ETRI-Knowledge Sharing Plaform

ENGLISH

성과물

논문 검색
구분 SCI
연도 ~ 키워드

상세정보

학술대회 Dynamic Object Recognition Using Precise Location Detection and ANN for Robot Manipulator
Cited 11 time in scopus Download 4 time Share share facebook twitter linkedin kakaostory
저자
김계경, 조재민, 표지형, 강상승, 김진호
발행일
201705
출처
International Conference on Control, Artificial Intelligence, Robotics and Optimization (ICCAIRO) 2017, pp.237-241
DOI
https://dx.doi.org/10.1109/ICCAIRO.2017.52
협약과제
17PS1300, 작업자 공간공유 및 스마트공장 적용을 위한 차세대 제조용 로봇, 강상승
초록
This paper presents vision based dynamic object recognition system for robot manipulation tasks by increasing the needs of automation machine vision. Object recognition or localization technology is used for pick and place task using robot. The dynamic object recognition system detects landmark features using neural network and provides grasping points of randomly located object, bin-picking object, visual servoing object to robot. The characteristic of dynamic object is free of posture, location, shape, distance, stacked form and is not restricted in illumination condition. This paper uses neural network based feature extractor and object classifier to recognize dynamic object. Dynamic object recognition system goes through image processing less impact to illumination effect, landmark feature extraction according to an object, coupled NN based object detection and recognition. We have evaluated performance of dynamic object recognition by testing detection of location, estimation of posture, distance to object and by identifying object type. And the other performance has evaluated by processing pick and place task using robot manipulator.
키워드
Aritificial neural network, Dynamic object recognition, Object detection, Robot manipulator
KSP 제안 키워드
Bin-picking, Feature extractioN, Illumination conditions, Illumination effect, Image processing, Located object, Neural networks, Object Detection and Recognition, Pick and Place, Precise location, Robot manipulation