로봇 물체 조작 작업을 위한 동적 3차원 장면 그래프 생성

정가영(Ga Young Jung),김인철(Incheol Kim) 제어로봇시스템학회 2021 제어·로봇·시스템학회 논문지 Vol.27 No.12

In this study, we proposed a novel dataset and a deep learning model that can generate three-dimensional (3D) dynamic scene graphs for robotic manipulation tasks. First, we defined a new 3D scene graph to effectively represent the dynamics of a robotic manipulation task environment. Subsequently, we collected a series of input sensory data by conducting multiple manipulation tasks in a simulated environment. Based on the collected sensory data and the corresponding 3D scene graphs, we constructed a dataset, namely, D3DSG, for training and validating a scene graph generation model. In addition, we proposed a ST-GCN based context reasoning module that can utilize both rich spatial and temporal contexts, after which an effective 3D scene graph generation model, namely, SG4RMT, which consisted of a 6DoF pose estimation module and a spatio-temporal context reasoning module, was presented. The superiority and high performance of the proposed SG4RMT model were demonstrated by performing multiple experiments using the D3DSG dataset.

유해가스 흡수처리를 위한 저비용 고효율 스크러버의 최적 설계기준 및 현장적용 방안

정가영(Ga-Young Jung),임경민(Kyung-Min Lim),마병철(Byung-Chol Ma) 한국가스학회 2021 한국가스학회지 Vol.25 No.3

사업장에서 유해가스를 처리하는 스크러버는 처리효율은 높으나 설계가 복잡하고 24시간 펌프를 가동시켜야 하므로 많은 비용이 소모될 수 있다. 따라서, 영세한 사업장에서는 스크러버를 설치하지 않거나 순환펌프의 작동을 중지한 상태로 스크러버를 운용하여 운영비를 최소화하고 있다. 이에, 본 연구에서는 소규모 사업장에서 경제적으로 활용할 수 있는 저비용 고효율 스크러버에 대한 적용방안을 연구하였다. 저비용 고효율 스크러버는 bubble column의 방식을 적용하여 유해화학물질 흡수처리 목적으로 장치를 활용하는 것으로, 이러한 스크러버의 개발을 위하여 실험을 통해 흡수성능을 검토하였으며 특정 조건에서 물리적 조건 변화에 따른 유해가스 흡수효율 변화 및 최적의 적용방안을 연구하였다. 그 결과, 저비용 고효율 스크러버로 유입되는 기체를 어느 정도 처리할 수는 있었으나 처리능력이 저하되는 문제가 발생하여 흡수액 공급을 통해 성능저하를 방지하고 순환량에 따라 일정 수준의 흡수율을 유지할 수 있는 것을 확인할 수 있었다. 이를 기반으로 최종적으로는 적정 순환시점 및 동결 방지방안을 포함한 저비용 고효율 스크러버의 현장 적용방안을 3가지 방식(type)으로 제시하였다. Scrubbers that treat hazardous materials at workplaces have high treatment efficiency; however, the design is complex, and pumps need to be operated 24 hours a day, which can be costly. Therefore, to minimize the operating costs, small businesses do not install scrubbers, or operate them while circulation pumps are suspended. Hence, this study investigated the application of low-cost, high-performance scrubbers that can be used economically in small businesses. Low-cost, high-efficiency scrubbers are applied to bubble columns to utilize devices for hazardous chemical absorption treatment purposes, and for the development of these scrubbers, absorption performance was reviewed and the optimal application method was studied under certain conditions. The changes in the absorption performance of hazardous gas were studied in certain environments by varying the physical conditions, and the optimal application methods were analyzed. The results showed that, while it was possible to treat some of the gas flowing into the low cost, high performance scrubber, the treatment capacity was reduced. Performance degradation was prevented by supplying an absorption liquid, and a certain level of absorption was maintained depending on the amount of circulation. Based on this, three types of site application methods of low cost, high performance scrubbers were presented. In addition, the appropriate timing of circulation and anti freezing measures were also discussed.

그래프신경망과 멀티 모달 장면 그래프 맥락 정보를 이용한 생성

정가영 ( Ga-young Jung ),김인철 ( In-cheol Kim ) 한국정보처리학회 2020 한국정보처리학회 학술대회논문집 Vol.27 No.1

본 논문에서는 입력 영상에 담긴 다양한 물체들과 그들 간의 관계를 효과적으로 탐지하여, 하나의 장면 그래프로 표현해내는 새로운 심층 신경망 모델을 제안한다. 제안 모델에서는 물체와 관계의 효과적인 탐지를 위해, 합성 곱 신경망 기반의 시각 맥락 특징들뿐만 아니라 언어 맥락 특징들을 포함 하는 다양한 멀티 모달 맥락 정보들을 활용한다. 또한, 제안 모델에서는 관계를 맺는 두 물체 간의 상호 의존성이 그래프 노드 특징값들에 충분히 반영되도록, 그래프 신경망을 이용해 맥락 정보를 임베딩한다. 본 논문에서는 Visual Genome 벤치마크 데이터 집합을 이용한 비교 실험들을 통해, 제안 모델의 효과와 성능을 입증한다.

모방범죄 방지를 위한 Scikit-learn 기반의 얼굴인식 도어락

정가영(Ga-Young Jung),이현정(Hyeon-Jung Lee),이창순(Chang-Soon Lee),조주연(Joo-Yeon Jo),황영준(Young-Jun Hwang),김남호(Nam-Ho Kim) 한국정보통신학회 2023 한국정보통신학회 종합학술대회 논문집 Vol.27 No.1

현대사회의 방범 시스템은 대부분 패스워드를 필요로 하는 전자 잠금 형태이다. 최근 비밀번호 키패드의 특성이 악용되는 모방 범죄가 지속적으로 발생되면서 도어락 관련 보안 문제가 사회적 문제로 대두되고 있다. 모방 범죄는 패스워드가 유출되거나 도어락의 남아있는 지문 등으로 암호를 유추하는 방식으로 발생한다. 이를 예방하기 위해 생체 기반의 바이오 인식과 일회성 인증 시스템을 활용하는 스마트 방범 시스템이 주목받고 있다. 따라서 본 논문에서는 라즈베리파이와 파이썬 머신러닝을 이용한 얼굴인식을 목표로 OpenCV를 통한 실시간 스트리밍 데이터베이스와 비교하여 흔적이 남지 않게 이용자 개인의 얼굴을 식별한다. 또한 아두이노 난수 코드를 활용해 임의로 생성되는 OTP(one time password)를 사용자에게 제공하여 일정 주기 동안만 유효한 일회성 패스워드로 보안성과 편의성을 확보한 도어락을 제안한다. Most of the home security systems in modern society are electronic locks that require passwords. Recently, as copycat crimes, which abuse the characteristics of a password keypad, continue to occur, security problems related to door locks have emerged as a social problem. Copycat crimes involve hacking or inferring passwords from fingerprints left on door locks. Smart crime prevention systems that utilize bio-based bio-recognition and one-time authentication systems are attracting attention as a means of preventing such crimes. Therefore, in this study, we develop a framework in which the face of the user is identified without leaving a trace by using a real-time streaming database via OpenCV with the goal of face recognition using Raspberry Pi and Python machine learning. In addition, we propose a door lock that ensures security and convenience by providing the users with a one time password that is generated randomly using Arduino random number code and is valid only for a limited time.

포스터 발표 : 상수처리 ; 순환식 광촉매 수처리 반응기 성능평가 연구

류정호 ( Jung Ho Ryu ),정가영 ( Ga Young Jung ),오제명 ( Jea Myung Oh ),김두수 ( Doo Soo Kim ),김동현 ( Dong Hyun Kim ) 대한상하수도학회 2007 공동추계 학술발표회 논문집 Vol.2007 No.-

상수처리 : 순환식 광촉매 수처리 반응기 성능평가 연구

류정호 ( Jung Ho Ryu ),정가영 ( Ga Young Jung ),오제명 ( Jea Myung Oh ),김두수 ( Doo Soo Kim ),김동현 ( Dong Hyun Kim ) 한국물환경학회 ( 구 한국수질보전학회 ) 2007 공동 추계학술발표회 Vol.2007 No.-

Michael Addition Reaction을 이용한 신규 Ester Quaternary Ammonium Salt 양이온 계면활성제의 합성

강은경(Eun kyung Kang),정선화(Seon Hwa Jung),정가영(Ga Young Jung),이병민(Byung Min Lee) 한국유화학회 2017 한국응용과학기술학회지 Vol.34 No.1

양이온 계면활성제는 살균력을 가지고 있으며 이를 개발하기 위한 연구는 다양한 산업분야에서 계속 진행되고 있다. 단순히 우수한 성능을 가질 뿐만 아니라 생분해성이 우수한 계면활성제를 개발하기 위한 연구는 증가되고 있다. 본 연구에서는 Michael addition 반응을 통하여 ester-type의 양이온 계면활성제를 합성하였다. 실온에서 촉매없이 alkyl acrylate와 1차 아민을 가지는 화합물을 합성한 뒤 dimethyl sulfate 로 4차화시킨다. 2개의 소수기와 한 개의 친수기를 갖는 HQ21과 2개의 소수기와 2개의 친수기를 갖는 HQ22를 합성하였다. 이들 합성화합물은 1H-NMR, HR-MS 와 FT-IR로 구조를 확인하였으며 생분해성을 측정하였다. Cationic surfactants have a bactericidal effect and the study for effective development of them became important parts in the industry. There have been increasing researches that focus on the development of products having not only outstanding features but also safety and biodegradability. In this work, novel ester-type cationic surfactants were obtained via Michael addition reaction. Intermediates were quantitatively prepared by the Michael addition reaction between alkyl acrylate and amine compounds under mild conditions without solvent and catalyst. The intermediates were quaternized with dimethyl sulfate. HQ21 with two hydrophobic groups and a hydrophilic group and HQ22 with two hydrophobic groups and two hydrophilic groups were obtained. The structures of the products were characterized by 1H-NMR, HR-MS and FT-IR and biodegradability of the products were tested.

자가혈위지압을 활용한 모바일 의료 어플에 대한 체계적 문헌고찰

박서경 ( Seokyung Park ),이조현 ( Johyun Lee ),정가영 ( Ga-young Jung ),장세린 ( Celine Jang ),김상호 ( Sang-ho Kim ) 경락경혈학회 2023 Korean Journal of Acupuncture Vol.40 No.1

Objectives : This study aims to provide information regarding the status and quality of mobile applications (MAs) using self-acupressure by performing a systematic review. Methods : We conducted comprehensive searching on five international databases and two app markets from inception to July 31, 2022 to identify MAs using self-acupressure. We analyzed the characteristics of each MA regarding the name of the MA, registered app markets, target symptoms, developers, the year and country of development, cost, target age, media function, and expertise. We assessed the quality of each MA using Mobile Application Rating Scale (MARS). Results : We identified a total of 30 MAs using self-acupressure (25 MAs from the app market and 6 MAs from clinical studies, with 1 MA in common). 17 out of 24 MAs from the app market provided self-acupressure regimens for various symptoms and the others provided regimens for specific symptoms such as memory, anxiety, depression, asthma, allergy, low back pain, and headache. 14 developers were reported. 23 MAs were developed after 2013. The largest number of MAs were developed in the United States. The target age group of 12 MAs was above the age of 3, and that of 11 MAs was above the age of 12. 14 MAs provided multimedia functions such as videos. 13 MAs provided information of expertise. From clinical studies, only 3 out of 6 MAs were accessible through the app market. 4 MAs were developed by the researchers of the study. In terms of MARS, the score of MAs from the app market was higher than that of MAs from clinical studies in both objective and subjective evaluation areas. Conclusions : This study summarizes the characteristics of MAs using self-acupressure. More MAs using self-acupressure should be developed and further clinical research for MA on each symptom and disease is warranted for the diversification of MA fields using self-acupressure.

가상현실(VR) 기반 화학사고 대응 다중협업 훈련 콘텐츠 개발에 관한 연구

안수빈(Su-Bin An),임경민(Kyung-Min Lim),고하은(Ha-Eun Go),정가영(Ga-Young Jung),마병철(Byung-Chol Ma) 한국디지털콘텐츠학회 2020 한국디지털콘텐츠학회논문지 Vol.21 No.1

As chemical accidents continue to occur, the importance of accident respond training is increasing. Accordingly, some response institutions are conducting mock response training, but it is difficult for responders to directly experience and train an accident scene similar to the actual one. In the midst of this, Virtual reality technology is emerging as a new alternative as responders can indirectly experience and respond to the accident scene by creating a virtual environment similar to a real chemical accident. Therefore, this study derived the role and composition of multi-collaboration training based on chemical accident response system and domestic and foreign chemical accident response training cases. In addition, by developing response and training scenarios and storyboards, contents for multi-collaboration training on chemical accident response based on virtual reality were developed. It is expected that the training applied with such training contents will contribute to the strengthening of the chemical accident responders’ ability and the improvement of the collaboration ability.

아파트 부엌 공간의 빌트인 가전에 관한 만족도 연구

박경난(Kyung Nan Park),임희민(Hee Min Lim),정가영(Ga Young Jung),전상미(Sang Mi Jeon) 충북대학교 생활과학연구소 2005 생활과학연구논총 Vol.9 No.1

  This study has been undertaken to research and analyse the use of built in kitchen appliances, the satisfaction brought upon by such appliances and the needs of the resident. The main aim is to provide the basic data needed when planning the construction of built-in kitchen appliances. The questionnaire survey was undertaken in the period May 7th to the 14th of 2004 with the target segment being the residents of apartment flats that have built-in kitchen appliances in Yong-in City. The SPSS11.5 package allowed the analysis of data using its system of descriptive statistics.   The satisfaction rate directly from built-in kitchen appliances was found to be at 3.35 and the satisfaction rate brought on to the home due these fixtures was found to be at an increased rate of 3.46. The positioning of the kitchen appliances was found to have the highest effect on satisfaction, while the type, size and the operations was found to have the least effect on satisfaction. It could also be seen that appliances, such as dishwashers imposed discomfort, such as difficulties during use and also showed a low rate of usage. The order of importance when selection of appliances were as such: type, brand name, size and colour, and it was recommended that fixtures that are not part of the basic kitchen fixtures, such as water filters and sidedish refrigerators, should be implemented.