• Proceedings of the Korea Information Processing Society Conference
• /
• 한국정보처리학회 2016년도 춘계학술발표대회
• /
• pp.627-628
• /
• 2016

해마다 발생하는 자연재해는 사망, 실종과 같은 인명 피해와 더불어 수억 원에 달하는 재산피해를 수반하기 때문에, 경제적, 사회적 손실을 최소화할 수 있는 ICT 기반 자연재난 대응 기술 개발에 대한 관심도가 높아지고 있다. 본 논문에서는 무인기 및 탑재 센서로부터 획득한 영상정보를 기반으로 재난 인지 및 대응을 위한 스마트아이 플랫폼을 소개하고, 재난 인지에 핵심요소 기술인 Deep Learning 기반 재난 영상 분석 기술을 제안한다. 제안하는 Deep Learning 기반 재난 영상 분석은 과거의 반복적으로 발생하는 영상 정보를 학습함으로써, 새로운 입력 영상에 대한 재난 여부를 사람과 유사한 판단이 가능하도록 한다. 제안하는 재난 영상 분석 기술에 대한 설계 방식 및 실험 성능 제시를 통하여 제안하는 기법에 대한 타당성을 검증한다.

• Korean Journal of Plant Taxonomy
• /
• 제47권2호
• /
• pp.124-131
• /
• 2017

Species identification is a fundamental and routine process in plant systematics, and linguistic-based dichotomous keys are widely used in the identification process. Recently, novel tools for species identification have been developed to improve the accuracy, ease to use, and accessibility related to these tasks for a broad range of users given the advances in information and communications technology. A visual identification key is such an approach, in which couplets consist of images of plants or a part of a plant instead of botanical terminology. We developed a visual identification key for 101 taxa of Orchidaceae in Korea and evaluated its performance. It uses short statements for image couplets to avoid misinterpretations by users. The key at the initial steps (couplets) uses relatively easy characters that can be determined with the naked eye. The final steps of the visual key provide images of species and information about distributions and flowering times to determine the species that best fit the available information. The number of steps required to identify a species varies, ranging from three to ten with an average of 4.5. A performance test with senior college students showed that species were accurately identified using the visual key at a rate significantly higher than when using a linguistic-based dichotomous key and a color manual. The findings presented here suggest that the proposed visual identification key is a useful tool for the teaching of biodiversity at schools, for the monitoring of ecosystems by citizens, and in other areas that require rapid, easy, and accurate identifications of species.

• Journal of the Optical Society of Korea
• /
• 제19권2호
• /
• pp.169-174
• /
• 2015

Muscle vibration measurement has long been an unique scientific study, in general, and the direct reaction of animals to feel pain (algesia), either from vascular or muscle contraction, is a complex perceptual experience. Thus this paper proposes a way to measure animal algesia quantitatively, by measuring the changes in muscle vibration due to a pinprick on the surface of the skin of a Spodoptera litura larva. Using the laser optical triangulation measurement principle, along with a CMOS image sensor, linear laser, software analysis, and other tools, we quantify the subtle object point displacement, with a precision of up to $10{\mu}m$, for our chosen Spodoptera litura larva animal model, in which it is not easy to identify the tiny changes in muscle contraction dynamics with the naked eye. We inject different concentrations of formalin reagent (empty needle, 12% formalin, and 37% formalin) to obtain a variety of different muscle vibration frequencies as the experimental results. Because of the high concentrations of reagent applied, we see a high frequency shift of muscle vibration, which can be presented as pain indices, so that the algesia can be quantified.

• Journal of the Korean Institute of Telematics and Electronics
• /
• 제27권11호
• /
• pp.32-41
• /
• 1990

This paper proposes a method of using image features in serving a robot manipulator. Specifically, the con-cept 'feature' is first mathematically defined and then differential relationship between the robot motion and feature vector is derived in terms of Feature Jacobian Matrix and its generalized inverse. Also, by using more features than the number of DOFs of the robot, the visual servoing performance is shown to be improv-ed. Via various examples, the method of feature-based servoing of a robot proposed in this paper is proved to be effective for conducting object-oriented robotic tasks.

• 전자공학회논문지 IE
• /
• 제48권1호
• /
• pp.16-24
• /
• 2011

In this paper, we developed a auto sorting system for T type welding nut using a vision inspector. We used edge and thread detection with histogram of image which is captured by machine vision camera. We also used a binary morphology operation for a detection of spot. As a result we performed numeric inspection of 0.1mm accuracy. This is impossible in old sorting system and inspector with naked eye. Also, we reduced the manufacturing unit cost to 25% and improved a production efficiency to 330%.

• Proceedings of the KOSOMBE Conference
• /
• 대한의용생체공학회 1996년도 추계학술대회
• /
• pp.90-94
• /
• 1996

We studied the problem of driver's low vigilance state which is related to the one reason of traffic accidents. In this paper, we developed the drowsiness warning system for a drowsy driver. To extract the eyes and mouth from the driver's facial image in real time, a computer vision method was used. The eye blink duration and yawning were used as measurement parameters of drowsiness detection. When the drowsy state of a driver was detected, the driver was refreshed by the scent generator and the alarm. Also, the driver's bio-signal was acquired and analyzed to measure the vigilance state.

• Proceedings of the KSRS Conference
• /
• 대한원격탐사학회 2002년도 Proceedings of International Symposium on Remote Sensing
• /
• pp.423-427
• /
• 2002

Contrary to the traditional text-based information, 4S(GIS,GNSS,SIIS,ITS) information can contribute to the citizen's welfare in upcoming era. Recently, GSIS(Geo-Spatial Information System) has been applied and stressed out in various fields. As analyzed the data from GSIS arena, the position information of objects and targets is crucial and critical. Therefore, several methods of getting and knowing position are proposed and developed. From this perspective, Position collection and processing are the heart of 4S technology. We develop 4S-Van that enables real-time acquisition of position and attribute information and accurate image data in remote site. In this study, the configuration of 4S-Van equipped with GPS, INS, CCD and eye-safe laser scanner is shown and the merits of DGPS/INS integration approach for geo-referencing is briefly discussed. The algorithm of DGPS/INS integration fur determination of six parameters of motion is eccential in the 4S-Van to avoid or simplify the complicated computation such as photogrammetric triangulation. 4S-Van has the application of Laser-Mobile Mapping System for three-dimensional data acquisition that merges the texture information from CCD camera. The technique is also applied in the fields of virtual reality, car navigation, computer games, planning and management, city transportation, mobile communication, etc.

• Journal of radiological science and technology
• /
• 제20권2호
• /
• pp.79-82
• /
• 1997

Image-based three dimensional radiation treatment planning(3D RTP) has a potential of generating superior treatment plans. Advances in computer technology and software developments quickly make 3D RTP a feasible choice for routine clinical use. However, it has become clear that an evaluation of a 3D plan is more difficult than a 2D plan. A number of tools have been developed to facilitate the evaluation of 3D RTP both qualitatively and quantitatively. For example, beam's eye view(BEV) is one of the most powerful and time-saving method as a qualitative tools. Dose-volume histogram(DVH) has been proven to be one of the most valuable methods for a quantitative tools. But it has a limitation to evaluate several different plans for biological effects of the tissue and critical organ. Therefore, there is a strong interest in developing quantitative models which would predict the likely biological response of irradiated organs and tissues, such as tumor control probability(TCP) and normal tissue complication probability(NTCP). DVH and NTCP of hepatoma were evaluated for three dimensional conformal radiotherapy(3D CRT). Also, 3D RTP was analysed as a dose optimization based on beam arrangement and beam modulation.

• Korean Journal of Clinical Laboratory Science
• /
• 제37권3호
• /
• pp.216-219
• /
• 2005

We report a case of narcolepsy. A 25-year-old man has had excessive daytime sleepiness of about 10 years durations. He awakens daily feeling exhausted and continually falls asleep during the day while engaged in such situation like reading and watching television. He has exhibited cataplexy, a sudden loss of muscular tone, brought on by emotion, usually laughter. Polysomnogram revealed increased sleep stage 1, 2 and decreased deep sleep. Multiple sleep latency test (MSLT) showed that sleep latency was 1.33 minutes and there were 3 noted sleep onset rapid eye movement (SOREM) on 5 trials. The epworth sleepiness scale (ESS) was 17/24. Typing of HLA haplotype that was positive for the $DQB1^{\ast}0602$ allele, and hypocretin-1 (orexin A) could not be detected in cerebrospinal fluid (CSF). Brain MRI showed normal image. We diagnosed his case as narcolepsy based on history of cataplexy, and three occurances of SOREM, and positive of HLA haplotype.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• 제12권1호
• /
• pp.196-204
• /
• 2008

In this paper, we propose an efficient method to classify human facial impression using frontal face image. The features that represent the shape of eye, jaw and face are used. The proposed method employs PCA, LDA and SVM in series. PCA is used to project the feature space to a low dimensional subspace. LDA produces well separated classes in a low dimensional subspace even under severe variation. This results in good discriminating power for classification. SVM is used to classify the data. Human face has been classified for 8 facial impressions. The experiments have been performed for many face images, and show encouraging result.