딥러닝 알고리즘을 이용한 머신 비전 기반 불량 검출 연구

김대현,부승빈,홍현철,여원구,이남용 한국비파괴검사학회 2020 한국비파괴검사학회지 Vol.40 No.1

최근 4차 산업혁명의 핵심기술인 딥러닝과 머신비전을 융합하여 제품의 불량을 검출하는 사례가 증가하고 있다. 본 논문에서는 케라스 (Keras) 오픈소스 라이브러리를 이용해 딥러닝 (Deap Learning)과 머신비전 (Machine vision) 기반의 불량 검출 소프트웨어를 개발하였고, 이를 이용해 정상품의 이미지를 기준으로 불량 유무를 판단하고 이후 불량 위치를 확률 분포로 찾아내도록 하였다. 또한 해당 소프트웨어의 성능을 검증 하기 위해, 이미지편집기로 제작한 이미지를 이용한 기본 검증실험과 실제 조립 블록을 이용한 검증실험 그리고 실제 전기 브레드 보드를 이용한 준 실제 적용실험으로 나누어 진행되었다. 이를 통해, 딥러닝 알고리 즘을 이용한 머신비전 기반의 불량 검출 시스템이 불량 유무와 불량 위치를 정확히 찾아낼 수 있음을 확인 하였다. Currently, there are numerous methods for detecting product defects by combining deep learning and machine vision, which are the core technologies of the fourth industrial revolution. In this study, we have developed a software that can identify defects, based on deep learning and machine vision, using the Keras open source library. The software was used to determine the defect based on an image of the regular product, and then identify its location using probability distribution. In addition, three verification experiments were carried out, the first which is a basic verification experiment, using an image produced by an image editor, the second, using an assembly block; and finally, a semi-real application experiment using an electric bread-board. Through these experiments, it was confirmed that machine vision-based defect detection system using deep learning algorithm could idetify the defects and pinpoint their locations.

Vision Sensor Technology Trends for Industrial Inspection System

김기수(Kisoo Kim),박준(June Park) Korean Society for Precision Engineering 2021 한국정밀공학회지 Vol.38 No.12

The fourth industrial revolution is rapidly emerging as a new innovation trend for industrial automation. Accordingly, the demand for inspection equipment is highly increasing and vision sensor technologies are continuously evolving. Machine vision algorithms applied to deep learning are also being rapidly developed to maximize the performance of inspection equipment. In this review, we highlight the recent progress of vision sensor technology for the industrial inspection system. In particular, inspection principles and industrial applications of a vision sensor are classified according to the vision scanning methods. We also discuss machine vision-based inspection techniques containing rule- and deep learning-based image processing algorithms. We believe that this review provides novel approaches for various inspection fields of agriculture, medicine, and manufacturing industries.

기계 시각과 포스트휴먼 주체: 박경근, 베르토프, 스노우의 카메라와 새로운 시지각

문혜진 현대미술사학회 2018 현대미술사연구 Vol.0 No.44

This paper started with interest in heterogeneous images and senses produced by machine vision. Since the invention of the camera, human vision has already been encroached by machines as it relies on machines to create and view images. However, the intervention of machines differing from the human eyes alters everything related to the visual system, including the nature of vision, the way of cognition, the attributes of the produced images, the observer's status, and the concept of subject. As a starting point for approaching this huge research topic, this paper cross-compares three film and video works dealing with machine vision based on camera, i.e. 1.6 sec(2016) by Kelvin Kyung Kun Park, La Région Centrale (1971) by Michael Snow and The Man with a Movie Camera (1929) by Dziga Vertov, and analyzes the heterogeneity of images, the relationship between human and machine, the possibility of new visibility and subject in these works. All of these works presents new perspectives that can not be grasped by the naked eye, which breaks the classical visual field and the humanistic visual system. Especially, the way that human factor is mixed with the production and appreciation of machine vision poses various issues about posthuman subject. Other visual machines including CCTV, smart phone, game and social media scattered everywhere are also changing current images and visual experiences. Changing visual perception and the development of alternative visual system depending on the varying concepts of machine will be left as a future research project. 박경근, 지가 베르토프, 마이클 스노우의 작업에 대한 비교 분석을 통해 이들 작업에서 드러나는 기계 시각과 이로 인한 새로운 시지각 및 포스트휴먼 주체의 가능성에 대해 논한다. 카메라를 통한 시각의 변형이 인간적 시각과 시각 체제, 나아가 인간-주체 개념에 어떤 영향을 끼치는지를 살펴본다.

세탁기 밸런스 방향 인식을 위한 머신 비전 시스템 개발

정상화,김태호 한국공작기계학회 2008 한국공작기계학회 추계학술대회논문집 Vol.2008 No.-

When the laundry rotates in the washing machine, it tends to lean toward one side. This tendency causes a serious vibration. The balance of washing machine plays an important role which reduces the vibration by injecting the sand or the salt water into the balance of washing machine. The hot plate welder is used to prevent from outflow of contents. The hot plate welder brings about many problems which is concerned with accident. The workers are protected by loading/unloading system for balance of washing machine in hot plate welder. This system is required direction recognition and location information in the balance. In this paper, development of machine vision system for recognition direction of balance in washing machine is studied. Pattern recognition techniques are used to extract an geometric characteristics from the balance of washing machine and to acquire the most similar form the next image of balance. The data acquired by the pattern matching is used for recognition direction of balance. The image processing software for the system is developed using LabVIEW.

머신 비젼을 이용한 구멍 정밀도 기상측정시스템

김민호,김태영 한국공작기계학회 2009 한국공작기계학회 춘계학술대회논문집 Vol.2009 No.-

Producing high quality products with a fully automated machine tool system has been investigated in industry and academia for five decades. The current study focuses on the on-line evaluation and control of drilling processes. On-line tool wear monitoring and tool replacement at the proper time are important techniques, as it is necessary to prevent damage of cutting tool, machine, and workpiece. Unfortunately, past research has been superficial regarding the quality of a drilled hole. One of the cause is difficulty in determining hole quality. The measurement of hole location and diameter would be performed by CMM(Coordinate Measurement Machine). However, the usage of CMM requires much time and cost. In order to overcome the difficulties, we have developed a hole location and diameter error measuring device using machine vision. The developed measurement device attached to a CNC machine can determine hole quality quickly and easily.

Machine vision in Process Systems Engineering

J. Jay Liu,Hyun-woo Cho 제어로봇시스템학회 2011 제어로봇시스템학회 국제학술대회 논문집 Vol.2011 No.10

Machine vision has been introduced to solve numerous problems in Process Systems Engineering (PSE) as well as in other discipline. In the field of PSE however, machine vision’s potentials have never been explored to the full extent yet. This is not only because the techniques involved in machine vision are still poorly-known to researchers and practitioners in PSE, but also because the characteristics of the scenes that the images acquired in process industries are often difficult to analyze due to their stochastic nature. On the other hand, data analysis has been known in PSE for several decades and extensively employed in process industries producing innumerous applications. The purpose of this article is to give an overview of methods and possible applications of machine vision from a data analysis perspective, which is more familiar to PSE community.

Cascade 안면 검출기와 컨볼루셔널 신경망을 이용한 얼굴 분류

유제훈(Je-Hun Yu),심귀보(Kwee-Bo Sim) 한국지능시스템학회 2016 한국지능시스템학회논문지 Vol.26 No.1

머신비전을 사용하여 사람의 얼굴을 인식하는 다양한 연구가 진행되고 있다. 머신비전은 기계에 시각을 부여하여 이미지를 분류 혹은 분석하는 기술을 의미한다. 본 논문에서는 이러한 머신비전 기술을 적용한 얼굴을 분류하는 알고리즘을 제안한다. 이 얼굴 분류 알고리즘을 구현하기 위해 컨볼루셔널 신경망(Convolution neural network)과 Cascade 안면 검출기를 사용하였고, 피험자들의 얼굴을 분류하였다. 구현한 얼굴 분류 알고리즘의 학습을 위해 한 피험자 당 이미지 2,000장, 3,000장, 40,00장을 10회와 20회 컨볼루셔널 신경망에 각각 반복하여 학습과 분류를 진행하였고, 학습된 컨볼루셔널 신경망과 얼굴 분류 알고리즘의 실효성을 테스트하기 위해 약 6,000장의 이미지를 분류하였다. 또한 USB 카메라 영상을 실험 데이터로 입력받아 실시간으로 얼굴을 검출하고 분류하는 시스템을 구현하였다. Nowadays, there are many research for recognizing face of people using the machine vision. the machine vision is classification and analysis technology using machine that has sight such as human eyes. In this paper, we propose algorithm for classifying human face using this machine vision system. This algorithm consist of Convolutional Neural Network and cascade face detector. And using this algorithm, we classified the face of subjects. For training the face classification algorithm, 2,000, 3,000, and 4,000 images of each subject are used. Training iteration of Convolutional Neural Network had 10 and 20. Then we classified the images. In this paper, about 6,000 images was classified for effectiveness. And we implement the system that can classify the face of subjects in realtime using USB camera.

비전 센서를 이용한 세탁기 밸런스 방향 인식 시스템에 관한 연구

정상화(Sang-Hwa Jeong),김태호(Tae-Ho Kim),김재상(Jae-Sang Kim) 한국기계가공학회 2008 한국기계가공학회 춘추계학술대회 논문집 Vol.2008 No.-

When the laundry rotates in the washing machine, it tends to lean toward one side. This tendency causes a serious vibration. The balance of washing machine plays an important role which reduces the vibration by injecting the sand or the salt water into the balance of washing machine. The hot plate welder is used to prevent from outflow of contents. The hot plate welder brings about many problems which is concerned with accident. The workers are protected by loading/unloading system for balance of washing machine in hot plate welder. This system is required direction recognition and location information in the balance. In this paper, development of machine vision system for recognition direction of balance in washing machine is studied. Pattern recognition techniques are used to extract an geometric characteristics from the balance of washing machine and to acquire the most similar form the next image of balance. The data acquired by the pattern matching is used for recognition direction of balance. The image processing software for the system is developed using LabVIEW.

비전 시스템을 이용한 인쇄형상의 오차 보정

이철수(Cheol-Soo Lee),이희승(Hee-Seung Lee),최인휴(In-Hugh Choi),허은영(Eun-Young Heo),김종민(Jong-Min Kim) (사)한국CDE학회 2012 한국 CAD/CAM 학회 학술발표회 논문집 Vol.2012 No.2

Today, machine vision is widely spread in manufacturing applications such as tool monitoring, workpiece recognition, collision detection, and measuring and inspection, offering digitized information to manufacturing systems. The main process of machine visioning consists of several steps: i) object image obtaining, ii) image processing and analysis, iii) feedback of image information to a control system, and iv) actuator manipulation. Especially, in the case of recognizing printed 2D shapes and generating machining data, the image processing and analysis step affects much to manufacturing quality and shop floor productivity. For instance, the patterns of a cover lens of a mobile phone camera are printed into a tempered glass, being measured through a vision system. Machining data for the camera lens is then generated by image processing and analysis. However, thermal effects and printing machine resolution force printed shapes to get errors such as irregular arrangement, rotated patterns, miss-printed patterns, and so on. Thus, this paper identifies the patterns of errors occurred in the image processing and analysis step, proposing an error compensation method based on error patterns for vision based manufacturing systems. A prototype vision-based manufacturing system is shown.

머신비전을 이용한 타원형 기어 검사 시스템에 관한 연구

박진주(Jin Joo Park),김기환(Gi hwan Kim),이응석(Eung Seok Lee) 대한기계학회 2014 大韓機械學會論文集A Vol.38 No.1

타원형 기어는 오발 유량계에 사용되는 특수기어로써 타원형 모양에 의해 생긴 공간으로 물을 흘려보내 계측한다. 본 연구의 목적은 타원형 기어의 가공정도를 판단하고 불량을 가려내는 시스템을 머신비전을 이용하여 개발하는 것이다. 각종 산업분야에서 자동화가 확산되면서 머신미전의 수요는 늘고 있으며 업계전반의 생산공정 중 검사 공정에서 빼놓을 수 없는 인자가 되었다. 하지만 머신비전을 이용한 기어측정은 기어의 형상이 복잡하다는 이유로 잘 사용되지 않고 있다. 본 연구에서는 머신비전을 이용한 검사 프로그램으로 타원형 기어의 측정이 가능함을 보였으며 본 연구에서 설계한 타원형 기어의 가공정도를 판단할 수 있었다. Elliptical gears are used in the oval flowmeter and oval flow meter inspects volume of water thanks to space by the elliptical shape. The purpose of this study is to judge accuracy of processing of the elliptical gear and develop inspection system using machine vision. Demand of machine vision is increasing while the factory automation is spreading and principle factor in-process inspection. But, gear inspection using the machine vision rarely used because of complex shape of gear. In this study, it seems possible that elliptical gear is inspected by inspection software using machine vision and inspection program can judge accuracy of processing of the elliptical gear designed this study.