Steel strip surface coating defect detection method

Abstract

The invention discloses a steel strip surface coating defect detection method, which relates to the technical field of image detection, and is characterized in that an acquisition device is assembled on a steel strip production line, and an acquisition controller is configured to acquire section steel strip images; and the processing unit is connected with the image processor and is used for carrying out pixel identification partitioning, executing Haar feature-based extraction and variance calculation partitioning by partitioning, constructing a texture distribution diagram, carrying out matching based on a plating defect library, determining defect features and identifying the texture distribution diagram, and taking the texture distribution diagram as a plating defect detection result. The technical problems that in the prior art, a detection mode is insufficient in adaptive capacity in a production environment, self-adaptive flexible adjustment cannot be effectively carried out, and defect detection completeness and accuracy are insufficient due to the fact that a treatment mode is limited are solved, the detection accuracy and efficiency of a steel strip coating are improved, and the detection efficiency is improved. Therefore, the overall quality and stability of steel belt production are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image detection, and particularly to a method for detecting coating defects on the surface of a steel strip. Background Art

[0002] With the expansion of the production scale of steel strips and the improvement of the automation level, how to ensure the stability and consistency of the coating quality during the production process has become one of the key technical challenges. Most of the traditional coating defect detection methods rely on manual visual inspection or simple surface quality detection, which are not only inefficient and inaccurate, but also difficult to meet the requirements of real-time and large-scale production in high-speed and high-precision production lines.

[0003] In the prior art, image processing technology and automated detection systems have been widely used in quality control during the production process of steel strips, but there are still some technical problems. For example, traditional image feature extraction methods, such as feature extraction and edge detection, cannot effectively identify small and diverse coating defects, and lack the ability to adapt to complex changes in the actual production environment, resulting in insufficient detection accuracy and flexibility.

[0004] In summary, the prior art still has deficiencies in the detection of steel strip coatings, such as insufficient adaptability of the detection method to the production environment, inability to effectively perform adaptive flexible adjustment, and insufficient completeness and accuracy of defect detection due to limitations in the processing method. Summary of the Invention

[0005] The present application provides a method for detecting coating defects on the surface of a steel strip, which is used to solve the technical problems existing in the prior art, such as insufficient adaptability of the detection method to the production environment, inability to effectively perform adaptive flexible adjustment, and insufficient completeness and accuracy of defect detection due to limitations in the processing method.

[0006] In view of the above problems, the present application provides a method for detecting coating defects on the surface of a steel strip.

[0007] The present application provides a method for detecting coating defects on the surface of a steel strip. The method includes: assembling a collection device on a steel strip production line and configuring a collection controller. Among them, an image sensor - light source is used as the collection device, a local section of the steel strip is used as the collection target, and relative position parameters are used as the collection control conditions; through steel strip conveying and positioning, the collection device is adaptively regulated in combination with the collection controller to collect an image of a section of the steel strip; connecting an image processor, performing pixel traversal and recognition zoning on the image of the section of the steel strip, and successively performing Haar feature extraction and variance calculation on each zone, and constructing a texture distribution map. Among them, the pixel change law is used as the zoning standard, and the image processor is built into the defect detection system; identifying the texture distribution map and performing matching based on a coating defect library to determine defect features, and marking the texture distribution map according to the defect features as the coating defect detection result.

[0008] Among them, the configuration of the collection controller includes: centering on the collection target, determining the standard relative position parameters of the collection device relative to the collection target, including relative position relationships and collection control parameters; aiming at the adjustment of real-time relative position parameters compared with the standard relative position parameters, constructing the collection controller and establishing a communication connection between the collection controller and the collection device.

[0009] Among them, the construction of the texture distribution map includes: constructing the image processor in a data-driven supervised training manner; the image processor receives the image of the section of the steel strip and performs grayscale processing to determine a grayscale steel strip image; performing pixel traversal on the grayscale steel strip image, zoning according to the pixel change law to determine N zoned grayscale images, where the zoning is irregular and the pixel change laws of each zoned grayscale image are the same; interacting with the standard pixel law to perform binarization processing on the N zoned grayscale images, screening M zoned grayscale images, where the standard pixel law is the pixel law in the state where the coating quality meets the standard, and M is a positive integer less than or equal to N; constructing the texture distribution map according to the M zoned grayscale images; among them, those consistent with the standard pixel law are binarized to 0, and those inconsistent with the standard pixel law are binarized to 1, and the M zoned grayscale images are the parts binarized to 1.

[0010] Among them, constructing the texture distribution map according to the M zoned grayscale images includes: traversing the M zoned grayscale images, successively performing Haar feature extraction to determine M zoned Haar features, where the Haar features are used to measure the grayscale change pattern; traversing the M zoned grayscale images, successively performing zoned pixel value variance calculation to determine M zoned variances, where the pixel value variance is used to measure the uniformity and complexity of the texture; constructing the texture distribution map according to the M zoned Haar features and the M zoned variances.

[0011] Among them, constructing the texture distribution map according to the M partition Haar features and the M partition variances includes: mapping the M partition Haar features and the M partition variances to determine M texture features; performing distribution positioning on the M texture features based on the distribution position of the section steel strip map, blanking the N - M partitions, and determining the texture distribution map through image reconstruction processing.

[0012] Among them, after identifying the texture distribution map according to the defect features, it includes: setting the production line sampling interval, controlling the acquisition device to perform acquisition parameter adjustment according to the production line sampling interval, performing interval image sampling and defect analysis, and adding them to the detection database, where the detection database is used to store the texture distribution map after defect feature identification; taking the preset sampling frequency as a constraint, performing texture distribution map splicing and global determination on the detection database to locate the process defect features; based on the process defect features, tracing the steel strip production process and guiding the process optimization of steel strip production.

[0013] Among them, when the detection database performs texture distribution map splicing and global determination to locate the process defect features, it includes: performing texture distribution map splicing to determine the spliced texture distribution map; traversing the spliced texture distribution map, and locating the frequent item defects based on the tangential position and longitudinal frequency of the coating defects, where the defect position is located by the tangential position and the frequent items are screened by the longitudinal frequency; locating the process defect features according to the frequent item defects.

[0014] Among them, the method further includes: performing sensing detection on the plating solution parameters to determine the real - time plating solution data, where the plating solution parameters at least include composition, temperature, and speed; constructing a plating solution state curve based on the production time series based on the real - time plating solution data; identifying the plating solution state curve, performing plating solution instability analysis according to the parameter variation vector, and generating defect warning information.

[0015] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0016] A method for detecting defects in the surface coating of a steel strip provided by an embodiment of the present application assembles a collection device on a steel strip production line and configures a collection controller. An image sensor-light source is used as the collection device, a local section of the steel strip is used as the collection target, and relative position parameters are used as the collection control conditions. By performing steel strip conveying and positioning, the collection device is adaptively adjusted in combination with the collection controller to collect images of the section of the steel strip. An image processor is connected to perform pixel traversal and recognition partitioning on the image of the section of the steel strip, and Haar feature extraction and variance calculation are performed for each partition to construct a texture distribution map. The pixel change law is used as the partitioning standard, the texture distribution map is recognized and matched based on a coating defect library to determine defect features, and the texture distribution map is marked according to the defect features as the detection result of the coating defect. It is used to solve the technical problems existing in the prior art that the detection method has insufficient adaptability to the production environment, cannot effectively perform flexible self-adaptive adjustment, and the limited processing method leads to insufficient completeness and accuracy of defect detection, improving the detection accuracy and efficiency of the steel strip coating, thereby improving the overall quality and stability of steel strip production. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 FIG. is a schematic flow chart of a method for detecting defects in the surface coating of a steel strip provided by the present application;

[0018] Figure 2 FIG. is a schematic flow chart of constructing a texture distribution map according to the gray-scale maps of M partitions in a method for detecting defects in the surface coating of a steel strip provided by the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] The present application provides a method for detecting defects in the surface coating of a steel strip, assembles a collection device on a steel strip production line, configures a collection controller, and collects images of a section of the steel strip. An image processor is connected to perform pixel recognition partitioning, Haar feature extraction and variance calculation are performed for each partition to construct a texture distribution map, and matching is performed based on a coating defect library to determine defect features and mark the texture distribution map as the detection result of the coating defect. It is used to solve the technical problems existing in the prior art that the detection method has insufficient adaptability to the production environment, cannot effectively perform flexible self-adaptive adjustment, and the limited processing method leads to insufficient completeness and accuracy of defect detection.

[0020] Embodiment: As Figure 1 shown, the present application provides a method for detecting defects in the surface coating of a steel strip, and the method includes:

[0021] S1: Assemble a collection device on a steel strip production line and configure a collection controller, where an image sensor-light source is used as the collection device, a local section of the steel strip is used as the collection target, and relative position parameters are used as the collection control conditions.

[0022] In this embodiment, to achieve efficient detection of coating defects on the steel strip surface, a collection device is first assembled in the steel strip production line, and a collection controller for scene adaptive parameter adjustment of the collection device is configured. Among them, the collection device mainly consists of an image sensor and a light source. The image sensor is used to capture the image information of the steel strip surface in real time, and the light source is used to ensure the uniformity and stability of illumination during image collection. Exemplarily, the image sensor can adopt a CCD or CMOS sensor, which can capture subtle surface texture changes, thereby providing high-resolution image data. The light source can select an LED light source, which has strong brightness and good color temperature control ability, thereby ensuring the contrast and detail performance of the image.

[0023] Furthermore, taking a local section of the steel strip as the collection target, precise image collection is carried out for this target. The local section of the steel strip refers to the section range that meets the preset size during the continuous production process of the steel strip. According to the production and processing characteristics of the steel strip, with a local section of the steel strip as the primary detection requirement, multiple intermittent detections are carried out under the same steps as the steel strip production. By setting the local section as the collection target, it is possible to focus on the key areas of the steel strip surface and reduce the interference of irrelevant areas. For example, a long strip-shaped local section on the steel strip surface can be selected for collection and analysis to ensure the efficiency and effectiveness of data collection.

[0024] To precisely control the collection process, a collection controller is configured, and the relative position parameter is used as the collection control condition. Specifically, the collection controller automatically adjusts the position and angle of the image sensor and the light source according to the real-time position and motion state of the steel strip, ensuring the stability and consistency of image collection. The relative position parameter refers to the relative spatial position relationship between the image sensor, the light source, and the local section of the steel strip. This position relationship is crucial for the clarity and quality of the image. During the actual operation of the collection controller, it dynamically adjusts the working state of the device based on the moving speed and direction of the steel strip on the production line, as well as the relative position change between the sensor and the steel strip. For example, if the steel strip undergoes a slight displacement, the controller can automatically adjust the position of the image sensor to ensure that the collection area is always aligned with the target section, preventing the image quality from deteriorating due to position deviation.

[0025] Through the above configuration and adjustment, the entire collection system can stably and efficiently adapt to the production line, providing reliable raw data support for subsequent image processing and defect analysis.

[0026] Furthermore, for the configuration of the collection controller, step S1 of this application includes:

[0027] Centered on the acquisition target, determine the standard relative position parameters of the acquisition device relative to the acquisition target, including relative position relationships and acquisition control parameters; aiming at adjusting the real-time relative position parameters compared with the standard relative position parameters, construct the acquisition controller and establish a communication connection between the acquisition controller and the acquisition device.

[0028] In this embodiment, first, taking the local section of the steel strip as the center, determine the standard relative position parameters of the acquisition device relative to the acquisition target. The standard relative position parameters refer to the spatial relationship between the acquisition device (including the image sensor and the light source) and the acquisition target (i.e., the local section of the steel strip) during the image acquisition process. This standard position relationship ensures that in an ideal situation, the image sensor is always aligned with the surface of the target section, guaranteeing the stability and accuracy of image acquisition. At the same time, with this as a constraint, the consistency of each acquisition can be ensured.

[0029] The standard relative position relationship not only includes the spatial position of the device but also parameters such as the working angle, focal length, and light source intensity of the acquisition device. To ensure stable image quality, the standard relative position relationship is usually precisely set according to the production line parameters of the steel strip during the equipment installation and commissioning stage.

[0030] Next, by monitoring the motion state on the steel strip production line in real time, obtain the real-time relative position parameters between the current acquisition device and the acquisition target. The real-time relative position parameters refer to the actual spatial position relationship between the acquisition device and the local section of the steel strip during the production of the steel strip. This parameter will change continuously with the conveyance of the steel strip, so it must be monitored in real time and dynamically adjusted. For example, when the steel strip is conveyed at high speed, the relative position of the image sensor may shift slightly, resulting in a decrease in the accuracy of image acquisition. Therefore, the device needs to be adjusted in a timely manner.

[0031] Therefore, an acquisition controller is constructed with the difference between the real-time relative position parameters and the standard relative position parameters as the adjustment target. The role of the acquisition controller is to automatically adjust the working state of the acquisition device according to the difference between the real-time relative position parameters and the standard relative position parameters to ensure that the device always maintains the predetermined standard relative position. For example, if the relative position of the acquisition device shifts, the controller will instruct the image sensor or the light source to make fine adjustments to restore to the standard position relationship, ensuring the quality and consistency of the acquired images. Through real-time data feedback, the controller continuously optimizes the position and state of the device, keeping the acquisition process in the best working state at all times.

[0032] Finally, a communication connection is established between the acquisition controller and the acquisition device. This communication connection is a real-time data transmission channel between the acquisition controller and devices such as image sensors and light sources. Through the above steps, the acquisition device can achieve stable and accurate image acquisition during the steel strip production process, providing high-quality original data support for subsequent defect detection and analysis.

[0033] S2: By positioning the steel strip transportation and adaptively regulating the acquisition device in combination with the acquisition controller, acquire the steel strip images in the acquisition section.

[0034] In this embodiment, to ensure the accuracy and stability of image acquisition, the transportation positioning of the steel strip is first carried out. The steel strip transportation positioning refers to determining the position of the target steel strip section according to the moving speed and direction of the steel strip. Then the acquisition controller starts to adaptively regulate the acquisition device. The adaptive regulation function of the acquisition controller means automatically adjusting the working states of the image sensor and the light source according to the real-time position and speed information of the target steel strip section, ensuring that the best imaging conditions are always met during the image acquisition process.

[0035] Among them, this process is based on the comparison between the feedback of the real-time relative position parameter and the standard relative position parameter. The acquisition controller calculates the adjustment amount through an algorithm and finely adjusts the acquisition device. For example, when the position of the steel strip changes or the steel strip transmission speed fluctuates, the acquisition controller can judge in real time whether the acquisition device needs to adjust parameters such as the angle of the image sensor and the brightness of the light source, so as to ensure the clarity of the image acquisition area and the image quality.

[0036] Combined with the regulation decision of the acquisition controller, determine the adjustment data of the acquisition device, and control the acquisition device to perform image acquisition on the local section of the steel strip according to the predetermined target. The image acquisition process includes capturing the image information on the surface of the steel strip in real time through the image sensor and ensuring that the details of the surface texture are clearly presented through the auxiliary illumination of the light source. During the acquisition process, the adaptive adjustment function of the acquisition device can dynamically adjust the imaging parameters according to the motion state and position change of the steel strip, ensuring that the image quality of each acquisition is not affected by factors such as the steel strip transmission speed and environmental light changes.

[0037] In summary, the stable acquisition of the steel strip images is ensured, enabling subsequent image processing and defect analysis to be carried out based on high-quality image data, thereby improving the detection accuracy and efficiency.

[0038] S3: Connect the image processor, perform pixel traversal and recognition zoning on the steel strip images in the section, and execute Haar feature extraction and variance calculation zone by zone to construct a texture distribution map, where the pixel change rule is used as the zoning standard, and the image processor is built into the defect detection system.

[0039] In this embodiment, the image processor, as part of the defect detection system, is responsible for analyzing and processing the collected steel strip images. The pixel recognition partition refers to dividing the steel strip image into multiple small regions according to certain rules. The image processor separates different regions of the steel strip image based on the changing trend of pixel values in the image, ensuring that the pixel characteristics within each partition are relatively consistent. For example, when there is an uneven coating on the surface of the steel strip, the fineness of the partition can help identify defects in local areas.

[0040] Next, Haar feature extraction and variance calculation are performed within each partition. Haar features are statistical features used to describe image textures, which are used to express the pattern of gray-scale changes and the complexity of local textures in the image. Within the partition, the image processor analyzes the changing trend of pixel gray-scale values by calculating the Haar features of each partition. This helps to identify texture changes within the region, such as the uniformity of the coating and defect characteristics. For example, when there is uneven coating on the surface of the steel strip, Haar feature extraction can identify abnormal patterns of gray-scale changes, providing a basis for defect identification.

[0041] After extracting the Haar features, the image processor performs pixel variance calculation for each partition. Pixel variance is an important indicator for measuring the uniformity and complexity of image textures. It can reflect the degree of dispersion of pixel values in the image, thereby helping to evaluate the consistency of the coating on the surface of the steel strip. Among them, a high variance usually indicates that there are large texture changes in the region, which may correspond to uneven coating or other defects, while a low variance indicates that the texture in the region is relatively uniform and the surface coating is relatively smooth.

[0042] Furthermore, based on the Haar features and variance features, a texture distribution map of the steel strip image is finally constructed. The texture distribution map can display the texture characteristics and gray-scale change patterns of each partition in the image, providing important image feature bases for subsequent defect detection and classification. Among them, the texture distribution map reflects the quality characteristics of the coating on the surface of the steel strip. Each partition in the image is assigned a texture feature value, which can effectively identify coating quality problems or defect areas. The construction of the texture distribution map greatly improves the accuracy and response speed of the defect detection system.

[0043] Further, for the construction of the texture distribution map, step S3 of this application includes:

[0044] The image processor is constructed in a data-driven supervised training manner; the image processor receives the sectional steel strip diagram and performs grayscale processing to determine a grayscale steel strip diagram; traverses the pixels of the grayscale steel strip diagram, partitions according to the pixel change trend rule, and determines N sectional grayscale diagrams, where the partition is an irregular partition, and the pixel change trend rule of each sectional grayscale diagram is the same; interacts with the standard pixel rule, performs binarization processing on the N sectional grayscale diagrams, and screens out M sectional grayscale diagrams, where the standard pixel rule is the pixel rule in the state where the coating quality meets the standard, and M is a positive integer less than or equal to N; constructs a texture distribution diagram according to the M sectional grayscale diagrams; where, those consistent with the standard pixel rule are binarized to 0, and those inconsistent with the standard pixel rule are binarized to 1, and the M sectional grayscale diagrams are the parts binarized to 1.

[0045] In this embodiment, the construction of the image processor is based on a data-driven supervised training method. Supervised training is to train an algorithm by using a labeled data set so that it can recognize and process new data. When constructing the image processor, a large number of steel strip image data that have been manually labeled are used to first determine the underlying logic of image processing, that is, grayscale processing - pixel-based traversal recognition and partitioning - partition screening - Haar feature extraction and variance calculation - image reconstruction. Based on the underlying logic, according to the labeled steel strip image data, that is, the training samples are trained until convergence to generate the image processor.

[0046] Next, the image processor receives the collected sectional steel strip diagram and performs grayscale processing on it. Grayscale processing is the process of converting a color image into a grayscale image. In a grayscale image, the value of each pixel represents its brightness. By performing grayscale processing, the color information in the image can be reduced, and the image can be converted into simple grayscale information, which helps to improve the subsequent processing efficiency and highlight the texture features. The obtained grayscale steel strip diagram can display the brightness changes in different areas of the steel strip surface and provide basic data for further texture analysis.

[0047] Traverse the pixels of the grayscale steel strip diagram and partition according to the pixel change trend rule. By analyzing the change trend of pixel values in the image, different areas in the image can be identified, and the image is divided into multiple partitions according to the texture features of these areas. The pixel change trend rule refers to the rule of how pixel values change with spatial position in the image, and common rules include smooth transition or mutation. In this step, the image processor traverses each pixel in the image, analyzes the grayscale changes between pixels, and thus determines different partitions of the image. The pixel value change rule of each partition tends to be the same, that is, their pixel change trend rules are similar.

[0048] Among them, the pixel change trend rule of the steel strip should be globally consistent. The differentiation of the pixel rule indicates that the surface texture of the coating is inconsistent, and targeted analysis is performed on it.

[0049] Based on the partitioning, the image processor divides the grayscale image into N irregular partitions. These partitions are not regular rectangular or square regions, but are flexibly divided according to the pixel change trend. The pixel change trend within each partition is consistent, meaning that within the same region, it corresponds to a certain type of texture characteristic. For example, a region with a uniform coating will form a relatively smooth pixel change trend, while a region with a non-uniform coating may exhibit larger gray-scale variations.

[0050] Subsequently, the image processor interacts and compares the standard pixel pattern with the grayscale image of each partition for binarization processing. The standard pixel pattern refers to the change pattern of pixel values in the image when the coating quality meets the standard. The image processor compares the pixel values of the partition with the standard pattern. If the gray-scale change of the partition conforms to the standard pixel pattern, it is marked as "0", indicating that the partition meets the quality requirements; if the gray-scale change of the partition does not conform to the standard pattern, it is marked as "1", indicating that there may be a defect in the partition. After binarization processing, M grayscale images of partitions that do not conform to the standard pattern are selected, where M is a positive integer less than or equal to N, representing the partitions that do not conform to the standard pixel pattern.

[0051] Among them, the M grayscale images of partitions are the parts with coating defects, and the defect characteristics of each partition are different.

[0052] Based on the M grayscale images of partitions after binarization, the image processor further constructs a texture distribution map. The texture distribution map reflects the texture characteristics and quality status of each region in the image. The partition with a binarization value of "1" indicates that there may be a defect in that region, and the partition with a binarization value of "0" indicates that the texture characteristics of that region conform to the standard. During the process of constructing the texture distribution map, the image processor synthesizes the binarization results of all partitions to form the final image texture characteristic map, providing a clear visual basis for subsequent defect detection and analysis, and thus providing accurate data support for the detection of coating defects.

[0053] Furthermore, as Figure 2 shown, based on the M grayscale images of partitions, constructing a texture distribution map, step S3 of the present application includes:

[0054] Traverse the M grayscale images of partitions, sequentially perform Haar feature extraction to determine the Haar features of the M partitions, where the Haar features are used to measure the gray-scale change pattern; traverse the M grayscale images of partitions, sequentially calculate the variance of the partition pixel values to determine the variances of the M partitions, where the pixel value variance is used to measure the uniformity and complexity of the texture; construct the texture distribution map according to the Haar features of the M partitions and the variances of the M partitions.

[0055] In this embodiment, the image processor first traverses the M partition grayscale images and sequentially performs Haar feature extraction to determine the Haar features of each partition. Among them, the Haar feature is a statistical feature for the grayscale change pattern in a local area of the image. By extracting the Haar feature, the grayscale change rule within each partition can be quantified, and then the texture feature of the partition can be identified. For example, if the grayscale values in a certain area change relatively smoothly, its Haar feature will reflect a smooth texture pattern; while if there are large fluctuations in the grayscale values within the area, the Haar feature will reflect complex texture changes. Therefore, the Haar feature can help effectively describe and distinguish different types of textures, especially in defect detection, which is helpful for identifying uneven coatings, scratches, or other abnormal textures.

[0056] Next, by traversing the M partition grayscale images, the variance of the partition pixel values is calculated sequentially to determine the variance value of each partition. Using all the pixel values in each partition grayscale image as data, the variance formula is sampled for calculation. The pixel value variance is a key indicator for measuring the texture uniformity and complexity of an image. The variance reflects the degree of dispersion of the pixel values in the image, that is, the change range of the grayscale values within the area. A high variance usually indicates that the texture within the area is relatively complex, and there may be defects or uneven coatings; while a low variance indicates that the texture change in this area is relatively smooth and may meet the standard coating quality. By calculating the pixel value variance of each partition, the image processor can further quantify the characteristics of the texture and distinguish the defective areas from the normal areas in the image. For example, in a uniform coating area, the variance is low, while in the area with defects, the variance is high.

[0057] After completing the Haar feature extraction and pixel value variance calculation, the image processor constructs the final texture distribution map based on the Haar feature and variance value of each partition. This texture distribution map can clearly reflect the overall texture structure of the steel strip image by integrating the Haar feature and variance data of each partition. The texture distribution map shows the texture features of each partition and their grayscale change patterns, which helps to further identify and locate the quality problems and defects of the coating.

[0058] For example, in the texture distribution map, areas with a high variance may be marked as defective areas, while areas with complex Haar features may indicate irregular changes in the coating. By constructing the texture distribution map, the image processor can provide a more accurate basis for subsequent defect analysis, helping to identify potential defects such as uneven plating, scratches, and bubbles.

[0059] In summary, by sequentially extracting Haar features and calculating pixel value variances, the image processor can not only quantify the texture features of each partition but also effectively construct an overall texture distribution map, providing strong data support for the detection and quality assessment of defects in the steel strip surface coating.

[0060] Further, according to the M partition Haar features and the M partition variances, construct the texture distribution map. Step S3 of the present application includes:

[0061] Map the M partition Haar features and the M partition variances to determine M texture features; based on the distribution positions of the sectional steel strip diagrams, perform distribution positioning on the M texture features, perform blank attribution on the N-M partitions, and determine the texture distribution map through image reconstruction processing.

[0062] In this embodiment, map the Haar features and variance values of the M partitions, that is, determine the Haar feature-variance values corresponding to the M partitions as texture features. By combining these two features, the texture characteristics of each partition can be described more comprehensively. For example, if the Haar features of a certain partition show complex texture changes and the variance of this partition is relatively high, there may be coating unevenness or defects in this area.

[0063] Next, the image processor locates the distribution positions of each M partition's texture features in the sectional steel strip diagram. The distribution positioning is to locate the distribution positions corresponding to the texture features of each partition in the sectional steel strip diagram to determine the distribution law of the texture features in the overall image. Through the distribution positioning, the image processor can accurately identify these areas from the overall image and provide basic data for subsequent defect identification and process improvement.

[0064] After the distribution positioning, determine the relative image positions of the M partitions that may have defects. At the same time, the image processor also needs to perform blank attribution on the remaining N-M partitions. The process of blank attribution is to mark the partitions that do not participate in the texture feature mapping (i.e., the N-M partitions) as "blank". These areas belong to the areas with qualified quality and do not need to be analyzed for defect identification. The blank attribution is to effectively remove the interference of these non-reference areas in the subsequent image reconstruction process, thereby ensuring the quality and accuracy of the final texture distribution map and reducing the amount of analysis data to improve the analysis efficiency.

[0065] The image reconstruction is to restore or reconstruct the complete texture structure of the steel strip image according to the previous texture features and the results of blank attribution. That is, according to the results of the distribution positioning, reasonably splice the texture features of each partition according to their positions in the image, perform blank attribution on the remaining partitions, and ensure that the reconstructed image can accurately reflect the overall distribution of the texture on the steel strip surface. For example, during the reconstruction process, the texture features of the defective areas will be highlighted, while the normal areas are blank. The finally generated texture distribution map will be able to clearly show the texture condition of the steel strip surface, thereby providing strong support for subsequent defect detection and quality assessment.

[0066] In summary, the image processor accurately constructs the texture distribution map of the steel strip image by mapping the Haar features and variances of M partitions, combined with distribution positioning, blank placement, and image reconstruction processing. On the basis of retaining valid data, invalid information is screened out, which not only improves the accuracy of defect detection but also makes the evaluation of coating quality more scientific and efficient.

[0067] S4: Identify the texture distribution map and match it based on the coating defect library to determine the defect features, and identify the texture distribution map according to the defect features as the coating defect detection result.

[0068] In this embodiment, each texture feature in the texture distribution map is identified to determine the texture pattern in each region. For example, by analyzing factors such as the change pattern of gray values, the uniformity of the texture, and the complexity, different texture features in the region are identified, such as smooth, rough, fluctuating, etc. And it is matched with the defect features in the coating defect library.

[0069] The coating defect library is a database that pre-stores various common coating defect features, including the texture feature information of different types of defects. For example, defects such as uneven coating, scratches, and bubbles all have corresponding texture patterns, and the coating defect library can provide the standard texture features of these defects. After identifying the texture features in the texture distribution map, they will be compared with the features in the coating defect library to determine whether there is a texture pattern that matches the defect features.

[0070] According to the matching result with the coating defect library, determine whether there are defect features and the specific types of these defects as the defect features. Furthermore, partition mapping identification of the defect features is performed in the texture distribution map. Exemplarily, the identification process is through highlighting or using specific marks in the texture distribution map. The identified defect areas can be further analyzed to understand the nature, severity, and possible impacts of the defects.

[0071] Finally, the texture distribution map after defect identification and marking becomes the coating defect detection result. This result provides an effective basis for subsequent production process optimization and quality control. Through this method, the detection system can efficiently and accurately identify the defects in the coating on the surface of the steel strip, thereby timely discovering and correcting possible problems in the production process to improve production efficiency and product quality.

[0072] In summary, by identifying the texture distribution map and matching it with the coating defect library, the image processor can accurately identify the defect features on the surface of the steel strip and mark them in the texture distribution map, providing an accurate basis for defect detection and subsequent process optimization.

[0073] Further, after identifying the texture distribution map according to the defect features, step S4 of the present application includes:

[0074] Set the production line sampling interval. According to the production line sampling interval, control the acquisition device to perform acquisition parameter adjustment, execute interval image sampling and defect analysis, and add them to the detection database, where the detection database is used to store the texture distribution map after defect feature identification; constrained by a preset sampling frequency, splice the texture distribution maps in the detection database and perform global determination to locate process defect features; based on the process defect features, trace the steel strip production process and provide guidance for optimizing the steel strip production process.

[0075] In this embodiment, the production line sampling interval refers to the time interval for the detection system to collect images of the steel strip surface during the production process, usually expressed in time units such as seconds and minutes. Setting an appropriate sampling interval can ensure that the detection system can obtain sufficient data for accurate defect detection without affecting production efficiency. If the sampling interval is too long, some minor defects may be missed; if the interval is too short, it may lead to data redundancy and increased processing burden. Therefore, it is crucial to reasonably set the sampling interval according to the actual production situation.

[0076] After determining the sampling interval, control the acquisition device to perform acquisition parameter adjustment. By precisely controlling the acquisition operation of the adjustment device, the quality and stability of each image acquisition process can be ensured. Exemplarily, the acquisition parameter adjustment of the acquisition device includes adjusting the position perspective, exposure time, light source intensity, focus distance, etc. of the image sensor to adapt to the surface state of the steel strip under different working conditions and ensure the clarity and accuracy of the acquired images. After each adjustment, the system will automatically trigger image acquisition and execute defect analysis according to the sampling interval.

[0077] After performing interval image sampling and defect analysis, obtain the identified defect features and add them to the detection database. The detection database is used to store the texture distribution map after defect identification and is archived according to the acquisition time series to ensure that each image can be associated with the corresponding production batch and process conditions. The role of the detection database is to provide support for subsequent data analysis, including process optimization and defect prediction.

[0078] Based on a preset sampling frequency, splice and globally determine the texture distribution maps in the detection database. The preset sampling frequency means that a set number of texture distribution maps are used as a group and spliced at a set time interval. That is, the texture distribution maps collected at different time points are arranged in time series to determine the complete steel strip map of the group, which is used as the determination global domain for analysis to locate common process defect features. For example, some defects may only occur within a specific time period, and the process of splicing and global determination can reveal the occurrence patterns and time rules of these defects.

[0079] Based on the identified process defect characteristics, the system further conducts a traceability analysis of the steel strip production process. Traceability analysis refers to tracing the possible causes of defects in each link of the process production, helping to identify whether a defect is caused by a specific process parameter (such as temperature, speed, etc.). This process locates the specific defect source by analyzing historical data, production parameters, and process conditions.

[0080] Finally, based on the process defect characteristics, that is, the generally applicable process traceability defects, it can provide optimization guidance for the steel strip production process. This includes adjusting or optimizing the process parameters during production to eliminate the occurrence of defects. The optimization suggestions can cover multiple aspects, such as adjusting the plating solution composition, optimizing the surface treatment process of the steel strip, or adjusting the working state of the equipment, etc., so as to improve production efficiency and product quality.

[0081] Furthermore, the detection database performs texture distribution map stitching and global determination to locate the process defect characteristics. Step S4 of this application includes:

[0082] Perform texture distribution map stitching to determine the stitched texture distribution map; traverse the stitched texture distribution map, and locate frequent item defects based on the tangential position and longitudinal frequency of the plating defects, where the defect position is located by the tangential position, and the frequent items are screened by the longitudinal frequency; based on the frequent item defects, locate the process defect characteristics.

[0083] In this embodiment, the purpose of stitching the texture distribution map is to combine the texture image data from different sampled steel strip sections into a complete image for more comprehensive defect analysis. Ensure the continuity and integrity of the image. For example, if image data is collected at regular intervals or at different steel strip positions during production, then these image data will be stitched according to certain rules (such as time sequence or spatial position) to form a large texture distribution map containing multiple regions.

[0084] Next, analyze the stitched texture distribution map one by one, paying particular attention to the tangential position and longitudinal frequency of the plating defects. The tangential position refers to the position of the defect on the surface of the steel strip in the transverse direction, usually located based on the production direction of the steel strip. The longitudinal frequency refers to the frequency of defect occurrence perpendicular to the production direction of the steel strip. By analyzing these tangential positions and longitudinal frequencies, the system can identify which positions have more frequent defects, thereby screening out the frequently occurring defect items. This process helps to discover those common or recurring defect types, which may be caused by problems in certain links of the production process.

[0085] During the process of locating frequently-occurring item defects, the tangential position is used to determine the specific position of the defects, that is, to clarify in which tangential area of the steel strip the defects are located. The steel strip is continuously produced along the vertical production direction. Defect frequency statistics are carried out along the long section direction, and defect position location is carried out along the short section side. The longitudinal frequency helps the system to screen out those defects that frequently occur in different sampling time periods. The screening of frequently-occurring items can help identify those defects that persist and may be associated with a certain specific process parameter.

[0086] Based on the information of frequently-occurring item defects, the location of process defect characteristics is further carried out. Process defect characteristics refer to the defect patterns caused by improper process parameters during the production process. By analyzing the located frequently-occurring item defects, the system can trace back to the root cause of the defects and determine their relationship with production process parameters. For example, if a specific type of defect frequently appears in certain areas, the system may identify that there are problems with the production process (such as plating solution temperature, steel strip speed, etc.) in that area, thereby providing guidance for the improvement of the production process.

[0087] Furthermore, the steps of this application also include:

[0088] Sensing and detecting the plating solution parameters to determine the real-time plating solution data, where the plating solution parameters at least include composition, temperature, and speed; based on the real-time plating solution data, constructing a plating solution state curve based on the production time series; identifying the plating solution state curve, and performing plating solution instability analysis according to the parameter variation vector to generate defect warning information.

[0089] In this embodiment, first, the plating solution parameters are sensed and detected to obtain real-time plating solution data. Plating solution parameters refer to the key factors that control the coating quality during the coating production process, including the composition, temperature, and speed of the plating solution, etc. Among them, the plating solution composition determines the chemical composition and structure of the coating, the plating solution temperature affects the adhesion and uniformity of the coating, and the plating solution speed affects the thickness and uniformity of the coating. These parameters directly affect the coating quality of the steel strip, so their real-time monitoring is necessary. By installing appropriate sensors, such as temperature sensors, flow sensors, and chemical composition sensors, etc., the plating solution parameter data can be obtained in real time.

[0090] Based on the real-time plating solution data, a plating solution state curve based on the production time series is constructed. The plating solution state curve is a graphical representation that reflects the change trend of plating solution parameters. It forms a continuously changing curve by associating time with the corresponding plating solution parameter data. For example, during the production process, as time goes by, the temperature, composition, and speed of the plating solution will fluctuate, forming a time series curve. This curve reflects the change law of the plating solution state and can help engineers monitor the stability of the plating solution and the rationality of the process.

[0091] Further identify the change patterns in the plating solution state curve, with particular attention to the gradient vectors of the plating solution parameters. The gradient vector refers to the rate and direction of change of the plating solution parameters over a certain period of time. Identify any abnormal fluctuations or instability phenomena in the plating solution state. For example, if the rate of change of the plating solution temperature suddenly increases, or the change in the plating solution composition is abnormal, it may indicate an unstable process condition, which may lead to the generation of coating defects.

[0092] Based on the analysis of the plating solution state curve, conduct an instability analysis of the coating processing during the production process and generate defect warning information. The process of instability analysis includes determining whether the change in the plating solution parameters exceeds the preset safety range and whether there is a trend that has an adverse impact on the coating quality. If an instability of a certain parameter of the plating solution is detected, the system will issue an alarm through the warning mechanism to prompt the operator to take timely corrective measures. For example, if the plating solution temperature is too high or too low, or the chemical composition of the plating solution does not meet the requirements, generate defect warning information and display the specific abnormal location and parameters to help production personnel intervene as early as possible, thereby reducing the occurrence of coating defects.

[0093] A method for detecting coating defects on the surface of a steel strip provided by this application has the following technical effects:

[0094] 1. By traversing the pixels of the grayscale steel strip image and binarizing the irregularly partitioned grayscale image, the construction process of the texture distribution map is optimized. It can automatically extract representative features from the image, reduce manual intervention, improve the image processing efficiency and quality, and enhance the adaptive ability and intelligent level of the defect detection system. Through pixel recognition and the extraction and variance calculation based on Haar features, construct a texture distribution map and match it with the coating defect library for defect detection. It can efficiently and accurately identify the coating defects on the surface of the steel strip, improving the detection speed and accuracy.

[0095] 2. By the acquisition controller adaptively regulating the acquisition device, based on the real-time steel strip conveying and positioning, adjust the relative position of the acquisition equipment to optimize the image acquisition process. Improve the quality and accuracy of the acquired images, reduce the errors caused by inaccurate equipment positions, ensure the stability and consistency of image acquisition, and thus improve the accuracy of subsequent image processing and defect detection.

[0096] 3. By the sensor real-time detecting the plating solution parameters, constructing the plating solution state curve based on the time series data, and conducting the plating solution instability analysis to generate defect warning information. It can real-time monitor the state of the plating solution, predict whether the plating solution is unstable, discover potential process defect problems in advance, and reduce the occurrence of coating quality problems.

[0097] 4. Through the splicing of texture distribution maps and global determination, combined with the tangential position and longitudinal frequency, the process defect features with universality are located, realizing the accurate positioning and analysis of process defects, optimizing the production process and quality control, and reducing the defect rate in production.

[0098] In summary, the defect detection efficiency, accuracy, and stability of the production process of steel strips are improved, and intelligent and automated process control is achieved.

[0099] Through the foregoing detailed description of a method for detecting defects in the surface coating of a steel strip in this specification, those skilled in the art can clearly know a method for detecting defects in the surface coating of a steel strip in this embodiment. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple. For related parts, reference may be made to the description in the method part.

[0100] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to these embodiments shown herein, but will be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for detecting defects in coating on a steel strip surface, characterized in that: The method comprises: The collection device is installed on the steel strip production line, and a collection controller is configured, wherein the image sensor-light source is used as the collection device, the local section of the steel strip is used as the collection target, and the relative position parameter is used as the collection control condition; By positioning the steel belt conveyor, the acquisition device is adaptively regulated in combination with the acquisition controller to acquire the steel belt image of the section; Connecting to an image processor, performing pixel traversal and identification partitioning on the steel strip image of the section, performing extraction and variance calculation based on Haar features on a partition-by-partition basis, and constructing a texture distribution map, wherein the pixel trend change rule is used as a partitioning standard, and the image processor is built into the defect detection system; The texture distribution map is identified and matched based on a coating defect library to determine defect features, and the texture distribution map is marked according to the defect features as a coating defect detection result.

2. A method for detecting defects in coating on a steel strip surface as claimed in claim 1, characterized in that: The configuration acquisition controller includes: Taking the acquisition target as the center, determining the standard relative position parameters of the acquisition device relative to the acquisition target, which include the relative position relationship and acquisition control parameters; With the goal of adjusting the real-time relative position parameter compared to the standard relative position parameter, the acquisition controller is constructed and a communication connection is established between the acquisition controller and the acquisition device.

3. A method for detecting defects in coating on a steel strip surface as claimed in claim 1, characterized in that: The constructing of the texture distribution map comprises: Constructing the image processor in a data-driven supervised training manner; The image processor receives the segment steel strip image and performs grayscale processing to determine a grayscale steel strip image; Perform pixel traversal on the grayscale steel strip image, partition it according to the pixel trend change rule, and determine N partition grayscale images, wherein the partition is irregular and the pixel trend change rule of each partition grayscale image is consistent; Interact with the standard pixel law, binarize the N partition grayscale images, and screen the M partition grayscale images, wherein the standard pixel law is the pixel law when the coating quality meets the standard, and M is a positive integer less than or equal to N; Constructing a texture distribution map according to the M-item partition grayscale map; Among them, the pixels that are consistent with the standard pixel rule are binarized as 0, and the pixels that are inconsistent with the standard pixel rule are binarized as 1. The M-item partition grayscale image is the part that is binarized as 1.

4. A method for detecting defects in coating on a steel strip surface as claimed in claim 3, characterized in that: According to the M-item partition grayscale map, a texture distribution map is constructed, including: Traversing the M-item partition grayscale images, extracting Haar features in sequence, and determining the M-item partition Haar features, wherein the Haar features are used to measure the grayscale change pattern; Traversing the M-item partition grayscale images, calculating the partition pixel value variances in turn, and determining the M-item partition variances, wherein the pixel value variance is used to measure the uniformity and complexity of the texture; The texture distribution map is constructed according to the M partition Haar features and the M partition variances.

5. A method for detecting defects in coating on a steel strip surface as claimed in claim 4, characterized in that: Constructing the texture distribution map according to the M partition Haar features and the M partition variances, including: Mapping the M partition Haar features and the M partition variances to determine M texture features; Based on the distribution position of the section steel strip map, the M texture features are distributed and located, the NM partitions are blanked, and the texture distribution map is determined by performing image reconstruction processing.

6. A method for detecting defects in coating on a steel strip surface as claimed in claim 1, characterized in that: After the texture distribution map is marked according to the defect feature, the method includes: Setting a production line sampling interval, and controlling the acquisition device to perform acquisition parameter adjustment according to the production line sampling interval, performing interval image sampling and defect analysis, and adding the images to the detection database, wherein the detection database is used to store the texture distribution map after the defect feature is identified; Based on the preset sampling frequency as a constraint, texture distribution map splicing and global determination are performed on the detection database to locate process defect features; Based on the process defect characteristics, the steel strip production process is traced and process optimization guidance for steel strip production is provided.

7. A method for detecting defects in coating on a steel strip surface as claimed in claim 6, characterized in that: The detection database performs texture distribution map splicing and global determination to locate process defect features, including: Perform texture distribution map splicing to determine the spliced ​​texture distribution map; Traversing the splicing texture distribution map, locating frequent defects based on the tangential position and longitudinal frequency of the coating defects, wherein the defect position is located based on the tangential position, and frequent items are screened based on the longitudinal frequency; Based on the frequent defects, the process defect characteristics are located.

8. A method for detecting defects in coating on a steel strip surface as claimed in claim 1, characterized in that: The method further comprises: Performing sensing detection on plating bath parameters to determine real-time plating bath data, wherein the plating bath parameters at least include composition, temperature, and speed; Based on the real-time plating solution data, construct a plating solution state curve based on a production time series; The plating solution state curve is identified, plating solution instability analysis is performed according to the parameter variation vector, and defect warning information is generated.

Images