KR20190063839A - Method and System for Machine Vision based Quality Inspection using Deep Learning in Manufacturing Process - Google Patents

Method and System for Machine Vision based Quality Inspection using Deep Learning in Manufacturing Process Download PDF

Info

Publication number
KR20190063839A
KR20190063839A KR1020170162907A KR20170162907A KR20190063839A KR 20190063839 A KR20190063839 A KR 20190063839A KR 1020170162907 A KR1020170162907 A KR 1020170162907A KR 20170162907 A KR20170162907 A KR 20170162907A KR 20190063839 A KR20190063839 A KR 20190063839A
Authority
KR
South Korea
Prior art keywords
learning
product
image
classifier
product image
Prior art date
Application number
KR1020170162907A
Other languages
Korean (ko)
Inventor
이승우
권영민
Original Assignee
전자부품연구원
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 전자부품연구원 filed Critical 전자부품연구원
Priority to KR1020170162907A priority Critical patent/KR20190063839A/en
Priority to PCT/KR2017/013922 priority patent/WO2019107614A1/en
Publication of KR20190063839A publication Critical patent/KR20190063839A/en

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84—Systems specially adapted for particular applications
    • G01N21/88—Investigating the presence of flaws or contamination
    • G06K9/3233—
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N99/00—Subject matter not provided for in other groups of this subclass
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/04—Manufacturing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/24—Aligning, centring, orientation detection or correction of the image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30168—Image quality inspection
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Business, Economics & Management (AREA)
  • Software Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Multimedia (AREA)
  • Tourism & Hospitality (AREA)
  • Computing Systems (AREA)
  • Human Resources & Organizations (AREA)
  • Marketing (AREA)
  • Primary Health Care (AREA)
  • Strategic Management (AREA)
  • Manufacturing & Machinery (AREA)
  • General Business, Economics & Management (AREA)
  • Mathematical Physics (AREA)
  • General Engineering & Computer Science (AREA)
  • Economics (AREA)
  • Medical Informatics (AREA)
  • Quality & Reliability (AREA)
  • Evolutionary Computation (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Chemical & Material Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)
  • Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)

Abstract

A method and a system for a machine vision-based quality inspection by utilizing deep learning in a manufacturing process are provided. According to an embodiment of the present invention, the method for a machine vision-based quality inspection comprises the steps of: generating a learning product image; learning a classifier for distinguishing good and bad products with the generated learning product image; and determining a product as a good or a bad product by using the learned classifier. Therefore, a feature value of data to be distinguished can be found by learning, and a machine vision-based inspection can be performed even for an inspection region dependent on a manual inspection as a defect is difficult to formalize.

Description

제조 공정에서 딥러닝을 활용한 머신 비전 기반 품질검사 방법 및 시스템{Method and System for Machine Vision based Quality Inspection using Deep Learning in Manufacturing Process}TECHNICAL FIELD The present invention relates to a machine vision based quality inspection method and system using deep running in a manufacturing process,

본 발명은 제조 관리 기술에 관한 것으로, 더욱 상세하게는 제조 공정에서 자동화된 머신 비전 기반의 품질검사 방법 및 시스템에 관한 것이다.TECHNICAL FIELD The present invention relates to manufacturing management techniques, and more particularly, to a method and system for automated machine vision-based quality inspection in a manufacturing process.

기존 제조 공정에서 발생하는 결함 검출은 대부분 작업자가 직접 관찰 수행하는 수동적 방법이다. 이에 따라, 작업자의 상태 및 판단 능력에 따라 정확도가 일관되지 못하고, 많은 시간이 소요되는바, 자동으로 결함을 검출할 수 있는 방법이 필요하다.Detection of defects in existing manufacturing processes is a passive method that most observers observe. Accordingly, there is a need for a method that can detect defects automatically because the accuracy is not consistent depending on the condition and judgment ability of the operator and takes a long time.

수동 검사의 문제점을 보완하려 자동 광학 검사(Automated Optical Inspection, AOI) 또는 머신 비전이라 불리는 자동화된 비전 기반 검사 시스템이 도입 되었다. 머신 비전 시스템은 결함 검사, 제품 분류 등의 여러 분야에서 수동 검사자를 대치하게 되었다.An automated vision-based inspection system called Automated Optical Inspection (AOI) or Machine Vision was introduced to overcome the problems of manual inspection. Machine vision systems have replaced manual inspectors in many areas such as defect inspection and product classification.

하지만, 머신 비전 검사 방법은 공정 특성 및 대상 품목에 맞춰 엔지니어가 규칙을 하나씩 설정하면서 설계하는 방법이다. 따라서, 머신 비전의 정확도는 엔지니어가 규칙을 얼마나 잘 만들었는지에 따라 성능이 좌우되는 한계와 함께, 분석 대상이 복잡한 경우 미세한 불량에 대한 검출 정확도가 떨어지게 되는 문제를 보이게 ehlsekHowever, the machine vision inspection method is a method of designing the engineers one by one according to the process characteristics and target items. Therefore, the accuracy of the machine vision is affected by the performance depending on how well the engineer made the rule, and the accuracy of the detection of the fine defect is lowered when the object to be analyzed is complex.

특히, 이미지가 불규칙하고 비정형인 경우엔 불량의 특징 벡터(Feature)를 엔지니어가 정의하기 어려운 경우, 머신 비전에 의한 장비 정확도가 낮기 때문에 여전히 사람의 육안검사에 의존하는 경우도 존재한다.Particularly, when the image is irregular and irregular, it is still difficult to define the feature vector of the defect when the engineer can hardly define the feature. Therefore, the accuracy of the machine vision is low.

일반적으로, 머신 비전에 사용되는 영상 처리 기법들은, 대상 품목의 변화가 없고 외부요인이 완전하게 동일하고 통제된 상황에서는 매우 높은 정확도와 불량률을 탐지가 가능하지만, 이렇게 엄격한 조건의 설정은 외부 환경의 변화 시에 엔지니어가 재학습을 하기 위해 새로운 특징점을 추출을 시도하거나 판별 알고리즘의 재설계가 필요하다는 한계를 가지고 있는 것이다.In general, image processing techniques used in machine vision can detect very high accuracy and defect rate in a situation where there is no change in target items and the external factors are completely the same and controlled. However, The engineer has a limitation in trying to extract new feature points or redesigning the discrimination algorithm in order to re-learn.

본 발명은 상기와 같은 문제점을 해결하기 위하여 안출된 것으로서, 본 발명의 목적은, 제조 공정에서 딥러닝을 활용한 머신 비전 기반 품질검사 방법 및 시스템을 제공함에 있다.SUMMARY OF THE INVENTION It is an object of the present invention to provide a machine vision-based quality inspection method and system using deep running in a manufacturing process.

상기 목적을 달성하기 위한 본 발명의 일 실시예에 따른, 품질 검사 방법은, 학습용 제품 영상을 생성하는 단계; 생성된 학습용 제품 영상으로 양품과 불량품을 구분하기 위한 분류기를 학습시키는 단계; 및 학습된 분류기를 이용하여, 제품을 양품 또는 불량품으로 판정하는 단계;를 포함한다.According to an aspect of the present invention, there is provided a quality checking method comprising: generating a product image for learning; Learning a classifier for classifying a good product and a defective product with the generated learning product image; And determining the product as a good product or a defective product using the learned classifier.

그리고, 학습용 제품 영상 생성단계는, 학습용 제품 영상을 조각화하는 단계; 및 조각화된 학습용 제품 영상을 확장하는 단계;를 포함할 수 있다.In addition, the learning product image generating step may include: segmenting the learning product image; And expanding the fragmented learning product image.

또한, 확장 단계는, 학습용 제품 영상에 얼라인먼트, 조명 효과 변환, 노이즈 효과 적용 중 적어로 하나를 이용하여 학습용 제품 영상을 확장할 수 있다.In addition, the expansion step can expand the learning product image using at least one of alignment, illumination effect conversion, and noise effect application to the learning product image.

그리고, 본 발명의 실시예에 따른 품질 검사 방법은, 제품을 촬영하는 단계; 촬영된 제품 영상에서 관심 영역을 추출하는 단계;를 더 포함하고, 판정 단계는, 추출된 관심 영역에서의 결함을 확인하여, 제품을 양품 또는 불량품으로 판정할 수 있다.A quality inspection method according to an embodiment of the present invention includes: photographing a product; And extracting a region of interest from the photographed product image, wherein the determining step can identify the defect in the extracted region of interest and determine the product as good or defective.

또한, 판정 단계는, 학습용 제품 영상들 중 추출된 관심 영역과 가장 근접한 학습용 제품 영상의 분류 번호 및 확률 값을 판정 결과로 출력할 수 있다.The determining step may output the classification number and the probability value of the learning product image that is the closest to the extracted interest area among the learning product images as the determination result.

그리고, 본 발명의 실시예에 따른 품질 검사 방법은, 잘못된 판정이 이루어진 제품 영상들을 수집하여, 분류기를 재학습시키는 단계;를 더 포함할 수 있다.The quality inspection method according to an embodiment of the present invention may further include collecting product images on which erroneous determination has been made, and re-learning the classifier.

또한, 재학습 단계는, 원격의 서버에 의해 수행될 수 있다.Further, the re-learning step can be performed by a remote server.

한편, 본 발명의 다른 실시예에 따른, 품질 검사 시스템은, 학습용 제품 영상을 생성하는 카메라; 및 생성된 학습용 제품 영상으로 양품과 불량품을 구분하기 위한 분류기를 학습시키고, 학습된 분류기를 이용하여 제품을 양품 또는 불량품으로 판정하는 프로세서;를 포함한다.According to another aspect of the present invention, there is provided a quality inspection system including: a camera for generating a product image of a learning product; And a processor that learns a classifier for classifying a good product and a defective product using the generated learning product image, and determines the product as a good product or a defective product using the learned classifier.

이상 설명한 바와 같이, 본 발명의 실시예들에 따르면, 기존의 머신 비전 검사 기법과 달리, 학습에 의해 판별 대상의 데이터의 특징값(Feature)을 스스로 찾을 수 있어, 결함의 정형화가 어려워 수동 검사에 의존하는 검사 영역에 대해서도 머신 비전 기반 검사가 가능해진다.As described above, according to the embodiments of the present invention, unlike the existing machine vision inspection technique, since the feature value of the data to be discriminated can be found by learning, it is difficult to formalize the defect, Machine vision-based inspection is also possible for dependent inspection areas.

또한, 본 발명의 실시예들에 따르면, 온라인 기반의 학습모델 업데이트 및 최적화를 지원하여, 새로운 불량 패턴이 들어오거나 학습의 정확도를 개선시키기 위해, 사용자 피드백을 통해 수집된 학습 데이터를 온라인을 통해 수집하고, 재학습 모델을 생성하여 교체가 가능한 서비스 제공이 가능하다.In addition, according to embodiments of the present invention, learning data collected through user feedback is collected on-line so as to support updating and optimization of an online-based learning model and to improve the accuracy of learning or introducing a new bad pattern. And a re-learning model can be created to provide a service that can be replaced.

뿐만 아니라, 본 발명의 실시예들에 따르면, 잦은 상품 변경이 필요한 경우에도 자동화된 학습 및 최적화 방안을 제공할 수 있어, 도입 비용을 줄이는 것이 가능해진다.In addition, according to the embodiments of the present invention, it is possible to provide an automated learning and optimization method even when frequent product changes are required, thereby reducing the introduction cost.

도 1에 학습 데이터 생성 과정의 설명에 제공되는 도면,
도 2는 딥러닝 기반 분류기 생성 과정의 설명에 제공되는 흐름도,
도 3은 머신 비전 기반 품질 검사 과정의 설명에 제공되는 흐름도,
도 4는 온라인 재학습 기반으로 최적화 모델을 생성하는 개념을 나타내었다.
도 5는 전술한 머신 비전 기반 품질검사 방법을 수행하기 위한 시스템의 블럭도이다.Fig. 1 is a block diagram of a learning data generation process,
Figure 2 is a flow chart provided in the description of the deep learning-based classifier generation process;
FIG. 3 is a flowchart provided in the explanation of the machine vision-based quality inspection process;
FIG. 4 shows a concept of generating an optimization model based on on-line re-learning.
5 is a block diagram of a system for performing the above-described machine vision-based quality checking method.

이하에서는 도면을 참조하여 본 발명을 보다 상세하게 설명한다.Hereinafter, the present invention will be described in detail with reference to the drawings.

1. 머신 비전 기반 품질검사 방법1. Machine vision-based quality inspection method

딥러닝은 인간의 신경망을 모방하여 심층 신경망(Deep Neural Network)을 만들어 입력과 출력 사이에 겹겹이 층(Layer)을 쌓아 만든 네트워크로 결합 가중치(Weight) 문제를 학습하고 해결한다.Deep Learning implements a human neural network and creates a Deep Neural Network to build up a layer between input and output to learn and solve the joint weight problem.

여기에 사용되는 심층 학습은 여러 비선형 변환 기법의 조합을 통하여, 기존의 머신 러닝의 기법에서 가장 취약점을 보이던 대상물의 높은 수준의 추상화에서 강점을 보인다.The in-depth learning used here is advantageous in the high level of abstraction of the objects that showed the weakest points in the existing machine learning technique through the combination of various nonlinear transformation techniques.

본 발명의 실시예에서는, 제조 공정에서 딥러닝을 활용한 머신 비전 기반 품질검사 방법을 제시한다. 제조 공장에서 생산되는 생산 제품을 분류하거나 외형적 불량을 검사 및 검출함에 있어, 딥러닝 기법을 사용하여 탐지 대상이 되는 대상물의 양품, 불량 데이터를 학습하고, 자동으로 판별하는 자동화된 비전 검사 방법이다.In the embodiment of the present invention, a machine vision-based quality inspection method utilizing deep running in a manufacturing process is presented. An automated vision inspection method that learns and automatically detects good products and bad data of objects to be detected by using a deep learning technique in sorting production products produced in a manufacturing factory or inspecting and detecting external defects .

본 발명의 실시예에 따른 품질검사 방법에서는, 학습 데이터(학습용 제품 영상과 양품/불량 여부)를 입력하여 자동으로 최적화된 판별 모델을 생성하는데, 이미지가 불규칙한 경우 비정형적 패턴에서도 학습을 통하여 판별이 가능하다.In the quality inspection method according to the embodiment of the present invention, learning data (learning product image and good product / badness) are input to generate an automatically optimized discrimination model. When the image is irregular, It is possible.

또한, 본 발명의 실시예에 따른 품질검사 방법에서는, 입력 데이터에 노이즈를 포함시켜 학습함으로써 외부 환경 요인의 변화에 강점을 보이고, 온라인으로 사용자 피드백 정보를 통한 재학습 및 검출 모델 교체가 가능하다.In addition, in the quality inspection method according to the embodiment of the present invention, noise is included in the input data and learning is performed to show strengths in changes in external environmental factors, and re-learning and detection model replacement can be performed online through user feedback information.

본 말명의 실시예에서는, 결함 탐지를 위한 이미지 분석 방법으로 심층 컨볼루션 신경망(Deep Convolutional Neural Networks, CNNs)을 활용하고, 1) 학습 데이터 생성, 2) 학습데이터 기반 딥러닝 기반 분류기 생성, (3) 딥러닝 기반 분류기를 사용한 결함 탐지 및 4) 기법과 분류기 모델의 재학습을 위한 온라인 기반 최적화를 수행한다. 이하에서 각각에 대해 상세히 설명한다.In the embodiment of the present invention, Deep Convolutional Neural Networks (CNNs) are used as an image analysis method for defect detection, and 1) generation of learning data, 2) generation of a deep learning based classifier based on learning data, and 3) ) Defect detection using deep learning based classifiers and 4) On-line based optimization for re-learning classifier models. Each will be described in detail below.

1) 학습 데이터 생성 1) Generate learning data

딥러닝 알고리즘에 기반한 검사 판별을 위해서는 데이터의 학습을 통해 컴퓨터 스스로가 특징 표현을 만들어 내는 특징 표현 학습(feature representation learning) 과정이 필요하므로, 양품의 영상 데이터와 불량품의 영상 데이터를 지도학습을 통해 학습된다. In order to discriminate the inspection based on the deep learning algorithm, it is necessary to perform a feature representation learning process in which the computer itself generates a characteristic expression through learning of data. Therefore, the image data of the good product and the image data of the defective product are learned do.

도 1에 학습 데이터 생성 과정의 설명에 제공되는 도면이다. 복잡한 영상 이미지를 판별할 때 분류 정확도를 높이고, 구체화된 레이블 정보를 학습시키기 위해, 도 1에 도시된 바와 같이, 먼저 대상 이미지를 일정 규격의 사이즈의 작은 이미지로 중첩이 되도록 자른다(Cropping).Fig. 1 is a diagram provided for explanation of a learning data generation process. As shown in FIG. 1, in order to increase the classification accuracy in discriminating a complicated video image and to learn the specified label information, first, the target image is cropped so as to overlap with a small image of a certain size.

잘라진(Cropped) 이미지와 함께 이미지 인덱스 및 결함 여부의 정보를 포함한 레이블링 정보가 딥러닝의 학습데이터로 사용된다.The labeling information including the image index and defect information together with the cropped image is used as the learning data of the deep learning.

나아가, 외부 환경 변수에 의한 영향을 줄이기 위하여, 수집된 데이터에 임의의 얼라인먼트, 조명 효과 변환, 노이즈 효과를 적용한 데이터를 각각 생성하여, 학습 데이터 수를 늘린다. 이는 환경 변화에 따른 노이즈를 강화킨다.Further, in order to reduce the influence by the external environmental variables, data in which arbitrary alignment, illumination effect conversion, and noise effect are applied to the collected data are respectively generated, and the number of learning data is increased. This enhances noise due to environmental changes.

뿐만 아니라, 컨볼루션 신경망 모델은 복잡한 만큼 학습을 위해 많은 데이터를 필요로 하기 때문에, 충분한 데이터 수집을 위해 데이터 셋을 확장을 위한 용도로도 사용한다.In addition, because the convolution neural network model requires a lot of data for learning as much as the complexity, the dataset is also used for expansion purposes for sufficient data collection.

2) 학습을 통한 딥러닝 기반 분류기 모델 생성2) Creating a deep learning based classifier model through learning

도 2는 딥러닝 기반 분류기 생성 과정의 설명에 제공되는 흐름도이다. 딥러닝 기반의 검사기를 만들기 위해, 도 2에 도시된 바와 같이, 먼저, 학습 과정을 거쳐야 한다. 즉, 생성된 많은 양의 학습 데이터를 기반으로 최적화를 통해 분류기가 결정된다.Figure 2 is a flow chart provided in the description of the deep learning based classifier generation process. In order to make a deep learning based checker, as shown in FIG. 2, a learning process must first be performed. That is, the classifier is determined through optimization based on a large amount of generated learning data.

학습에 사용되는 데이터는 영상 데이터와 함께, 인덱스와 결함 정보를 분류값으로 사용하여 학습하며, 최적화 과정을 통해 학습 결과로 나온 분류기 모델은 검사장치에 적용된다.The data used for learning are learned by using index and defect information together with image data as classification values, and the classifier model obtained as a result of learning through the optimization process is applied to the inspection apparatus.

구체적으로, 학습 데이터(학습용 제품 영상)를 조각화하고(S110), 조각화된 학습 데이터로부터 레이블 데이터 생성 및 확장 후에(S120), 분류기를 학습시켜(S130), 비전 검사 장치에 적용하게 된다(S140).Specifically, the learning data (learning product image) is fragmented (S110), the label data is generated and expanded from the fragmented learning data (S120), the classifier is learned (S130) .

3) 딥러닝 기반의 결함 탐지 과정3) Deep learning based defect detection process

도 3은 머신 비전 기반 품질 검사 과정의 설명에 제공되는 흐름도이다.3 is a flow chart provided in the description of the machine vision-based quality inspection process.

품질 검사를 통해 제품을 양품 또는 불량으로 판정하기 위해, 먼저, 검사 카메라의 위치를 정렬하여 검사 대상 제품을 촬영하고(S210), 촬영을 통해 생성된 영상에서 관심 영역을 추출하여 저장한다(S220).In order to determine whether the product is good or defective through quality inspection, first, the position of the inspection camera is aligned and the inspection target product is photographed (S210), and the ROI is extracted from the image generated through the photographing and stored (S220) .

그리고, S220단계에서 저장된 대상 이미지를 학습 모델 생성의 이미지 전처리 과정에서와 같이, 일정 규격의 사이즈의 작은 이미지로 중첩되도록 자른다(Cropping)(S230).In step S220, the stored target image is cropped so as to overlap with a small image of a predetermined size, as in the image preprocessing process of generating a learning model (S230).

다음, 품질 검사 장치는, 각각의 조각 이미지 각각에 대해 분류기를 이용하여 검사를 수행한다(S240). 이때, 심층신경망 알고리즘은 분류 결과로 학습을 통해 분류기를 생성할 때 사용된 학습 데이터들 중 가장 근접한 이미지의 분류 번호와 확률 값을 결과로 출력한다.Next, the quality inspection apparatus performs inspection using a classifier for each piece image (S240). In this case, the neural network algorithm outputs the classification number and the probability value of the closest image among the learning data used when generating the classifier through the learning as the classification result.

이에, 각각의 조각된 이미지를 딥러닝 기반 분류기를 통해, 해당 인덱스에서 가장 높은 확률로 분류표가 양품인지 결함인지 확인하여(S250, S260, S270), 판정 결과값을 표시한다(S280). 이때, 결함 제품인 경우에는 결함으로 판명하고 영역을 표시하여 줄 수 있다.Then, each of the sculptured images is checked (S250, S260, S270) whether the classification table is good or defective with the highest probability in the corresponding index through the deep learning-based classifier, and the determination result is displayed (S280). At this time, in the case of a defective product, it is determined that the defect is defective and the area can be displayed.

4) 온라인 기반 재학습을 통한 개선된 분류기 적용 과정4) Improved classifier application process through on-line re-learning

비젼 검사 장치를 생산 공정의 검사 장치로 활용하고 있는 상황에서, 기존에 학습되지 않은 새로운 타입의 결함이 발생하거나, 이전에 학습한 판별 모델에서 오류가 발생하는 경우 재학습이 필요한 경우 새로운 학습데이터를 추가로 활용하여 개선된 모델을 다음과 같은 절차로 만들어 낼 수 있다. If a new type of defect that has not been learned occurs or an error occurs in the previously learned discrimination model in the situation where the vision inspection apparatus is used as an inspection apparatus of the production process, Further, the improved model can be created by the following procedure.

공장 내 작업자가 사용하는 HMI(Human-Machine Interface) 장치를 통해서 잘못 판별한 정보(양품 데이터를 불량으로 판별하거나 또는, 불량 제품을 양품으로 판별하는 경우)에 대해, 해당 영상 데이터와 해당 영상데이터의 레이블 정보를 사용자로부터 피드백 받아서 따로 수집을 한다. (In the case of determining that the good data is defective or the defective product is judged as good) through the HMI (Human-Machine Interface) device used by the factory operator, the image data and the corresponding image data The label information is fed back from the user and collected separately.

딥러닝 알고리즘의 학습 모델은 가중치를 최적화하기 위해 상당한 연산량과 학습 시간이 필요하며 하며, 특히 많은 GPU(Graphics Processing Unit) 연산을 필요로 한다.The learning model of the deep learning algorithm requires a significant amount of computation and learning time to optimize weights, and in particular requires many GPU (Graphics Processing Unit) operations.

따라서, 원격지에 고성능의 GPU를 탑재한 서버에서, 사용자의 피드백 정보로 따로 수집된 오류 데이터를 새로운 판별 모델을 만들기 위한 학습 데이터로 사용된다.Therefore, in a server equipped with a high-performance GPU at a remote site, error data collected separately as user feedback information is used as learning data for creating a new discrimination model.

원격지에서 딥러닝 재학습을 통해 새로운 만들어진 모델은, 공장 내 검사 장치의 새로운 분류기로 대체되어 검사 장치의 판별 모델을 개선시킬 수 있다.The newly created model can be replaced with a new classifier of the in-house inspection apparatus through the deep learning re-learning at the remote site, thereby improving the discrimination model of the inspection apparatus.

도 4에는 온라인 재학습 기반으로 최적화 모델을 생성하는 개념을 나타내었다.FIG. 4 shows the concept of creating an optimization model based on on-line re-learning.

2. 머신 비전 기반 품질검사 시스템2. Machine vision-based quality inspection system

도 5는 전술한 머신 비전 기반 품질검사 방법을 수행하기 위한 시스템의 블럭도이다. 도시된 머신 비전 기반 품질검사 시스템은, 카메라(310), 통신부(320), 프로세서(330), 저장부(340) 및 디스플레이(350)를 포함하여 구성되는 컴퓨팅 시스템이다.5 is a block diagram of a system for performing the above-described machine vision-based quality checking method. The illustrated machine vision based quality inspection system is a computing system that includes a camera 310, a communication unit 320, a processor 330, a storage unit 340 and a display 350.

카메라(310)는 검사 대상 제품을 촬영하여 영상을 생성하기 위한 영상 생성 장치다.The camera 310 is an image generation device for generating an image by photographing a product to be inspected.

프로세서(330)는 카메라(310)에 의해 생성된 영상에서 관심 영역을 추출하고 영상 처리를 수행하며, 분류기를 학습시켜 생성하고 구동시켜 검사를 수행하는 수단으로 CPU와 GCU들로 구성된다.The processor 330 includes a CPU and GCUs as means for extracting a region of interest from an image generated by the camera 310, performing image processing, and generating and generating a classifier to perform an inspection.

저장부(340)는 프로세서(330)가 위 절차들을 수행함에 있어 필요한 저장공간을 제공한다.The storage unit 340 provides the necessary storage space for the processor 330 to perform the above procedures.

디스플레이(350)에는 제품 검사에 따른 결함 판정 여부 및 판정 값이 표시된다.On the display 350, whether or not a defect is determined according to a product inspection and a determination value are displayed.

통신부(320)는 온라인 기반 재학습을 위한 수단으로, 서버가 재학습을 통해 개선한 모델을 수신하기 위한 수단이다.The communication unit 320 is a means for on-line re-learning, and is means for receiving a model that the server has improved through re-learning.

3. 변형예3. Variations

지금까지, 제조 공정에서 딥러닝을 활용한 머신 비전 기반 품질검사 방법 및 시스템에 대해 바람직한 실시예들을 들어 상세히 설명하였다.Up to now, preferred embodiments have been described in detail for a machine vision-based quality inspection method and system utilizing deep running in the manufacturing process.

본 발명의 실시예에서는, 딥러닝 알고리즘을 적용하여 제조 공장에서 생산품의 분류 및 불량 판별을 수행하며, 딥러닝을 통하여 영상 데이터 학습을 통한 불량을 검출 모델을 자동으로 만들어 단순 패턴이 아닌 비정형화되고 복잡한 영상 이미지까지 분석을 가능한 검사가 가능하게 하였다.In the embodiment of the present invention, a deep-learning algorithm is applied to perform classification and defect discrimination in a manufacturing factory, and a defect model is automatically generated through learning of image data through deep learning to create a non-standard pattern Which makes it possible to perform analysis for complex images.

또한, 본 발명의 실시예들에서는, 학습 데이터에 의해서 자동으로 검출 모델이 만들 수 있고, 사용자 피드백을 통한 온라인 학습을 통한 자가-최적화(self optimization)가 가능하여 높은 유연성을 제공함으로써, 비전 검사 시스템 도입 비용과 변경 비용을 낮출 수 있다.Further, in the embodiments of the present invention, a detection model can be automatically created by learning data, self-optimization through online learning through user feedback is possible, and by providing high flexibility, The introduction cost and the change cost can be lowered.

한편, 본 실시예에 따른 장치와 방법의 기능을 수행하게 하는 컴퓨터 프로그램을 수록한 컴퓨터로 읽을 수 있는 기록매체에도 본 발명의 기술적 사상이 적용될 수 있음은 물론이다. 또한, 본 발명의 다양한 실시예에 따른 기술적 사상은 컴퓨터로 읽을 수 있는 기록매체에 기록된 컴퓨터로 읽을 수 있는 코드 형태로 구현될 수도 있다. 컴퓨터로 읽을 수 있는 기록매체는 컴퓨터에 의해 읽을 수 있고 데이터를 저장할 수 있는 어떤 데이터 저장 장치이더라도 가능하다. 예를 들어, 컴퓨터로 읽을 수 있는 기록매체는 ROM, RAM, CD-ROM, 자기 테이프, 플로피 디스크, 광디스크, 하드 디스크 드라이브, 등이 될 수 있음은 물론이다. 또한, 컴퓨터로 읽을 수 있는 기록매체에 저장된 컴퓨터로 읽을 수 있는 코드 또는 프로그램은 컴퓨터간에 연결된 네트워크를 통해 전송될 수도 있다.It goes without saying that the technical idea of the present invention can also be applied to a computer-readable recording medium having a computer program for performing the functions of the apparatus and method according to the present embodiment. In addition, the technical idea according to various embodiments of the present invention may be embodied in computer-readable code form recorded on a computer-readable recording medium. The computer-readable recording medium is any data storage device that can be read by a computer and can store data. For example, the computer-readable recording medium may be a ROM, a RAM, a CD-ROM, a magnetic tape, a floppy disk, an optical disk, a hard disk drive, or the like. In addition, the computer readable code or program stored in the computer readable recording medium may be transmitted through a network connected between the computers.

또한, 이상에서는 본 발명의 바람직한 실시예에 대하여 도시하고 설명하였지만, 본 발명은 상술한 특정의 실시예에 한정되지 아니하며, 청구범위에서 청구하는 본 발명의 요지를 벗어남이 없이 당해 발명이 속하는 기술분야에서 통상의 지식을 가진자에 의해 다양한 변형실시가 가능한 것은 물론이고, 이러한 변형실시들은 본 발명의 기술적 사상이나 전망으로부터 개별적으로 이해되어져서는 안될 것이다.While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments, but, on the contrary, It will be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the spirit and scope of the present invention.

310 : 카메라
320 : 통신부
330 : 프로세서
340 : 저장부
350 : 디스플레이310: camera
320:
330: Processor
340:
350: Display

Claims (8)

학습용 제품 영상을 생성하는 단계;
생성된 학습용 제품 영상으로 양품과 불량품을 구분하기 위한 분류기를 학습시키는 단계; 및
학습된 분류기를 이용하여, 제품을 양품 또는 불량품으로 판정하는 단계;를 포함하는 것을 특징으로 하는 품질 검사 방법.
Generating a learning product image;
Learning a classifier for classifying a good product and a defective product with the generated learning product image; And
And determining the product as good or defective by using the learned classifier.
청구항 1에 있어서,
학습용 제품 영상 생성단계는,
학습용 제품 영상을 조각화하는 단계; 및
조각화된 학습용 제품 영상을 확장하는 단계;를 포함하는 것을 특징으로 하는 품질 검사 방법.
The method according to claim 1,
The learning product image generation step includes:
Segmenting the learning product image; And
And expanding the fragmented learning product image.
청구항 2에 있어서,
확장 단계는,
학습용 제품 영상에 얼라인먼트, 조명 효과 변환, 노이즈 효과 적용 중 적어로 하나를 이용하여 학습용 제품 영상을 확장하는 것을 특징으로 하는 품질 검사 방법.
The method of claim 2,
In the expansion step,
A method of quality inspection comprising the steps of: expanding a learning product image using at least one of alignment, lighting effect conversion, and noise application to a learning product image.
청구항 1에 있어서,
제품을 촬영하는 단계;
촬영된 제품 영상에서 관심 영역을 추출하는 단계;를 더 포함하고,
판정 단계는,
추출된 관심 영역에서의 결함을 확인하여, 제품을 양품 또는 불량품으로 판정하는 것을 특징으로 하는 품질 검사 방법.
The method according to claim 1,
Photographing a product;
And extracting a region of interest from the photographed product image,
In the determination step,
And checking defects in the extracted region of interest to determine that the product is good or defective.
청구항 4에 있어서,
판정 단계는,
학습용 제품 영상들 중 추출된 관심 영역과 가장 근접한 학습용 제품 영상의 분류 번호 및 확률 값을 판정 결과로 출력하는 것을 특징으로 하는 품질 검사 방법.
The method of claim 4,
In the determination step,
And outputting the classification number and the probability value of the learning product image that is closest to the extracted interest region among the learning product images as the determination result.
청구항 4에 있어서,
잘못된 판정이 이루어진 제품 영상들을 수집하여, 분류기를 재학습시키는 단계;를 더 포함하는 것을 특징으로 하는 품질 검사 방법.
The method of claim 4,
Collecting product images for which a false determination has been made, and re-learning the classifier.
청구항 6에 있어서,
재학습 단계는,
원격의 서버에 의해 수행되는 것을 특징으로 하는 품질 검사 방법.
The method of claim 6,
The re-
Wherein the quality inspection is performed by a remote server.
학습용 제품 영상을 생성하는 카메라;
생성된 학습용 제품 영상으로 양품과 불량품을 구분하기 위한 분류기를 학습시키고, 학습된 분류기를 이용하여 제품을 양품 또는 불량품으로 판정하는 프로세서;를 포함하는 것을 특징으로 하는 품질 검사 시스템.
A camera that generates a training product image;
And a processor that learns a classifier for classifying a good product and a defective product using the generated learning product image and determines the product as a good product or a defective product using the learned classifier.
KR1020170162907A 2017-11-30 2017-11-30 Method and System for Machine Vision based Quality Inspection using Deep Learning in Manufacturing Process KR20190063839A (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
KR1020170162907A KR20190063839A (en) 2017-11-30 2017-11-30 Method and System for Machine Vision based Quality Inspection using Deep Learning in Manufacturing Process
PCT/KR2017/013922 WO2019107614A1 (en) 2017-11-30 2017-11-30 Machine vision-based quality inspection method and system utilizing deep learning in manufacturing process

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
KR1020170162907A KR20190063839A (en) 2017-11-30 2017-11-30 Method and System for Machine Vision based Quality Inspection using Deep Learning in Manufacturing Process

Publications (1)

Publication Number Publication Date
KR20190063839A true KR20190063839A (en) 2019-06-10

Family

ID=66664462

Family Applications (1)

Application Number Title Priority Date Filing Date
KR1020170162907A KR20190063839A (en) 2017-11-30 2017-11-30 Method and System for Machine Vision based Quality Inspection using Deep Learning in Manufacturing Process

Country Status (2)

Country Link
KR (1) KR20190063839A (en)
WO (1) WO2019107614A1 (en)

Cited By (26)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR102108956B1 (en) 2019-11-13 2020-05-12 주식회사 만컴퍼니 Apparatus for Performing Inspection of Machine Vision and Driving Method Thereof, and Computer Readable Recording Medium
WO2021033791A1 (en) 2019-08-19 2021-02-25 엘지전자 주식회사 Ai-based new learning model generation system for vision inspection on product production line
WO2021033792A1 (en) 2019-08-19 2021-02-25 엘지전자 주식회사 Ai-based pre-training model determination system, and ai-based vision inspection management system using same for product production lines
WO2021049868A1 (en) * 2019-09-11 2021-03-18 (주) 넥스트랩 System and method for product quality management
KR20210071882A (en) 2019-12-06 2021-06-16 주식회사 아이디알인비전 Monitoring method for prouct faults in manufacturing process of ball joint for car
WO2021187683A1 (en) * 2020-03-20 2021-09-23 라온피플 주식회사 Method and system for testing quality of new product by using deep learning
KR102323607B1 (en) * 2020-05-19 2021-11-08 부산대학교 산학협력단 Method and system for detecting anomaly based images
KR102358948B1 (en) * 2021-08-18 2022-02-08 (주)썬컴퍼니 Automated system for manufacturing paper boxes for packaging based on ai
KR102362615B1 (en) * 2020-11-27 2022-02-15 (주)리눅스아이티 Plc system for supporting ai-based vision inspection and control method thereof
KR20220038858A (en) 2020-09-21 2022-03-29 주식회사 스프링클라우드 Deep learning-based defect data detection apparatus and method of processed autonomous driving scenario data
KR20220042916A (en) 2020-09-28 2022-04-05 (주)미래융합정보기술 Vision inspection system by using remote learning of product defects image
WO2022097775A1 (en) * 2020-11-05 2022-05-12 위즈코어 주식회사 5g-based production, logistics management, and cloud-oriented machine vision service providing method
KR20220065545A (en) * 2020-11-13 2022-05-20 주식회사 트윔 Apparatus for checking defect of object having atypical surface
KR102402685B1 (en) * 2021-08-31 2022-05-26 오성컨버전스 주식회사 Box manufacturing method that provides IoT-based services
WO2022108249A1 (en) * 2020-11-17 2022-05-27 (주)자비스 Method, apparatus, and program for generating training data, and method for detecting foreign substance using same
KR20220087360A (en) 2020-12-17 2022-06-24 연세대학교 산학협력단 Method and device for classifying and quantity of empty bottles through multi-camera based sequential machine vision
WO2022158628A1 (en) * 2021-01-20 2022-07-28 주식회사 솔루션에이 System for determining defect in display panel on basis of machine learning model
KR20220118140A (en) 2021-02-18 2022-08-25 (주)라임 Deep learning based injection quality inspection system
KR20220132220A (en) 2021-03-23 2022-09-30 정원기계 (주) Real-time vision inspection device for gear parts
WO2022250190A1 (en) * 2021-05-27 2022-12-01 주식회사 솔루션에이 Defect determination system for image inspection object by using deep learning model
KR102527061B1 (en) 2021-11-15 2023-04-28 (주)미래융합정보기술 Manufacturing intelligence service system connected to MES in smart factory
WO2023080587A1 (en) * 2021-11-05 2023-05-11 한국기술교육대학교 산학협력단 Deep learning-based mlcc stacked alignment inspection system and method
KR102541166B1 (en) 2022-10-31 2023-06-14 주식회사 인터엑스 AI vision inspection system using robot
KR20230115024A (en) 2022-01-26 2023-08-02 부경대학교 산학협력단 Deep Learning based Vision System for Inspecting Atypical Defect and Method for Controlling the same
KR102660626B1 (en) * 2023-11-07 2024-04-25 주식회사 파인더아이 System and method for vision inspection based on AI that can be verified on site
KR20240067639A (en) 2022-11-09 2024-05-17 이재준 Vision alignment system and error correction method for vibration and vibration during stage movement using the same

Families Citing this family (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
RU2686257C1 (en) * 2018-04-27 2019-04-24 Ационерное общество "РОТЕК" (АО "РОТЕК") Method and system for remote identification and prediction of development of emerging defects of objects
EP3837535B1 (en) * 2018-09-19 2024-04-17 Edwards Lifesciences Corporation Medical implant inspection using imaging
KR102271418B1 (en) * 2019-06-17 2021-07-01 라온피플 주식회사 Method and system for auto training for product inspection
KR20210057518A (en) * 2019-11-12 2021-05-21 라온피플 주식회사 Apparatus and method for generating a defect image
KR102298708B1 (en) * 2020-06-15 2021-09-07 한국생산기술연구원 Method of inspecting defects present in the inspection object and apparatuses performing the same
CN111912759A (en) * 2020-07-15 2020-11-10 天津日博工业技术有限公司 Method for testing air permeability of waterproof breathable film
KR20220085589A (en) * 2020-12-15 2022-06-22 현대모비스 주식회사 Method and System for Defect Inspection Based on Deep Learning
CN113111946A (en) * 2021-04-15 2021-07-13 宁波九纵智能科技有限公司 Quality control method and system integrating hands, eyes and brains
CN114113119A (en) * 2021-12-03 2022-03-01 浙江伟星实业发展股份有限公司 Product visual quality inspection method, system, equipment and medium based on defect grade

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2005156334A (en) * 2003-11-25 2005-06-16 Nec Tohoku Sangyo System Kk Pseudo defective image automatic creation device and imaging inspection device
KR100786823B1 (en) * 2005-12-30 2007-12-20 한이진 A quality monitoring system of butchered chicken line and method thereof
KR101010136B1 (en) * 2009-01-30 2011-01-24 (주) 티아이에스 System for intelligent quality judgment based on HTM
KR20160105082A (en) * 2015-02-27 2016-09-06 삼성전기주식회사 Apparatus for testing board and testing method thereof
KR101688458B1 (en) * 2016-04-27 2016-12-23 디아이티 주식회사 Image inspection apparatus for manufactured articles using deep neural network training method and image inspection method of manufactured articles thereby

Cited By (30)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2021033791A1 (en) 2019-08-19 2021-02-25 엘지전자 주식회사 Ai-based new learning model generation system for vision inspection on product production line
WO2021033792A1 (en) 2019-08-19 2021-02-25 엘지전자 주식회사 Ai-based pre-training model determination system, and ai-based vision inspection management system using same for product production lines
WO2021049868A1 (en) * 2019-09-11 2021-03-18 (주) 넥스트랩 System and method for product quality management
KR20220014898A (en) * 2019-09-11 2022-02-07 (주) 넥스트랩 Product quality management system and method
KR102108956B1 (en) 2019-11-13 2020-05-12 주식회사 만컴퍼니 Apparatus for Performing Inspection of Machine Vision and Driving Method Thereof, and Computer Readable Recording Medium
KR20210071882A (en) 2019-12-06 2021-06-16 주식회사 아이디알인비전 Monitoring method for prouct faults in manufacturing process of ball joint for car
WO2021187683A1 (en) * 2020-03-20 2021-09-23 라온피플 주식회사 Method and system for testing quality of new product by using deep learning
KR102323607B1 (en) * 2020-05-19 2021-11-08 부산대학교 산학협력단 Method and system for detecting anomaly based images
KR20220038858A (en) 2020-09-21 2022-03-29 주식회사 스프링클라우드 Deep learning-based defect data detection apparatus and method of processed autonomous driving scenario data
KR20220042916A (en) 2020-09-28 2022-04-05 (주)미래융합정보기술 Vision inspection system by using remote learning of product defects image
WO2022097775A1 (en) * 2020-11-05 2022-05-12 위즈코어 주식회사 5g-based production, logistics management, and cloud-oriented machine vision service providing method
KR20220065545A (en) * 2020-11-13 2022-05-20 주식회사 트윔 Apparatus for checking defect of object having atypical surface
WO2022108249A1 (en) * 2020-11-17 2022-05-27 (주)자비스 Method, apparatus, and program for generating training data, and method for detecting foreign substance using same
KR102362615B1 (en) * 2020-11-27 2022-02-15 (주)리눅스아이티 Plc system for supporting ai-based vision inspection and control method thereof
KR20220087360A (en) 2020-12-17 2022-06-24 연세대학교 산학협력단 Method and device for classifying and quantity of empty bottles through multi-camera based sequential machine vision
WO2022158628A1 (en) * 2021-01-20 2022-07-28 주식회사 솔루션에이 System for determining defect in display panel on basis of machine learning model
KR20220118140A (en) 2021-02-18 2022-08-25 (주)라임 Deep learning based injection quality inspection system
KR20220132220A (en) 2021-03-23 2022-09-30 정원기계 (주) Real-time vision inspection device for gear parts
WO2022250190A1 (en) * 2021-05-27 2022-12-01 주식회사 솔루션에이 Defect determination system for image inspection object by using deep learning model
KR102463064B1 (en) * 2021-08-18 2022-11-04 (주)썬컴퍼니 Paper box manufacturing system for detecting errors in manufacturing modules in real time based on convolutional neural network
KR102358948B1 (en) * 2021-08-18 2022-02-08 (주)썬컴퍼니 Automated system for manufacturing paper boxes for packaging based on ai
KR102454707B1 (en) * 2021-08-31 2022-10-14 오성컨버전스 주식회사 Environmentally Friendly Box Manufacturing Method Using IOT-based Box Manufacturing System
KR102402685B1 (en) * 2021-08-31 2022-05-26 오성컨버전스 주식회사 Box manufacturing method that provides IoT-based services
WO2023080587A1 (en) * 2021-11-05 2023-05-11 한국기술교육대학교 산학협력단 Deep learning-based mlcc stacked alignment inspection system and method
KR20230065558A (en) * 2021-11-05 2023-05-12 한국기술교육대학교 산학협력단 Deep learning-based MLCC stacked alignment inspection system and method
KR102527061B1 (en) 2021-11-15 2023-04-28 (주)미래융합정보기술 Manufacturing intelligence service system connected to MES in smart factory
KR20230115024A (en) 2022-01-26 2023-08-02 부경대학교 산학협력단 Deep Learning based Vision System for Inspecting Atypical Defect and Method for Controlling the same
KR102541166B1 (en) 2022-10-31 2023-06-14 주식회사 인터엑스 AI vision inspection system using robot
KR20240067639A (en) 2022-11-09 2024-05-17 이재준 Vision alignment system and error correction method for vibration and vibration during stage movement using the same
KR102660626B1 (en) * 2023-11-07 2024-04-25 주식회사 파인더아이 System and method for vision inspection based on AI that can be verified on site

Also Published As

Publication number Publication date
WO2019107614A1 (en) 2019-06-06

Similar Documents

Publication Publication Date Title
KR20190063839A (en) Method and System for Machine Vision based Quality Inspection using Deep Learning in Manufacturing Process
CN108805170B (en) Forming data sets for fully supervised learning
CN109919934B (en) Liquid crystal panel defect detection method based on multi-source domain deep transfer learning
WO2019188040A1 (en) Image processing device, image processing method, and image processing program
CN111693534B (en) Surface defect detection method, model training method, device, equipment and medium
CN115082683A (en) Injection molding defect detection method based on image processing
Jiang et al. A machine vision-based realtime anomaly detection method for industrial products using deep learning
CN111079638A (en) Target detection model training method, device and medium based on convolutional neural network
CN113642474A (en) Hazardous area personnel monitoring method based on YOLOV5
KR20220085589A (en) Method and System for Defect Inspection Based on Deep Learning
US20220076404A1 (en) Defect management apparatus, method and non-transitory computer readable medium
JP2016110290A (en) Information processing apparatus and information processing method
CN113516146A (en) Data classification method, computer and readable storage medium
KR20210122429A (en) Method and System for Artificial Intelligence based Quality Inspection in Manufacturing Process using Machine Vision Deep Learning
CN111758117A (en) Inspection system, recognition system, and learning data generation device
CN117036271A (en) Production line quality monitoring method and system thereof
CN117030129A (en) Paper cup on-line leakage detection method and system thereof
CN115908988A (en) Defect detection model generation method, device, equipment and storage medium
CN117557784B (en) Target detection method, target detection device, electronic equipment and storage medium
CN112967292B (en) Automatic cutout and scoring method and system for E-commerce products
JP6988995B2 (en) Image generator, image generator and image generator
WO2020137228A1 (en) Image determination device, image determination method, and image determination program
CN114202544B (en) Complex workpiece defect detection method based on self-encoder
CN114638989A (en) Fault classification visualization method based on target detection and fine-grained identification
JP7206892B2 (en) Image inspection device, learning method for image inspection, and image inspection program