Steel surface defect online detection method, device and system and storage medium

Abstract

The invention provides a steel surface defect online detection method, device and system and a storage medium, and relates to the technical field of defect detection. The method comprises the following steps: acquiring a preset reference spot map; wherein the reference spot map is obtained by establishing a coordinate system for a first aluminum nitride spot image on the surface of a defect-free steel sample under a scanning electron microscope; after the target steel leaves the annealing furnace, generating raised aluminum nitride spots on the surface of the target steel through a pulse nitriding process by utilizing the residual temperature of the target steel; collecting a second aluminum nitride spot image on the surface of the target steel, converting the second aluminum nitride spot image into the coordinate system, and generating an online spot map; and taking the reference spot map as priori knowledge of a pre-trained defect identification model, inputting the online spot map into the defect identification model, and identifying the defect of the target steel. According to the invention, the interference of the steel surface condition and the image quality can be overcome, and the online detection accuracy and efficiency of the steel are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of defect detection, and in particular to a steel surface defect online detection method, device, system and storage medium. BACKGROUND

[0002] The surface quality of steel directly determines the mechanical properties and service life of the steel, especially for special steels such as electrical steel, surface defects can seriously affect the core performance of the steel. With the development of steel production towards high speed and continuity, efficient and accurate surface defect online detection technology has become the core requirement of quality control.

[0003] Current steel surface defect detection mainly adopts manual visual inspection or traditional machine vision detection. Manual detection is highly dependent on experience, and traditional machine vision detection is based on basic defect shape recognition algorithm. Due to the non-contact feature, it is widely used in electrical steel detection.

[0004] However, manual detection is low in efficiency and high in error and omission rate, and cannot adapt to high-speed production lines; traditional machine vision detection is easily affected by surface conditions such as oil film and image quality, and lacks precision in identifying small defects, and relies on a large number of samples, making it difficult to achieve stable and reliable recognition when defect samples are scarce.

[0005] Therefore, there is an urgent need for a steel surface defect online detection method that can overcome the interference of steel surface conditions and image quality to improve the online detection accuracy and efficiency of steel. SUMMARY

[0006] Therefore, the embodiments of the present application provide a steel surface defect online detection method, device, system and storage medium to improve the online detection accuracy and efficiency of steel surface defects.

[0007] In a first aspect, the embodiments of the present application provide a steel surface defect online detection method, comprising: obtaining a preset reference spot map; wherein the reference spot map is obtained by establishing a coordinate system through a first aluminum nitride spot image of a defect-free steel sample surface under a scanning electron microscope; After the target steel leaves the annealing furnace, the residual temperature of the target steel is used to generate protruding aluminum nitride spots on the surface of the target steel through pulse nitriding process; Collecting a second aluminum nitride spot image of the target steel surface, and converting the second aluminum nitride spot image into the coordinate system and generating an online spot map; The reference spot map is used as prior knowledge of a pre-trained defect recognition model, the online spot map is input into the defect recognition model, and the defects of the target steel are recognized.

[0008] In a possible implementation, the collecting the second aluminum nitride spot image of the target steel material surface comprises: The target steel material surface is irradiated by a coaxial white light source, and a line array camera system is used to scan the surface of the target steel material. Whenever the target steel material moves a certain distance, the line array camera system is controlled to synchronously collect images of the target steel material surface by using blue light and near-infrared light, and the blue light image and the near-infrared light image are subjected to difference calculation, so as to retain the aluminum nitride spot signal, thereby obtaining the second aluminum nitride spot image.

[0009] In a possible implementation, the converting the second aluminum nitride spot image into the coordinate system and generating an online spot map comprises: A third aluminum nitride spot image of the first batch of offline steel material surface on the same day under a scanning electron microscope is obtained in the coordinate system; The second aluminum nitride spot image is converted into the coordinate system according to the third aluminum nitride spot image; The position coordinates and size parameters of the aluminum nitride spots in the second aluminum nitride spot image in the coordinate system are detected, thereby obtaining the online spot map.

[0010] In a possible implementation, the converting the second aluminum nitride spot image into the coordinate system according to the third aluminum nitride spot image comprises: According to the grain boundary network of the steel material, the grain boundary contrast feature points in the third aluminum nitride spot image and the second aluminum nitride spot image are enhanced and extracted, respectively; The second aluminum nitride spot image is converted into the coordinate system according to the matching relationship of the grain boundary contrast feature points of the third aluminum nitride spot image and the second aluminum nitride spot image.

[0011] In a possible implementation, the obtaining a preset reference spot map comprises: The upper left corner of the first aluminum nitride spot image is taken as an origin, and a pixel is taken as a unit length, so as to establish the coordinate system; According to a preset gray threshold, the position coordinates and size parameters of each aluminum nitride spot in the first aluminum nitride spot image are extracted, thereby obtaining the preset reference spot map.

[0012] In a possible implementation, in the training process of the defect identification model, the penalty weight of the reference spot map identification area defect identification error in the model loss function is set to be greater than 1.

[0013] In a possible implementation, the generating the convex aluminum nitride spots on the surface of the target steel material by the pulse nitriding process comprises: The high-purity ammonia gas is controlled by an electromagnetic valve to pulse spray the target steel material, so as to generate active nitrogen atoms; The active nitrogen atoms combine with aluminum elements in the target steel material to generate aluminum nitride spots in situ on the surface of the target steel material, which are convex and firmly attached to the base body of the target steel material.

[0014] In a second aspect, an embodiment of the present application provides a steel surface defect online detection device, comprising: A reference map acquisition module is configured to acquire a preset reference spot map, wherein the reference spot map is obtained by establishing a coordinate system based on a first aluminum nitride spot image on the surface of a defect-free steel sample under a scanning electron microscope; A surface spot marking module is configured to generate convex aluminum nitride spots on the surface of the target steel material by pulse nitriding process by using the residual temperature of the target steel material after the target steel material leaves the annealing furnace; An online map generation module is configured to acquire a second aluminum nitride spot image on the surface of the target steel material, convert the second aluminum nitride spot image into the coordinate system, and generate an online spot map; A defect detection module is configured to use the reference spot map as prior knowledge of a pre-trained defect recognition model, input the online spot map into the defect recognition model, and identify defects of the target steel material.

[0015] In a third aspect, an embodiment of the present application provides a steel surface defect online detection system, comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes steps of the method in the first aspect or any one of the implementation manners of the first aspect.

[0016] In a fourth aspect, an embodiment of the present application provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement steps of the method in the first aspect or any one of the implementation manners of the first aspect.

[0017] Compared with the prior art, the embodiment of the present application has the following beneficial effects: In the embodiment of the present application, the reference spot map is obtained by establishing a coordinate system for the first aluminum nitride spot image of the surface of the defect-free steel sample under the scanning electron microscope, which can provide a spot standard reference for the defect-free steel, and clearly define the reference position and distribution characteristics of the aluminum nitride spot; the raised aluminum nitride spot is generated on the surface of the target steel through the pulse nitriding process, which forms a raised and visually recognizable physical marker without damaging the performance of the steel, thereby avoiding the impact of online detection on production efficiency; the second aluminum nitride spot image of the surface of the target steel is collected, and the second aluminum nitride spot image is converted into the coordinate system to obtain the online spot map, which quickly generates the online spot map of the target steel under the coordinate system of the scanning electron microscope, realizes the coordinate system of the spot position, and guarantees the accurate comparison of the spot; the reference spot map is used as the prior knowledge of the pre-trained defect recognition model, and the online spot map is input into the defect recognition model to obtain the defect classification result of the target steel, which provides standard data for defect recognition and enhances the recognition pertinence of the model to the abnormal state of the spot. The embodiment of the present application realizes the online accurate recognition of the surface defects of the steel without damaging the performance of the steel, and provides reliable technical support for the real-time quality control of the high-speed production line of electrical steel. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 is the implementation flow diagram of the steel surface defect online detection method provided by an embodiment of the present application; Figure 2 is the structural schematic diagram of the steel surface defect online detection device provided by an embodiment of the present application; Figure 3 is the schematic diagram of the steel surface defect online detection system provided by an embodiment of the present application. DETAILED DESCRIPTION

[0019] The present application will be described in more detail below with reference to specific embodiments. The following embodiments will help those skilled in the art to further understand the role of the present application, but do not limit the present application in any form. It should be noted that, for those skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made. These all belong to the protection scope of the present application.

[0020] It should be understood that when used in the specification and the appended claims of the present application, the term "comprising" indicates the presence of the described features, whole, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, whole, steps, operations, elements, components and / or sets thereof.

[0021] It should also be understood that the term "and / or" used in the specification and the appended claims of the present application means any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0022] In the description of the present application and the appended claims, the terms "first", "second", "third", etc. are used only to distinguish descriptions and cannot be understood as indicating or implying relative importance.

[0023] In the present application, the reference "one embodiment" or "some embodiments" means that the specific features, structures or characteristics described in connection with the embodiment are included in one or more embodiments of the present application. Therefore, the statements "in one embodiment", "in some embodiments", "in other some embodiments", "in further some embodiments" and the like appearing in different places in the specification are not necessarily all referring to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized. The terms "include", "contain", "have" and their variants mean "include but not limited to", unless otherwise specifically emphasized.

[0024] In addition, "a plurality of" mentioned in the embodiments of the present application should be interpreted as two or more.

[0025] Electrical steel as the core soft magnetic material of new energy vehicles, high-end transformers and other equipment, the surface quality of the equipment is of great importance to the performance and reliability of the terminal product. Surface scratches, indentation and other small defects will destroy the magnetic domain continuity and affect the energy efficiency of the equipment. The high-speed characteristics of modern electrical steel production lines have put forward real-time and accurate requirements for defect detection.

[0026] Traditional machine vision detection can realize non-contact detection by combining camera image acquisition and algorithm recognition of defects, but it is easily affected by optical interference caused by oil film and other substances remaining on the surface of the steel, it is difficult to accurately distinguish between small defects and false features, and the high-resolution equipment and a large number of samples required for model training also result in high application costs. The detection speed of non-destructive testing technologies such as eddy current detection and ultrasonic detection cannot match the production line, and the cost of equipment purchase and maintenance is high, which is difficult for small and medium-sized enterprises to bear. The shortcoming of these existing technologies is that they cannot balance detection accuracy, speed and cost, and it is difficult to directly conduct comprehensive detection on the production line, so there is an urgent need for a detection scheme that can adapt to high-speed production lines, has strong anti-interference ability and controllable cost. Therefore, it is of great practical significance and application value to develop a steel surface defect online detection method that can overcome the interference of the surface condition of the steel and the image quality.

[0027] Referring to Figure 1 , the embodiments of the present application provide a steel surface defect online detection method, which is described in detail as follows: Step S101, a preset reference spot map is obtained; wherein the reference spot map is obtained by establishing a coordinate system through a first aluminum nitride spot image of the surface of a defect-free steel sample under a scanning electron microscope.

[0028] In the embodiment of the present application, the reference spot map is obtained by establishing a coordinate system for the first aluminum nitride spot image of the surface of the defect-free steel sample under the scanning electron microscope, which can provide a spot standard reference for the defect-free steel, and clearly define the reference position and distribution characteristics of the aluminum nitride spot.

[0029] In step S102, after the target steel leaves the annealing furnace, the raised aluminum nitride spots are generated on the surface of the target steel by the pulse nitriding process using the residual temperature of the target steel.

[0030] In the embodiment of the present application, the raised aluminum nitride spots are generated on the surface of the target steel by the pulse nitriding process, which forms a physical mark that is raised and visually recognizable without damaging the performance of the steel, thereby avoiding the impact of online detection on production efficiency.

[0031] In step S103, the second aluminum nitride spot image of the surface of the target steel is collected, and the second aluminum nitride spot image is converted into the coordinate system to obtain the online spot map.

[0032] In the embodiment of the present application, the online spot map of the target steel under the scanning electron microscope coordinate system is quickly generated, which realizes the coordinate system of the spot position and guarantees the accurate comparison of the spot.

[0033] In step S104, the reference spot map is used as the prior knowledge of the pre-trained defect recognition model, and the online spot map is input into the defect recognition model to obtain the defect classification result of the target steel.

[0034] In the embodiment of the present application, the reference spot map provides standard data for defect recognition, and enhances the recognition pertinence of the model to the abnormal state of the spot.

[0035] The defect recognition model is a machine learning model specially established for online detection of the high-speed production line of the steel, which can adapt to the real-time recognition of the small real defects on the surface of the electrical steel. In the steel detection scene, there may be rolling marks, oil films, uneven lighting, camera shaking or dust shielding, and other non-real defects. Such surface appearance differences do not affect the product function and service life, but are easy to interfere with traditional detection and judgment.

[0036] The recognition basis of the defect recognition model is that the real defect will cause the change of the material properties on the surface of the steel (such as the appearance of crystal defects on the surface of the steel), and then the aluminum nitride spots generated by the pulse nitriding process will appear abnormal in the distribution area, position, size, etc. The non-real defect will not have a substantial impact on the spot characteristics. The embodiment of the present application focuses on the change of the spot characteristics based on this difference, realizes the accurate recognition of the real defect, and excludes the interference of the non-real defect.

[0037] Exemplarily, the defect recognition model of the embodiment of the present application adopts a convolutional neural network and deeply fuses the prior knowledge of the benchmark spot map. The defect recognition model of the embodiment of the present application is described in detail as follows. I. Constructing a defect detection model The model adopts a double-branch parallel input structure, and two inputs are simultaneously sent into a convolutional neural network for processing, which specifically includes: An online spot map branch (3-channel input), which inputs 3 types of images related to the actual spots of the target steel and directly splices them into a 3-channel format to input the model, providing the real spot features of the steel to be detected: Channel 1: original online spot map (spot original image of the steel to be detected); Channel 2: gray-scale normalized online spot map (gray value calibration is performed on the original online spot map to eliminate the interference of light and device acquisition differences); Channel 3: edge-enhanced online spot map (edge highlighting is performed on the original online spot map to make the outlines of the spots and potential defects clearer).

[0038] A benchmark spot map branch (2-channel input), which inputs 2 types of reference images related to the spots of the defect-free steel sample and directly splices them into a 2-channel format to input the model, providing the spot prior standard of the defect-free steel, which specifically includes: Channel 1: spot position distribution map (spot space distribution reference map of the defect-free steel, which clearly shows the positions of normal spots); Channel 2: spot size normalized distribution map (spot size standard reference map of the defect-free steel, which clearly shows the size range of normal spots).

[0039] After the model obtains the input data, it first extracts features through 3 types of conventional convolution kernels, and then collects the features, which specifically includes: The double-branch inputs are synchronously extracted through the following 3 types of convolution kernels without interference: Size feature convolution kernel: filtering out miscellaneous points and noise that do not conform to the size of the spot, and only retaining the features of normal-sized spots; Position feature convolution kernel: capturing the arrangement rule of the spots and clearly showing the positions and adjacent distances of the spots; Topological feature convolution kernel: obtaining the topological relationship of the spots.

[0040] After the convolution kernel extraction, the two-branch input data respectively obtains corresponding spot feature information (the online branch is the spot feature of the steel to be detected, and the benchmark branch is the spot feature of the defect-free steel sample), and the difference features of the two groups of features are obtained according to the position correspondence relationship of the two groups of features, which are input into the activation layer, the pooling layer and the full connection layer of the convolutional neural network.

[0041] II. Training defect detection model With the benchmark spot map as prior knowledge, combined with the defect label corresponding to the online spot map, through label and multi-task training, the dependence of the model on defect samples is reduced, and the specific process is as follows: First, the defect label is normalized: the defect type is multi-classified and one-hot encoded, and the defect position and defect size are normalized.

[0042] With the online spot map, the benchmark spot map and the defect label as a complete sample, a multi-task parallel output head is designed after the full connection layer, which corresponds to the defect type, position, size and confidence respectively.

[0043] The joint loss function of the model is established, including classification loss, position loss, size regression loss and topological consistency loss, and the Adam optimizer is used, and the output result is verified once every 100 rounds, when the classification accuracy, positioning deviation and size error all meet the preset model precision threshold, the training is stopped, and the model weight is saved.

[0044] III. Using the trained defect detection model to identify defects of the steel to be detected The online spot map and the corresponding benchmark spot map of the steel to be detected are obtained and input into the defect detection model. The model generates a prediction result from the multi-task output head, restores the actual defect type, defect position, defect size from the output result, and outputs the prediction confidence to prompt the reliability of the defect identification result.

[0045] The embodiment of the present application realizes online accurate identification of surface defects of steel without damaging the performance of the steel, and provides reliable technical support for real-time quality control of high-speed production lines of electrical steel.

[0046] In one possible implementation, a preset benchmark spot map is obtained, including: A coordinate system is established with the upper left corner of the first aluminum nitride spot image as the origin and with pixels as the unit length; According to the preset gray scale threshold, the position coordinates and size parameters of each aluminum nitride spot in the first aluminum nitride spot image are extracted to obtain the preset benchmark spot map.

[0047] The embodiment of the present application accurately constructs the preset benchmark spot map by establishing a standardized coordinate system and extracting the position coordinates and size parameters of the spots, and provides a reliable standard for spot image comparison and defect identification.

[0048] In one possible implementation, the pulsed nitriding process is used to generate convex aluminum nitride spots on the surface of the target steel, including: The high-purity ammonia gas is pulsed sprayed on the target steel through the electromagnetic valve to generate active nitrogen atoms; By combining the active nitrogen atom with the aluminum element in the target steel material, aluminum nitride spots in relief and firmly attached to the target steel material substrate are generated in situ on the surface of the target steel material.

[0049] The embodiment of the present application provides stable and identifiable physical markers for steel surface defect identification by generating aluminum nitride spots in relief in situ on the surface of the steel.

[0050] In some embodiments, when collecting the image of the aluminum nitride spots on the surface of the target steel material online, the oil film and other impurities remaining on the surface of the steel material can easily cause optical interference in the imaging process. It is difficult to separate the interference signal from the spot signal using conventional lighting equipment and shooting methods, resulting in low recognition of the spot image. Therefore, collecting the second aluminum nitride spot image on the surface of the target steel material can include: A coaxial white light source is used to irradiate the surface of the target steel material, and a line array camera system provided with a polarizer is used to scan the surface of the target steel material; Whenever the target steel material moves a certain distance, the line array camera system is controlled to synchronously collect the images of the surface of the target steel material under blue light and near-infrared light, and the blue light image and the near-infrared light image are differentially calculated to retain the aluminum nitride spot signal, thereby obtaining the second aluminum nitride spot image.

[0051] In the embodiment of the present application, by irradiating with a coaxial white light source, scanning with a line array camera system equipped with a polarizer, synchronously collecting the blue light and near-infrared light images of the surface of the steel material and performing differential calculation, the optical interference signal of the oil film and other impurities on the surface of the steel material can be effectively separated, the aluminum nitride spot signal can be accurately retained, and the recognition of the second aluminum nitride spot image can be significantly improved, thereby providing high-quality data support for coordinate conversion and defect identification.

[0052] In some embodiments, in the process of converting the second aluminum nitride spot image to the coordinate system, since the scales of the online detection image and the scanning electron microscope image are different, directly matching the contours of the two may cause deviation and affect the accuracy of the online spot map. Therefore, converting the second aluminum nitride spot image to the coordinate system and generating the online spot map can include: Obtaining a third aluminum nitride spot image of the first batch of offline steel material on the surface of the day in the coordinate system under the scanning electron microscope; Converting the second aluminum nitride spot image to the coordinate system according to the third aluminum nitride spot image; Detecting the position coordinates and size parameters of the aluminum nitride spots in the second aluminum nitride spot image in the coordinate system to obtain the online spot map.

[0053] In the embodiment of the present application, by virtue of the consistency of the first batch of offline steel material and the target steel material in material and process, the scanning electron microscope image thereof is used as a coordinate template, and the target steel material does not need to be observed by the scanning electron microscope, so that the coordinates of the spots on the surface of the target steel material in the coordinate system can be obtained conveniently and quickly.

[0054] In some embodiments, when converting the second aluminum nitride spot image coordinates according to the third aluminum nitride spot image, if only spot features are considered for coordinate matching, it is easy to cause insufficient matching accuracy due to uneven spot distribution or local image loss, etc., and it is difficult to achieve accurate alignment of coordinates. Therefore, converting the second aluminum nitride spot image into the coordinate system according to the third aluminum nitride spot image can include: According to the grain boundary network of the steel, the grain boundary contrast feature points in the third aluminum nitride spot image and the second aluminum nitride spot image are enhanced and extracted respectively; According to the matching relationship of the grain boundary contrast feature points of the third aluminum nitride spot image and the second aluminum nitride spot image, the second aluminum nitride spot image is converted into the coordinate system.

[0055] In the embodiments of the present application, by matching the grain boundary contrast feature points of the third aluminum nitride spot image and the second aluminum nitride spot image, the precise alignment of the second aluminum nitride spot image to the coordinate system is realized, and the coordinate conversion accuracy is improved.

[0056] In some embodiments, during training of the defect recognition model, if the same penalty weight is used for the area with spots and the area without spots in the reference spot map, the model may not pay enough attention to the defects in the key area with spots, and the recognition accuracy of the key area may be low. Therefore, during the training process of the defect recognition model, the penalty weight of the model loss function for the defect recognition error of the identified area in the reference spot map can be set to be greater than 1.

[0057] In the embodiments of the present application, the defect recognition error penalty weight of the key area with spots in the reference spot map in the model loss function is set to be greater than 1, which can strengthen the model's learning attention to the key area and significantly improve the defect recognition accuracy of the area with spots.

[0058] In the input data of the defect recognition model, the reference spot map and the online spot map are aligned by coordinates, which can realize pixel-level correspondence; the difference between the reference spot map and the online spot map can reflect the surface defect condition of the steel, and strengthen the correlation judgment of the spot distribution. According to the adaptation requirement, the loss function of the defect recognition model applies higher penalty weight to the recognition error of the key area, so as to ensure that the model learning direction meets the detection requirements.

[0059] In this embodiment of the invention, stable and distinguishable aluminum nitride spots are generated in situ on the steel surface through a pulsed nitriding process. Optical interference signals are separated by an image acquisition scheme using multiple light sources and polarizers. Using the scanning electron microscope (SEM) images of the first batch of offline steel samples taken that day as coordinate templates, a unified coordinate system can be quickly achieved without requiring individual SEM inspection of each target steel sample. The accuracy of coordinate transformation is improved by matching grain boundary feature points in the steel, and the identification of defects in key areas is enhanced by adjusting the weights of the model loss function. This achieves accurate online identification of steel surface defects without compromising steel performance. This embodiment of the invention is adapted to the real-time detection requirements of high-speed production lines for electrical steel, providing efficient and reliable technical support for quality control during the production process.

[0060] This invention provides an online detection method for surface defects in steel, detailed below: (1) Obtain the baseline spot map.

[0061] Defect-free steel samples were selected and subjected to pulse nitriding marking followed by polishing. A light etching process was then performed using 4% nitric acid alcohol. Utilizing the difference in corrosion resistance between AlN and the steel matrix, the AlN spots appeared raised under a scanning electron microscope (SEM). The spot images of the defect-free steel samples were observed using SEM at 5000x magnification. Spots were extracted based on a preset grayscale threshold, and their location coordinates and diameters were determined to obtain a baseline spot map.

[0062] The process of extracting the spot coordinates is as follows: with the top left corner of the SEM image as the origin (0,0) and pixels as the unit length, set the SEM image coordinate system and record the coordinates of each AlN spot in the SEM image coordinate system.

[0063] (2) Perform pulse nitriding marking on the annealed steel.

[0064] Approximately 5 meters behind the annealing furnace outlet, utilizing the residual temperature of the target steel (820±10℃) after annealing, ammonia gas with a purity greater than 99.99% is controlled by a solenoid valve to... A pulse jet is applied to the target steel at a flow rate of 2-3 seconds. By controlling the temperature, time, and flow rate parameters, a diameter of [missing information] is generated on the surface of the target steel. The AlN spots have a distribution density of [missing information]. about.

[0065] The spot growth process is: the surface of high-temperature steel as a catalyst, prompting ammonia decomposition, generating a highly active nitrogen atom [N], active nitrogen atom [N] is adsorbed by the steel surface, and through short-range surface diffusion, it rapidly combines with Al elements in the steel to generate AlN clusters in situ on the surface of the steel. Because the ammonia injection duration is only 2-3 seconds, the active nitrogen atom [N] has not diffused to the deep part of the steel, so the AlN clusters formed only exist on the surface of the steel, with a diameter of , in the form of raised spots.

[0066] After pulse nitriding marking, the treated target steel continues to cool down relying on its own residual heat and air cooling, so that the AlN spots on the surface of the steel are firmly attached to the matrix. Ammonia pulse injection is carried out in a closed chamber with inert gas to prevent explosion and ensure process stability.

[0067] (3) Obtain the online spot map of the target steel.

[0068] After pulse nitriding marking and polishing treatment of the first batch of offline steels of the day, light etching is carried out with 4% nitric acid alcohol, and the spot images of the first batch of offline steels of the day are observed under SEM at 5000 times, and the position coordinates and diameter of the spots are collected according to the SEM image coordinate system.

[0069] A coaxial white light source with a hue angle of 20-40° is used to irradiate the surface of the target steel, and a line array camera system with a polarizer is used to scan the surface of the target steel. The coaxial light can highlight the edges and contours of the raised AlN spots on the surface of the steel, avoid shadow interference, and enhance the light and dark contrast of the edges of the AlN spots.

[0070] Whenever the target steel moves a certain distance, the line array camera system synchronously collects a row of images of the target steel surface using blue light and near-infrared light, and performs difference calculation on the collected blue light images and near-infrared light images to remove oil film interference fringes and retain aluminum nitride spot signals, obtaining the aluminum nitride spot image of the target steel.

[0071] For the same kind of steel produced under the same process conditions, the grain boundary network of the steel is stable and unchanged, therefore, through image processing algorithm, the aluminum nitride spot image of the target steel and the grain boundary contrast feature points in the SEM image of the first batch of offline steels of the day are enhanced and extracted respectively, according to the matching relationship of the grain boundary contrast feature points in the two images, the coordinates of the aluminum nitride spot image of the target steel are converted to the SEM image coordinate system, the position coordinates and diameter of the aluminum nitride spot of the target steel are determined, and the online spot map of the target steel is obtained.

[0072] The image processing algorithm does not directly match the AlN spots which are easy to change, but matches the stable and unchanging grain boundary contrast feature points. By aligning the two images, the position of the AlN spots attached in or on the grain boundary in the SEM image coordinate system is determined, so as to realize accurate coordinate conversion. The edge enhancement processing is performed on the SEM image and the optical image, the line features shared by the two are highlighted, the modal difference is suppressed, and the grain boundary profile is more clear and prominent.

[0073] (4) Surface defect recognition based on spot map.

[0074] The reference spot map is input as prior knowledge into the defect recognition model to be trained, and the defect recognition model is trained according to the defect steel spot map with a defect label, to obtain a pre-trained defect recognition model. In the loss function of the model training, a defect recognition error penalty weight of 1.3 is applied to the area in the reference spot map where the AlN spot exists, so that the model can sensitively identify the phenomena such as spot breakage and loss caused by real defects such as scratches and indentations.

[0075] The online spot map of the target steel is input into the trained defect recognition model to obtain the defect classification result of the target steel.

[0076] The traditional defect detection model needs to learn the characteristics of defects from a large amount of data, but in the industrial scene, the steel defect samples are relatively few, and the surface texture changes caused by rolling marks, oil film, uneven illumination, camera shaking, dust shielding and other reasons are harmless surface appearance differences that do not affect the product function and service life. The traditional defect detection model is only trained according to the steel surface image, which is easy to cause the model to misjudge.

[0077] The reference spot map identifies the position and size of the AlN spots under the condition of no defects, so the possible causes of the disappearance or deformation of these AlN spots are real defects that affect the performance of the target steel. In the training process, the model will be guided by the improved loss function to pay attention to the changes in these key areas.

[0078] In the embodiment of the present application, pulse nitriding is performed with the help of the annealing afterheat of steel, in-situ AlN spots are generated on the surface of the steel, no additional process equipment is needed, and the performance of the steel is not damaged, which provides stable and distinguishable physical markers for defect detection; through the imaging scheme of coaxial white light and polarizer, images are synchronously collected by double light and difference calculation is performed to remove interference, the spot coordinates are matched by combining the grain boundary contrast feature points, accurate coordinate system conversion is realized, and the online spot map under the SEM coordinate system can be obtained without SEM detection of the target steel.

[0079] Referring to Figure 2 The embodiment of the present application provides a steel surface defect online detection device 2, which comprises: A reference map acquisition module 21 is configured to acquire a preset reference spot map; wherein the reference spot map is obtained by establishing a coordinate system based on a first aluminum nitride spot image of a surface of a defect-free steel sample under a scanning electron microscope; A surface spot marking module 22 is configured to generate protruding aluminum nitride spots on the surface of the target steel by pulse nitriding process using the afterheat of the target steel after the target steel leaves the annealing furnace; An online map generation module 23 is configured to acquire a second aluminum nitride spot image of the surface of the target steel, convert the second aluminum nitride spot image into the coordinate system, and generate an online spot map; A defect detection module 24 is configured to use the reference spot map as prior knowledge of a pre-trained defect recognition model, input the online spot map into the defect recognition model, and recognize the defects of the target steel.

[0080] In a possible implementation, the reference map acquisition module 21 is configured to establish a coordinate system with the upper left corner of the first aluminum nitride spot image as the origin and with pixels as the unit length; According to the preset gray threshold, the position coordinates and size parameters of each aluminum nitride spot in the first aluminum nitride spot image are extracted, and the preset reference spot map is obtained.

[0081] In a possible implementation, the surface spot marking module 22 is configured to pulse spray high-purity ammonia gas on the target steel through an electromagnetic valve to generate active nitrogen atoms; The active nitrogen atoms combine with aluminum elements in the target steel to generate protruding and firmly adhered aluminum nitride spots on the surface of the target steel in-situ.

[0082] In a possible implementation, the online map generation module 23 is configured to irradiate the surface of the target steel material with a coaxial white light source and scan the surface of the target steel material with a line array camera system; Whenever the target steel material moves a certain distance, the control line array camera system synchronously collects images of the surface of the target steel material with blue light and near-infrared light, performs difference calculation on the blue light image and the near-infrared light image, retains the aluminum nitride spot signal, and obtains a second aluminum nitride spot image.

[0083] In a possible implementation, the online map generation module 23 is further configured to obtain a third aluminum nitride spot image of the surface of the first batch of offline steel materials on the same day under a scanning electron microscope in the coordinate system; According to the third aluminum nitride spot image, the second aluminum nitride spot image is converted into the coordinate system; The position coordinates and size parameters of the aluminum nitride spots in the second aluminum nitride spot image in the coordinate system are detected, and an online spot map is obtained.

[0084] In a possible implementation, the online map generation module 23 is further configured to enhance and extract grain boundary contrast feature points in the third aluminum nitride spot image and the second aluminum nitride spot image according to the grain boundary network of the steel material. According to the matching relationship between the grain boundary contrast feature points of the third aluminum nitride spot image and the second aluminum nitride spot image, the second aluminum nitride spot image is converted into the coordinate system.

[0085] In the embodiment of the application, the reference map acquisition module establishes a standardized reference, and the surface spot marking module generates firm and protruding aluminum nitride spots without damaging the steel material substrate, thereby greatly reducing the detection cost. The online map generation module adopts a coordinate conversion mode of multi-light source collection and grain boundary feature point matching, effectively suppresses optical interference, ensures accurate alignment of the online spot image and the reference, and provides reliable data support for defect identification. The device realizes online real-time detection of the surface defects of the steel material through the cooperation of multiple modules, improves the pertinence of defect identification according to the prior knowledge of the reference spots, and completes the online detection of the high-speed production line.

[0086] Referring to Figure 3 , a schematic diagram of a steel surface defect online detection system 3 provided by an embodiment of the application is shown, and the details are as follows: As shown in Figure 3 , the steel surface defect online detection system 3 of this embodiment includes a processor 30 and a memory 31. The memory 31 stores a computer program 32. The processor 30 implements the steps in each of the method embodiments described above when executing the computer program 32. Alternatively, the processor 30 implements the functions of each module in each of the device embodiments described above when executing the computer program 32.

[0087] For example, the computer program 32 can be divided into one or more modules / units, which are stored in the memory 31 and executed by the processor 30 to complete the present application. The one or more modules / units can be a series of computer program instruction segments capable of completing a specific function, which are used to describe the execution process of the computer program 32 in the steel surface defect online detection system 3.

[0088] The steel surface defect online detection system 3 can include, but is not limited to, the processor 30 and the memory 31. Those skilled in the art can understand that the steel surface defect online detection system 3 can further include other components, such as an input / output device, a network access device, a bus, etc. Figure 3 The steel surface defect online detection system 3 shown in the figure is only an example and does not constitute a limitation on the steel surface defect online detection system 3. The steel surface defect online detection system 3 can include more or fewer components than those shown in the figure, or combine some components, or different components, for example, the steel surface defect online detection system 3 can also include an input / output device, a network access device, a bus, etc.

[0089] The processor 30 can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0090] The memory 31 can be an internal storage unit of the steel surface defect online detection system 3, such as a hard disk or a memory of the steel surface defect online detection system 3. The memory 31 can also be an external storage device of the steel surface defect online detection system 3, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the steel surface defect online detection system 3. Further, the memory 31 can include both the internal storage unit and the external storage device of the steel surface defect online detection system 3. The memory 31 is used to store the computer program 32 and other programs and data required by the steel surface defect online detection system 3. The memory 31 can also be used to temporarily store data that has been output or will be output.

[0091] For the convenience and brevity of description, only the above-mentioned division of each functional module / unit is exemplified, and in actual application, the above-mentioned functions can be completed by different functional modules / units according to needs. The above-mentioned modules / units can be realized in the form of hardware, software or a combination of hardware and software.

[0092] The embodiment of the present application further provides a computer readable storage medium, which stores a computer program. When the computer program is executed by a processor, the method in each method embodiment described above is realized.

[0093] The embodiment of the present application further provides a computer program product, which comprises a computer program. When the computer program is executed by a processor, the method in each method embodiment described above is realized.

[0094] The computer program comprises computer program code, which can be in the form of source code, object code, executable file or some intermediate form. The computer readable medium can include any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), electric carrier wave signal, telecommunication signal and software distribution medium, etc.

[0095] In the above embodiments, the description of each embodiment has its own emphasis, and the parts not described or recorded in a certain embodiment can be referred to the relevant description of other embodiments. If there is no special description and logical conflict, the terms and / or descriptions of different embodiments are consistent and can be mutually referred to, and the technical features in different embodiments can be combined to form new embodiments according to their inherent logical relationship.

[0096] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. An online detection method for surface defects in steel, characterized in that, include: A preset reference spot map is obtained; wherein, the reference spot map is obtained by establishing a coordinate system based on the first aluminum nitride spot image on the surface of a defect-free steel sample under a scanning electron microscope; After the target steel leaves the annealing furnace, the residual heat of the target steel is used to generate raised aluminum nitride spots on the surface of the target steel through a pulse nitriding process. Acquire images of second aluminum nitride spots on the surface of the target steel, transform the second aluminum nitride spot images into the coordinate system, and generate an online spot map; The baseline speckle map is used as prior knowledge for the pre-trained defect recognition model. The online speckle map is then input into the defect recognition model to identify defects in the target steel.

2. The online detection method for steel surface defects according to claim 1, characterized in that, The acquisition of the second aluminum nitride spot image on the surface of the target steel includes: The surface of the target steel is illuminated by a coaxial white light source, and the surface of the target steel is scanned by a line scan camera system; Whenever the target steel moves a certain distance, the linear array camera system is controlled to simultaneously acquire images of the target steel surface using blue light and near-infrared light. The blue light image and the near-infrared light image are differentially calculated, and the aluminum nitride spot signal is retained to obtain the second aluminum nitride spot image.

3. The online detection method for steel surface defects according to claim 1, characterized in that, The step of converting the second aluminum nitride spot image to the coordinate system and generating an online spot map includes: Obtain an image of the third aluminum nitride spot on the surface of the first batch of offline steel materials on the same day in the coordinate system using a scanning electron microscope; Based on the third aluminum nitride spot image, the second aluminum nitride spot image is transformed into the coordinate system; The position coordinates and size parameters of aluminum nitride spots in the second aluminum nitride spot image under the coordinate system are detected to obtain the online spot map.

4. The online detection method for steel surface defects according to claim 3, characterized in that, The step of transforming the second aluminum nitride spot image to the coordinate system based on the third aluminum nitride spot image includes: Based on the grain boundary network of the steel, the grain boundary contrast feature points are enhanced and extracted in the third aluminum nitride spot image and the second aluminum nitride spot image, respectively. Based on the grain boundary contrast feature point matching relationship between the third aluminum nitride spot image and the second aluminum nitride spot image, the second aluminum nitride spot image is transformed into the coordinate system.

5. The online detection method for steel surface defects according to any one of claims 1 to 4, characterized in that, The process of obtaining a preset baseline spot map includes: The coordinate system is established with the top left corner of the first aluminum nitride spot image as the origin and pixels as the unit length. Based on a preset grayscale threshold, the position coordinates and size parameters of each aluminum nitride spot in the first aluminum nitride spot image are extracted to obtain the preset reference spot map.

6. The online detection method for steel surface defects according to claim 1, characterized in that, During the training process of the defect recognition model, the penalty weight for incorrect defect recognition in the benchmark spot map identification area is set to be greater than 1 in the model loss function.

7. The online detection method for steel surface defects according to claim 1, characterized in that, The process of generating raised aluminum nitride spots on the surface of the target steel using pulsed nitriding includes: High-purity ammonia gas is pulsed and injected onto the target steel using a solenoid valve to generate active nitrogen atoms. The active nitrogen atoms combine with the aluminum element in the target steel to generate raised aluminum nitride spots on the surface of the target steel that are firmly attached to the target steel matrix.

8. An online detection device for steel surface defects, characterized in that, include: The benchmark map acquisition module is used to acquire a preset benchmark spot map; wherein, the benchmark spot map is obtained by establishing a coordinate system from the first aluminum nitride spot image on the surface of a defect-free steel sample under a scanning electron microscope; The surface spot marking module is used to generate raised aluminum nitride spots on the surface of the target steel by means of the residual heat of the target steel after it leaves the annealing furnace through a pulse nitriding process. An online map generation module is used to acquire images of second aluminum nitride spots on the surface of the target steel, convert the second aluminum nitride spot images to the coordinate system, and generate an online spot map; The defect detection module is used to use the benchmark spot map as prior knowledge for the pre-trained defect recognition model, input the online spot map into the defect recognition model, and identify the defects of the target steel.

9. An online detection system for steel surface defects, characterized in that, It includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method as described in any one of claims 1 to 7.

Images