Darkfield light type automated inspection system for metal surface defect using artifical intelligence model and operating method thereof

Abstract

An automated metal surface defect inspection system using a darkfield light method according to one embodiment of the present invention may include a lighting module that illuminates a metal surface using a darkfield light method, a shooting module that captures a metal surface to acquire metal surface image data, an artificial intelligence module that uses a deep learning artificial intelligence model to extract surface pattern data from the metal surface image data and compares and analyzes defect outliers calculated from the surface pattern data with preset reference defect outliers, and an alarm module that provides alarm information including surface pattern data and defect outliers based on the results of the comparison analysis.

Description

Technology Field

[0001] The present invention relates to an automated metal surface defect inspection system using a darkfield light method utilizing an artificial intelligence model and a method for operating the same. Background Technology

[0002] Various defects can occur on the metal surface during the metal production process. This not only leads to a deterioration in the quality of steel products but is also a major cause of significant scrap costs incurred during post-production quality inspections.

[0003] The reality is that an average of approximately 1 to 4% of defects occur daily, and the resulting scrap disposal costs amount to about 20 million won per month. Furthermore, the problem of increasing customer claims due to quality degradation continues to occur.

[0004] Conventional visual inspection and simple image analysis methods have limitations, such as inspection accuracy varying depending on operator skill level and the difficulty of real-time response. Additionally, existing inspection systems suffer from the problem of being unable to accurately identify various types of defects due to the influence of lighting conditions and shooting angles.

[0005] In particular, for small and medium-sized steel companies, the cost burden of process automation is significant, and the maintenance and management costs of existing inspection systems are also high, making its adoption difficult.

[0006] Therefore, an inspection system integrating process and management technologies capable of automatically detecting surface defects in steel products is required, and technology is needed to realize automation in the steel contract manufacturing industry and improve the efficiency of quality control. The problem to be solved

[0007] The technical problem that the present invention aims to solve is to streamline the metal surface defect detection process by applying artificial intelligence-based automation technology and to minimize errors occurring in existing visual inspection and simple image analysis methods.

[0008] In addition, the technical problem that the present invention aims to solve is to enable consistent quality inspection by utilizing a deep learning-based artificial intelligence model to analyze metal transported on a conveyor belt in real time and accurately identify the same metal even when it passes through different shooting modules.

[0009] In addition, the technical problem that the present invention aims to solve is to realize faster and more objective quality inspection than existing manual defect detection methods by utilizing a deep learning-based artificial intelligence model to precisely analyze metal surface patterns and automatically calculate defect anomalies, thereby reducing scrap costs and improving productivity.

[0010] In addition, the technical problem that the present invention aims to solve is to determine in real time whether there are defects on a metal surface based on defect data analyzed using a deep learning-based artificial intelligence model, and to automatically provide identification information and location information for metals in which defects are detected to support the decision-making of operators and to improve the quality control level of the steel processing process. means of solving the problem

[0011] The present invention may include a lighting module that illuminates a metal surface using a darkfield light method according to one embodiment of the present invention, a shooting module that captures the metal surface to acquire metal surface image data, an artificial intelligence module that utilizes a deep learning artificial intelligence model to extract surface pattern data from the metal surface image data and compares and analyzes defect outliers calculated from the surface pattern data with preset reference defect outliers, and an alarm module that provides alarm information including surface pattern data and defect outliers based on the results of the comparison analysis.

[0012] In addition, a shooting module according to one embodiment of the present invention is at least one CCTV and can shoot at least one metal surface being transported on a conveyor belt.

[0013] In addition, an artificial intelligence model according to one embodiment of the present invention can track and identify the same metal surface captured by different capturing modules on a conveyor belt using a re-identification model utilizing a deep learning-based deep appearance descriptor, and calculate defect outliers by analyzing the surface pattern data of the metal surface tracked and identified using an RNN-based LSTM algorithm.

[0014] In addition, an artificial intelligence module according to one embodiment of the present invention can determine whether a calculated defect outlier is greater than a reference defect outlier, and can calculate identification information and location information regarding a metal determined to be greater than the reference defect outlier and provide them additionally to an alarm module.

[0015] In addition, a method for operating an automated metal surface defect inspection system using a darkfield light method utilizing an artificial intelligence model according to one embodiment of the present invention may include the steps of: illuminating a metal surface using darkfield light by a lighting module; capturing a metal surface by a shooting module to acquire metal surface image data; extracting surface pattern data from the metal surface image data using a deep-learned artificial intelligence model by an artificial intelligence module, comparing and analyzing defect outliers calculated from the surface pattern data with preset reference defect outliers; and providing alarm information including surface pattern data and defect outliers based on the comparison analysis results by an alarm module. Effects of the invention

[0016] An automated metal surface defect inspection system using a darkfield light method utilizing an artificial intelligence model according to one embodiment of the present invention and a method for operating the same can streamline the metal surface defect detection process by applying artificial intelligence-based automation technology and minimize errors occurring in existing visual inspection and simple image analysis methods.

[0017] In addition, the automated metal surface defect inspection system using a darkfield light method utilizing an artificial intelligence model according to one embodiment of the present invention and the method of operating the same can enable consistent quality inspection by utilizing a deep learning-based artificial intelligence model to analyze metal transported on a conveyor belt in real time and accurately identifying the same metal even when it passes through different shooting modules.

[0018] In addition, the automated metal surface defect inspection system and method of operation using a darkfield light method utilizing an artificial intelligence model according to one embodiment of the present invention utilizes a deep learning-based artificial intelligence model to precisely analyze metal surface patterns and automatically calculate defect abnormalities, thereby realizing faster and more objective quality inspection than the existing manual defect detection method, and thereby reducing scrap costs and improving productivity.

[0019] In addition, the automated metal surface defect inspection system and the method of operating the same using a darkfield light method utilizing an artificial intelligence model according to one embodiment of the present invention can determine in real time whether there is a defect on the metal surface based on defect data analyzed by a deep learning-based artificial intelligence model, automatically provide identification information and location information for the metal in which a defect is detected to support the decision-making of the operator, and improve the quality control level of the steel processing process. Brief explanation of the drawing

[0020] FIG. 1 is a drawing of the components of a Darkfield light type automated metal surface defect inspection system utilizing an artificial intelligence model according to one embodiment of the present invention. FIG. 2 is a drawing showing the appearance of photographing a metal surface being transported on a conveyor belt using a darkfield light method according to one embodiment of the present invention. FIG. 3 is a drawing relating to a method for determining metal surface defects by an artificial intelligence model according to one embodiment of the present invention. FIG. 4 is a diagram relating to alarm information output to a control system according to one embodiment of the present invention. FIG. 5 is a flowchart illustrating a method of operating an automated metal surface defect inspection system using a darkfield light method utilizing an artificial intelligence model according to one embodiment of the present invention. Specific details for implementing the invention

[0021] Hereinafter, various embodiments of the present invention will be described in detail with reference to the attached drawings so that those skilled in the art can easily implement the present invention. The present invention may be embodied in various different forms and is not limited to the embodiments described herein.

[0022] To clearly explain the present invention, parts unrelated to the explanation have been omitted, and the same reference numerals are assigned to identical or similar components throughout the specification. Accordingly, the reference numerals described above may also be used in other drawings.

[0023] Furthermore, the size and thickness of each component shown in the drawings are depicted arbitrarily for convenience of explanation, and thus the present invention is not necessarily limited to what is illustrated. Thickness may be exaggerated in the drawings to clearly represent various layers and regions.

[0024] Furthermore, the expression "identical" in the explanation may mean "substantially identical." In other words, it may be an identicality to the extent that a person with ordinary knowledge would accept it as identical. Other expressions may also be those in which "substantially" has been omitted.

[0025] Furthermore, when a part in the description is described as 'including' a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components. As used in this specification, 'part' refers to a unit that processes at least one function or operation, and may, for example, mean software, FPGA, or hardware components. The function provided by the 'part' may be performed separately by multiple components or integrated with other additional components. The 'part' in this specification is not necessarily limited to software or hardware, and may be configured to reside in an addressable storage medium or configured to run one or more processors. Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

[0027] FIG. 1 is a drawing of the components of a Darkfield light type automated metal surface defect inspection system utilizing an artificial intelligence model according to one embodiment of the present invention.

[0028] An automated metal surface defect inspection system (1) using a darkfield light method utilizing an artificial intelligence model according to one embodiment of the present invention may include a lighting module (10), a shooting module (11), a sensor module (12), an artificial intelligence module (13), and an alarm module (14).

[0029] However, to operate the automated metal surface defect inspection system (1) using the Darkfield light method with an artificial intelligence model, it may include fewer or more components than the components shown in FIG. 1.

[0030] The lighting module (10) can illuminate the metal surface using a darkfield light method.

[0031] Dark-field illumination refers to a method of observing a metal surface by illuminating it with light at an oblique angle and collecting only the scattered or diffracted light. Dark-field illumination can be useful for detecting metal surface defects because light scatters and appears bright in rough or defective parts of the metal surface.

[0032] At least one lighting module (10) may be installed at the metal production site, and more specifically, may be located on or near the conveyor belt where the metal is transported.

[0033] For example, the lighting module (10) can be positioned at multiple points where light can be incident at oblique angles on a number of metal surfaces being transported on a conveyor belt. The positions can be set in various ways by a manager or operator.

[0034] The shooting module (11) can capture the metal surface and obtain metal surface image data.

[0035] Specifically, the imaging module (11) can collect light scattered or diffracted from a metal surface, thereby obtaining metal surface image data for the metal surface.

[0036] The shooting module (11) may be at least one CCTV, and more specifically, may be located on or near a conveyor belt where metal is transported.

[0037] For example, the shooting module (11) can be located at multiple points facing a number of metal surfaces being transported on a conveyor belt, and the locations can be set in various ways by a manager or operator.

[0038] During the process of transporting metal by a conveyor belt, any metal may come into contact with the imaging module (11). At this time, the imaging module (11) can collect light scattered or diffracted from the metal surface to obtain metal surface image data.

[0039] The sensor module (12) can sense environmental information. Environmental information sensed by the sensor module (12) may include temperature, humidity, illuminance, ultraviolet rays, rainfall, wind volume, wind speed, etc., but the present invention is not limited thereto.

[0040] Meanwhile, at least one sensor module (12) may be installed, and more specifically, it may be located on or near or outside the conveyor belt where the metal is transported.

[0041] The artificial intelligence module (13) can receive metal surface image data from the shooting module (11) and environmental information from the sensor module (12).

[0042] The artificial intelligence module (13) can extract surface pattern data from metal surface image data by utilizing a deep learning artificial intelligence model. The artificial intelligence module (13) can compare and analyze defect outliers calculated from surface pattern data with preset reference defect outliers by utilizing a deep learning artificial intelligence model.

[0043] At this time, the artificial intelligence model may be in a deep learning state using an RNN (Recurrent Neural Network)-based LSTM (Long Short-Term Memory) algorithm.

[0044] That is, the artificial intelligence module (13) can extract surface pattern data from metal surface image data by utilizing an RNN-based LSTM algorithm, and can calculate defect outliers by analyzing the surface pattern data. Specific details will be described later in FIG. 3.

[0045] The artificial intelligence module (13) can determine whether the calculated defect outlier is greater than or equal to the reference defect outlier. The artificial intelligence module (13) can calculate identification information and location information regarding the metal determined to be greater than the reference defect outlier and provide them to the alarm module (14).

[0046] The artificial intelligence module (13) can use a deep-learned artificial intelligence model to track and identify the same metal surface captured by different shooting modules (11) on the conveyor belt.

[0047] In this case, the artificial intelligence model may be a re-identification model utilizing a deep learning-based deep appearance descriptor.

[0048] That is, the artificial intelligence module (13) can analyze surface pattern data by utilizing a re-identification model that utilizes a deep learning-based deep appearance descriptor, and can track and identify the same metal surface among multiple metals by analyzing the surface pattern data. Specific details will be described later in FIG. 3.

[0049] The alarm module (14) can provide alarm information including surface pattern data, defect outliers, environmental information, and identification information and location information regarding the metal.

[0050] Specifically, the alarm module (14) can be linked with the control system (2) and can provide alarm information to the control system (2). Through this, a manager or worker located at a distance from the site has the advantage of being able to monitor the quality and production of the metal in real time through the control system (2).

[0052] FIG. 2 is a drawing showing the appearance of photographing a metal surface being transported on a conveyor belt using a darkfield light method according to one embodiment of the present invention.

[0053] Referring to Fig. 2, a number of manufactured metals can be sequentially transported from one point to another on a conveyor belt.

[0054] At this time, at least one lighting module (10) may be installed on the conveyor belt, and when multiple lighting modules (10) are installed, they may be installed with a predetermined spacing from each other.

[0055] Meanwhile, as described above in FIG. 1, the lighting module (10) illuminates the metal surface using a dark-field lighting method, and at the site, external light other than the lighting module (10) may be blocked.

[0056] Additionally, the lighting module (10) can be installed in a position where it can irradiate the metal surface being transported on the conveyor belt at an oblique angle.

[0057] At least one CCTV, which is a shooting module (11), can be installed in a position facing a plurality of metal surfaces being transported on a conveyor belt. When multiple shooting modules (11) are installed, they can be installed with a predetermined spacing from each other.

[0058] The first metal can be transported from one point to another on a conveyor belt, and the surface of the first metal can face the first imaging module (11(1)). The first imaging module (11(1)) can acquire first-1 metal surface image data by collecting light scattered or diffracted from the surface of the first metal by the lighting module (10). At this time, the first-1 metal surface image data may include surface defects of the first metal.

[0059] The surface of the first metal may pass through the first imaging module (11(1)) and come into contact with the second imaging module (11(2)) by the conveyor belt. The second imaging module (11(2)) can acquire first-second metal surface image data by collecting light scattered or diffracted from the first metal surface by the lighting module (10). At this time, the first-second metal surface image data may include surface defects of the first metal.

[0060] By the above method, the first metal can sequentially come into contact with the first to nth imaging modules (11(1), 11(2), , 11(n), n is a natural number) on a conveyor belt, and each of the first to nth imaging modules (11(1), 11(2), , 11(n)) can acquire image data of the first-1 to first-n metal surface.

[0061] At least one sensor module (12) may be installed on the conveyor belt. The sensor module (12) can sense environmental information such as temperature, humidity, illuminance, ultraviolet rays, rainfall, wind volume, wind speed, etc.

[0062] Meanwhile, although only a single conveyor belt is shown in FIG. 2, the present invention is not limited thereto. That is, multiple conveyor belts may be installed at the site, and each conveyor belt may be equipped with a lighting module (10), a shooting module (11), and a sensor module (12) so that the above process can be performed simultaneously.

[0064] FIG. 3 is a drawing relating to a method for determining metal surface defects by an artificial intelligence model according to one embodiment of the present invention.

[0065] As described above in FIG. 1, the artificial intelligence module (13) can analyze surface pattern data by utilizing an artificial intelligence model that is a re-identification model utilizing a deep learning-based appearance descriptor, and can track and identify the same metal surface among a number of metals.

[0066] Specifically, the artificial intelligence module (13) can sequentially receive metal surface image data from the first to nth shooting modules (11(1) to 11(n)).

[0067] As the first metal is transported on the conveyor belt, the first shooting module (11(1)) can photograph the surface of the first metal to obtain first-1 metal surface image data and transmit it to the artificial intelligence module (13).

[0068] The artificial intelligence module (13) can analyze the first-1 metal surface image data and express the unique features of the first metal surface (first-1 metal surface pattern data) in a vector form by utilizing an artificial intelligence model that is a re-identification model using a deep learning-based appearance descriptor.

[0069] At this time, the unique features may be surface defects of the first metal, such as scratches, patterns, colors, textures, etc. The artificial intelligence module (13) may assign an arbitrary ID (e.g., ID 1) to the first-1 metal surface image data.

[0070] As the first metal is transported on the conveyor belt, the second shooting module (11(2)) can photograph the surface of the first metal to obtain first-second metal surface image data and transmit it to the artificial intelligence module (13).

[0071] The artificial intelligence module (13) can analyze the first-second metal surface image data and express the unique features of the first metal surface (first-second metal surface pattern data) in a vector form by using an artificial intelligence model that is a re-identification model utilizing a deep learning-based appearance descriptor.

[0072] The artificial intelligence module (13) can determine whether the vector expressed from the first-1 metal surface pattern data and the vector expressed from the first-2 metal surface pattern data are the same metal surface by comparing and analyzing them.

[0073] Specifically, the artificial intelligence module (13) can perform a similarity analysis (e.g., vector similarity analysis) between a vector represented from the first-1 metal surface pattern data and a vector represented from the first-2 metal surface pattern data.

[0074] If the artificial intelligence module (13) determines that the first-1 metal surface image data and the first-2 metal surface image data are surface image data of the same metal when the similarity analysis result is determined to be greater than the reference similarity, it can determine that the first-1 metal surface image data and the first-2 metal surface image data are surface image data of the same metal.

[0075] The artificial intelligence module (13) can assign the same ID to the first-2 metal surface image data as ID 1 assigned to the first-1 metal surface image data.

[0076] Meanwhile, if the artificial intelligence module (13) determines that the similarity analysis result is less than the standard similarity, it can determine that the first-1 metal surface image data and the first-2 metal surface image data are different metal surface image data.

[0077] The artificial intelligence module (13) can assign an ID different from ID 1 assigned to the first-1 metal surface image data to the first-2 metal surface image data.

[0078] The above process can be repeated while the metal is being transported on the conveyor belt and can be terminated when it is photographed by the nth photographing module (11(n)).

[0079] Specifically, the artificial intelligence module (13) can perform a similarity analysis between a vector represented from the first-n metal surface pattern data and a vector represented from the first-(n-1) metal surface pattern data.

[0080] If the artificial intelligence module (13) determines that the first-n metal surface image data and the first-(n-1) metal surface image data are the same metal surface image data when the similarity analysis result is determined to be greater than the reference similarity, the first-n metal surface image data and the first-(n-1) metal surface image data can be determined to be the same metal surface image data.

[0081] At this time, the artificial intelligence module (13) can assign the same ID to the first-n metal surface image data as ID 1 assigned to the first-(n-1) metal surface image data.

[0082] Meanwhile, if the artificial intelligence module (13) determines that the first-n metal surface image data and the first-(n-1)) metal surface image data are different metal surface image data when the similarity analysis result is less than the standard similarity.

[0083] Through the above process, the artificial intelligence module (13) can analyze image data of multiple metal surfaces being transported on a conveyor belt and can track and identify the same metal surface among multiple metals.

[0085] Meanwhile, as described above in FIG. 1, the artificial intelligence module (13) can analyze metal surface image data using an artificial intelligence model equipped with an RNN-based LSTM algorithm and can calculate defect outliers on the metal surface.

[0086] Specifically, the artificial intelligence module (13) can extract image data of metal surfaces 1-1 to 1-n identified as the same metal surface. At this time, the extracted image data of metal surfaces 1-1 to 1-n may be in the form of time series data.

[0087] The artificial intelligence module (13) can analyze the change in the surface pattern of the first metal by comparing the first-2 metal surface pattern data with the first-1 metal surface pattern data.

[0088] At this time, if a defect at a specific location is maintained in the first-2 metal surface pattern data and the first-1 metal surface pattern data (if the pattern is constant), the artificial intelligence module (13) can determine that it is the same defect.

[0089] Alternatively, if a defect at a specific location in the first-2 metal surface pattern data and the first-1 metal surface pattern data has increased, the artificial intelligence module (13) can determine that the defect is progressing (corrosion, damage due to external impact, etc.).

[0090] Alternatively, if a defect at a specific location disappears or is not detected in the first-2 metal surface pattern data and the first-1 metal surface pattern data, the artificial intelligence module (13) may determine that noise is generated by the external environment.

[0091] Meanwhile, the artificial intelligence module (13) can calculate defect outliers based on pattern changes. Specifically, the artificial intelligence module (13) can calculate defect outliers that arithmetically represent the pattern change between the first-2 metal surface pattern data and the first-1 metal surface pattern data.

[0092] The artificial intelligence module (13) can compare the defect abnormal value with a preset reference defect abnormal value. In this case, if it is determined that the defect abnormal value is greater than the reference defect abnormal value, the artificial intelligence module (13) can determine that there is a defect on the metal surface, calculate identification information and location information regarding the metal, and provide them to the alarm module (14).

[0093] The above process may be terminated only by the analysis of pattern changes between the 1-2 metal surface pattern data and the 1-1 metal surface pattern data, or after the analysis of pattern changes between the 1-n metal surface pattern data and the 1-(n-1) surface pattern data has been fully completed.

[0094] For example, the artificial intelligence module (13) can compare the defect outlier between the second metal surface pattern data and the first metal surface pattern data, or the defect outlier that arithmetically represents the pattern change between the first-n metal surface pattern data and the first-(n-1)) metal surface pattern data, with a comprehensively preset reference defect outlier.

[0095] At this time, if it is determined that the standard defect is greater than the abnormal value, the artificial intelligence module (13) can calculate identification information and location information regarding the metal and provide them to the alarm module (14).

[0096] Through the above process, the artificial intelligence module (13) can analyze multiple metal surface image data being transported on a conveyor belt and inspect metal defects based on pattern changes.

[0098] FIG. 4 is a diagram relating to alarm information output to a control system according to one embodiment of the present invention.

[0099] Figure 4 is a screen showing alarm information including metal surface pattern data, defect anomalies, environmental information, and identification information and location information regarding the metal, which can be checked by a worker or manager through a control system (2).

[0100] Referring to Fig. 4, a tab for setting a period may be displayed. That is, a worker or manager can set the desired period using the corresponding tab. Additionally, a tab for selecting environmental information (temperature, humidity, illuminance, ultraviolet radiation, rainfall, wind volume, wind speed) may be displayed. That is, a worker or manager can set the desired environmental information using the corresponding tab.

[0101] If the search button is clicked after setting the corresponding period and environment information, surface pattern data within the period can be displayed in the form of a line graph.

[0102] Referring to Fig. 4, surface pattern data output by setting environmental information for temperature and humidity during the period from March 22, 2022 to March 24, 2022 is shown.

[0103] The red graph represents metal surface pattern data for the period from March 22, 2022 to March 24, 2022, when the temperature is 7 degrees, and the blue graph represents metal surface pattern data for the period from March 22 to March 24, 2022, when the humidity is 77.9 percent.

[0104] An operator or manager can visually check metal surface pattern data according to the corresponding period and set environment information through the screen of Fig. 4. In addition, identification information (e.g., LoRa.wth.01) and location information regarding the metal can be checked together. Through this, it is easy to identify which metal on which conveyor belt has a surface defect.

[0105] That is, the automated metal surface defect inspection system (1) using a darkfield light method utilizing an artificial intelligence model according to one embodiment of the present invention can realize faster and more objective quality inspection than the existing manual-based defect detection method, reduce scrap costs, and improve productivity. In addition, it has the advantage of determining the presence of defects on the metal surface in real time, automatically providing identification information and location information for metals in which defects are detected to support the decision-making of the operator, and improving the quality control level of the steel processing process.

[0107] FIG. 5 is a flowchart illustrating a method of operating an automated metal surface defect inspection system using a darkfield light method utilizing an artificial intelligence model according to one embodiment of the present invention.

[0108] In step (S10), the metal surface can be illuminated using a darkfield light.

[0109] Specifically, the lighting module (10) can be positioned at multiple points where light can be incident at an oblique angle on a plurality of metal surfaces being transported on a conveyor belt.

[0110] In step (S12), metal surface image data can be obtained by photographing the metal surface.

[0111] Specifically, the imaging module (11) can collect light scattered or diffracted from a metal surface, thereby obtaining metal surface image data for the metal surface.

[0112] In step (S14), surface pattern data can be extracted from metal surface image data, and defect outliers and preset reference defect outliers can be compared and analyzed.

[0113] Specifically, the artificial intelligence module (13) can sequentially receive metal surface image data from the first to nth shooting modules (11(1) to 11(n)).

[0114] The artificial intelligence module (13) can analyze the metal surface image data from the first to the first-n using an artificial intelligence model that is a re-identification model utilizing a deep learning-based appearance descriptor, and can express the unique features of the metal surface (metal surface pattern data from the first to the first-n) in a vector form.

[0115] The artificial intelligence module (13) can perform a similarity analysis between a vector represented from the first-n metal surface pattern data and a vector represented from the first-(n-1) metal surface pattern data.

[0116] If the artificial intelligence module (13) determines that the similarity analysis result is greater than the standard similarity, it can determine that the first-n metal surface image data and the first-(n-1) metal surface image data are surface image data of the same metal and can assign the same ID.

[0117] Through the above process, the artificial intelligence module (13) can analyze image data of multiple metal surfaces being transported on a conveyor belt and can track and identify the same metal surface among multiple metals.

[0118] Meanwhile, the artificial intelligence module (13) can analyze metal surface image data using an artificial intelligence model equipped with an RNN-based LSTM algorithm and can calculate defect anomalies on the metal surface.

[0119] The artificial intelligence module (13) can extract image data of metal surfaces identified as the same metal surface from 1-1 to 1-n. The artificial intelligence module (13) can analyze changes in the metal surface pattern by comparing the 1-n metal surface pattern data with the 1-(n-1) metal surface pattern data.

[0120] The artificial intelligence module (13) can calculate a defect outlier value that arithmetically represents the pattern change between the first-n metal surface pattern data and the first-(n-1) metal surface pattern data.

[0121] The artificial intelligence module (13) can compare the defect abnormal value with a preset reference defect abnormal value. In this case, if it is determined that the defect abnormal value is greater than the reference defect abnormal value, the artificial intelligence module (13) can calculate identification information and location information regarding the metal and provide them to the alarm module (14).

[0122] Through the above process, the artificial intelligence module (13) can analyze multiple metal surface image data being transported on a conveyor belt and inspect metal defects based on pattern changes.

[0123] In step (S16), alarm information including surface pattern data, defect outliers, environmental information, and identification information and location information regarding the metal can be provided.

[0124] Specifically, the alarm module (14) can provide alarm information including metal surface pattern data, defect outliers, environmental information, and identification information and location information regarding the metal to the control system (2).

[0125] Operators or managers can set period and environmental information, and visually check surface pattern data within the period.

[0126] Through this, the automated metal surface defect inspection system (1) using a darkfield light method utilizing an artificial intelligence model according to one embodiment of the present invention can realize faster and more objective quality inspection than the existing manual-based defect detection method, reduce scrap costs, and improve productivity. In addition, it has the advantage of determining the presence of defects on the metal surface in real time, automatically providing identification information and location information for metals in which defects are detected to support the decision-making of the operator, and improving the quality control level of the steel processing process.

[0128] The drawings and detailed description of the invention referenced so far are merely exemplary of the invention and are used only for the purpose of explaining the invention, not to limit the meaning or the scope of the invention as defined in the claims. Therefore, those skilled in the art will understand that various modifications and equivalent alternative embodiments are possible therefrom. Accordingly, the true technical scope of protection of the invention should be determined by the technical spirit of the appended claims.

[0129] The embodiments described above may be implemented as hardware components, software components, and / or combinations of hardware components and software components. For example, the devices, methods, and components described in the embodiments may be implemented using one or more general-purpose computers or special-purpose computers, such as, for example, a processor, a controller, an Arithmetic Logic Unit (ALU), a Digital Signal Processor (DSP), a microcomputer, a Field Programmable Gate Array (FPGA), a Programmable Logic Unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions.

[0130] The processing unit may execute an operating system and one or more software applications executed on said operating system. Additionally, the processing unit may access, store, manipulate, process, and generate data in response to the execution of the software. For convenience of understanding, the processing unit may be described as being used as a single unit, but a person of ordinary skill in the art will understand that the processing unit may include multiple processing elements and / or multiple types of processing elements.

[0131] For example, the processing unit may include multiple processors or one processor and one controller. Additionally, other processing configurations, such as a parallel processor, are also possible. Software may include a computer program, code, instructions, or a combination of one or more of these, and may configure the processing unit to operate as desired or command the processing unit independently or collectively.

[0132] Software and / or data may be embodied in any type of machine, component, physical device, virtual equipment, computer storage medium, or device so as to be interpreted by a processing device or to provide instructions or data to a processing device. Software may be distributed over networked computer systems and stored or executed in a distributed manner. Software and data may be stored on one or more computer-readable recording media.

[0133] The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc., either alone or in combination. The program instructions recorded on the medium may be those specifically designed and configured for the embodiment, or they may be those known and available to those skilled in the art of computer software.

[0134] Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, and flash memory. Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The hardware devices described above may be configured to operate as one or more software modules to perform the operation of the embodiments, and vice versa.

[0135] Although the embodiments have been described above with reference to limited examples and drawings, those skilled in the art can make various modifications and variations from the description above. For example, suitable results may be achieved even if the described techniques are performed in a different order than described, and / or if the components of the described system, structure, device, circuit, etc. are combined or assembled in a form different from described, or replaced or substituted by other components or equivalents. Therefore, other implementations, other embodiments, and equivalents to the claims are also included within the scope of the claims set forth below. Explanation of the symbols

[0136] 10; lighting module 11: Shooting module 12: Sensor Module 13: Artificial Intelligence Module 14: Alarm Module

Claims

Claim 1 An automated metal surface defect inspection system using an artificial intelligence model based on a darkfield light method, comprising: a lighting module that illuminates a metal surface using a darkfield light method; a shooting module that captures the metal surface to acquire metal surface image data; an artificial intelligence module that utilizes a deep learning artificial intelligence model to extract surface pattern data from the metal surface image data and compares and analyzes defect outliers calculated from the surface pattern data with preset reference defect outliers; and an alarm module that provides alarm information including surface pattern data and defect outliers based on the comparison analysis results. Claim 2 In claim 1, the shooting module is at least one CCTV, and the automated metal surface defect inspection system using a Darkfield light method utilizing an artificial intelligence model captures at least one metal surface being transported on a conveyor belt. Claim 3 In claim 2, the artificial intelligence model tracks and identifies the same metal surface captured by different capturing modules on the conveyor belt using a re-identification model utilizing a deep learning-based deep appearance descriptor, and calculates the defect outlier value by analyzing the surface pattern data of the tracked and identified metal surface using an RNN-based LSTM algorithm, thereby forming an automated inspection system for metal surface defects using a darkfield light method utilizing an artificial intelligence model. Claim 4 In claim 2, the artificial intelligence module determines whether the calculated defect outlier is greater than the reference defect outlier, and calculates identification information and location information regarding the metal determined to be greater than the reference defect outlier, and additionally provides this information to the alarm module, thereby forming an automated metal surface defect inspection system using a darkfield light method utilizing an artificial intelligence model. Claim 5 A method for operating an automated metal surface defect inspection system using a darkfield light method utilizing an artificial intelligence model, comprising: a step of illuminating a metal surface using darkfield light by a lighting module; a step of acquiring metal surface image data by photographing the metal surface by a photographing module; a step of extracting surface pattern data from the metal surface image data using a deep-learned artificial intelligence model by an artificial intelligence module, and comparing and analyzing defect outliers calculated from the surface pattern data with preset reference defect outliers; and a step of providing alarm information including the surface pattern data and defect outliers based on the comparison analysis results by an alarm module.

Images