marked as anomalies. All anomalies are then integrated to form an anomalous coupling region. An anomalous coupling region is an area where the coupling relationship of coating surface parameters is unbalanced. These areas may exhibit quality problems such as mismatch between gloss and thickness, or inconsistency between color difference and gloss. In the inspection of automotive center console coatings, some areas show normal gloss but abnormally large thickness gradients and excessive color differences. This three-parameter imbalance is identified as an anomalous coupling region, requiring in-depth analysis of its causes.

[0070] In some embodiments, the step of generating imbalance compensation parameters by performing similarity matching between the abnormal coupling region and the morphological similarity features includes: constructing a morphological deviation mapping map based on the abnormal coupling region; locating key similarity points on the morphological deviation mapping map and the morphological similarity features to obtain matching benchmark values; extending the matching benchmark values ​​according to the defect level to form a compensation adjustment domain; and correcting the boundary of the compensation adjustment domain to generate imbalance compensation parameters.

[0071] A morphology deviation mapping map is constructed based on anomalous coupling regions. For each identified anomalous coupling region, the morphology deviation characteristics of the coating surface within that region are analyzed. Morphology deviation refers to the degree of difference between the actual morphology and the standard morphology. Surface morphology data, including surface roughness, contour height, and microtexture, are extracted from the anomalous coupling regions. By comparing with a standard morphology template, the morphology deviation values ​​at each location are obtained. Locations with large morphology deviation values ​​show significant surface morphology anomalies, while locations with small morphology deviation values ​​have surface morphologies close to the standard state. A morphology deviation mapping map is drawn, displaying the distribution of morphology deviations at each location within the anomalous coupling region in a two-dimensional plane. Color gradients are used to represent the magnitude of morphology deviations, with warm colors representing large deviations and cool colors representing small deviations. The spatial distribution pattern of the morphology deviation mapping map is analyzed to identify morphology deviation clusters and transition regions. Morphology deviation clusters represent the core locations where surface morphology anomalies are most severe, while transition regions are the edges where the morphology gradually changes from anomalous to normal.

[0072] Similarity key points are located using morphology deviation mapping and morphology similarity features to obtain matching benchmark values. Morphology similarity features reflect the consistency of defect morphology within defect clusters, while morphology deviation mapping reflects the degree of surface morphology anomaly within abnormal coupling regions. In the overlapping areas of the two types of data, locations with large morphology deviations and high morphology similarity are identified as similarity key points. Similarity key points are typical locations with significant morphology anomalies and highly consistent defect morphologies; parameter imbalances at these locations have a clear causal relationship with defect morphology. Morphology deviation and morphology similarity values ​​are extracted from each similarity key point. The matching benchmark value B_matching = (D_morphology deviation + S_morphology similarity) / 2 is obtained, where B_matching is the matching benchmark value, D_morphology deviation is the standardized value of the key point's morphology deviation, and S_morphology similarity is the standardized value of the key point's morphology similarity. The matching benchmark value comprehensively reflects the degree of morphology anomaly and the degree of morphology consistency. Key points with high matching benchmark values ​​indicate significant morphology anomalies and prominent defect morphology features at that location, representing typical quality problems. Locations with low matching baseline values ​​show inconspicuous morphological features or mild anomalies. In the inspection of automotive door handle coatings, the similarity key point matching baseline value is high in the area surrounding the handle mounting holes. This area has large morphological deviations, and the defect morphologies of multiple handle hole areas are highly similar, indicating the existence of systematic process defects.

[0073] The matching benchmark values ​​are extended according to defect levels to form a compensation adjustment domain. Based on the matching benchmark values ​​of each similarity key point, and combined with defect level identification data, the matching benchmark values ​​are spatially extended. From the defect level identification data output in step S110, the level information of each defect is extracted, including Grade A severe defects, Grade B moderate defects, and Grade C minor defects. Different extension radii and extension intensities are set for different defect levels. A larger extension radius is set for the similarity key points corresponding to Grade A severe defects, typically 30 to 50 mm, indicating a wide impact range for severe defects. The extension radius for Grade B moderate defects is set to 15 to 30 mm. The extension radius for Grade C minor defects is set to 5 to 15 mm. With each similarity key point as the center, the domain is expanded outward according to the corresponding extension radius to form a compensation adjustment domain. The compensation adjustment domain is the spatial range within which imbalance compensation adjustments are performed. Within the compensation adjustment domain, the matching benchmark value decreases with distance from the key point. The compensation intensity S_compensation = B_matching × e^(-r / R) is obtained at each location within the adjustment domain, where S_compensation is the compensation intensity at that location, B_matching is the matching baseline value of the keypoint, r is the distance from that location to the keypoint, and R is the extension radius. The compensation intensity is maximum at the keypoint, gradually decreases with increasing distance, and approaches zero at the boundary of the extension radius. The compensation adjustment domains of each similarity keypoint are superimposed to form a complete compensation adjustment domain distribution.

[0074] Imbalance compensation parameters are generated by correcting the boundary of the compensation adjustment domain. For the formed compensation adjustment domain, boundary correction is performed to optimize the compensation range and effect. The boundary positions of the compensation adjustment domain are analyzed to identify whether the compensation intensity at the boundary reaches a preset boundary threshold. The boundary threshold is typically set to 10% to 20% of the maximum compensation intensity. For boundary positions where the compensation intensity does not reach the boundary threshold, the boundary is contracted inward to reduce the range of the compensation adjustment domain. For boundary positions where the compensation intensity significantly exceeds the boundary threshold, the boundary is maintained or appropriately extended to ensure that the compensation range covers all areas requiring adjustment. Through boundary correction, the problem of over-extension or under-extension of the compensation adjustment domain is eliminated. The corrected compensation adjustment domain boundary is more accurate, and the compensation range is more reasonable. The compensation intensity values ​​at each position within the corrected compensation adjustment domain are extracted; these values ​​constitute the spatial distribution of the imbalance compensation parameters. An imbalance compensation parameter dataset is generated, containing the spatial coordinates of each compensation position, the compensation intensity, the corresponding defect level, and the compensation direction. The compensation direction indicates whether the weight of a certain quality parameter should be increased or decreased. Imbalance compensation parameters provide a targeted adjustment basis for quality judgment, and can dynamically adjust the quality evaluation criteria based on the defect morphology characteristics and parameter coupling imbalance.

[0075] In some embodiments, generating a quality level determination result based on the imbalance compensation parameter and the partition detection strategy includes: identifying detection response differences for the partition detection strategy; establishing a constraint response distribution field based on the imbalance compensation parameter and the detection response differences; using the constraint response distribution field to evaluate the imbalance transformation characteristics and generate an imbalance utilization factor; and dynamically adjusting the imbalance utilization factor to generate a quality level determination result.

[0076] This study identifies differences in detection response based on a zoned detection strategy. Detection parameter settings and standards for the core, affected, and peripheral zones are extracted from the zoned detection strategy data. Detection sensitivity, accuracy, and judgment thresholds differ across detection zones. The actual number of defects detected, defect types, and parameter deviations in each zone are analyzed. The actual detection results are compared with the expected detection responses for each zone to identify detection response differences. Detection response difference refers to the degree of deviation between the actual detection results and the expected detection level. The detection response difference R_difference = |N_actual - N_expected| / N_expected, where R_difference is the detection response difference, N_actual is the actual number of defects or parameter deviations detected, and N_expected is the expected number of detections based on the zone characteristics. Zones with large detection response differences indicate a significant deviation between actual and expected quality, possibly due to process fluctuations, material changes, or improper detection parameter settings. Zones with small detection response differences indicate that the actual quality meets expectations. The type of detection response difference is analyzed to identify whether it is a positive or negative difference. A positive difference indicates that more problems were detected than expected, while a negative difference indicates that fewer problems were detected than expected. In the inspection of automotive trunk lid coatings, the detection response difference in the corner-affected area of ​​a certain batch of products was significantly higher than expected, with the actual number of defects being 1.8 times that of the expectation, indicating that the quality control in this area was fluctuating.

[0077] A constraint response distribution field is established based on imbalance compensation parameters and detection response differences. Spatial coordinates and compensation intensity values ​​for each compensation location are extracted from the imbalance compensation parameter dataset, and response difference values ​​and region boundary coordinates for each detection area are extracted from the detection response difference data. The measurement points of the two types of data are spatially aligned, and the coating surface is divided into uniform grid cells. For each grid cell, the imbalance compensation parameters and detection response difference values ​​within or covering that cell are extracted. For missing data, a distance-weighted interpolation method is used to estimate the values, ensuring that each grid cell simultaneously possesses both imbalance compensation parameters and detection response difference attributes. The relative importance of the two types of data is analyzed to determine weighting coefficients. Based on the focus of coating quality control, w_compensation and w_difference are set, with the weighting coefficients satisfying the normalization constraint condition w_compensation + w_difference = 1. The imbalance compensation parameters and detection response differences are standardized, normalizing the values ​​to the range of 0 to 1, ensuring that the two types of parameters are weighted under the same dimensions. A constraint response distribution field is constructed, with the spatial coordinates of the grid cells as indices and the combined effect of the imbalance compensation parameters and detection response differences as the field value. For each grid cell, the constraint response field value is obtained as F_constraint = w_compensation × C_compensation + w_difference × R_difference, where F_constraint is the constraint response field value, C_compensation is the standardized value of the imbalance compensation parameter for that cell, and R_difference is the standardized value of the detection response difference for that cell. The constraint response field value comprehensively reflects the overall requirements for quality judgment adjustment at that location.

[0078] Imbalance transformation characteristics are assessed using the constraint response distribution field to generate an imbalance utilization factor. Imbalance transformation characteristics reflect the actual impact of parameter imbalance on the final product quality grade. High field value regions in the constraint response distribution field are analyzed to identify whether parameter imbalances in these regions can be compensated through process adjustments or optimization of usage conditions. Compensable imbalances have a relatively small long-term impact on product quality, while uncompensable imbalances pose a continuous threat to product quality. The usability of imbalances in each region is assessed; highly usable imbalances can be transformed into an acceptable quality state through appropriate technical means. An imbalance utilization factor, U_imbalance = F_constraint × (1 - K_compensable), is generated, where U_imbalance is the imbalance utilization factor, F_constraint is the constraint response field value, and K_compensable is the compensability coefficient of the imbalance. The compensability coefficient is determined by analyzing the causes, development trends, and feasibility of compensation techniques for the imbalance. The imbalance utilization factor reflects the actual impact weight of the imbalance on quality judgment after considering the possibility of compensation. Regions with high imbalance utilization factors require more stringent evaluation in quality grade determination, while regions with low imbalance utilization factors can have their evaluation criteria appropriately relaxed.

[0079] The quality grade determination results are generated through dynamic grading adjustments based on the imbalance utilization factor. The quality grade of each testing area is dynamically adjusted based on the imbalance utilization factor data. Areas with high imbalance utilization factors have their quality grade lowered, while areas with low imbalance utilization factors maintain or appropriately increase their grade. Dynamic adjustment rules are set: areas with an imbalance utilization factor exceeding 1.5 have their quality grade lowered by one level; areas between 1.0 and 1.5 maintain their original grade; and areas below 1.0 may be considered for grade increase. The overall quality grade of the product is determined by combining the adjusted quality grades of each testing area. The overall quality grade score is calculated as Q_overall = w_core × Q_core + w_influence × Q_influence + w_periphery × Q_periphery, where the core area has the highest weight, the influence area is in the middle, and the periphery area has the lowest weight. The final quality grade is determined based on the overall quality score: excellent products score above 90 points, good products score 75 to 90 points, qualified products score 60 to 75 points, and unqualified products score below 60 points. A quality grade determination report is generated, which includes evaluations of each area, imbalance analysis, the basis for dynamic adjustments, and the final quality grade conclusion. The quality grade determination results are fully recorded according to the regional grade, adjustment records, and overall conclusions, thus completing the quality monitoring of automotive component coatings.

[0080] To implement the above-described method embodiments, a method for monitoring the coating quality of automotive parts is provided to achieve the corresponding functions and technical effects. See also... Figure 2 , Figure 2 This diagram illustrates a structural block diagram of an automotive parts coating quality monitoring system 200 according to an embodiment of this application. For ease of explanation, only the parts relevant to this embodiment are shown. The automotive parts coating quality monitoring system 200 provided in this embodiment includes:

[0081] Data acquisition module 201 is used to acquire optical images and process parameter information of the coating surface, perform dark field enhancement processing on the optical images of the coating surface to generate scattering enhancement spectrum, and generate defect identification weights based on the scattering enhancement spectrum;

[0082] The defect localization module 202 is used to locate key defect clusters through the defect identification weights, extract the shape similarity features of the key defect clusters, and perform multi-dimensional similarity analysis on the shape similarity features to determine potential defect risk areas;

[0083] The coupling mapping module 203 is used to acquire the gloss distribution characteristics and color difference spatial distribution of the potential defect risk area, identify thickness fluctuation characteristics from the process parameter information to generate thickness gradient parameters, perform three-dimensional correlation mapping between the gloss distribution characteristics, the thickness gradient parameters and the color difference spatial distribution to generate a gloss-thickness-color difference coupling map, and determine the defect judgment boundary based on the gloss-thickness-color difference coupling map.

[0084] The detection optimization module 204 is used to determine the detection sensitivity by combining the defect judgment boundary and the scattering enhancement spectrum, perform influence range analysis on the key defect aggregation area to generate influence radius parameters, and construct a partitioned detection strategy based on the influence radius parameters and the detection sensitivity.

[0085] The determination output module 205 is used to identify abnormal coupling regions for the gloss-thickness-color difference coupled spectrum, perform similarity matching between the abnormal coupling regions and the morphological similarity features to generate imbalance compensation parameters, and generate quality level determination results based on the imbalance compensation parameters and the partition detection strategy.

[0086] The aforementioned automotive component coating quality monitoring system 200 can implement one of the automotive component coating quality monitoring methods described in the above-described method embodiments. The options described in the above method embodiments are also applicable to this embodiment and will not be detailed here. The remaining content of this application's embodiments can be referred to the content of the above method embodiments, and will not be repeated in this embodiment.

[0087] The above embodiments are not an exhaustive list based on the present invention, and there may be many other embodiments not listed. Any substitutions and improvements made without departing from the concept of the present invention are within the protection scope of the present invention.

Claims

1. A method of monitoring the quality of a coating on an automotive component, characterised by, The method comprises the following steps: Obtaining a coating surface optical image and process parameter information, performing dark field enhancement processing on the coating surface optical image to generate a scattering enhancement map, and generating a defect recognition weight based on the scattering enhancement map; Locating a key defect aggregation area through the defect recognition weight, extracting a topographic similarity feature of the key defect aggregation area, and performing multidimensional similarity analysis on the topographic similarity feature to determine a potential defect risk area; Obtaining a glossiness distribution feature and a color difference spatial distribution of the potential defect risk area, identifying a thickness fluctuation feature from the process parameter information to generate a thickness gradient parameter, performing three-dimensional correlation mapping on the glossiness distribution feature, the thickness gradient parameter and the color difference spatial distribution to generate a gloss-thickness-color difference coupling map, and determining a defect judgment limit based on the gloss-thickness-color difference coupling map; The three-dimensional correlation mapping on the glossiness distribution feature, the thickness gradient parameter and the color difference spatial distribution to generate a gloss-thickness-color difference coupling map comprises: performing hierarchical analysis on the glossiness distribution feature to generate a gloss level structure; projecting the color difference spatial distribution to the gloss level structure to form a color difference mapping layer; pairing and fusing the color difference mapping layer with the thickness gradient parameter to form a three-dimensional correlation structure; and constructing a gloss-thickness-color difference coupling map according to the three-dimensional correlation structure; Combining the defect judgment limit with the scattering enhancement map to determine a detection sensitivity, performing influence range analysis on the key defect aggregation area to generate an influence radius parameter, and constructing a partition detection strategy according to the influence radius parameter and the detection sensitivity; Performing coupling anomaly recognition on the gloss-thickness-color difference coupling map to detect an abnormal coupling area, generating an imbalance compensation parameter according to the similarity matching between the abnormal coupling area and the topographic similarity feature, and generating a quality grade judgment result based on the imbalance compensation parameter and the partition detection strategy.

2. The method of claim 1, wherein, The method comprises the following steps: Performing defect feature extraction on the scattering enhancement map to generate a defect feature set; Classifying the defect feature set according to severity to form a defect level identification; Calculating a recognition priority according to the defect level identification; Generating a defect recognition weight through the recognition priority.

3. The method of claim 1, wherein, The method comprises the following steps: Decomposing the defect recognition weight into primary defects and secondary defects; Tracing the primary defects to the secondary defects to form a defect correlation path; Detecting aggregation intensity of the defect correlation path to obtain an aggregation position set; Determining a key defect aggregation area according to the aggregation position set.

4. The method of claim 1, wherein, The method comprises the following steps: Performing diffusion tracking on the key defect aggregation area to obtain a diffusion trajectory curve; Identifying an accelerated diffusion paragraph from the diffusion trajectory curve to generate a diffusion intensification factor; Performing dynamic boundary processing on the diffusion trajectory curve by using the diffusion intensification factor to obtain a boundary parameter set; Determining an influence range to generate an influence radius parameter according to the boundary parameter set.

5. The method of claim 1, wherein, The imbalance compensation parameter is generated by similarity matching of the abnormal coupling region and the topographic similarity feature, including: A topographic deviation map is constructed based on the abnormal coupling region; A matching reference value is obtained by locating the similarity of the topographic deviation map and the topographic similarity feature; The matching reference value is extended to form a compensation adjustment domain according to the defect level; The imbalance compensation parameter is generated by correcting the boundary of the compensation adjustment domain.

6. The method of claim 1, wherein, The quality level judgment result is generated based on the imbalance compensation parameter and the partition detection strategy, including: The detection response difference is identified for the partition detection strategy; A constraint response distribution field is established based on the imbalance compensation parameter and the detection response difference; The imbalance utilization factor is generated by evaluating the imbalance conversion characteristics using the constraint response distribution field; The quality level judgment result is generated by dynamic grading adjustment according to the imbalance utilization factor.

7. The method of claim 1, wherein, The color difference mapping layer is formed by projecting the color difference spatial distribution to the gloss level structure, including: The maximum gradient position is identified by calculating the gloss difference of the highlight layer and the low light layer of the gloss level structure; The color difference reversal area of the color difference spatial distribution is located at the maximum gradient position; The decay utilization value is generated by obtaining the color difference decay rate from the color difference reversal area; The color difference mapping layer is formed by projecting the decay utilization value.

8. The method of claim 4, wherein, The diffusion strengthening factor is generated by identifying the accelerated diffusion paragraph from the diffusion trajectory curve, including: The curvature analysis is performed on the diffusion trajectory curve to determine the curvature jump point; The change rate reference is generated by locating the gradient change rate at the curvature jump point; The accelerated diffusion paragraph is divided by comparing the change rate reference with the diffusion trajectory curve; The diffusion strengthening factor is extracted according to the accelerated diffusion paragraph.

9. An automotive parts coating quality monitoring system characterized by comprising: It includes: A data acquisition module is used to acquire coating surface optical image and process parameter information, and to generate a scattering enhanced graph by dark field enhancement processing of the coating surface optical image, and to generate a defect recognition weight based on the scattering enhanced graph; A defect positioning module is used to locate a key defect aggregation area through the defect recognition weight, and to extract a topographic similarity feature of the key defect aggregation area, and to determine a potential defect risk area by multidimensional similarity analysis of the topographic similarity feature; A coupling mapping module is used to acquire a gloss distribution feature and a color difference spatial distribution of the potential defect risk area, to identify a thickness gradient parameter from the process parameter information, to generate a gloss-thickness-color difference coupling graph by three-dimensional correlation mapping of the gloss distribution feature, the thickness gradient parameter and the color difference spatial distribution, and to determine a defect judgment limit based on the gloss-thickness-color difference coupling graph; The three-dimensional correlation mapping of the glossiness distribution feature, the thickness gradient parameter and the color difference spatial distribution generates a gloss-thickness-color difference coupling atlas, which comprises: performing hierarchical analysis on the glossiness distribution feature to generate a gloss level structure; projecting the color difference spatial distribution to the gloss level structure to form a color difference mapping layer; pairing and fusing the color difference mapping layer with the thickness gradient parameter to form a three-dimensional correlation structure; and constructing a gloss-thickness-color difference coupling atlas according to the three-dimensional correlation structure; The detection optimization module is used for determining the detection sensitivity in combination with the defect determination limit and the scattering enhancement atlas, performing influence range analysis on the key defect aggregation area to generate an influence radius parameter, and constructing a partition detection strategy according to the influence radius parameter and the detection sensitivity; The determination output module is used for performing coupling anomaly recognition on the gloss-thickness-color difference coupling atlas to detect an abnormal coupling area, performing similarity matching according to the abnormal coupling area and the topographic similarity feature to generate an imbalance compensation parameter, and generating a quality level determination result based on the imbalance compensation parameter and the partition detection strategy.

Images