Aluminum pigment thickness uniformity analysis and detection method and system

Abstract

The invention relates to the technical field of coating detection, and particularly provides an aluminum pigment thickness uniformity analysis and detection method and system.The method comprises the steps that a multispectral image and a polarization image of an aluminum pigment coating are obtained, and geometrical shape information of a base material is obtained; analyzing whether the aluminum pigment coating has a non-uniformity defect or not according to the multispectral image and the polarization image; when the non-uniformity defect exists, acquiring the defect type and the optical characteristic of the non-uniformity defect according to the multispectral image and the polarization image; acquiring a target light source parameter and an image sensor target receiving angle according to the defect type, the optical characteristic and the geometrical shape information; adjusting a light source and an image sensor of the aluminum pigment thickness uniformity analysis and detection equipment according to the target light source parameter and the target receiving angle of the image sensor; the method comprises the following steps: acquiring a coating image by using aluminum pigment thickness uniformity analysis and detection equipment, and acquiring non-uniformity defect characteristics according to the coating image; according to the method, adaptive adjustment of detection parameters can be realized.

Description

Technical Field

[0001] This application relates to the field of coating inspection technology, and more specifically, to a method and system for analyzing and detecting the thickness uniformity of aluminum pigments. Background Technology

[0002] In the production of aluminum pigment coatings, particularly in high-end automotive bodies, consumer electronics casings, and architectural decorative panels, automated optical inspection systems have become a core component of quality control. These systems, through integrated formula parameter libraries, support rapid switching between different coating formulations, significantly improving production efficiency and the level of inspection automation. With the development of industrial intelligence, inspection technologies are evolving towards greater efficiency and integration to meet diverse product demands and quality standards.

[0003] However, existing inspection systems face significant challenges in identifying coating non-uniformity defects. Relying on preset, fixed optical geometry configurations, these systems cannot adaptively adjust inspection parameters based on dynamic changes in real-time coating characteristics, defect types, or substrate geometry. For example, different defects, such as pigment agglomeration, runs, or streaks, only exhibit optimal optical characteristics at specific viewing angles, but fixed angle settings cannot comprehensively capture all defects, increasing the risk of missed detections. Furthermore, batch variations in raw materials, minor fluctuations in manufacturing processes, or geometric variations in curved substrates can alter the optical scattering characteristics of the coating, rendering preset inspection angles ineffective and leading to misjudgments. This lack of real-time adaptability limits the comprehensiveness and accuracy of inspections, failing to ensure the effective identification and quantification of all potential defects, especially in complex and variable production environments, impacting the reliability of product quality control.

[0004] Currently, there is no effective technical solution to the above-mentioned problems. It should be noted that the information disclosed in this section is only for understanding the background of the present invention and therefore may include information that does not constitute prior art. Summary of the Invention

[0005] The purpose of this application is to provide a method and system for analyzing and detecting the uniformity of aluminum pigment thickness, which can achieve adaptive adjustment of detection parameters to effectively solve the problems of missed detection and misjudgment in existing fixed optical configuration detection systems in complex production environments.

[0006] In a first aspect, this application provides a method for analyzing and detecting the thickness uniformity of aluminum pigments, which includes the following steps: S1. Obtain multispectral and polarization images of the aluminum pigment coating, and obtain geometric information of the substrate on which the aluminum pigment coating is located; S2. Analyze whether there are non-uniform defects in the aluminum pigment coating based on multispectral and polarization images; S3. When non-uniform defects exist, obtain the defect type and optical characteristics of the non-uniform defects based on multispectral and polarization images; S4. Obtain the target light source parameters and the target receiving angle of the image sensor based on the defect type, optical characteristics, and geometric shape information; S5. Adjust the light source and image sensor of the aluminum pigment thickness uniformity analysis and detection equipment according to the target light source parameters and the target receiving angle of the image sensor. S6. Use aluminum pigment thickness uniformity analysis and detection equipment to collect coating images, and obtain non-uniformity defect characteristics based on the coating images.

[0007] Secondly, this application also provides an aluminum pigment thickness uniformity analysis and detection system, which includes: The coating data acquisition module is used to acquire multispectral and polarization images of the aluminum pigment coating and to acquire geometric information of the substrate on which the aluminum pigment coating is located. The defect analysis module is used to analyze whether there are non-uniform defects in the aluminum pigment coating based on multispectral and polarization images; The defect information acquisition module is used to acquire the defect type and optical characteristics of non-uniform defects based on multispectral and polarization images when non-uniform defects are present. The parameter acquisition module is used to acquire the target light source parameters and the target receiving angle of the image sensor based on the defect type, optical characteristics and geometric shape information; The parameter adjustment module is used to adjust the light source and image sensor of the aluminum pigment thickness uniformity analysis and detection equipment according to the target light source parameters and the target receiving angle of the image sensor. The defect feature acquisition module is used to acquire coating images using aluminum pigment thickness uniformity analysis and detection equipment, and to acquire non-uniform defect features based on the coating images.

[0008] As can be seen from the above, the aluminum pigment thickness uniformity analysis and detection method and system provided in this application acquires multispectral images, polarization images and geometric information of the coating and the substrate, and based on the type and optical characteristics of the non-uniformity defects obtained from the analysis, combined with the substrate geometric information, dynamically acquires and adjusts the light source parameters of the detection equipment and the receiving angle of the image sensor to achieve adaptive adjustment of the detection parameters. Therefore, this application can effectively solve the problems of missed detection and misjudgment in existing fixed optical configuration detection systems in complex production environments, thereby improving the accuracy and reliability of aluminum pigment thickness uniformity detection. Attached Figure Description

[0009] Figure 1 This is a flowchart of a method for analyzing and detecting the thickness uniformity of aluminum pigments, provided in an embodiment of this application.

[0010] Figure 2 This is a schematic diagram of the structure of an aluminum pigment thickness uniformity analysis and detection system provided in an embodiment of this application.

[0011] Reference numerals in the attached figures: 1. Coating data acquisition module; 2. Defect analysis module; 3. Defect information acquisition module; 4. Parameter acquisition module; 5. Parameter adjustment module; 6. Defect feature acquisition module. Detailed Implementation

[0012] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0013] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0014] Firstly, such as Figure 1 As shown, this application provides a method for analyzing and detecting the thickness uniformity of aluminum pigments, which includes the following steps: S1. Obtain multispectral and polarization images of the aluminum pigment coating, and obtain geometric information of the substrate on which the aluminum pigment coating is located; S2. Analyze whether there are non-uniform defects in the aluminum pigment coating based on multispectral and polarization images; S3. When non-uniform defects exist, obtain the defect type and optical characteristics of the non-uniform defects based on multispectral and polarization images; S4. Obtain the target light source parameters and the target receiving angle of the image sensor based on the defect type, optical characteristics, and geometric shape information; S5. Adjust the light source and image sensor of the aluminum pigment thickness uniformity analysis and detection equipment according to the target light source parameters and the target receiving angle of the image sensor. S6. Use aluminum pigment thickness uniformity analysis and detection equipment to collect coating images, and obtain non-uniformity defect characteristics based on the coating images.

[0015] Multispectral images refer to image data acquired across multiple discrete wavelengths or bands. This embodiment can utilize a multispectral camera or a broadband camera with tunable filters to acquire multispectral images, which can include images in the visible, near-infrared, or ultraviolet bands. This embodiment can capture the reflection, absorption, or scattering characteristics of the coating under different spectra by acquiring multispectral images to reveal differences in pigment distribution or thickness. Polarized images refer to images that record information about the polarization state of light. This embodiment can utilize a polarization camera or a combination of a polarization filter and a standard camera to acquire polarized images, which can be linearly polarized or circularly polarized. This embodiment can use polarization images to analyze the microstructure of the coating surface, pigment alignment direction, or anisotropic characteristics to assess uniformity. The geometric information of the substrate refers to the physical contour data of the material to which the coating is attached. This can be acquired using a 3D scanner, a structured light projection system, or a laser rangefinder. It primarily provides the geometric context of the light reflection and scattering paths on the coating surface to compensate for visual differences caused by shape variations. Non-uniformity defects refer to areas in the aluminum pigment coating that deviate from the expected uniformity in thickness, which can manifest as pigment agglomeration, runs, streaks, or color differences. Defect type refers to the specific classification or nature of the non-uniformity defect, which can be distinguished based on its morphology, cause, or optical appearance. This defect type can be agglomeration, runs, streaks, or orange peel texture. Optical properties refer to the response attributes of the non-uniformity defect area to light, which can include reflectivity, transmittance, absorptivity, degree of polarization, or spectral curves. These optical properties quantify the optical performance of the defect and provide a basis for adaptive adjustments. Target light source parameters refer to the operating attributes of the light source set to optimize defect detection. These can include wavelength, illumination intensity, polarization state, incident angle, or light source type, such as the brightness, color, or polarization direction of a specific LED light source. Their main purpose is to ensure that light illuminates the defect area in an optimal manner, enhancing defect contrast. The image sensor target receiving angle refers to the observation direction of the image sensor relative to the coating surface set to optimize defect image acquisition. This can include azimuth, zenith angle, or distance, such as the angle of the camera relative to the coating normal. Its main purpose is to ensure that the image sensor captures defect features from an optimal viewing angle. Adjusting the light source and image sensor of the aluminum pigment thickness uniformity analysis and testing equipment refers to modifying the hardware configuration of the testing equipment according to the acquired target parameters (target light source parameters and target receiving angle of the image sensor). This can be done using an electric adjustment mechanism, a programmable light source controller, or a robotic arm. For example, it can change the physical position of the light source or adjust the focal length or aperture of the image sensor. The main purpose is to enable the testing equipment to acquire images under optimal optical conditions.

[0016] The core innovation of this application lies in acquiring multispectral images and polarization images of the coating, as well as the geometric information of the substrate. Based on the analysis of the type and optical characteristics of non-uniform defects, and combined with the geometric information of the substrate, the light source parameters of the detection equipment and the receiving angle of the image sensor are dynamically acquired and adjusted to achieve adaptive adjustment of the detection parameters. Therefore, this application can effectively solve the problems of missed detection and misjudgment in existing fixed optical configuration detection systems in complex production environments, thereby improving the accuracy and reliability of aluminum pigment thickness uniformity detection.

[0017] Specifically, this method provides an adaptive scheme for analyzing and detecting the uniformity of aluminum pigment coating thickness. First, multispectral and polarization images of the coating are acquired, along with the geometric information of the substrate on which the coating is located, laying the data foundation for subsequent analysis. Multispectral images capture the optical response of the coating at different wavelengths, polarization images provide information on the microstructure of the coating surface and the orientation of the pigment arrangement, while the substrate geometry provides the geometric context of light reflection and scattering paths. Subsequently, the system uses this multidimensional optical information to perform preliminary defect screening of the coating, identifying regions that differ from normal coating areas in spectral response or polarization characteristics (non-uniform defects). Once non-uniform defects are confirmed, the system extracts the optical features of the defect region from the multispectral and polarization images and combines these features to determine the specific type and optical characteristics of the defect. This step is crucial for subsequent optimization of detection parameters, as different types of non-uniform defects only exhibit detectability under specific optical conditions. Furthermore, the system comprehensively utilizes the defect type, defect optical characteristics, and substrate geometry information to determine the appropriate light source parameters and image sensor receiving angle for the current defect detection. For example, defects of specific types and optical characteristics may exhibit high optical contrast at specific wavelengths, incident angles, and viewing angles. Combining this with substrate geometry information can compensate for viewing angle changes caused by curved surfaces, ensuring optimized optical paths. This dynamic parameter acquisition based on defect and substrate characteristics enables the detection system to specifically identify different defects. Subsequently, the system automatically adjusts the light source and image sensor of the detection equipment based on the calculated light source parameters and the image sensor's receiving angle. This real-time automated hardware adjustment capability ensures that subsequent image acquisition is performed under optical conditions conducive to defect identification and quantification. Finally, after the equipment parameters are adaptively adjusted, the system re-acquires coating images using the optimized detection equipment. Because the layer images acquired at this time are under optimized optical conditions, defect features are clearly presented in the images. Therefore, the system can extract accurate non-uniformity defect features from the coating images, providing data support for subsequent quality assessment and process improvement. The entire process forms a closed loop, ensuring that defects can be effectively identified and quantified in complex production environments.

[0018] As a preferred embodiment, the solution of this application is implemented as follows: In practical applications, a composite imaging system integrating a multispectral camera and a polarization camera is used to acquire multispectral and polarization images of the coating. The multispectral camera is preferably configured with multiple narrowband filters covering the visible and near-infrared bands, and the polarization camera integrates a polarization filter array. Simultaneously, a laser triangulation sensor or structured light scanner is used to acquire three-dimensional point cloud data of the substrate, and the geometric shape information of the substrate (e.g., radius of curvature or surface normal vector) is calculated based on the three-dimensional point cloud data. After acquiring the images and geometric information, the image processing unit runs an image analysis algorithm. This algorithm first preprocesses the multispectral and polarization images (e.g., denoising and calibration), and then identifies potential non-uniform defects in the coating through image segmentation and feature extraction techniques. Once non-uniform defects are identified, the system can use a machine learning model to further analyze the spectral response, polarization response, and spatial morphological characteristics of these defects based on the multispectral and polarization images to identify whether the defects are pigment agglomeration, sagging, or streaks, and quantify their optical properties (e.g., reflectance spectral curves or polarization distribution). Next, the parameter optimization module utilizes a rule-based or deep learning-based decision-making system to calculate the light source parameters and image sensor receiving angle based on the identified defect type, optical characteristics, and substrate geometry. For example, for sagging defects, the system suggests using linearly polarized light with a specific incident angle and positioning the image sensor at a specific reflection angle to the incident light. The light source can be an LED array with adjustable wavelength, intensity, and polarization state, and its incident angle can be adjusted using a robotic arm or motorized pan-tilt unit. The image sensor can be mounted on a multi-axis robotic arm to adjust its position and angle. After the light source and image sensor are adjusted, the inspection device re-acquires an image of the coating. Since this image is acquired under optimized optical conditions, defect features will be visible. The image processing unit then analyzes the newly acquired image again to extract non-uniform defect features (such as defect size, shape, location, and severity) and generates an inspection report.

[0019] In some preferred embodiments, the target light source parameters include wavelength, illumination intensity, polarization state, and incident angle. Wavelength is an inherent property of light waves and can be generated by a combination of a tunable laser, a multispectral LED array, or a filter and a broadband light source. Illumination intensity is the light power received per unit area and can be controlled by an illumination intensity adjuster, a variable aperture, or a neutral density filter. Polarization state is a characteristic of the direction of the electric field vibration of the light wave and can be changed by a polarizer, a liquid crystal polarizer, or a polarizing fiber. The incident angle is the angle of the light ray relative to the normal to the surface being tested and can be set by a light source position adjustment mechanism, a mirror system, or a fiber array. Specifically, this embodiment can enhance the contrast between highly non-uniform defects and the background by selecting or adjusting light of a specific wavelength based on the spectral response characteristics of the defects. This embodiment can avoid overexposure or underexposure of the image and ensure complete capture of defect features by dynamically adjusting the light intensity based on the reflection or scattering intensity of the coating and defects. This embodiment can utilize the polarization characteristics of light by adjusting the polarization state of the light to detect polarization-sensitive defects, thereby revealing defect information that is difficult to detect with unpolarized light. This embodiment can also maximize the detectability of defects by selecting the incident angle of the light according to the geometry of the defect, the optical scattering characteristics, and the geometry of the substrate to ensure that the light irradiates the coating in a manner conducive to defect manifestation.

[0020] In some preferred embodiments, step S4 includes: S41. Obtain the optical response information of the first non-defect region adjacent to the non-uniformity defect based on the multispectral image and polarization image; S42. Based on the difference between optical characteristics and optical response information, the optical characteristics are corrected to obtain the corrected defect optical characteristics; S43. Obtain the target light source parameters and the target receiving angle of the image sensor based on the defect type, the corrected defect optical characteristics, and the geometric shape information.

[0021] The optical response information of the first non-defect region refers to the optical properties (e.g., reflectivity, transmittance, or scattering characteristics) of the normal region immediately adjacent to the non-uniform defect in the coating. This embodiment obtains the optical response information of the first non-defect region by analyzing pixel data within a specific range around the defect based on multispectral and polarization images. For example, the spectral intensity or polarization state of pixels in the non-defect region near the defect boundary can be averaged, or the background optical behavior can be characterized using a statistical model. This embodiment provides a benchmark for evaluating the optical properties of the defect itself by acquiring the optical response information of the first non-defect region. Furthermore, correcting the optical properties based on the difference between the optical properties and the optical response information refers to adjusting or optimizing the optical properties of the defect by comparing the initial optical properties of the non-uniform defect with the optical response information of adjacent non-defect regions. This correction can be achieved using various methods, such as calculating the difference or ratio between the two, or applying specific filtering algorithms to highlight the optical contrast between the defect and the normal background. The corrected defect optical properties refer to the dataset that more accurately reflects the optical characteristics of the defect itself after the above correction process. Since the corrected defect optical properties in this embodiment eliminate the interference of background environment, measurement noise or coating inherent variability on the evaluation of defect optical properties, this embodiment can provide a cleaner input for subsequent parameter determination.

[0022] This application incorporates consideration of optical information from non-defect regions surrounding the defect because, in coating inspection, the optical properties of a defect are often influenced by the surrounding normal coating background, and the initially acquired optical properties may contain measurement errors or environmental interference. Due to this background influence and potential errors, directly using the initially acquired optical properties to determine the detection parameters may result in suboptimal parameters, thus affecting the accurate capture of defect features. To address this issue, this scheme first acquires the optical response information of a first non-defect region adjacent to the non-uniform defect based on multispectral and polarization images. This step provides an important benchmark for subsequent corrections, enabling the system to understand the optical behavior of the defect against a normal coating background. Subsequently, the optical properties of the defect are corrected based on the difference between the optical properties of the non-uniform defect and the optical response information of the adjacent non-defect region, resulting in corrected defect optical properties. This correction process effectively eliminates or reduces interference from the background environment, measurement noise, or inherent coating variability on the evaluation of defect optical properties, making the corrected optical properties more accurately reflect the true optical signature of the defect itself. Finally, based on the corrected defect optical properties, combined with defect type and geometric information, the target light source parameters and the target receiving angle of the image sensor are obtained. Because the optical properties used to determine the detection parameters are modified and optimized, this embodiment enables more accurate detection parameters. This allows the aluminum pigment thickness uniformity analysis and detection equipment to acquire coating images under optimized lighting and observation conditions during subsequent inspections, thereby more effectively capturing and identifying the characteristics of non-uniform defects and significantly improving the accuracy and reliability of the inspection. This refined processing of the defect optical properties overcomes the limitations of traditional fixed-parameter detection, ensuring high-precision defect detection even in complex and variable production environments. It also solves the problem of poor detection results or missed detections due to inaccurate initial optical properties.

[0023] In one specific embodiment, to obtain the optical response information of a first non-defect region adjacent to the non-uniform defect, the boundary of the non-uniform defect is first identified in a multispectral image and a polarization image. Then, a ring-shaped region of a predetermined width, for example, 5 to 10 pixels wide, is defined outside the defect boundary as the first non-defect region. For each pixel within this region, its spectral response data at different wavelengths is extracted from the multispectral image, and its polarization response data (including polarization degree and polarization angle) is extracted from the polarization image. These pixel-level optical response data can be averaged or statistically analyzed to obtain an average spectral response curve and average polarization characteristics (optical response information) representing the first non-defect region. Further, to correct the optical characteristics based on the difference between the optical characteristics and the optical response information, the initially obtained defect optical characteristics are compared with the optical response information of the first non-defect region. For example, the difference between the defect spectral response curve and the non-defect region spectral response curve at various wavelengths, and the difference between the defect polarization degree and the non-defect region polarization degree, can be calculated; these differences are considered as difference features. The optical properties of the original defects are adjusted based on these differences using a correction algorithm (which can be a weighted average model or a threshold-based adjustment rule). For example, if the reflectivity of a non-uniform defect is significantly higher than that of a non-defect area at a certain wavelength, the optical properties of the non-uniform defect at that wavelength can be enhanced or normalized based on this difference, thus obtaining the corrected defect optical properties. Finally, when acquiring the target light source parameters and the target receiving angle of the image sensor, the corrected defect optical properties, defect type, and substrate geometry information are input into a pre-established parameter mapping model. This model can be a lookup table based on empirical data or a predictive model trained through machine learning. This model outputs the target light source parameters and the target receiving angle of the image sensor that maximize the contrast of the defect features based on these input parameters. In this way, the detection device can be precisely configured to capture defect features under optimal conditions.

[0024] In some preferred embodiments, step S42 includes: S421. Obtain the coating formulation information for aluminum pigment coatings; S422. Determine the difference analysis rules based on the coating formulation information and defect types; S423. Obtain difference characteristics based on optical properties and optical response information using difference analysis rules; S424. Correct the optical properties based on the difference characteristics to obtain the corrected defect optical properties.

[0025] Coating formulation information refers to the specific material composition data used in aluminum pigment coatings. This can include detailed proportions or specifications of pigment types, particle size distributions, resin types, curing agents, additives, and solvents. This information can be obtained from product specifications obtained from coating suppliers, internal production database records, or through chemical analysis. Difference analysis rules refer to a set of logic or algorithms used to guide the comparison and feature extraction between optical properties and optical response information. This embodiment can employ predefined mathematical models, empirically based lookup tables, machine learning models (e.g., decision trees, support vector machines, or neural networks), or expert system rule bases to determine difference analysis rules based on coating formulation information and defect types. Difference features refer to indicators that, after processing by difference analysis rules, can quantify and characterize specific differences between defective optical properties and the optical response of non-defective areas. These can be represented as normalized difference values, reflectance or polarization ratios in specific wavelength bands, or multidimensional feature vectors.

[0026] This method, when correcting the optical properties of defects, first obtains the coating formulation information of the aluminum pigment coating. Since the inherent material properties of the coating, such as pigment type, resin type, and additives, significantly affect the optical scattering characteristics of the coating and the optical performance of defects, this embodiment lays the foundation for subsequent difference analysis tailored to specific material systems by introducing coating formulation information. Based on this, the system determines difference analysis rules according to the obtained coating formulation information and defect type. Because this step closely links the specificity of the coating material and the type of defect with the difference analysis of optical properties and optical response information—for example, for a specific coating formulation, a certain type of defect may exhibit a unique optical difference pattern at a specific wavelength or polarization state—this embodiment is equivalent to establishing customized difference feature analysis rules based on the coating formulation information and defect type, avoiding the use of a one-size-fits-all difference feature analysis method and making subsequent difference feature extraction more targeted and accurate. Subsequently, using the customized difference analysis rules, the original defect optical properties and the optical response information of non-defect areas are compared and analyzed to obtain more representative and diagnostic difference features. These discrepancies are no longer simple numerical differences, but rather rule-guided quantitative indicators that reflect the true optical differences under specific formulations and defect types, providing a more reliable basis for subsequent corrections. Ultimately, based on these customized analysis-derived discrepancies, the original defect optical properties are corrected, resulting in corrected defect optical properties. This correction method fully considers the inherent properties of the coating material and the specific manifestations of the defects, enabling the corrected defect optical properties to more accurately reflect the true optical behavior of the defects. Through this refined correction process, this method significantly improves the accuracy and universality of defect optical property correction, thus providing a solid foundation for the accurate acquisition of target light source parameters and image sensor target receiving angles, thereby enhancing the accuracy and reliability of the entire detection method. This refinement of defect optical property correction allows for a more precise match between the actual optical performance of the defects when acquiring target light source parameters and image sensor target receiving angles, thereby optimizing the adaptability and reliability of the overall detection process.

[0027] In one specific embodiment, firstly, the coating formulation information for the aluminum pigment coating is obtained by querying the unique identifier of the current batch of coating from the production management system or material database, and retrieving detailed data such as its chemical composition, pigment particle size distribution, resin type, and additive content. For example, if the coating formulation is identified as "XYZ-123", the system can automatically load the optical scattering characteristic curve and optical fingerprint data of known defects associated with the formulation. Subsequently, difference analysis rules are determined by querying a pre-built multidimensional lookup table based on the coating formulation information and defect type, or by inputting the coating formulation information and defect type into a pre-trained machine learning model. This lookup table or model takes the coating formulation information and defect type as input and outputs a set of specific weighting coefficients, thresholds, or transformation functions to guide the comparison between optical properties and optical response information. For example, for the "pigment agglomeration" defect under the "XYZ-123" formulation, the rule can specify that the reflectivity difference is weighted within a specific wavelength range (such as 450nm-550nm) and a specific threshold considering the polarization difference is taken into account. When acquiring difference features using difference analysis rules, the system can calculate normalized difference values, ratios, or vectors between defective and non-defective regions in these specific dimensions based on the wavelength, polarization state, or angle range specified by the rules. For example, if the rule indicates a focus on the reflectivity difference at a wavelength of 500 nm and the transmittance difference of horizontally polarized light, the system will accurately calculate these values ​​as difference features. Finally, a linear or nonlinear correction model based on these difference features is used to correct the optical properties. The corrected defect optical properties can be equal to the original defect optical properties plus a compensation term determined by the difference features and a preset correction coefficient. For example, if the difference features show that the reflectivity of the defective region is low at a specific wavelength, the correction process can add a compensation value based on the degree of low reflectivity and the characteristics of the coating formulation, making the corrected reflectivity closer to the true optical performance of a normal defect under that formulation.

[0028] By incorporating coating formulation information of aluminum pigment coatings and determining difference analysis rules based on the coating formulation information and defect types, this method can customize optical property correction for different material systems and defect types. Utilizing these customized difference analysis rules to obtain difference features makes the extracted features more representative and diagnostic. Correcting optical properties based on these refined difference features yields more accurate and universally applicable corrected defect optical properties. This significantly improves the accuracy of subsequent acquisition of target light source parameters and image sensor target receiving angles, thereby effectively enhancing the overall accuracy and reliability of aluminum pigment coating non-uniformity defect detection and solving the problem of insufficient correction accuracy in the face of complex materials and diverse defects using traditional methods.

[0029] In some preferred embodiments, step S422 includes: A1. Determine preliminary analysis rules based on paint formulation information and defect types; A2. Obtain the spraying process parameters of the aluminum pigment coating; A3. Adjust the preliminary analysis rules based on the spraying process parameter information to obtain the difference analysis rules.

[0030] Spraying process parameters refer to various process variables that affect the formation and performance of aluminum pigment coatings during the spraying process. These can include spraying pressure, spraying distance, spraying speed, curing temperature, curing time, ambient humidity, or coating viscosity. This information can be obtained through real-time sensor data acquisition, production management system recording, or manual input. Difference analysis rules refer to the final model or algorithm that more accurately reflects the true difference between the optical characteristics of defects and the optical response of non-defect areas under specific process conditions, after modification and optimization based on preliminary analysis rules and actual spraying process parameters. This embodiment can obtain difference analysis rules by performing parameter correction, weight adjustment, model retraining, or dynamic selection of rule sets on the preliminary analysis rules.

[0031] This method first determines preliminary analysis rules based on coating formulation information and defect types. Then, it acquires the spraying process parameters of the aluminum pigment coating. These parameters (e.g., spraying pressure, spraying distance, curing temperature) directly affect the microstructure of the coating (e.g., the alignment direction and density distribution of the aluminum pigment, and the smoothness of the coating surface), thus significantly altering the coating's light scattering and reflection characteristics. Therefore, these process parameters are key factors influencing the coating's optical performance. After acquiring the process parameter information, the preliminary analysis rules are adjusted based on this information to obtain the final difference analysis rules. This adjustment mechanism allows the difference analysis rules to dynamically adapt to changes in the coating's optical properties under different production batches and process conditions. Since the true difference in optical response between defective and non-defective areas is influenced not only by the coating itself and the defect type but also by the direct modulation of the production process, this approach can more accurately capture the true difference in optical response between defective and non-defective areas by incorporating spraying process parameters into the determination process of the difference analysis rules. This more precise difference analysis rule provides a more reliable basis for subsequent correction of defective optical properties, thereby improving the accuracy of defect correction. The corrected defect optical properties, combined with defect type and substrate geometry information, can more accurately guide the acquisition of target light source parameters and image sensor target receiving angle, ensuring that the detection equipment works under optimal optical configuration, thereby significantly improving the accuracy and reliability of aluminum pigment thickness uniformity analysis and detection.

[0032] In one specific embodiment, a pre-established database is used to determine preliminary analysis rules. This database stores optical difference models corresponding to different coating formulations (e.g., a specific color formulation of a certain automotive paint) and common defect types (e.g., sagging, orange peel texture, pinholes). When the system identifies a certain coating formulation and defect type, it retrieves the corresponding preliminary analysis rules from this database. The spraying process parameters of the aluminum pigment coating can be obtained using a sensor network on the production line. For example, spraying pressure can be monitored in real time using pressure sensors, the temperature inside the curing oven can be monitored using infrared thermometers, and the moving speed and distance of the spraying robot can be recorded using encoders. This data is then transmitted to the processing unit. After receiving this process parameter information, the processing unit adjusts the preliminary analysis rules based on this information. For example, if the spraying pressure is detected to be too high, it may cause the aluminum pigment to be more densely packed, thus affecting its specular reflection characteristics. In this case, the weight of the reflectivity difference in the preliminary analysis rules can be increased. If the curing temperature is too low, it may cause insufficient drying of the coating, affecting surface smoothness. In this case, the threshold for scattering characteristic differences in the preliminary analysis rules can be adjusted. In this way, the preliminary analysis rules can be dynamically modified according to actual production conditions, thereby obtaining difference analysis rules that are more consistent with the current coating condition.

[0033] In some preferred embodiments, step S41 includes: S411. Acquire optical response data of the first non-defect region adjacent to the non-uniformity defect at multiple viewing angles and / or multiple illumination angles; S412. Obtain the first local geometric information of the first non-defect region; S413. Obtain the optical anisotropy characteristics of the first non-defect region based on the first local geometric information and all optical response information; S414. Perform angle compensation on the optical response data based on the optical anisotropy characteristics to obtain optical response information.

[0034] Optical response data refers to the quantitative measurement values ​​of the response of a coating surface to incident light, such as reflection, scattering, or transmission, under specific lighting and observation conditions. These values ​​can include reflectivity, transmittance, scattering intensity, or degree of polarization. First local geometric information refers to the shape characteristics of a local region of the substrate where the first non-defect region is located. This can be represented using surface normal vectors, curvature parameters, or three-dimensional point cloud data. Optical anisotropy refers to the inherent property of a material's optical properties changing with direction. This embodiment utilizes an optical model (physical model, empirical model, or data-driven model) to obtain the optical anisotropy of the first non-defect region based on first local geometric information and all optical response information. This model takes the multi-angle optical response data obtained in step S411 and the first local geometric information obtained in step S412 as input, and fits the anisotropy parameters through optimization algorithms or machine learning methods. For example, an anisotropy model based on Mie scattering theory or geometric optics theory can be used to accurately identify and quantify its optical anisotropy. The optical anisotropy of this embodiment can be expressed as parameters of the bidirectional reflectance distribution function (BRDF) or bidirectional transmission distribution function (BTDF), or the variation of polarization characteristics with angle. Angle compensation refers to the process of mathematically correcting the original optical response data based on the optical anisotropy of the coating material. This can be achieved using ray tracing algorithms based on physical models, empirical correction factors, or machine learning models to eliminate deviations introduced by measurement angles and local geometric changes, thereby obtaining standardized optical response information.

[0035] This approach employs a series of meticulously processed steps to ensure that the acquired optical response information accurately reflects the true optical baseline of the non-defect region, effectively addressing the inaccuracies in traditional methods caused by angle dependence, optical anisotropy, and local geometric variations. Specifically, firstly, by acquiring optical response data from multiple observation angles and / or multiple illumination angles, the complete optical behavior of the aluminum pigment coating under different lighting and observation conditions is comprehensively captured, providing rich and multi-dimensional data for subsequent precise analysis. Simultaneously, the first local geometric information of the first non-defect region is acquired to quantify the local shape characteristics of the substrate on which the coating is located. Since the substrate geometry directly affects the actual incident and exit angles of light, the first local geometric information provides necessary spatial reference for the analysis of optical response data. Based on this, by combining the first local geometric information and all optical response data, the optical anisotropy characteristics of the first non-defect region are analyzed and acquired in depth. This allows the understanding of the coating's optical behavior to deepen from apparent phenomena to the inherent properties of the material itself and reveals the differences in optical response of the coating material in different directions. Finally, the acquired optical anisotropy characteristics are used to perform angle compensation on the original optical response data. This compensation mechanism effectively eliminates the influence of measurement angle variations and local geometric differences on optical response data, resulting in a standardized, accurate optical response information that truly reflects the optical characteristics of non-defect areas. This finely processed optical response information serves as the basis for subsequent correction of defect optical characteristics and for obtaining target light source parameters and image sensor target receiving angles, significantly improving the accuracy and robustness of the entire detection method. By providing more reliable baseline data, this scheme enables more precise correction of defect optical characteristics, thereby ensuring the accuracy of the determined target light source parameters and image sensor target receiving angles. This provides a solid foundation for subsequent equipment adjustment and defect feature acquisition, effectively improving the reliability and accuracy of aluminum pigment thickness uniformity analysis and detection.

[0036] In one specific embodiment, a multi-angle measurement system is first used to acquire optical response information of a first non-defect region adjacent to the non-uniformity defect. The system includes a rotatable sample stage, or multiple image sensors and light source arrays fixed at different angles. For example, a robotic arm carries the image sensors and light sources to scan the first non-defect region and acquire multispectral and polarization images at multiple preset observation angles (e.g., 0 degrees, 30 degrees, 60 degrees relative to the surface normal) and / or multiple illumination angles (e.g., 20 degrees, 45 degrees, 70 degrees), thereby obtaining optical response data of the region under different geometric conditions. Next, a 3D scanning device (such as a laser triangulation instrument or a structured light projector) is used to scan the substrate surface where the first non-defect region is located to obtain high-precision 3D point cloud data. This point cloud data is then processed to calculate the surface normal vector and local curvature of each pixel in the region, thus obtaining the first local geometric information. Subsequently, an optical model (such as an improved Phong model or a Torrance-Sparrow model) is used to obtain the optical anisotropy characteristics based on the first local geometric information and all optical response data. This model can use the least squares method to calculate the optical response data at different angles and the first local geometric information. By fitting local geometric information, the specular reflection component, diffuse reflection component, and anisotropic parameters of the coating are obtained, thus acquiring its optical anisotropic properties. Finally, by normalizing all the original optical response data to a certain standard viewing and illumination angle using the optical anisotropic properties, angle compensation is performed on the optical response data based on the optical anisotropic properties to eliminate deviations caused by measurement angles and surface curvature. In this way, a standardized and accurate optical response information can be obtained, which can truly reflect the inherent optical properties of non-defect areas, providing a reliable benchmark for subsequent defect analysis and parameter optimization.

[0037] This approach acquires optical response data from multiple viewing angles and / or multiple illumination angles, and combines this data with local geometric information to obtain the optical anisotropy characteristics of the coating. Angle compensation is then applied to the optical response data, effectively overcoming the complexity, strong angle dependence, and inaccuracies caused by variations in the local geometry of the substrate in the optical response of aluminum pigment coatings. This ensures that the acquired optical response information accurately reflects the true optical baseline of non-defect areas, significantly improving the accuracy of subsequent defect correction and the acquisition of target light source parameters and image sensor target receiving angles.

[0038] In some preferred embodiments, step S3 includes: S31. When there is a non-uniformity defect, identify the first non-uniformity defect region from the multispectral image and polarization image, and obtain the pixel-level spectral response data and polarization response data of the first non-uniformity defect region. S32. Extract the spectral and polarization features of the first non-uniformity defect region based on pixel-level spectral response data and polarization response data; S33. Perform image segmentation and morphological analysis on the first non-uniform defect region to obtain the spatial morphological features of the first non-uniform defect region. The spatial morphological features include the shape, size, edge characteristics, or linearity of the non-uniform defect. S34. Obtain the defect type and optical properties of non-uniform defects based on spectral characteristics, polarization characteristics and spatial morphology characteristics.

[0039] Obtaining pixel-level spectral and polarization response data for the first non-uniform defect region means directly extracting the intensity value (response at different wavelengths) of each pixel within the defect region in the multispectral image and its polarization information (e.g., Stokes parameters, degree of polarization, polarization angle) in the polarization image once the defect region is identified. Spectral features refer to the quantitative information extracted from pixel-level spectral response data that characterizes the optical behavior of the defect at different wavelengths. This can be achieved using feature vectors obtained through dimensionality reduction methods such as the shape parameters of the spectral curve, the intensity ratio of specific bands, the position and intensity of absorption or reflection peaks, or principal component analysis. Polarization features refer to the quantitative information extracted from pixel-level polarization response data that characterizes the effect of the defect on the polarization state of light. This can be achieved using the degree of polarization, polarization angle, ellipticity, or specific elements of the Mueller or Jones matrix. Image segmentation refers to the process of separating regions with specific attributes from the background in an image. This can be achieved using thresholding, region growing, edge detection, clustering algorithms, or deep learning-based semantic segmentation models. Morphological analysis refers to the process of extracting or modifying the structure of objects in an image by performing a series of shape-based operations. These operations can include erosion, dilation, opening, closing, skeleton extraction, or connected component analysis. Spatial morphological features refer to the quantitative information obtained through image segmentation and morphological analysis, describing the geometric shape and structural properties of non-uniform defects in two-dimensional space. These features can be achieved using the area, perimeter, aspect ratio, circularity, rectangularity, convexity, Euler number of the defect region, or shape descriptors based on Fourier descriptors or Zernike moments.

[0040] To obtain a more comprehensive and accurate understanding of the types and optical properties of non-uniform defects, this method refines the defect information acquisition steps. First, the first non-uniform defect region is identified from multispectral and polarization images, and pixel-level spectral and polarization response data for this region are acquired, laying the data foundation for subsequent in-depth analysis. This process ensures the direct capture of original, fine-grained information about the defect region, avoiding information loss or judgment bias that may result from macroscopic image analysis. Based on this, the spectral and polarization features of the first non-uniform defect region are further extracted using the acquired pixel-level spectral and polarization response data. This step refines and abstracts the original data, transforming massive amounts of pixel data into more discriminative quantitative features. These features effectively characterize the absorption and reflection of light at different wavelengths and the influence of the defect on the polarization state, thereby revealing the essential optical properties of the defect. Simultaneously, to overcome the limitations of single optical feature analysis, this method also performs image segmentation and morphological analysis on the first non-uniform defect region to obtain its spatial morphological features (including the shape, size, edge characteristics, or linearity of the defect). Image segmentation technology precisely separates defect areas from the background, while morphological analysis quantifies the physical and geometric properties of the defects. These spatial morphological features are closely related to the formation mechanism of defects; for example, agglomeration defects may present as irregular blocks, while sagging defects may present as strips. Finally, by comprehensively utilizing the extracted spectral features, polarization features, and acquired spatial morphological features, the defect type and optical properties of non-uniform defects are obtained. This multi-dimensional, multi-modal feature fusion analysis overcomes the shortcomings of single feature analysis, making defect identification and classification more accurate and the description of defect optical properties more refined and comprehensive. It is precisely because of this comprehensive analysis that this method can more accurately identify different types of non-uniform defects and precisely capture their unique optical properties, thus providing high-quality input for subsequent adaptive adjustment of detection parameters and significantly improving the overall accuracy and reliability of the detection. This meticulous defect information acquisition method works closely with other steps in the entire aluminum pigment thickness uniformity analysis and detection method, enabling the detection system to dynamically optimize according to the actual characteristics of the defects, thereby achieving more comprehensive and accurate defect detection in complex and ever-changing production environments.

[0041] In a specific embodiment, to obtain the type and optical properties of non-uniform defects in an aluminum pigment coating, the following steps can be taken. First, using a multispectral camera equipped with a tunable filter and a polarization camera, images of the aluminum pigment coating are acquired, obtaining multispectral images in the visible and near-infrared bands, as well as polarization images at 0°, 45°, 90°, and 135° polarization directions. In these images, a method based on local contrast enhancement and threshold segmentation (e.g., Otsu thresholding combined with connected component analysis) is used to automatically identify the first non-uniform defect region in the image. For each identified defect region, the system can extract the reflection intensity value of each pixel in the multispectral image at different wavelengths, forming pixel-level spectral response data; simultaneously, the polarization degree, polarization angle, and other parameters of each pixel are calculated from the polarization image, forming pixel-level polarization response data. Next, based on these pixel-level data, the spectral and polarization features of the defects are further extracted. For example, for spectral features, the average slope of the pixel spectral curve within the defect region, the reflection intensity ratio of a specific band, or principal component analysis is used to extract the first few principal components as spectral feature vectors. For polarization features, the average value of pixel polarization degree and the variance of polarization angle distribution within the defect region are calculated, or specific elements of the Stokes vector or Mueller matrix are used for characterization. Simultaneously, to obtain the spatial morphological features of the defect, image segmentation and morphological analysis can be performed on the identified first non-uniform defect region. For example, a region growing algorithm can be used to accurately segment the defect region, followed by morphological operations (such as erosion and dilation) to smooth defect edges and remove noise. Subsequently, geometric parameters such as area, perimeter, aspect ratio, and circularity of the segmented defect region are calculated as spatial morphological features. For example, the linearity of a long, narrow defect can be calculated, and the circularity of a clustered defect can be calculated. Finally, the extracted spectral features, polarization features, and spatial morphological features are input into a pre-trained classification model (such as a support vector machine or convolutional neural network). This model can output the specific type of non-uniform defect (e.g., "pigment agglomeration," "sag," "scratches," or "orange peel texture") based on the combination of these multimodal features. Simultaneously, the model can also output optical characteristic parameters corresponding to the defect type (such as its scattering coefficient at a specific wavelength, the peak position of the reflectance spectrum, or the degree of anisotropy of polarization scattering). In this way, comprehensive and accurate identification and characterization of non-uniform defects in aluminum pigment coatings can be achieved.

[0042] This method, by refining the defect information acquisition process, effectively addresses the problem of similar or overlapping optical features in multispectral and polarization images of non-uniform defects in aluminum pigment coatings, which makes it difficult to accurately distinguish defect types and precisely capture their subtle optical properties. By identifying defect regions and acquiring pixel-level spectral and polarization response data, a raw, fine-grained data foundation is provided for subsequent analysis. Based on this, extracting spectral and polarization features allows for a more effective characterization of the inherent optical properties of the defects. Simultaneously, combining spatial morphological features obtained through image segmentation and morphological analysis overcomes the limitations of single optical features, resulting in a more comprehensive description of the defects. Finally, the comprehensive utilization of multi-dimensional features for fusion analysis significantly improves the accuracy of non-uniform defect type identification and enables a more refined and comprehensive acquisition of defect optical properties. This provides a more reliable and accurate input for the adaptive adjustment of subsequent detection parameters, thereby improving the overall accuracy and reliability of the detection.

[0043] In some preferred embodiments, step S34 includes: S341. Obtain the coating formulation information and spraying process parameter information of the aluminum pigment coating, and obtain the second local geometric information of the first non-uniformity defect area; S342. Perform geometric correction on spectral features, polarization features and spatial morphological features based on the second local geometric information; S343. Weight the spectral characteristics, polarization characteristics and spatial morphological characteristics according to the coating formulation information and spraying process parameter information; S344. Obtain the defect type and optical properties of non-uniform defects based on spectral characteristics, polarization characteristics and spatial morphological characteristics.

[0044] The second local geometric information refers to the local geometric morphology data of the substrate surface where the first non-uniform defect region is located. This data can be acquired using techniques such as 3D scanning, structured light measurement, or image-based depth estimation. Geometric correction refers to eliminating or reducing feature distortion caused by the object's geometric shape during image acquisition through mathematical transformations. This can be achieved using perspective transformation, nonlinear deformation correction, or correction algorithms based on physical models. Weighted processing refers to adjusting the relative importance of different features in the decision-making process based on specific factors. This can be achieved using linear weighting, nonlinear weighting, or feature weight learning based on machine learning.

[0045] This solution, when acquiring the defect type and optical properties of non-uniform defects, first obtains the coating formulation information and spraying process parameters of the aluminum pigment coating, and simultaneously acquires the second local geometric information of the first non-uniform defect region. This information forms the basis for subsequent feature processing, as the intrinsic properties of the coating, the external conditions during coating formation, and the local geometry of the defect region all directly affect the optical performance and image features of the defect. Based on this auxiliary information, the spectral features, polarization features, and spatial morphological features extracted from multispectral and polarization images are processed. Since image acquisition on curved substrates introduces perspective distortion and projection effects, leading to distortion of the original features, this embodiment can effectively eliminate the influence of these geometric factors on the features by geometrically correcting the spectral features, polarization features, and spatial morphological features according to the second local geometric information. This restores the features to a state closer to the true physical properties of the defect, thereby improving the accuracy and comparability of the features. Based on this, the corrected spectral features, polarization features, and spatial morphological features are weighted according to the coating formulation information and spraying process parameters. The coating formulation and spraying process parameters have a decisive influence on the manifestation and optical properties of defects. Therefore, this embodiment can dynamically adjust the importance of different features in defect identification and classification based on the specific coating formulation and production conditions through weighted processing, making the defect identification and characteristic acquisition process more targeted and adaptive. Ultimately, after geometric correction and weighted processing, the spectral features, polarization features, and spatial morphological features used become more accurate and representative. Based on these optimized features, the specific type of non-uniform defects and their precise optical properties can be identified more reliably. This series of steps forms a complete and adaptive feature processing flow, ensuring more accurate defect judgment results and providing high-quality input for subsequent adaptive adjustment of detection equipment parameters. This scheme, combined with the previous steps of identifying the first non-uniform defect region, acquiring pixel-level spectral response data and polarization response data, and extracting spectral features, polarization features, and spatial morphological features, forms a more complete defect analysis system. By introducing refined processing of these features after feature extraction, features that might otherwise be distorted due to environmental complexity are optimized, thereby achieving higher accuracy and robustness in the final defect type and optical property judgment stage. This combination enables the entire detection method to better adapt to complex substrate geometries and varying production conditions, significantly improving the accuracy and reliability of defect identification and effectively solving the problem of misjudgment and missed detection that traditional methods are prone to in complex environments.

[0046] In one specific embodiment, preset data such as paint batch number, pigment type, solid content, viscosity, spraying pressure, spraying distance, baking temperature, and time are read from the production management system or product database to obtain paint formulation information and spraying process parameters. Simultaneously, a laser triangulation sensor or structured light scanner is used to perform a local 3D scan of the defect area, generating a depth map or point cloud data of the area. This data is then used to calculate the local curvature, normal vector, or tilt angle to obtain the second local geometric information of the first non-uniform defect area. Subsequently, a pre-constructed perspective correction model based on the local normal vector of the defect area is used to perform geometric correction on spectral features, polarization features, and spatial morphological features according to the second local geometric information. For example, if the defect is located on a curved surface, its image may have radial distortion. A nonlinear distortion correction algorithm can be applied based on the local curvature information to map the image pixel coordinates to the corrected planar coordinates, thereby eliminating optical and spatial morphological distortions caused by changes in viewing angle. For polarization features, the polarization degree calculation formula can be adjusted according to the local tilt angle to compensate for the influence of surface tilt on the polarization response. A weighted model based on rules or machine learning is used to weight spectral, polarization, and spatial morphology features according to paint formulation information and spraying process parameters. For example, if the paint formulation contains high-reflectivity aluminum pigments and the spraying process parameters indicate a thin coating, the weight of the spectral reflectance feature can be increased, as the spectral response is more sensitive to defect type under these conditions. If the process parameters indicate a risk of sagging, the weight of linearity or edge smoothness in the spatial morphology feature can be increased. This weighted model can be a multilayer perceptron network, whose input includes the paint formulation and process parameters, and whose output is the weight coefficients of each feature. Finally, based on the geometrically corrected and weighted spectral, polarization, and spatial morphology features, a multimodal fusion classifier is used to obtain the defect type and optical properties of non-uniform defects. This classifier can be a support vector machine or a deep learning network, whose training data includes corrected and weighted feature samples of different defect types under different formulation, process, and geometric conditions. For example, for pigment agglomeration defects, the corrected spectral characteristics may show absorption peaks in specific wavelength bands, the weighted polarization characteristics may show low polarization degree, and the spatial morphology characteristics may show irregular clumps. By integrating these optimized features, the classifier can accurately determine the specific type of defect and output its corresponding optical property parameters, such as local reflectivity, polarization degree, or scattering coefficient.

[0047] This application acquires coating formulation information, spraying process parameters, and local geometric information of defect areas. Based on this, it performs geometric correction and weighting processing on spectral features, polarization features, and spatial morphological features. This effectively eliminates the distortion effects of complex substrate geometry on image features and dynamically adjusts the importance of features according to the coating's inherent properties and production conditions. This makes the acquired defect types and optical properties of non-uniform defects more accurate and robust, thus solving the problems of insufficient accuracy and poor robustness when directly using original features to determine defect types and optical properties under complex substrates and variable production conditions. It provides high-quality input for the adaptive adjustment of subsequent detection parameters.

[0048] In some preferred embodiments, step S6 includes: S61. Collect coating images using aluminum pigment thickness uniformity analysis and detection equipment; S62. Identify the second non-uniformity defect region and the second non-defect region based on the coating image; S63. Obtain pixel-level optical response data of the second non-defective region; S63. Analyze the local optical response mode and noise characteristics of the background of the aluminum pigment coating based on the pixel-level optical response data of the second non-defect region; S64. Based on the local optical response mode and noise characteristics, perform background suppression and noise filtering on the optical response data of the second non-uniform defect region to obtain the defect optical response data. S65. Extract non-uniform defect features based on defect optical response data.

[0049] Local optical response mode refers to the distribution of optical properties such as light intensity, color, and texture exhibited by pixels in the defect-free area of ​​an aluminum pigment coating under specific lighting and observation conditions. This embodiment can use statistical analysis, texture analysis, or machine learning models to obtain the local optical response mode. Noise characteristics refer to random or systematic interference introduced during image acquisition, such as sensor noise, ambient light fluctuations, or speckle caused by the coating's microstructure. These can be quantified using noise model analysis or frequency domain analysis. Background suppression refers to eliminating or reducing the interference of the optical response of non-defective areas in the image on the defect signal through algorithms, thereby highlighting the visual characteristics of the defect itself. This can be achieved using background subtraction, adaptive thresholding, or image registration techniques. Noise filtering refers to removing irrelevant random noise from the image using digital image processing techniques to improve the signal-to-noise ratio of the defect signal and image clarity. This can be achieved using methods such as Gaussian filtering, median filtering, or wavelet transform.

[0050] This solution, after acquiring coating images using an aluminum pigment thickness uniformity analysis and detection device, overcomes the challenges of defect feature extraction posed by the complexity of the coating background and noise interference. First, a preliminary analysis of the acquired coating images is performed to identify a second non-uniformity defect region and a second non-defect region. This distinction forms the basis for subsequent processing, dividing the image into potential defect regions to be analyzed and reference regions that can be used to model the normal background. Subsequently, pixel-level optical response data is obtained from the identified second non-defect region, representing the true optical performance of the coating in a defect-free state. Based on this pixel-level optical response data, the system can deeply analyze and establish local optical response patterns and noise characteristic models of the aluminum pigment coating background. This analysis allows subsequent background suppression and noise filtering to be targeted, rather than simply applying general filters. Specifically, by modeling the background patterns, background components can be accurately extracted from the optical response of the defect region; by quantifying the noise characteristics, random interference introduced during image acquisition can be effectively filtered out. Finally, using the local optical response patterns and noise characteristics obtained from these analyses, background suppression and noise filtering are performed on the optical response data of the second non-uniformity defect region, thereby obtaining clean defect optical response data. The processed data significantly improves the signal-to-noise ratio of the defect signal, enabling more reliable extraction of non-uniform defect features, such as defect shape, size, edge sharpness, or intensity. Through this refined processing, the proposed solution effectively addresses the challenge of accurately extracting defect features in complex backgrounds and noisy environments. Combined with the previously mentioned steps of acquiring images and obtaining defect features using adjusted detection equipment, this method for analyzing the uniformity of aluminum pigment thickness not only captures defect information using specific optical parameters but also, through image processing, precisely separates the defect signal from complex backgrounds and noise. This significantly improves the accuracy and reliability of defect feature acquisition, thereby enhancing the stability and practicality of the entire detection method.

[0051] In a specific embodiment, the method for analyzing and detecting the uniformity of aluminum pigment thickness can be implemented as follows: First, after adjusting the aluminum pigment thickness uniformity analysis and detection equipment, the equipment is used to acquire images of the aluminum pigment coating surface, obtaining coating images. Next, preliminary image segmentation processing is performed on the acquired coating images. For example, methods based on region growing or cluster analysis are used to identify second non-uniformity defect regions in the image that exhibit abnormal brightness, color, or texture, while simultaneously identifying second non-defect regions that appear normal and uniform. Subsequently, optical response data is extracted pixel-by-pixel from these second non-defect regions. This data may include the RGB value, brightness value, or spectral intensity value of each pixel. Based on this pixel-level optical response data, a local background model is constructed. For example, the local optical response pattern of the coating background is characterized by calculating the average brightness, standard deviation, or texture descriptor of different local windows within the second non-defect regions. Simultaneously, statistical analysis (e.g., through histogram analysis or Fourier transform) can be performed on the pixel values ​​of these regions to determine the noise characteristics of the image (e.g., the distribution of Gaussian noise or salt-and-pepper noise). After obtaining the local optical response pattern and noise characteristics, the optical response data of the second non-uniform defect region undergoes refined processing. For example, an adaptive background subtraction algorithm can be used to subtract background components from the defect region based on the local optical response pattern, making the defect signal more prominent. Simultaneously, non-local mean filtering or wavelet thresholding denoising algorithms can be applied to remove random noise from the image based on noise characteristics, obtaining cleaner defect optical response data. Finally, based on these defect optical response data processed by background suppression and noise filtering, image processing algorithms, such as morphological operations, connected component analysis, or feature point detection, can be used to extract the geometric features (such as area, perimeter, aspect ratio), optical intensity features (such as average brightness, contrast), or texture features of the non-uniform defects, providing a basis for subsequent defect assessment.

[0052] This scheme first identifies potential defect areas and normal non-defect areas after acquiring coating images. Pixel-level optical response data is then obtained from the non-defect areas to analyze the local optical response patterns and noise characteristics of the aluminum pigment coating background. Based on this, background suppression and noise filtering are applied to the optical response data of defect areas to obtain clean defect optical response data. This processing method effectively eliminates the inherent high reflectivity of the aluminum pigment coating, angle-dependent color variations, and background texture interference caused by the complex arrangement of pigment flakes, while also removing random noise introduced during image acquisition. Therefore, this scheme can more accurately identify and quantify true defects in images, avoiding misjudging the coating's normal optical performance or background noise as defects. It significantly improves the accuracy and stability of non-uniform defect feature extraction, providing high-quality input for subsequent defect classification, quantification, and diagnosis.

[0053] Secondly, such as Figure 2 As shown, this application also provides an aluminum pigment thickness uniformity analysis and detection system, which includes: The coating data acquisition module 1 is used to acquire multispectral and polarization images of the aluminum pigment coating and to acquire geometric information of the substrate on which the aluminum pigment coating is located. Defect analysis module 2 is used to analyze whether there are non-uniform defects in the aluminum pigment coating based on multispectral and polarization images; The defect information acquisition module 3 is used to acquire the defect type and optical characteristics of non-uniform defects based on multispectral images and polarization images when non-uniform defects exist. Parameter acquisition module 4 is used to acquire target light source parameters and image sensor target receiving angle based on defect type, optical characteristics and geometric shape information; Parameter adjustment module 5 is used to adjust the light source and image sensor of the aluminum pigment thickness uniformity analysis and detection equipment according to the target light source parameters and the target receiving angle of the image sensor; The defect feature acquisition module 6 is used to acquire coating images using aluminum pigment thickness uniformity analysis and detection equipment, and to acquire non-uniform defect features based on the coating images.

[0054] The aluminum pigment thickness uniformity analysis and detection system provided in this application includes a coating data acquisition module 1, a defect analysis module 2, a defect information acquisition module 3, a parameter acquisition module 4, a parameter adjustment module 5, and a defect feature acquisition module 6. The aluminum pigment thickness uniformity analysis and detection system provided in this embodiment is used to perform the steps in the aluminum pigment thickness uniformity analysis and detection method provided in the first aspect above. The principle of the aluminum pigment thickness uniformity analysis and detection system provided in this embodiment is the same as the principle of the aluminum pigment thickness uniformity analysis and detection method provided in the first aspect above, and will not be discussed in detail here.

[0055] As can be seen from the above, the aluminum pigment thickness uniformity analysis and detection method and system provided in this application acquires multispectral images, polarization images and geometric information of the coating and the substrate, and based on the type and optical characteristics of the non-uniformity defects obtained from the analysis, combined with the substrate geometric information, dynamically acquires and adjusts the light source parameters of the detection equipment and the receiving angle of the image sensor to achieve adaptive adjustment of the detection parameters. Therefore, this application can effectively solve the problems of missed detection and misjudgment in existing fixed optical configuration detection systems in complex production environments, thereby improving the accuracy and reliability of aluminum pigment thickness uniformity detection.

[0056] In the embodiments provided in this application, it should be understood that the disclosed apparatus and method can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of the above units is only a logical functional division, and there may be other division methods in actual implementation. Furthermore, multiple units or components may be combined or integrated into another robot, or some features may be ignored or not executed. Additionally, the coupling or direct coupling or communication connection shown or discussed may be through some communication interface; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0057] In addition, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0058] In this document, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, without necessarily requiring or implying any such actual relationship or order between these entities or operations.

[0059] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for analyzing and detecting the thickness uniformity of aluminum pigment, characterized in that, The method for analyzing and detecting the uniformity of aluminum pigment thickness includes the following steps: S1. Obtain multispectral and polarization images of the aluminum pigment coating, and obtain geometric information of the substrate on which the aluminum pigment coating is located; S2. Analyze whether there are non-uniform defects in the aluminum pigment coating based on the multispectral image and the polarization image; S3. When the non-uniformity defect exists, obtain the defect type and optical characteristics of the non-uniformity defect based on the multispectral image and the polarization image; S4. Obtain the target light source parameters and the target receiving angle of the image sensor based on the defect type, the optical characteristics, and the geometric shape information; S5. Adjust the light source and image sensor of the aluminum pigment thickness uniformity analysis and detection device according to the target light source parameters and the target receiving angle of the image sensor. S6. Use the aluminum pigment thickness uniformity analysis and detection equipment to acquire coating images, and obtain non-uniformity defect characteristics based on the coating images.

2. The method for analyzing and detecting the thickness uniformity of aluminum pigment according to claim 1, characterized in that, The target light source parameters include wavelength, illumination intensity, polarization state, and incident angle.

3. The method for analyzing and detecting the uniformity of aluminum pigment thickness according to claim 1, characterized in that, Step S4 includes: S41. Obtain the optical response information of the first non-defect region adjacent to the non-uniformity defect based on the multispectral image and the polarization image; S42. Correct the optical characteristics based on the difference between the optical characteristics and the optical response information to obtain corrected defect optical characteristics; S43. Obtain the target light source parameters and the target receiving angle of the image sensor based on the defect type, the optical characteristics of the corrected defect, and the geometric shape information.

4. The method for analyzing and detecting the thickness uniformity of aluminum pigment according to claim 3, characterized in that, Step S42 includes: S421. Obtain the coating formula information of the aluminum pigment coating; S422. Determine the difference analysis rules based on the coating formulation information and the defect type; S423. Obtain difference features based on the optical properties and optical response information using the difference analysis rules; S424. The optical properties are corrected according to the difference characteristics to obtain corrected defect optical properties.

5. The method for analyzing and detecting the thickness uniformity of aluminum pigment according to claim 4, characterized in that, Step S422 includes: A1. Determine preliminary analysis rules based on the paint formulation information and the defect type; A2. Obtain the spraying process parameter information of the aluminum pigment coating; A3. Adjust the preliminary analysis rules based on the spraying process parameter information to obtain the difference analysis rules.

6. The method for analyzing and detecting the thickness uniformity of aluminum pigment according to claim 3, characterized in that, Step S41 includes: S411. Obtain optical response data of the first non-defect region adjacent to the non-uniformity defect under multiple viewing angles and / or multiple illumination angles; S412. Obtain the first local geometric information of the first non-defect region; S413. Obtain the optical anisotropy characteristics of the first non-defect region based on the first local geometric information and all the optical response information; S414. Perform angle compensation on the optical response data according to the optical anisotropy characteristics to obtain optical response information.

7. The method for analyzing and detecting the thickness uniformity of aluminum pigment according to claim 1, characterized in that, Step S3 includes: S31. When the non-uniformity defect exists, identify the first non-uniformity defect region from the multispectral image and the polarization image, and obtain pixel-level spectral response data and polarization response data of the first non-uniformity defect region. S32. Extract the spectral and polarization features of the first non-uniformity defect region based on the pixel-level spectral response data and the polarization response data; S33. Perform image segmentation and morphological analysis on the first non-uniform defect region to obtain the spatial morphological features of the first non-uniform defect region, wherein the spatial morphological features include the shape, size, edge characteristics or linearity of the non-uniform defect. S34. Obtain the defect type and optical properties of the non-uniformity defect based on the spectral characteristics, polarization characteristics and spatial morphology characteristics.

8. The method for analyzing and detecting the thickness uniformity of aluminum pigment according to claim 7, characterized in that, Step S34 includes: S341. Obtain the coating formula information and spraying process parameter information of the aluminum pigment coating, and obtain the second local geometric information of the first non-uniformity defect region; S342. Perform geometric correction on the spectral features, polarization features and spatial morphological features based on the second local geometric information; S343. Weight the spectral features, polarization features, and spatial morphological features according to the coating formulation information and the spraying process parameter information; S344. Obtain the defect type and optical properties of the non-uniformity defect based on the spectral characteristics, polarization characteristics and spatial morphology characteristics.

9. The method for analyzing and detecting the thickness uniformity of aluminum pigment according to claim 1, characterized in that, Step S6 includes: S61. Acquire coating images using the aluminum pigment thickness uniformity analysis and detection equipment; S62. Identify the second non-uniformity defect region and the second non-defect region based on the coating image; S63. Obtain pixel-level optical response data of the second non-defective region; S63. Analyze the local optical response mode and noise characteristics of the background of the aluminum pigment coating based on the pixel-level optical response data of the second non-defect region; S64. Based on the local optical response mode and the noise characteristics, perform background suppression and noise filtering on the optical response data of the second non-uniform defect region to obtain defect optical response data. S65. Extract non-uniformity defect features based on the defect optical response data.

10. A system for analyzing and detecting the thickness uniformity of aluminum pigments, characterized in that, The aluminum pigment thickness uniformity analysis and detection system includes: The coating data acquisition module is used to acquire multispectral and polarization images of the aluminum pigment coating and to acquire geometric information of the substrate on which the aluminum pigment coating is located. The defect analysis module is used to analyze whether there are non-uniform defects in the aluminum pigment coating based on the multispectral image and the polarization image; The defect information acquisition module is used to acquire the defect type and optical characteristics of the non-uniformity defect based on the multispectral image and the polarization image when the non-uniformity defect exists. The parameter acquisition module is used to acquire target light source parameters and image sensor target receiving angle based on the defect type, optical characteristics and geometric shape information; The parameter adjustment module is used to adjust the light source and image sensor of the aluminum pigment thickness uniformity analysis and detection device according to the target light source parameters and the target receiving angle of the image sensor. The defect feature acquisition module is used to acquire coating images using the aluminum pigment thickness uniformity analysis and detection equipment, and to acquire non-uniformity defect features based on the coating images.

Images