On-line detection method and system for coating quality of automobile skylight glass

Abstract

The invention relates to the technical field of coating detection, and provides an on-line detection method and system for the coating quality of automobile skylight glass. The method comprises the following steps: capturing a surface image of the automobile skylight through a high-resolution industrial camera to obtain a coating image data set; positioning the automobile sunroof, and performing defect identification on the coating image data set according to the position of the sunroof to obtain an image defect label group; performing classification according to the image defect label group, constructing a defect list, performing automatic sorting according to the defect list, and determining a coating sorting result; and carrying out quality detection on the automobile skylight glass coating according to the coating process information in combination with the coating sorting result, and generating a coating quality report. The technical problems of insufficient defect identification precision and inaccurate quality detection caused by glass reflection and shadow interference in automobile skylight glass coating quality detection are solved, and the technical effects of high-precision and high-efficiency automobile skylight glass coating quality detection and automatic sorting, and improvement of process stability and product quality are achieved.

Description

Technical Field

[0001] The present application relates to the technical field of coating detection, and in particular to an online detection method and system for coating quality of automobile sunroof glass. Background Art

[0002] As an important part of the car, the coating quality of the car sunroof glass directly affects the light transmittance, weather resistance and aesthetics of the sunroof. With the rapid development of the automobile industry, consumers have increasingly higher requirements for the quality of car sunroof glass, especially in terms of coating uniformity and defect detection accuracy. Traditional coating quality inspection methods mainly rely on manual visual inspection or offline sampling inspection, which has the problems of low efficiency, strong subjectivity and high missed detection rate, and it is difficult to meet the high-precision and high-efficiency requirements of modern automobile manufacturing.

[0003] In addition, the surface of the car sunroof glass is easily disturbed by environmental factors such as reflections and shadows, resulting in insufficient defect recognition accuracy, further affecting the accuracy of the test results. At the same time, the optimization and feedback mechanism of the coating process parameters is not perfect, making it difficult to achieve real-time monitoring and process adjustment of quality problems, increasing the quality risk in the production process.

[0004] Therefore, there is an urgent need for a method and system that can detect the coating quality of automobile sunroof glass online in real time. Through high-precision image acquisition, intelligent defect recognition, automatic sorting and process optimization, the shortcomings of traditional detection methods can be solved, the accuracy and efficiency of coating quality detection can be improved, and product quality and process stability can be ensured. Summary of the invention

[0005] The present application provides an online detection method and system for the coating quality of automobile sunroof glass, aiming to solve the technical problems of insufficient defect recognition accuracy and inaccurate quality detection caused by glass reflection and shadow interference in the coating quality detection of automobile sunroof glass.

[0006] The first aspect disclosed in the present application provides an online detection method for the coating quality of automobile sunroof glass, the method comprising: capturing a surface image of an automobile sunroof by a high-resolution industrial camera to obtain a coating image data set; positioning the automobile sunroof, and performing defect identification on the coating image data set according to the sunroof position information to obtain an image defect label group; classifying according to the image defect label group to construct a defect list, automatically sorting according to the defect list, and determining a coating sorting result; performing quality inspection on the automobile sunroof glass coating according to the coating process information combined with the coating sorting result, and generating a coating quality report for the automobile sunroof glass.

[0007] Another aspect disclosed in the present application provides an online detection system for the coating quality of automobile sunroof glass, the system comprising: an image capture module: capturing the surface image of the automobile sunroof through a high-resolution industrial camera to obtain a coating image data set; a defect recognition module: locating the automobile sunroof, and performing defect recognition on the coating image data set according to the sunroof position information to obtain an image defect label group; an automatic sorting module: classifying according to the image defect label group, constructing a defect list, automatically sorting according to the defect list, and determining a coating sorting result; a quality inspection module: performing quality inspection on the automobile sunroof glass coating according to the coating process information combined with the coating sorting result, and generating a coating quality report for the automobile sunroof glass.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages: The above-mentioned method for online detection of coating quality of automobile sunroof glass firstly captures the surface image of the automobile sunroof through a high-resolution industrial camera to obtain a coating image data set, then locates the automobile sunroof, and performs defect recognition on the coating image data set according to the sunroof position information to obtain an image defect label group, then classifies according to the image defect label group, constructs a defect list, automatically sorts according to the defect list, determines the coating sorting result, and finally, performs quality detection of automobile sunroof glass coating according to the coating process information combined with the coating sorting result to generate a coating quality report of the automobile sunroof glass. The method achieves high-precision and high-efficiency automobile sunroof glass coating quality detection and automatic sorting, and improves process stability and product quality.

[0009] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0011] Figure 1 The figure is a flow chart of an online detection method for coating quality of automobile sunroof glass in one embodiment.

[0012] Figure 2 The present invention is an architecture diagram of an online detection system for coating quality of automobile sunroof glass in one embodiment.

[0013] Explanation of the reference numerals: image capturing module 11 , defect recognition module 12 , automatic sorting module 13 , quality inspection module 14 . DETAILED DESCRIPTION

[0014] The embodiments of the present application provide a method and system for online detection of automobile sunroof glass coating quality, thereby solving the technical problems of insufficient defect recognition accuracy and inaccurate quality detection caused by glass reflection and shadow interference in automobile sunroof glass coating quality detection.

[0015] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0016] It should be noted that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or inherent to these processes, methods, products or devices.

[0017] Embodiment 1, as Figure 1 As shown, the present application provides a method for online detection of coating quality of automobile sunroof glass, the method comprising: The surface images of the car sunroof are captured by a high-resolution industrial camera to obtain the coating image dataset.

[0018] In the embodiment of the present application, in the process of online detection of the coating quality of the automobile sunroof glass, it is first necessary to obtain high-precision surface image data, which is the basis for subsequent defect identification and quality inspection. In order to achieve this goal, a high-resolution industrial camera is used for image capture. Specifically, in order to ensure the clarity of image acquisition and reduce environmental interference, a multi-angle LED light source is first equipped. These light sources can be adjusted according to different lighting conditions to optimize the image acquisition effect, and then based on the preset multiple lighting conditions, an initial lighting condition is selected to activate the high-resolution industrial camera. When the automobile sunroof enters the detection area, the coating area is covered from multiple angles by a high-resolution industrial camera to obtain preliminary image data. This process ensures the comprehensiveness and detail capture ability of the image data. Subsequently, according to the preliminary collected image data, the interference factors in the image, such as shadows and reflections, are analyzed, and based on these analysis results, the angle and intensity of the light source are adjusted to optimize the lighting conditions to further improve the image quality. Under the optimized lighting conditions, the high-resolution industrial camera again covers the coating area with multiple angles, and finally obtains a high-quality coating image data set. These image data will serve as the basis for subsequent defect identification and quality inspection. Through the above steps, high-quality image data of the coating area of ​​the automobile sunroof glass can be obtained efficiently and accurately, providing reliable data support for subsequent defect identification and quality inspection.

[0019] Furthermore, the present application provides a method for capturing a surface image of a car sunroof by a high-resolution industrial camera to obtain a coating image dataset, the method comprising: A multi-angle LED light source is used to set multiple lighting conditions, and a first lighting condition is selected based on the multiple lighting conditions to activate a high-resolution industrial camera; when a car sunroof enters a detection area, the surface of the car sunroof is traversed to perform coating marking and determine a coating area; a high-resolution industrial camera is used to perform multi-angle coverage acquisition of the coating area to obtain a first acquisition result; the first lighting condition of the multi-angle LED light source is adjusted according to the first acquisition result, a target lighting condition is set, and a high-resolution industrial camera is activated according to the target lighting condition to perform multi-angle coverage acquisition of the coating area to obtain a coating image data set.

[0020] Preferably, in order to ensure the accuracy and comprehensiveness of image acquisition, a multi-angle LED light source is used in combination with a high-resolution industrial camera for image capture. Specifically, the equipped multi-angle LED light source can provide a variety of different lighting conditions. These lighting conditions are the default working modes of the multi-angle LED light source. It is designed to ensure that the surface of the car sunroof glass can obtain a uniform lighting effect at different angles and intensities, thereby reducing the interference of shadows and reflections on image acquisition, and has a certain universality. In addition, the multi-angle LED light source also has an adjustment function. By adjusting the angle and intensity of the light source, different lighting environments can be simulated, providing flexibility and adaptability for subsequent image acquisition. At the beginning of the detection, an initial lighting condition will be selected from multiple preset lighting conditions as the first lighting condition according to the current business needs and the surrounding environment, and a high-resolution industrial camera will be activated. The purpose of this step is to quickly obtain preliminary image data of the sunroof glass surface at the beginning of the detection, providing a basis for subsequent optimization. When the car sunroof enters the detection area, the sunroof surface will be fully scanned by the image sensor, and then the collected image will be analyzed by image processing algorithms (such as edge detection, region segmentation, etc.) to identify the boundary between the coated area and the non-coated area. In this process, the Canny or Sobel algorithm will be used to extract the edge, preliminarily locate the boundary of the coated area, and then the threshold segmentation will be used to divide the image into multiple areas. For each area, the color (HSV or Lab space can be used for extraction), texture (GLCM or Gabor filter can be used for extraction, and contrast, energy, entropy, etc. can be obtained), reflection characteristics (grayscale value or RGB brightness component can be used for extraction, and brightness value, highlight area ratio, diffuse reflection area, etc. can be obtained) and other features are extracted. By comparing these features with the preset coating area feature range, it is determined whether the features of each area are within the coating area feature range. If the features of a certain area are all within the coating area feature range, it means that the area is a coating area. At this time, this area will be marked. This process ensures the pertinence and comprehensiveness of subsequent image acquisition to avoid missing any potential defective areas. After determining the coating area, the high-resolution industrial camera will perform coverage acquisition of the coating area from multiple angles to obtain preliminary image data, namely the first acquisition result. This step ensures the comprehensiveness and detail capture capability of the image data, and provides a basis for the subsequent optimization of lighting conditions. Subsequently, the first acquisition result will be analyzed, focusing on interference factors such as shadows and reflections in the image to determine whether the current lighting conditions meet the current detection requirements. In most cases, due to differences in the surrounding environment, the default working mode of the multi-angle LED light source will have a certain deviation. Therefore, based on the analysis results, the angle and intensity of the multi-angle LED light source will be adjusted to optimize the lighting conditions, and the optimized first lighting condition will be set as the target lighting condition. The purpose of this step is to eliminate or reduce interference in the image and improve image quality.Under the optimized target lighting conditions, the high-resolution industrial camera once again covers the coating area at multiple angles, and finally obtains a high-quality coating image data set, which will serve as the basis for subsequent defect identification and quality inspection to ensure the accuracy and reliability of the inspection results. Through the above process, high-quality image data of the coating area of ​​the automobile sunroof glass can be obtained efficiently and accurately under different lighting conditions, providing reliable data support for subsequent defect identification and quality inspection.

[0021] Furthermore, the present application provides a method for adjusting the first illumination condition of the multi-angle LED light source according to the first acquisition result to set a target illumination condition, the method comprising: Based on the first acquisition result, interference analysis is performed to determine multiple light source interference data, which include shadow interference data and reflection interference data; according to the shadow interference data combined with the coating characteristics, the high-resolution industrial camera is collected and adjusted to determine the lighting adjustment angle information; according to the reflection interference data combined with the coating characteristics, the high-resolution industrial camera is collected and adjusted to determine the lighting adjustment intensity information; based on the lighting adjustment angle information and the lighting adjustment intensity information, the first lighting condition is adjusted to set the target lighting condition.

[0022] Optionally, in order to optimize the image acquisition effect, interference analysis is performed based on the first acquisition result, and the target lighting condition is set by adjusting the lighting condition. Specifically, the first acquisition result acquired by the high-resolution industrial camera is analyzed in detail to identify and quantify multiple light source interference data, which include at least shadow interference data and reflection interference data. For shadow interference data, it is achieved by identifying areas in the image where the brightness is significantly lower than that of the surrounding areas, that is, the result of the brightness of the surrounding area minus the brightness of the position is greater than the brightness deviation tolerance, indicating that there is a shadow at that position. At this time, the center position, area, light intensity and other parameters of the shadow are recorded to form the shadow interference data; for reflection interference data, it is achieved by identifying areas in the image where the brightness is significantly higher than that of the surrounding areas, that is, the result of the brightness of the position minus the brightness of the surrounding area is greater than the brightness deviation tolerance, indicating that there is a reflection at that position. At this time, the center position, area, light intensity and other parameters of the reflection are recorded to form the reflection interference data. Subsequently, the coordinates of the center position of the shadow area are obtained from the shadow interference data, and the distance d between the center of the shadow and the center of the coating area is calculated by the Euclidean distance, and then the formula is used to calculate the distance d between the center of the shadow and the center of the coating area. Calculate the angle that needs to be adjusted, where is the angle increment, and D is the distance between the light source and the coating area. Afterwards, the adjustment direction is determined in combination with the coating characteristics, that is, the positional relationship between the shadow and the coating area is analyzed. If the shadow is on the left side of the coating area, it means that the light source needs to be adjusted to the right. If the shadow is on the right side of the coating area, it means that the light source needs to be adjusted to the left. If the shadow is above or below the coating area, it means that the light source needs to be adjusted downward or upward. By combining the angle increment and the adjustment direction, the lighting adjustment angle information is generated. Then, the reflected light intensity is extracted from the reflection interference data, and the ratio of the reflected light intensity to the target brightness value is calculated to obtain the light intensity adjustment ratio, and then the light intensity adjustment ratio is multiplied by the current light intensity to obtain the light adjustment intensity information. Finally, based on the illumination adjustment angle information and illumination adjustment intensity information, the first illumination condition is optimized and adjusted. In this process, the position and angle of the light source are adjusted according to the illumination adjustment angle information, for example, the light source is adjusted from 30° to 45°. The brightness of the light source is also adjusted according to the illumination adjustment intensity information, for example, the light source intensity is reduced from 1000lux to 800lux. After the adjustment is completed, the final illumination angle and intensity are recorded as the target illumination condition for subsequent image acquisition. Through the above process, interference analysis can be performed based on the first acquisition result, and the target illumination condition can be set by adjusting the illumination angle and intensity, thereby optimizing the image acquisition effect and providing high-quality data support for subsequent defect identification and quality inspection.

[0023] The automobile sunroof is located, and defects are identified on the coating image data set according to the sunroof position information to obtain an image defect label group.

[0024] In one embodiment, after locating the car sunroof through infrared sensors and laser sensors, defects are identified on the coating image dataset based on the position information of the sunroof. In this process, deep learning is performed on the coating image dataset in combination with the three-dimensional model of the sunroof to generate image learning results. The historical coating defect log is then retrieved and compared with the image learning results to identify possible defects. By matching and dividing these defects, an image defect label group is finally formed to provide a basis for subsequent classification and sorting.

[0025] Furthermore, the present application provides a method for obtaining the sunroof position information, comprising: A first sensor is started to perform sensor analysis on a coating area to obtain a first sensor data set; a second sensor is started to perform sensor analysis on the coating area to obtain a second sensor data set; a spatial rectangular coordinate system is constructed based on the coating area, and the first sensor data set and the second sensor data set are mapped to the spatial rectangular coordinate system to obtain a three-dimensional model of the coating of the automobile sunroof glass, wherein the three-dimensional model includes the sunroof position information.

[0026] Optionally, before defect identification, the first sensor and the second sensor are activated to perform sensing analysis on the coating area, wherein the first sensor is an infrared sensor and the second sensor is a laser sensor. For the first sensor, this sensor is used to scan the coating area of ​​the automobile sunroof glass, and by detecting the difference in thermal radiation between the coating material and the glass, the range of the coating area is analyzed to determine whether there is coating on the upper part of the sunroof or whether the entire sunroof is completely coated, thereby generating the position information of the coating area and forming a first sensing data set. For the second sensor, this sensor is used to scan the coating area, and by emitting a laser beam and receiving a reflected signal, the thickness information of the sunroof glass is measured to form a second sensing data set. Subsequently, based on the physical space of the coating area, a three-dimensional rectangular coordinate system is constructed in the 3D modeling software. The X, Y, and Z axes of the coordinate system correspond to the length, width, and height of the coating area, respectively. The coating area position information in the first sensing data set is then mapped to the spatial rectangular coordinate system to determine the spatial distribution of the coating area in the X and Y directions. The thickness information in the second sensing data set is then mapped to the spatial rectangular coordinate system to determine the thickness change of the coating area in the Z direction. Afterwards, based on the mapping results in the X, Y, and Z axes, a three-dimensional model of the coating of the automobile sunroof glass is generated through the 3D modeling software. The three-dimensional model contains the spatial distribution of the coating area, the thickness change, and the position information of the sunroof, providing an accurate spatial reference for subsequent defect identification and quality inspection.

[0027] Furthermore, the present application provides a method for performing defect recognition on the coating image data set according to the skylight position information to obtain an image defect label group, the method comprising: Based on the skylight position information, the coating image dataset and the three-dimensional model are deeply learned to generate image learning results; the historical coating defect log is retrieved, the image learning results are traversed and combined with the historical coating defect log to perform defect identification to generate multiple defect identification results; the coating image dataset is matched and divided according to the multiple defect identification results to obtain the image defect label group.

[0028] Optionally, based on the skylight position information, the coating image dataset is aligned with the three-dimensional model to ensure that the image data is spatially consistent with the coating area in the three-dimensional model. The alignment process is performed using camera calibration technology (such as Zhang Zhengyou calibration method). After alignment, the aligned coating image dataset and the three-dimensional model are used as input data, and analyzed using a pre-built convolutional neural network (CNN) or other deep learning model. The coating image dataset is inferred through the learned mapping relationship to generate image learning results, including the texture, thickness change, surface features, etc. of the coating area. Taking the convolutional neural network as an example, multiple sets of historical coating image datasets, multiple historical three-dimensional models, and multiple sets of historical feature analysis results are collected, and these data are divided to form training sets and validation sets. Subsequently, CNN is used to construct an initial network structure, including input layer, output layer, convolution layer, pooling layer, full connector, etc., and then the weights of CNN are initialized using random numbers, etc., and the training set is input into the initialized CNN for forward propagation. Through layer-by-layer transmission, the prediction results including multiple features are calculated. After that, the mean square error loss function is used to calculate the loss value between the prediction result and the historical feature analysis result, and the gradient of the loss to the weight of each layer is calculated layer by layer through back propagation. Then the Adam optimizer is used to optimize the network parameters, and the weights are adjusted to minimize the value of the loss function. The above process is repeated until the maximum number of iterations is reached. After the training is completed, the validation set is used to test the network performance and evaluate the accuracy of the network. If the accuracy meets the expected accuracy, the current convolutional neural network is output. Otherwise, the hyperparameters such as the learning rate and the number of training batches are adjusted to further improve the effect of the convolutional neural network. After obtaining the image learning results, the historical coating defect logs are retrieved from the database, including the type, location, feature description and other information of defects such as uneven coating, bubbles, scratches, and foreign objects. The image learning results are then traversed, and the traversal results are compared and analyzed with the historical defect logs. The defects that may exist in the image are identified through similarity calculation. Specifically, the features of the image learning results are spliced ​​into image feature vectors, and the features of each coating defect in the historical coating defect logs are spliced ​​into defect feature vectors. The cosine similarity calculation formula is then used to calculate the similarity between the image feature vector and each defect feature vector. If the calculated similarity is greater than or equal to the defect recognition threshold, it means that such defects exist in the image learning results. At this time, the defect will be recorded as a defect recognition result. After comparing with each defect, the defect recognition results obtained will be summarized to obtain multiple defect recognition results, including the defect type (such as uneven coating, bubbles, scratches, foreign objects, etc.) and its location and features in the image. Then, according to the defect recognition results, the images in the coating image dataset are divided into different defect categories, and a corresponding defect label is generated for each image to form an image defect label group.Through the above process, the specific process from deep learning to defect recognition to label generation is completed, providing basic data for subsequent image classification and quality inspection.

[0029] Classification is performed according to the image defect label group, a defect list is constructed, automatic sorting is performed according to the defect list, and a coating sorting result is determined.

[0030] In one embodiment, after obtaining the image defect label group, cluster analysis and impact analysis are performed on the image defect label group to construct a defect list, which contains the type and severity of each defect. Subsequently, the automobile sunroof glass and coating image data set are classified according to the defect list to determine the coating sorting result. The coating sorting result records the next process that each automobile sunroof glass needs to enter, providing data support for subsequent quality inspection.

[0031] Furthermore, the present application provides a method for classifying according to the image defect label group, constructing a defect list, automatically sorting according to the defect list, and determining a coating sorting result, the method comprising: The coating image data set is clustered and analyzed according to the image defect label group to determine a plurality of coating defect classes; an impact analysis is performed on the automobile sunroof glass according to the plurality of coating defect classes to set a plurality of defect impact levels; the plurality of coating defect classes are associated with the plurality of defect impact levels to construct a defect list; the automobile sunroof glass is classified according to the defect list to determine a target automobile sunroof glass, and target coating image data of the target automobile sunroof glass is extracted according to the coating image data set; the target automobile sunroof glass is automatically sorted according to the target coating image data to obtain a coating sorting result.

[0032] Preferably, after obtaining the image defect label group, the coating image data set will be clustered according to the image defect label group, that is, the coating image data with the same label will be stored in one data set to form multiple coating defect classes. Subsequently, an impact analysis is performed on each coating defect class, that is, according to the relevant industry standards for automotive glass manufacturing and coating processes (such as ISO, ASTM, etc.), combined with expert decisions, the impact level of each coating defect class is determined, and a defect impact level is assigned to it, for example, the impact level of bubble defects is medium and the impact level of stain defects is severe. Afterwards, each coating defect class is associated with its corresponding defect impact level, such as bubble defect → medium, stain defect → severe, etc., and summarized into a defect list for subsequent classification and sorting. Then, the automobile sunroof glass is classified according to the defect type in the defect list, that is, the automobile sunroof glass corresponding to the coating image data with the same label is put into one set, and the automobile sunroof glass without label (no defect) is put into another set. After the division of the automobile sunroof glass is completed, these automobile sunroof glasses with clear classification will be used as target automobile sunroof glasses. In addition, according to the target automobile sunroof glass, the image data in the coating image data set is classified to form target coating image data, and this target coating image data corresponds to the target automobile sunroof glass one by one. Finally, according to the defect impact level and target coating image data in the defect list, the target automobile sunroof glass is sorted, and the target automobile sunroof glass is transmitted to the first process area or the second process area by automated equipment (such as a robotic arm, a conveyor belt, etc.), and the sorting results are recorded to form coating sorting results, thereby improving the detection efficiency of automobile sunroof glass coating.

[0033] Furthermore, the present application automatically sorts the target automobile sunroof glass according to the target coating image data to obtain a coating sorting result, and the method includes: Determine whether the target coating image data is in the defect list; if the target coating image data is in the defect list, generate a primary sorting instruction, and sort the target automobile sunroof glass to a first process area through the primary sorting instruction; if the target coating image data is not in the defect list, generate a secondary sorting instruction, and sort the target automobile sunroof glass to a second process area through the secondary sorting instruction.

[0034] Optionally, when performing automatic sorting, it is first detected whether the target coated image data has an image defect label. If so, the image defect label of the coated image data will be matched with the defect list to obtain the defect impact level corresponding to the defect, and a first-level sorting instruction is generated according to the defect impact level. If the defect impact level is severe, then the first-level sorting instruction will instruct the automated equipment to sort the target automobile sunroof glass to the scrap area in the first process area for scrapping. If the defect impact level is slight or medium, then the first-level sorting instruction will instruct the automated equipment to sort the target automobile sunroof glass to the rework area in the first process area for repair. If the target coating image data does not have an image defect label, it means that the target car sunroof glass is a normal car sunroof glass. At this time, a secondary sorting instruction will be generated. This secondary sorting instruction will first instruct the automated equipment to randomly sample the target car sunroof glass, and sort the sampled target car sunroof glass to the manual re-inspection area of ​​the second process area. The quality inspectors will re-inspect the product to confirm whether it is qualified. When the quality inspectors feedback that it is qualified, the target car sunroof glass will be sorted to the next process area of ​​the second process area, enter the subsequent production process, and continue processing or assembly. This method combines automated inspection and manual re-inspection, which not only improves inspection efficiency but also ensures product quality.

[0035] The coating quality of the automobile sunroof glass is inspected according to the coating process information and the coating sorting result, and a coating quality report of the automobile sunroof glass is generated.

[0036] In one embodiment, according to the target automobile sunroof glass in the first process area, the coating process information is subjected to a risk assessment of process consistency, and whether there is a deviation or abnormality is identified, and coating process risk data is generated. These risk data include coating thickness deviation, abnormal curing temperature, insufficient curing time, and other issues, providing a basis for subsequent analysis. Subsequently, the coating sorting results are continuously statistically analyzed according to the coating process risk data to determine the high-frequency defect mode, and this high-frequency defect mode is associated with the coating machine parameter timing data. Afterwards, by analyzing the operating parameters of the coating machine when the defect occurs (such as nozzle pressure, conveyor belt speed, coating material flow, etc.), the parameter combination that may cause the defect is extracted as a parameter abnormal combination, and the parameter abnormal combination is added to the coating quality report as the core analysis result of the quality inspection. At the same time, it will also determine whether to generate an alarm signal based on the optimization of the parameter abnormal combination, so as to prompt the operator to deal with the abnormal situation in time and avoid further defects. Through this process, the coating quality report can not only fully reflect the current production quality status, but also provide a scientific basis for process optimization, thereby improving the overall quality level of automobile sunroof glass coating.

[0037] Furthermore, the present application performs quality inspection on the coating of the automobile sunroof glass according to the coating process information combined with the coating sorting result, and generates a coating quality report of the automobile sunroof glass. The method includes: A risk assessment of process consistency is performed on the coating process information according to the target automobile sunroof glass in the first process area to generate coating process risk data; the coating sorting results are continuously counted according to the coating process risk data to determine the high-frequency defect mode; the high-frequency defect mode is associated with the coating machine parameter timing data to obtain a plurality of parameter combinations, and the plurality of parameter combinations are mined through a random forest algorithm to obtain abnormal parameter combinations; the abnormal parameter combinations are added to the coating quality report, and the coating process information is feedback optimized according to the abnormal parameter combinations to generate an alarm signal.

[0038] Optionally, the target automobile sunroof glass in the first process area is taken as the object, and its actual coating process information is extracted, including key parameters such as coating thickness, curing temperature, curing time, coating material ratio, nozzle pressure, and coating material flow rate. The actual coating process information is then compared with the coating process information specified according to business needs to identify whether there are deviations or anomalies, for example, whether the coating thickness is within the allowable range, whether the curing temperature deviates from the standard temperature range, whether the curing time meets the standard, etc. Based on the comparison results, the coating process risk data is generated, which records which parameters have deviations. Subsequently, the sorting results of N consecutive batches of inspection data are collected, including defect types, quantities and distributions, and the frequency of occurrence of each defect type is counted to identify high-frequency defect patterns, such as the number and proportion of stains, bubbles, scratches and other defects in N consecutive batches. Then, the high-frequency defect patterns are determined based on the statistical results. For example, if the bubble defect occurs frequently in multiple batches, it is marked as a high-frequency defect pattern. In addition, the high-frequency defect pattern is associated with the coating process risk data, and expert decision-making and historical experience are combined to determine which process parameter deviations may cause these defects. For example, bubble defects may be related to improper coating material ratio or unstable coating machine nozzle pressure. After that, the parameter time series data corresponding to the occurrence time of high-frequency defects are extracted from the coating machine operation record, such as curing temperature, curing time, coating material ratio, nozzle pressure, conveyor belt speed, etc., depending on which process parameter deviations analyzed may cause the generation of high-frequency defect patterns, and then the high-frequency defect patterns are associated with the parameter time series data to generate multiple parameter combinations. Then, the generated parameter combinations are analyzed using the pre-constructed random forest to identify abnormal parameter combinations that are significantly related to high-frequency defects. Then, the abnormal parameter combination and its corresponding defect mode are integrated into the coating quality report. Based on the abnormal parameter combination, the coating process information is optimized. Based on the optimization result, an alarm signal is generated in real time to prompt the operator to deal with it in time, ensuring the continuous improvement of the coating quality of automobile sunroof glass.

[0039] For the pre-built random forest, it is trained based on the historical parameter combination set, which has an abnormal identification. During the training process, the model is first initialized using the random forest classifier, and key parameters are set, such as the number of decision trees (such as 100 or 200), the maximum depth of each decision tree (such as 10 or None, indicating no depth limit), the minimum number of samples required for node splitting (such as 2 or 5), and the random seed (such as 42 to ensure that the results are reproducible). Subsequently, the historical parameter combination set is divided into a training set and a test set, usually divided in a ratio of 80% training set and 20% test set, to ensure that the ratio of normal and abnormal samples in the training set and test set is consistent with the original data set. After that, the random forest model is trained using the training set data. The model will construct multiple decision trees, each of which is trained based on a randomly selected feature subset and sample subset, and then the samples are classified through a voting mechanism. After the training is completed, the test set is used to evaluate the model performance, and indicators such as accuracy, recall, and F1 score are calculated to determine whether the model can effectively distinguish between normal and abnormal parameter combinations. Then, analyze the feature importance of the model, output the contribution ranking of each parameter to the classification result, identify the parameters that have the greatest impact on the high-frequency defect mode, and if the model performance is not ideal, adjust the parameters, for example, increase the number of decision trees, adjust the maximum depth, or reselect features, and then retrain the model. Finally, save the trained model for subsequent abnormal parameter combination prediction and analysis.

[0040] Furthermore, the present application performs quality inspection on the coating of the automobile sunroof glass according to the coating process information combined with the coating sorting result, and generates a coating quality report of the automobile sunroof glass. The method includes: Establish a coating quality-process parameter mapping matrix, extract the defect rate data of the abnormal parameter combination; traverse the defect rate data of the abnormal parameter combination to select the minimum value as the preferred parameter; gradually adjust the coating process information to the preferred parameter, and monitor the change of the defect rate data in real time during the adjustment process. If the defect rate data after adjustment does not drop to a preset threshold, trigger the alarm signal.

[0041] Optionally, a coating quality-process parameter mapping matrix is ​​established based on historical data. The rows of the matrix represent different process parameter combinations (such as coating thickness, nozzle pressure, conveyor belt speed, curing temperature, etc.), and the columns represent corresponding coating quality indicators (such as defect rate, thickness uniformity, etc.). By analyzing historical data, each parameter combination is associated with the corresponding defect rate data to form a complete mapping matrix, and then the defect rate data of each group of parameters in the abnormal parameter combination is extracted from the mapping matrix. Subsequently, by traversing the defect rates of the abnormal parameter combinations, the parameter group with a lower defect rate is obtained and used as the preferred parameter. The goal of the preferred parameter is to reduce the defect rate as much as possible while ensuring that other indicators of coating quality (such as thickness uniformity) are within an acceptable range. After that, starting from the current coating process parameters, each parameter (such as coating thickness, nozzle pressure, etc.) is gradually adjusted to be close to the value of the preferred parameter. After the adjustment is completed, the defect rate data is recorded and compared with the defect rate before the adjustment. If the defect rate after adjustment does not reach the preset threshold (such as a 5% or 10% reduction in defect rate), an alarm signal is triggered, indicating that the process parameter adjustment has failed to effectively improve the coating quality. During the entire adjustment process, real-time monitoring and data analysis are key. By continuously tracking changes in defect rates, the effects of process parameter adjustments can be quickly identified, and alarm signals can be triggered when necessary to avoid further deterioration of coating quality due to improper parameter adjustments. Ultimately, by optimizing process parameters, coating quality can be steadily improved.

[0042] In summary, the embodiments of the present application have at least the following technical effects: The embodiment of the present application captures the surface image of the car sunroof through a high-resolution industrial camera to obtain a coating image data set; locates the car sunroof, identifies defects in the coating image data set according to the sunroof position information, and obtains an image defect label group; classifies according to the image defect label group, constructs a defect list, automatically sorts according to the defect list, and determines the coating sorting result; performs quality inspection on the car sunroof glass coating according to the coating process information combined with the coating sorting result, and generates a coating quality report for the car sunroof glass. These technical effects jointly solve the technical problems of insufficient defect identification accuracy and inaccurate quality inspection due to glass reflection and shadow interference in the coating quality inspection of the car sunroof glass, and achieve high-precision and high-efficiency car sunroof glass coating quality inspection and automatic sorting, and improve process stability and product quality.

[0043] Embodiment 2, based on the same inventive concept as the method for online detection of coating quality of automobile sunroof glass in the above embodiment, Figure 2 As shown, the present application provides an online detection system for coating quality of automobile sunroof glass, the system comprising: Image capture module 11: capture the surface image of the car sunroof through a high-resolution industrial camera to obtain a coating image data set; defect recognition module 12: locate the car sunroof, perform defect recognition on the coating image data set according to the sunroof position information, and obtain an image defect label group; automatic sorting module 13: classify according to the image defect label group, build a defect list, perform automatic sorting according to the defect list, and determine the coating sorting result; quality inspection module 14: perform quality inspection on the car sunroof glass coating according to the coating process information and the coating sorting result, and generate a coating quality report for the car sunroof glass.

[0044] Furthermore, the image capturing module 11 is also used to execute the following method: A multi-angle LED light source is used to set multiple lighting conditions, and a first lighting condition is selected based on the multiple lighting conditions to activate a high-resolution industrial camera; when a car sunroof enters a detection area, the surface of the car sunroof is traversed to perform coating marking and determine a coating area; a high-resolution industrial camera is used to perform multi-angle coverage acquisition of the coating area to obtain a first acquisition result; the first lighting condition of the multi-angle LED light source is adjusted according to the first acquisition result, a target lighting condition is set, and a high-resolution industrial camera is activated according to the target lighting condition to perform multi-angle coverage acquisition of the coating area to obtain a coating image data set.

[0045] Furthermore, the image capturing module 11 is also used to execute the following method: Based on the first acquisition result, interference analysis is performed to determine multiple light source interference data, which include shadow interference data and reflection interference data; according to the shadow interference data combined with the coating characteristics, the high-resolution industrial camera is collected and adjusted to determine the lighting adjustment angle information; according to the reflection interference data combined with the coating characteristics, the high-resolution industrial camera is collected and adjusted to determine the lighting adjustment intensity information; based on the lighting adjustment angle information and the lighting adjustment intensity information, the first lighting condition is adjusted to set the target lighting condition.

[0046] Furthermore, the defect identification module 12 is also used to perform the following method: A first sensor is started to perform sensor analysis on a coating area to obtain a first sensor data set; a second sensor is started to perform sensor analysis on the coating area to obtain a second sensor data set; a spatial rectangular coordinate system is constructed based on the coating area, and the first sensor data set and the second sensor data set are mapped to the spatial rectangular coordinate system to obtain a three-dimensional model of the coating of the automobile sunroof glass, wherein the three-dimensional model includes the sunroof position information.

[0047] Furthermore, the defect identification module 12 is also used to perform the following method: Based on the skylight position information, the coating image dataset and the three-dimensional model are deeply learned to generate image learning results; the historical coating defect log is retrieved, the image learning results are traversed and combined with the historical coating defect log to perform defect identification to generate multiple defect identification results; the coating image dataset is matched and divided according to the multiple defect identification results to obtain the image defect label group.

[0048] Furthermore, the automatic sorting module 13 is also used to perform the following method: The coating image data set is clustered and analyzed according to the image defect label group to determine a plurality of coating defect classes; an impact analysis is performed on the automobile sunroof glass according to the plurality of coating defect classes to set a plurality of defect impact levels; the plurality of coating defect classes are associated with the plurality of defect impact levels to construct a defect list; the automobile sunroof glass is classified according to the defect list to determine a target automobile sunroof glass, and target coating image data of the target automobile sunroof glass is extracted according to the coating image data set; the target automobile sunroof glass is automatically sorted according to the target coating image data to obtain a coating sorting result.

[0049] Furthermore, the automatic sorting module 13 is also used to perform the following method: Determine whether the target coating image data is in the defect list; if the target coating image data is in the defect list, generate a primary sorting instruction, and sort the target automobile sunroof glass to a first process area through the primary sorting instruction; if the target coating image data is not in the defect list, generate a secondary sorting instruction, and sort the target automobile sunroof glass to a second process area through the secondary sorting instruction.

[0050] Furthermore, the quality detection module 14 is also used to perform the following method: A risk assessment of process consistency is performed on the coating process information according to the target automobile sunroof glass in the first process area to generate coating process risk data; the coating sorting results are continuously counted according to the coating process risk data to determine the high-frequency defect mode; the high-frequency defect mode is associated with the coating machine parameter timing data to obtain a plurality of parameter combinations, and the plurality of parameter combinations are mined through a random forest algorithm to obtain abnormal parameter combinations; the abnormal parameter combinations are added to the coating quality report, and the coating process information is feedback optimized according to the abnormal parameter combinations to generate an alarm signal.

[0051] Furthermore, the quality detection module 14 is also used to perform the following method: Establish a coating quality-process parameter mapping matrix, extract the defect rate data of the abnormal parameter combination; traverse the defect rate data of the abnormal parameter combination to select the minimum value as the preferred parameter; gradually adjust the coating process information to the preferred parameter, and monitor the change of the defect rate data in real time during the adjustment process. If the defect rate data after adjustment does not drop to a preset threshold, trigger the alarm signal.

[0052] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0053] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0054] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A method for online detection of coating quality of automobile sunroof glass, characterized in that: The method comprises: Capture the surface image of the car sunroof with a high-resolution industrial camera to obtain a coating image dataset; Locating the sunroof of the car, and performing defect recognition on the coating image data set according to the sunroof position information to obtain an image defect label group; Classify according to the image defect label group, construct a defect list, automatically sort according to the defect list, and determine the coating sorting result; The coating quality of the automobile sunroof glass is inspected according to the coating process information and the coating sorting result, and a coating quality report of the automobile sunroof glass is generated.

2. The method for online detection of coating quality of automobile sunroof glass according to claim 1, characterized in that: The surface image of the car sunroof is captured by a high-resolution industrial camera to obtain a coating image dataset, and the method includes: A multi-angle LED light source is used to set multiple lighting conditions, and a first lighting condition is selected based on the multiple lighting conditions to activate a high-resolution industrial camera; When the car sunroof enters the detection area, the surface of the car sunroof is traversed for coating marking to determine the coating area; Perform multi-angle coverage acquisition of the coating area by using a high-resolution industrial camera to obtain a first acquisition result; The first lighting condition of the multi-angle LED light source is adjusted according to the first acquisition result, the target lighting condition is set, and the high-resolution industrial camera is activated according to the target lighting condition to perform multi-angle coverage acquisition on the coating area to obtain a coating image data set.

3. The method for online detection of coating quality of automobile sunroof glass according to claim 2, characterized in that: The first illumination condition of the multi-angle LED light source is adjusted according to the first acquisition result to set a target illumination condition, the method comprising: Perform interference analysis based on the first acquisition result to determine multiple light source interference data, where the multiple light source interference data include shadow interference data and reflection interference data; According to the shadow interference data combined with coating characteristics, the high-resolution industrial camera is collected and adjusted to determine the illumination adjustment angle information; According to the reflection interference data combined with the coating characteristics, the high-resolution industrial camera is collected and adjusted to determine the light adjustment intensity information; The first lighting condition is adjusted based on the lighting adjustment angle information and the lighting adjustment intensity information, and the target lighting condition is set.

4. The method for online detection of coating quality of automobile sunroof glass according to claim 2, characterized in that: The method for obtaining the sunroof position information includes: Starting a first sensor to perform sensing analysis on the coating area to obtain a first sensing data set; Starting the second sensor to perform sensing analysis on the coating area to obtain a second sensing data set; A spatial rectangular coordinate system is constructed based on the coating area, and the first sensor data set and the second sensor data set are mapped to the spatial rectangular coordinate system to obtain a three-dimensional model of the automobile sunroof glass coating, wherein the three-dimensional model includes the sunroof position information.

5. The method for online detection of coating quality of automobile sunroof glass according to claim 4, characterized in that: Defect recognition is performed on the coating image data set according to the skylight position information to obtain an image defect label group, the method comprising: Based on the skylight position information, deep learning is performed on the coating image data set and the three-dimensional model to generate an image learning result; Retrieving historical coating defect logs, traversing the image learning results and combining the historical coating defect logs to perform defect recognition, and generating multiple defect recognition results; The coating image data set is matched and divided according to the multiple defect recognition results to obtain the image defect label group.

6. The method for online detection of coating quality of automobile sunroof glass according to claim 1, characterized in that: Classifying according to the image defect label group, constructing a defect list, automatically sorting according to the defect list, and determining the coating sorting result, the method includes: Performing cluster analysis on the coating image data set according to the image defect label group to determine a plurality of coating defect classes; Performing an impact analysis on the automobile sunroof glass according to the multiple coating defect types, and setting multiple defect impact levels; Associating the plurality of coating defect classes with the plurality of defect impact levels to construct a defect list; Classifying the automobile sunroof glass according to the defect list, determining the target automobile sunroof glass, and extracting the target coating image data of the target automobile sunroof glass according to the coating image data set; The target automobile sunroof glass is automatically sorted according to the target coating image data to obtain a coating sorting result.

7. The method for online detection of coating quality of automobile sunroof glass according to claim 6, characterized in that: Automatically sorting the target automobile sunroof glass according to the target coating image data to obtain a coating sorting result, the method comprising: Determining whether the target coating image data is in the defect list; If the target coating image data is in the defect list, a primary sorting instruction is generated, and the target automobile sunroof glass is sorted to a first process area through the primary sorting instruction; If the target coating image data is not in the defect list, a secondary sorting instruction is generated, and the target automobile sunroof glass is sorted to the second process area through the secondary sorting instruction.

8. The method for online detection of coating quality of automobile sunroof glass according to claim 7, characterized in that: The coating quality of the automobile sunroof glass is inspected according to the coating process information and the coating sorting result, and a coating quality report of the automobile sunroof glass is generated. The method includes: Performing a risk assessment of process consistency on the coating process information according to the target automobile sunroof glass in the first process area to generate coating process risk data; Continuously counting the coating sorting results according to the coating process risk data to determine a high-frequency defect mode; Associating the high-frequency defect mode with the coating machine parameter time series data to obtain multiple parameter combinations, and mining the multiple parameter combinations through a random forest algorithm to obtain abnormal parameter combinations; The abnormal parameter combination is added to the coating quality report, and the coating process information is feedback optimized according to the abnormal parameter combination to generate an alarm signal.

9. The method for online detection of coating quality of automobile sunroof glass according to claim 8, characterized in that: The abnormal parameter combination is added to the coating quality report, and the coating process information is optimized based on the abnormal parameter combination to generate an alarm signal, the method comprising: Establishing a coating quality-process parameter mapping matrix and extracting defect rate data of abnormal parameter combinations; Traversing the defect rate data of the abnormal parameter combination, selecting the minimum value as the preferred parameter; The coating process information is gradually adjusted to the preferred parameters, and the change of the defect rate data is monitored in real time during the adjustment process. If the defect rate data after adjustment does not drop to a preset threshold, the alarm signal is triggered.

10. An online detection system for coating quality of automobile sunroof glass, characterized in that: The system is used to execute the method for online detection of coating quality of automobile sunroof glass according to any one of claims 1 to 9, comprising: Image capture module: Capture the surface image of the car sunroof through a high-resolution industrial camera to obtain a coating image dataset; Defect recognition module: locates the car sunroof, performs defect recognition on the coating image data set according to the sunroof position information, and obtains an image defect label group; Automatic sorting module: classify according to the image defect label group, build a defect list, automatically sort according to the defect list, and determine the coating sorting result; Quality inspection module: Perform quality inspection on the automobile sunroof glass coating according to the coating process information and the coating sorting result, and generate a coating quality report for the automobile sunroof glass.

Images