KR102077281B1 - SoC based steel surface defect detection system - Google Patents

Abstract

Surface defect detection apparatus for detecting a defect on the surface of the steel sheet according to the present invention is an image photographing unit using a camera, an image signal processing unit for filtering the image, video or image captured by the image photographing unit, the data passed through the image signal processing unit It consists of a video analysis unit for analyzing, the MAC processing the data, a host processor and Ethernet, the video analysis unit and the image signal processing unit is composed of one FPGA, the FPGA, the host processor and Ethernet is a single chip (SoC, System on Chip) is characterized in that all provided.

Description

SoC based steel surface defect detection system {SoC based steel surface defect detection system}

The present invention relates to an apparatus and a system for detecting surface defects of a steel sheet, and more particularly, to an apparatus and system for detecting a surface defect of a steel sheet which improves the detection and classification performance of surface defects by using embedded and deep learning technology based on SoC.

Defect detection technology of steel sheet includes the method of visual detection by the operator or by image, Ultrasonic Test, Magnetic Flux Leakage, Magnetic Particle Inspection and Eddy Current Detection Etc.

In particular, the simplest method is for the operator to observe the surface of the steel sheet during the operation. In this method, when the steel plate is traveling at high speed, the operator cannot observe the surface accurately and can inspect the upper and lower surfaces of the steel sheet at the same time. There are no drawbacks.

Accordingly, the surface defect detection method of steel sheet is used through the image analysis using the image capturing apparatus. The conventional method using the CCD camera is mainly used in most surface defect detection apparatus because the operator can confirm the image of the surface defect. do. Furthermore, by using an infrared LED light and a CCD camera, the unevenness defect detecting device generates parallel light by using an infrared LED light, and arranges the CCD camera at a 45 degree position of the illumination diffuse reflection, that is, the incident angle of illumination, to detect the irregularly reflected defect signal. Uneven defects may be detected by camera sensors.

By the way, such a detection device and system is composed of a conventional PC-based image processing system, such a configuration has a problem of high cost, high power, low efficiency and large size.

Publication No. 2003-0043506 (Published June 2, 2003)

The present invention has been made in view of the above-described problems, and an object of the present invention is to convert a PC-based system into a SoC system to embed a surface defect detection device, and to apply deep learning technology to increase accuracy of defect detection. To provide a low cost steel sheet surface defect detection apparatus and system.

Surface defect detection apparatus for detecting a defect on the surface of the steel sheet according to the present invention is an image photographing unit using a camera, an image signal processing unit for filtering the image, video or image captured by the image photographing unit, the data passed through the image signal processing unit It consists of a video analysis unit for analyzing, the MAC processing the data, a host processor and Ethernet, the video analysis unit and the image signal processing unit is composed of one FPGA, the FPGA, the host processor and Ethernet is a single chip (SoC, System on Chip) is characterized in that all provided.

In addition, the MAC is characterized in that any one of the GigE MAC, USB MAC or Coax Press.

In addition, the processing speed of the image or video is characterized in that 20 MB / sec to 40 MB / sec.

In addition, the image capturing unit is provided with a CCD camera, and by attaching and detaching a separate filter unit, by selecting and sensing the image according to the infrared, ultraviolet, and visible light region, the separate correction unit is provided when sensing the infrared region. And correcting the detection signal offset value and the gain value so as to reduce the detection deviation of the infrared detection signal.

In addition, the single chip (SoC) is characterized in that the integration using the MEMS (Micro Electro Mechanical Systems) technology.

The steel plate surface defect detection system using the SoC-based steel plate surface defect detection device according to the present invention is connected to an image collection server, a process control unit, a computer, or the image collection server collecting image data obtained by the steel plate surface defect detection device. A management server for controlling and managing data and at least one learning platform, wherein the learning platform includes deep learning technology.

In addition, the image acquisition server, management server and learning platform is characterized in that the IoT is applied by communicating with the mobile terminal.

The steel sheet surface defect detection apparatus and system according to the present invention provides a high efficiency, low cost steel sheet surface defect detection apparatus and system.

1 is an explanatory diagram of a conventional PC-based steel sheet surface defect detection device and system,
2 is a block diagram of a steel sheet surface defect detection apparatus according to the present invention,
3 is an explanatory diagram of a steel sheet surface defect detection system according to the present invention;
4 is a block diagram of a steel sheet surface defect detection system grafted IoT according to the present invention,
5 is a conceptual configuration of the steel sheet surface defect detection apparatus and system according to the present invention,
6 is a comparison of the SoC system applied to the present invention,
7 is a conceptual diagram of a deep learning technology applied to the present invention,
8 is a structural diagram of a SoC system applied to the present invention;
9 is an explanatory photograph of the type of steel plate surface defects.

Hereinafter, an apparatus and a system for detecting a surface defect of a steel sheet according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings so that a person skilled in the art may easily implement the present invention. The present invention is not limited to the embodiments described herein, and may be implemented in various different forms.

 In order to clearly describe the present invention, parts irrelevant to the description are omitted, and like reference numerals designate like elements throughout the specification. Apparatus and system according to the present invention is not limited to the name and can be applied in various ways, and need not be limited to the terms or names described below.

2 and 3 are schematic diagrams illustrating the apparatus and system for detecting surface defects of the steel sheet according to the present invention. The surface defect detecting apparatus for detecting defects on the surface of the steel sheet according to the present invention includes an image photographing unit 300 using a camera. The image signal processor 110 filters the image, video, or image captured by the image capturing unit 300, the video analyzer 130 analyzes the data passed through the image signal processor 110, and processes the data. MAC, host processor 150 and Ethernet 170.

The video analyzer 130 and the image signal processor 110 are configured as one FPGA 190, and the FPGA 190, the host processor 150, and the Ethernet 170 are connected to one single chip (SoC, System on Chip, 100) is characterized in that all provided.

That is, as shown in FIG. 1, in the conventional PC 200-based steel plate surface defect detection apparatus and system, the functions of the PC 20 and other components are integrated into the single chip 100 (SoC), which is one chip, and replaced. Will be done.

Through this, the use efficiency is increased by occupying a relatively small cost and a small volume, and can be a cornerstone for implementing the smart factory of the fourth industrial revolution, and grafting and deepening the IoT through the single chip 100 Learning technology can be implemented together.

In the case of using the single chip (100, SoC) according to the present invention, the processing speed can be increased approximately 2 to 6 times compared to the existing PC-based devices and systems, and the installation cost is reduced by more than 30%, The power consumption of use can be lowered by more than 20%.

The MAC may be provided as one of GigE MAC, USB MAC, or Coax Press, or may combine or omit the above-described components, if necessary, and include one FPGA together with the video analyzer 130 and the image signal processor 110. A field programmable gate array 190 may be provided.

 In the steel sheet surface defect detecting apparatus according to the present invention, the processing speed of an image or an image may be provided in a range of 20 MB / sec to 40 MB / sec. That is, the speed of the conventional PC-based steel sheet surface defect detecting apparatus may be up to 10 MB / sec. Compared to the image processing speed of the steel plate surface defect detection apparatus according to the present invention is very excellent.

 The image capturing unit 300 is provided as a CCD camera, and by attaching and detaching a separate filter unit, the image capturing unit 300 separately selects and senses the images according to the infrared, ultraviolet, and visible light regions, and separately corrects the sensing of the infrared region. A portion may be provided to correct the detection signal offset value and the gain value so as to reduce the detection deviation of the infrared detection signal.

Steel sheet surface defect detection apparatus according to the present invention may further include a light source (not shown). The light source unit includes a plurality of light sources disposed at different angles. The plurality of Kangwon can irradiate light of different wavelengths to one region of the steel sheet. In some cases, the light source unit may include a first light source for irradiating light of a first wavelength, and a second light source disposed at a predetermined angle with the first light source and irradiating light of a second wavelength.

The number of the light sources or the angle between the light sources may be set differently according to embodiments. The camera or image capturing unit 300 may obtain a captured image of a region of the steel sheet to which the light sources are radiating light. The picked-up image is provided to the image processing unit and used to determine whether a defect exists.

In addition, the image capturing unit 300 may be a color camera. The first light source may emit light of any one of RGB (Red, Green, Blue), and the second light source may emit light of another of RGB (Red, Green, Blue).

The camera or image capturing unit 300 may obtain a captured image by blocking light of one of RGB (Red, Green, Blue) using a filter (not shown).

For example, the first light source may be a green light source, the second light source may be a blue light source, and the camera may apply a red filter. In this way, the camera may block light corresponding to the red wavelength and pass light corresponding to the green and blue wavelengths. Therefore, it is easy to separate the captured image generated by such a camera by the wavelength of each light source.

In addition, the image capturing unit 300 or the camera may photograph the surface of the steel sheet to generate an original image. The camera may photograph the steel sheet in real time, and the camera may be installed in an appropriate number so as to photograph the entire width of the steel sheet in consideration of the imaging angle of the camera. As previously described, the camera may select only the wavelength of the blue system, thereby excluding the effect of self-luminescence of the steel sheet.

The single chip (SoC, 100) may be integrated using MEMS (Micro Electro Mechanical Systems) technology. MEMS is called a microelectromechanical system, a microelectronic control technology, and the like, and refers to a micron (μm) or millimeter-sized micromachined manufacturing technology based on semiconductor process technology. In particular, it is a technology to make ultra-fine mechanical structures such as ultra-high density integrated circuits, micro gears half the thickness of hair, and nail-sized hard disks by processing silicon, crystal, and glass.

Micromachines made of MEMS have a precision of less than a micrometer (1 million meters). Structurally, it is possible to mass-produce micro products at low cost by applying semiconductor micro process technology that repeats processes such as deposition and etching, and driving power generates electric current by using electrostatic force and surface tension, which are pulling force between charges. Apply the principle of significantly lowering consumption.

The SoC (System on Chip) technology is a technology that integrates a system from a processor, a memory, and various sensors onto a single chip, thereby providing a compact and simple manufacturing process.

In addition, in the present invention, the at least one infrared sensor (not shown) and the signal processor may be integrated in a single chip (SoC, 100) so that the image capturing unit 300 may photograph infrared rays.

The infrared sensor of the present invention collects infrared rays (IR) emitted from an object existing in the space in the image capturing unit 300 or the lens unit, and transmits only a specific wavelength of the focused infrared rays in the filter unit. The infrared rays of a specific wavelength transmitted in this way are incident on the infrared sensing element. At this time, preferably, the infrared sensing element is integrated in the driving circuit unit for driving the infrared sensing element and one single chip (SoC) 100.

In addition, the single chip 100 may further include a housing (not shown) for configuring a plurality of components for implementing each function into one single package to prevent external shock or contamination.

As shown in FIG. 3, the steel plate surface defect detection system using the SoC-based steel plate surface defect detection device according to the present invention includes an image collector to an image collection server 400 for collecting image data obtained by the steel plate surface defect detection device. And a management server 500 and one or more learning platforms 700 connected to the process control unit, the computer, or the image collection unit 400 to control and manage data.

The SoC detects a defect image through image processing, classifies the detected defect image in the learning platform applying the trained AI model, and transmits the detected defect image to the management server or the data storage.

In addition, the SoC may apply the trained AI model to classify the defect image at the same time as the detection and transmit the image to the image collecting unit, and then to the management server or the data storage unit.

The learning platform 700 includes an AI model and deep learning technology. Deep learning is based on an artificial neural network (ANN) in order to enable a computer to learn as a human being using various data. One is machine learning technology.

Deep learning mimics the information processing method that separates objects after the human brain discovers patterns in a lot of data, and then trains machines to make computers distinguish things. Deep learning technology allows computers to recognize, reason, and judge on their own, without having to set all the criteria.

In particular, since it is widely used for voice, image recognition and photo analysis, it can be applied to a detection apparatus and system for detecting surface defects of steel sheets through analysis by image as in the present invention.

In addition, the image acquisition server, management server and learning platform of the surface defect detection apparatus and system of the steel sheet according to the present invention can be applied to the IoT by communicating with the mobile terminal.

In other words, the technology, service, and next-generation Internet of hyper-connected society are defined as the Internet of Things or the Internet of Objects, and the environment in which information generated by things that can be uniquely identified is shared through the Internet. This is a concept developed from the existing USN (Ubiquitous Sensor Network) and M2M (Machine to Machine), and can be recognized as being extended to IoT, Internet of Everything (IoE).

Components according to the present invention may be applied independently and may express their function by combination or combination. In addition, the present invention may be changed and used to apply various standards. The present invention described above is not limited to the above-described embodiments and the accompanying drawings, and simple substitutions, modifications, and changes within the technical idea of the present invention will be apparent to those skilled in the art.

Steel sheet surface defect detection apparatus and system according to the present invention can be used in high efficiency, low-cost detection apparatus and system by utilizing the SoC and deep learning technology.

100: System on Chip (SoC) 110: Image signal processor
130: video analysis unit 150: host processor
170: Ethernet 190: FPGA
200: PC 300: video recording unit
400: image acquisition unit (server) 500: management server
600: mobile terminal 700: learning platform
10: detection unit 30: classification unit
50: operating unit

Claims (9)

In the surface defect detection device for detecting defects on the surface of the steel sheet,
Image capturing unit using a camera;
An image signal processor for filtering an image, a video, or an image captured by the image capturing unit;
A video analyzer which analyzes the data passing through the image signal processor;
A MAC for processing the data;
A host processor; And
Ethernet; Consists of
The image capturing unit is provided as a CCD camera and a separate filter unit is detachably provided to separately detect and select an image according to an infrared, ultraviolet, and visible light region. The detection signal offset value and the gain value are corrected to reduce the detection deviation of the detection signal.
The video analyzer and the image signal processor are configured as one FPGA, and the FPGA, the host processor, and the Ethernet are all provided in one single chip (SoC, System on Chip), and the single chip (SoC) is a MEMS ( Surface defect detection device for steel sheet, characterized in that the integration using the Micro Electro Mechanical Systems) technology.
The method of claim 1,
The MAC is a surface defect detection apparatus for a steel sheet, characterized in that any one of GigE MAC, USB MAC or Coax Press. The method of claim 1,
An apparatus for detecting a surface defect of a steel sheet, wherein the processing speed of the image or the image is 20 MB / sec to 40 MB / sec. delete delete The method of claim 1,
An image collecting unit collecting image data obtained by the steel plate surface defect detecting apparatus; A management server connected to a process control unit, a computer, or the image collection unit to control and manage data; And one or more learning platforms; The steel plate surface defect detection apparatus further comprises a deep learning technique.
The method of claim 6
The image collecting unit, the management server and the learning platform is a steel plate surface defect detection device, characterized in that using the IoT technology by communicating with the mobile terminal.
The method of claim 6,
The SoC detects a defect image through image processing, classifies the detected defect image in the learning platform applying the trained AI model, and transmits the detected defect image to the management server or the data storage unit. .
The method of claim 6
The SoC applies a trained AI model, classifies and simultaneously detects a defect image to the image collecting unit, and transmits it to the management server or data storage unit.

Images