• Restorative Dentistry and Endodontics
• /
• 제21권1호
• /
• pp.227-241
• /
• 1996

The effect of collagen dissolution in acid conditioned dentin was morphologically examined by both scanning and transmission electron microscopy. 18 freshly extracted human molars and dentin bonding systems of All Bond 2, Scotchbond Multipurpose, Superbond D-Liner were used in this study. For SEM preparation, each 3 of ~ exposed dentin surfaces were acid conditioned by using various acids within the above three bonding systems respectively. After acid conditioning of the other 3 exposed dentin surfaces as above, they were treated with 1.7% NaOCl for 2 minutes. The remaining 3 dentin surfaces were acid conditioned and treated with 3.3 % NaOCl for 2 minutes. All of the specimens were then fixed in 4 % glutaraldehyde for 12 h at $4^{\circ}C$ and dehydrated in ethanols grades from 50 % to 100 %, then surface changes of the specimens were observed by using SEM. For TEM preparation, exposed dentin surfaces were acid conditioned with the same acid as SEM specimens and treated with 1.7%, 3.3 % NaOCl respectively, then applied with corresponding bonding agents. After the procedures were finished, composite resin were applied on the dentin surfaces and light cured. Small, rectangular sticks with end dimensions of approximately 1 by 1 mm were sectioned and further sample preparative techniques for transmission electron microscopy were performed in accordance with the procedures used for ultrastructural TEM observations of calcified tissues. The results were as follows : 1. In the 1.7 % NaOCl retreated specimens after acid conditioning, the porous dentin surface of intertubular dentin and wide opening of dentinal tubules were appeared. And there were fine irregularities on the intertubular dentin, indicating a clear difference as compared with the acid conditioned specimens. 2. In the 3.3% NaOCl retreated specimens after acid conditioning, the intertubular dentin was further eroded causing a more porous and wider opening of dentinal tubules. Moreover, sharp irregularities on the intertubular dentin were more evident than those of acid conditioned and 1.7% NaOCl retreated specimens. 3. In all of the acid conditioned specimens, the resin-dentin hybrid layer of approximately 3.5mm thickness was formed and the collapsed collagen layer was observed on the uppermost part of hybrid layer in the specimens applied with All Bond 2. The collgen fibrils of intertubular dentin in specimens applied with Scotchbond Multipurpose were running perpendicular to the interface, and electron dense black layer demarcated from the deep unaltered dentin was more evident in the specimen applied with Superbond D-Liner than any other specimens. 4. In the 1.7 % NaOCl retreated specimens after acid conditioning, the resin-dentin hybrid layer of approximately 2.5-3.0mm thickness was formed and the collapsed collagen layer and longitudinally running collagen fibrils as shown in the acid conditioned specimens were observed in the specimens applied with All Bond 2 and Superbond D-Liner. 5. In all of the 3.3% NaOCl retreated specimens after acid conditioning, the evidence of resin-dentin hybrid layer was not identified ; nevertheless, the longitudinally running collagen fibrils remained slightly in the specimens applied with All Bond 2.

• 한국융합학회논문지
• /
• 제9권8호
• /
• pp.1-8
• /
• 2018

얼굴인식, 홍채인식과 같은 생체보안 분야에서 눈, 코, 입술 등 얼굴특징을 추출하는 과정은 필수적이다. 본 논문은 초고속(faster) R-CNN을 이용하여 얼굴영상에서 눈 및 입술영역을 검출하는 방법을 연구하였다. 초고속 R-CNN은 딥러닝을 이용한 물체검출 방법으로 기존의 특징기반 방법에 비해 성능이 우수한 것으로 알려져 있다. 본 논문에서는 얼굴영상에 콘볼루션, 선형정류과정, max pooling과정을 차례로 적용하여 특징맵을 추출하고 이로부터 제안영역(region proposal)을 검출하는 RPN(region proposal network)을 학습한다. 그리고 제안영역과 특징맵을 이용하여 눈 및 입술 검출기(detector)를 학습한다. 제안방법의 성능을 검토하기 위해 남녀한국인 얼굴영상 800장으로 실험하였다. 학습을 위해 480장을 이용했으며 테스트용으로 320장을 사용하였다. 컴퓨터모의 실험결과 눈 및 입술영역 검출의 평균정확도는 50 에포치일 때 각각 97.7%, 91.0%를 얻을 수 있었다.

• 대한한의진단학회지
• /
• 제15권1호
• /
• pp.1-28
• /
• 2011

Objectives: A study on the importance and consistency of pulse-daignosis in the Dongeuibogam Methods: We used Deyeuk Dongeuibogam of Dongeuibogam publishing company from original photographic edition. Results: The frequency of 27 mek (pulse condition) in Dongeuibogam is as in the following. Bumek (Floating pulse) appeared 120(8.9%) times, wanmek (moderate pulse) appeared 28(2%) times, chokmek (running pulse) appeared 7 (0.5%) times, gyumek (hollow pulse) appeared 19 (1.4%) times, saekmek (uneven pulse) appeared 33 (2.4%) times, sapmek (uneven pulse) appeared 51 (3.8%) times, kyulmek (knotted pulse) appeared 18 (1.3%) times, whalmek (slippery pulse) appeared 69 (5.1%) times, chimek (slow pulse) appeared 43 (3.2%) times, demek (intermittent pulse) appeared 13 (1%) times, silmek (replete pulse) appeared 45 (3.3%) times, bokmek (deep-sited pulse) appeared 29 (2.1%) times, neomek (firm pulse) appeared 4 (0.3%) times, hyunmek (taut pulse) appeared 110 (8.1%) times, yumek (soft pulse) appeared 20 (1.5%) times, dongmek (short and rapid pulse) appeared 16 (1.2%) times, kinmek (tense pulse) appeared 67 (5%) times, yakmek (weak pulse) appeared 46 (3.4%) times, semek (thready pulse) appeared 62 (4.6%) times, hongmek (full pulse) appeared 50 (3.7%) times, jangmek (long pulse) appeared 14 (1%) times, sakmek (rapid pulse) appeared 103 (7.6%) times, mimek (indistinctive pulse) appeared 65 (4.8%) times, danmek (short pulse) appeared 16 (1.2%) times, demek (large pulse) appeared 106 (7.9%) times, chimmek (deep pulse) appeared 112 (8.3%) times, heomek appeared 70 (5.2%) times, sanmek (scattered pulse) appeared 14(1%)times. Conclusions: We can know Donguibogam is given on the basis 27mek (pulse condition), because the frequency of 27mek (pulse condition) is high. But there are another expressions. So we can not say that Donguibogam is consistent in expressing mekbub(the way of pulse-diagnosis).

• 천문학회지
• /
• 제48권1호
• /
• pp.21-55
• /
• 2015

If the Universe is dominated by cold dark matter and dark energy as in the currently popular ${\Lambda}CDM$ cosmology, it is expected that large scale structures form gradually, with galaxy clusters of mass $M{\geq}10^{14}M_{\odot}$ appearing at around 6 Gyrs after the Big Bang (z ~ 1). Here, we report the discovery of 59 massive structures of galaxies with masses greater than a few times $10^{13}M_{\odot}$ at redshifts between z = 0.6 and 4.5 in the Great Observatories Origins Deep Survey fields. The massive structures are identified by running top-hat filters on the two dimensional spatial distribution of magnitude-limited samples of galaxies using a combination of spectroscopic and photometric redshifts. We analyze the Millennium simulation data in a similar way to the analysis of the observational data in order to test the ${\Lambda}CDM$ cosmology. We find that there are too many massive structures (M > $7{\times}10^{13}M_{\odot}$) observed at z > 2 in comparison with the simulation predictions by a factor of a few, giving a probability of < 1/2500 of the observed data being consistent with the simulation. Our result suggests that massive structures have emerged early, but the reason for the discrepancy with the simulation is unclear. It could be due to the limitation of the simulation such as the lack of key, unrecognized ingredients (strong non-Gaussianity or other baryonic physics), or simply a difficulty in the halo mass estimation from observation, or a fundamental problem of the ${\Lambda}CDM$ cosmology. On the other hand, the over-abundance of massive structures at high redshifts does not favor heavy neutrino mass of ~ 0.3 eV or larger, as heavy neutrinos make the discrepancy between the observation and the simulation more pronounced by a factor of 3 or more.

• 한국가스학회지
• /
• 제23권5호
• /
• pp.52-58
• /
• 2019

본 연구에서는 국내 도시가스 인수량에 대한 예측 모델을 개발하였다. 국내의 도시가스 회사는 KOGAS에 차년도 수요를 예측하여 보고해야 하므로 도시가스 인수량 예측은 도시가스 회사에 중요한 사안이다. 도시가스 사용량에 영향을 미치는 요인은 용도구분에 따라 다소 상이하나, 인수량 데이터는 용도별 구분이 어렵기 때문에 특정 용도에 관계없이 영향을 주는 요인으로 외기온도를 고려하여 모델개발을 실시하였다.실험 및 검증은 JB주식회사의 2008년부터 2018년까지 총 11년 치 도시가스 인수량 데이터를 사용하였으며, 전통적인 시계열 분석 중 하나인 ARIMA(Auto-Regressive Integrated Moving Average)와 딥러닝 기법인 LSTM(Long Short-Term Memory)을 이용하여 각각 예측 모델을 구축하고 두 방법의 단점을 최소화하기 위하여 다양한 앙상블(Ensemble) 기법을 사용하였다. 본 연구에서 제안한 일별 예측의 오차율 절댓값 평균은 Ensemble LSTM 기준 0.48%, 월별 예측의 오차율 절댓값 평균은 2.46%, 1년 예측의 오차율 절댓값 평균은 5.24%임을 확인하였다.

• 한국정보과학회논문지:시스템및이론
• /
• 제26권12호
• /
• pp.1458-1467
• /
• 1999

VLSI 공정 기술이 발달하면서 이웃한 전선 간의 간격이 점점 더 가까워 지고 있으며, 그에 따라 인접 전선 간의 혼신 문제가 심각해지고 있다. 본 논문에서는 3층 그리드 채널 배선에 적용 가능한 혼신을 최소화시키는 배선층 할당 방법을 제안한다. 이 방법은 선형 의사 불린 최적화 기법에 맞도록 고안되었으며, 적절한 변수 선택 휴리스틱과 상한값 추정 방법을 통하여 최적의 결과를 짧은 시간 안에 찾아낸다. 실험 결과를 통하여, 일반적인 0/1 정수 선형 프로그래밍 기법과 비교하여 성능과 수행시간 면에서 우수함을 보인다. Abstract Current deep-submicron VLSI technology appears to cause crosstalk problem severe since it requires adjacent wires to be placed closer and closer. In this paper, we deal with a horizontal layer assignment problem for three layer HVH channel routing to minimize coupling capacitance, a main source of crosstalk. It is formulated in a 0/1 integer linear programming problem which is then solved by a linear pseudo boolean optimization technique. Experiments show that accurate upper bound estimation technique effectively reduces crosstalk in a reasonable amount of running times.

• 한국정보통신학회:학술대회논문집
• /
• 한국정보통신학회 2018년도 춘계학술대회
• /
• pp.671-674
• /
• 2018

본 연구에서는 모형 자동차를 이용한 YOLO 운전 보조 시스템을 구현 하였다. YOLO는 최근에 잇슈가 되고 있는 딥 러닝을 사용하는 물체 감지 및 인식 알고리즘입니다. 이 시스템은 카메라를 통해 획득한 영상에 영상처리 기술을 적용하여 차선 이탈을 경고하고, YOLO를 이용하여 객체를 인식하며 객체 유형 및 차량 사이의 거리에 따라 다양한 기능을 수행한다. 기존 물체 검출 및 인식 알고리즘 보다 우수한 YOLO는 추가 장비 없이 주행 보조 시스템 성능을 향상시킨다. YOLO를 이용한 주행 보조 시스템은 적은 비용으로 운전자의 안전성을 확보할 수 있을 것이다.

• 한국컴퓨터정보학회논문지
• /
• 제21권7호
• /
• pp.1-8
• /
• 2016

Nowadays, distributed denial of service (DDoS) attacks on web sites reward attackers financially or politically because our daily lifes tightly depends on web services such as on-line banking, e-mail, and e-commerce. One of DDoS attacks to web servers is called HTTP-GET flood attack which is becoming more serious. Most existing techniques are running on the application layer because these attack packets use legitimate network protocols and HTTP payloads; that is, network-level intrusion detection systems cannot distinguish legitimate HTTP-GET requests and malicious requests. In this paper, we propose a practical detection technique against HTTP-GET flood attacks, based on the access behavior of inline objects in a webpage using NetFlow data. In particular, our proposed scheme is working on the network layer without any application-specific deep packet inspections. We implement the proposed detection technique and evaluate the ability of attack detection on a simple test environment using NetBot attacker. Moreover, we also show that our approach must be applicable to real field by showing the test profile captured on a well-known e-commerce site. The results show that our technique can detect the HTTP-GET flood attack effectively.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제14권10호
• /
• pp.4080-4097
• /
• 2020

In recent years, convolutional neural network (CNN) has become the primary method for face detection. But its shortcomings are obvious, such as expensive calculation, heavy model, etc. This makes CNN difficult to use on the mobile devices which have limited computing and storage capabilities. Therefore, the design of lightweight CNN for face detection is becoming more and more important with the popularity of smartphones and mobile Internet. Based on the CPU real-time face detector FaceBoxes, we propose a multi-task lightweight face detector, which has low computing cost and higher detection precision. First, to improve the detection capability, the squeeze and excitation modules are used to extract attention between channels. Then, the textual and semantic information are extracted by shallow networks and deep networks respectively to get rich features. Finally, the landmark detection module is used to improve the detection performance for small faces and provide landmark data for face alignment. Experiments on AFW, FDDB, PASCAL, and WIDER FACE datasets show that our algorithm has achieved significant improvement in the mean average precision. Especially, on the WIDER FACE hard validation set, our algorithm outperforms the mean average precision of FaceBoxes by 7.2%. For VGA-resolution images, the running speed of our algorithm can reach 23FPS on a CPU device.

• 한국제4기학회지
• /
• 제19권1호
• /
• pp.41-46
• /
• 2005

This study discusses the soil erosion on hillslopes and sediment deposition in lakes within catchments in South Korea. In order to determine seasonal variations of sedimentation in Yeongcheon and Seondong lakes, the sediment traps were set in the deep part of both lakes and lake sediments have been sampledmonthly from July 2004 to August 2005. Some properties such as highmineral content, fine particle size and high particle density in the Yeongcheon Lake indicate intensive soil erosion, sediment transportation and deposition throughout the catchment for a long time. The high sediment yield in the Seondong Lake is related with higher weathering intensity and extreme soil erosion by running water due to higher seasonal rainfall amount. Rates of erosion and sedimentation in the Seondong Lake are estimated to be higher than those of the Yeongcheon Lake, suggesting that the Seondong Lake is associated with higher precipitation, smaller catchment area, and extreme soil vulnerability to ephemeral erosion by overland flow during the heavy rainfall event. Consequently, both catchments are characterized by different erosion and sedimentation processes, as well as different geomorphic factors (bedrock, soil structure, rainfall intensity and catchment area).