• 한국컴퓨터정보학회논문지
• /
• 제26권1호
• /
• pp.45-55
• /
• 2021

본 논문은 고해상도를 가지는 영상에서 겹쳐져있는 소형 물체를 효과적으로 검출하고 추적하는 알고리즘을 제안한다. Coarse to Fine 방식을 기본으로 하는 두 개의 Deep-Learning Network을 앙상블 형태로 구성하여 차량이 존재할 위치를 미리 판단하고 서브영역으로 선택한 이미지로부터 차량을 정확하게 검출한다. Coarse 단계에서는 서로 다른 다수의 Deep-Learning Network 에 대한 각각의 결과로 Voting Space를 생성한다. 각 Voting Space 의 조합을 통해 Voting Map을 만들고 차량이 존재할 위치를 선택한다. Fine 단계에서는 Coarse 단계에서 선택된 영역을 기준으로 서브영역을 추출하고 해당 영역을 최종 Deep-Learning Network 에 입력한다. 서브 영역은 Voting Map을 이용하여 영상에서의 높이에 적합한 크기의 동적 윈도우를 생성함으로써 정의되며, 본 논문에서는 원거리에서 근거리로 접근하는 도로의 이미지를 대상으로 미리 계산된 매핑테이블을 적용하였다. 각 서브 영역 간 이동하는 차량의 동일성 판단은 검출된 영역의 하단 중심점에 대한 근접성을 기반으로 하였으며, 이를 통해 이동하는 차량의 정보를 트래킹 하였다. 실제 주야간 도로 CCTV를 통해 획득한 실시간 영상에서 처리 속도 및 검출 성능을 비교 실험하여 제안한 알고리즘을 평가하였다.

• 한국멀티미디어학회논문지
• /
• 제25권8호
• /
• pp.1075-1084
• /
• 2022

Internet of Things(IoT) has become a key available technology for efficiently implementing device to device(D2D) services in various domains such as smart home, healthcare, smart city, agriculture, energy, logistics, and transportation. A lightweight publish/subscribe(Pub/Sub) messaging protocol not only establishes data dissemination pattern but also supports connectivity between IoT devices and their applications. Also, a Pub/Sub broker is deployed to facilitate data exchange among IoT devices. A scalable edge-based publish/subscribe (Pub/Sub) broker overlay networks support latency-sensitive IoT applications. In this paper, we design a hypercube grid quorum(HGQ) for distributed Pub/Sub systems based IoT applications. In designing HGQ, the network of hypercube structures suitable for the publish/subscribe model is built in the edge layer, and the proposed HGQ is designed by embedding a mesh overlay network in the hypercube. As their applications, we propose an HGQ-based mechansim for dissemination of the data of sensors or the message/event of IoT devices in IoT environments. The performance of HGQ is evaluated by analytical models. As the results, the latency and load balancing of applications based on the distributed Pub/Sub system using HGQ are improved.

• 한국전자통신학회논문지
• /
• 제11권7호
• /
• pp.665-670
• /
• 2016

본 논문에서는 IoT(Internet of Things) 시스템의 기본 구성이 되는 센서 네트워크에 사용될 수 있는 MOSFET(Metal Oxide Semiconductor Field Effect Transistor)의 Sub-threshold 동작을 이용하는 OP-AMP(Operational amplifier) 설계를 제안한다. MOSFET의 Sub-threshold 동작은 전원전압을 낮추는 효과로 회로 시스템을 초저전력으로 유도할 수 있는 특징이 있기 때문에 배터리를 사용하는 IoT의 센서 네트워크 시스템의 초저전력화에 매우 유용한 회로설계 기술이라고 할 수 있다. $0.35{\mu}m$ 공정을 이용한 시뮬레이션 결과, VDD를 0.6 V로 설계할 수 있었으며, OP-AMP 의 Open-loop Gain은 43 dB, 또한 설계한 OP-AMP의 소비전력은 $1.3{\mu}W$가 계산되었다. 또한, Active Layout 면적은 $64{\mu}m{\times}105{\mu}m$이다. 제안한 OP-AMP는 IoT의 저전력 센서 네트워크에 다양한 응용이 가능할 것으로 기대된다.

• 한국실내디자인학회논문집
• /
• 제25권3호
• /
• pp.31-39
• /
• 2016

This study is aimed to increase the possibility of using the Christopher Alexander's pattern language. The methodology of this study is (i) to analyze the pattern language by using the network analysis tool in order to understand the complicate network structure of the pattern language, and (ii) to apply the Alexander's method of using the pattern language by using the network analysis tool (Gephi) and to examine the feasibility of the network analysis tool as a tool for using the pattern language. Firstly, as a result of analysing the pattern language, (i) the pattern language classified by pattern number is distinguished by the patterns of towns, buildings and construction, among which the pattern of buildings plays a key function in the networks; (ii) the buildings functions a medium connecting between the towns and the construction; and (iii) the pattern language is divided into 6 sub-modules, through which the user can select a pattern. Secondly, the result of using the network analysis tool as a tool for using the pattern language (i) suggests the new method of using the pattern language by using the network analysis tool (Gephi); (ii) makes it possible to easily figure out the characteristics of the links between the patterns; and (iii) increases the completeness of the pattern language by making it easy to find out the sub-patterns in selecting a pattern.

• ETRI Journal
• /
• 제32권2호
• /
• pp.241-247
• /
• 2010

We propose and experimentally demonstrate a service-aware optical transport system. The proposed service-aware optical transport system makes a flow based on service type and priority of traffic. The generated flow is mapped to a corresponding sub-${\lambda}$ for transport over an optical network. Using sub-${\lambda}$ provided by the centralized control plane, we could effectively provide quality-of-service guaranteed Ethernet service and best-effort service simultaneously in a single link. The committed information rate (CIR) traffic and best-effort traffic are assigned to different sub-${\lambda}s$. The bandwidth of the CIR traffic is guaranteed without being affected by violation traffic because the bandwidth is managed per each sub-${\lambda}$. The failure detection time and restoration time from a link failure is measured to be about 60 ${\mu}s$ and 22 ms, respectively, in the ring network. The measured restoration time is much smaller than the 50 ms industry requirement for real-time services. The fast restoration time allows the proposed service-aware optical transport system to offer high availability and reliability which is a requirement for transport networks.

• 정보처리학회논문지A
• /
• 제17A권3호
• /
• pp.113-120
• /
• 2010

이븐 연결망은 고장허용 다중컴퓨터에 대한 하나의 모형으로 제안된 연결망으로, 간단한 라우팅 알고리즘, 최대고장허용도, 노드 중복 없는 경로와 같은 여러 가지 유용한 성질과 알고리즘들이 분석되었다. 기존에 발표된 라우팅 알고리즘과 노드 중복 없는 경로를 구성하는 알고리즘은 최적임이 증명되었다. 하지만 아직까지 이븐 연결망에서 에지 중복 없는 스패닝 트리를 구성하는 기법은 소개되지 않았다. 에지 중복 없는 스패닝 트리는 상호연결망의 고장허용도의 성능 향상과 효율적인 방송 기법을 분석하기 위해서 사용되는 매우 유용한 기법이다. 기존에 발표된 라우팅 알고리즘 또는 노드 중복 없는 경로를 구성하는 알고리즘은 라우팅 또는 노드 중복 없는 경로를 위한 알고리즘으로 에지 중복 없는 스패닝 트리를 구성하기 위해 적용될 수 없는 알고리즘이다. 본 논문에서는 이븐 연결망 $E_d$에서 에지 중복 없는 스패닝 트리를 구성하는 알고리즘을 제안한다.

• 농업과학연구
• /
• 제47권4호
• /
• pp.1109-1122
• /
• 2020

The aim of this study was to develop a marbling classification and prediction model using small parts of sirloin images based on a deep learning algorithm, namely, a convolutional neural network (CNN). Samples were purchased from a commercial slaughterhouse in Korea, images for each grade were acquired, and the total images (n = 500) were assigned according to their grade number: 1++, 1+, 1, and both 2 & 3. The image acquisition system consists of a DSLR camera with a polarization filter to remove diffusive reflectance and two light sources (55 W). To correct the distorted original images, a radial correction algorithm was implemented. Color images of sirloins of Hanwoo (mixed with feeder cattle, steer, and calf) were divided and sub-images with image sizes of 161 × 161 were made to train the marbling prediction model. In this study, the convolutional neural network (CNN) has four convolution layers and yields prediction results in accordance with marbling grades (1++, 1+, 1, and 2&3). Every single layer uses a rectified linear unit (ReLU) function as an activation function and max-pooling is used for extracting the edge between fat and muscle and reducing the variance of the data. Prediction accuracy was measured using an accuracy and kappa coefficient from a confusion matrix. We summed the prediction of sub-images and determined the total average prediction accuracy. Training accuracy was 100% and the test accuracy was 86%, indicating comparably good performance using the CNN. This study provides classification potential for predicting the marbling grade using color images and a convolutional neural network algorithm.

• Animal Bioscience
• /
• 제36권1호
• /
• pp.10-18
• /
• 2023

Objective: In this study, we aimed to position the Hungarian Merino among other Merinoderived sheep breeds, explore the characteristics of our sampled animals' genetic similarity network within the breed, and highlight single nucleotide polymorphisms (SNPs) associated with daily weight-gain. Methods: Hungarian Merino (n = 138) was genotyped on Ovine SNP50 Bead Chip (Illumina, San Diego, CA, USA) and positioned among 30 Merino and Merino-derived breeds (n = 555). Population characteristics were obtained via PLINK, SVS, Admixture, and Treemix software, within-breed network was analysed with python networkx 2.3 library. Daily weight gain of Hungarian Merino was standardised to 60 days and was collected from the database of the Association of Hungarian Sheep and Goat Breeders. For the identification of loci associated with daily weight gain, a multi-locus mixed-model was used. Results: Supporting the breed's written history, the closest breeds to Hungarian Merino were Estremadura and Rambouillet (pairwise F_ST values are 0.035 and 0.036, respectively). Among Hungarian Merino, a highly centralised connectedness has been revealed by network analysis of pairwise values of identity-by-state, where the animal in the central node had a betweenness centrality value equal to 0.936. Probing of daily weight gain against the SNP data of Hungarian Merinos revealed five associated loci. Two of them, OAR8_17854216.1 and s42441.1 on chromosome 8 and 9 (-log₁₀P>22, false discovery rate<5.5e-20) and one locus on chromosome 20, s28948.1 (-log₁₀P = 13.46, false discovery rate = 4.1e-11), were close to the markers reported in other breeds concerning daily weight gain, six-month weight, and post-weaning gain. Conclusion: The position of Hungarian Merino among other Merino breeds has been determined. We have described the similarity network of the individuals to be applied in breeding practices and highlighted several markers useful for elevating the daily weight gain of Hungarian Merino.

• Journal of the Korean Society for Industrial and Applied Mathematics
• /
• 제3권2호
• /
• pp.69-80
• /
• 1999

Recently in telecommunication, BISDN ( Broadband Integrated Service Digital Network ) has received considerable attention for its capability of providing a common interface for future communication needs including voice, data and video. Since all information in BISDN are statistically multiplexed and are transported in high speed by means of discrete units of 53-octet ATM ( asynchronous Transfer Mode ) cells, appropriate traffic control needs. For traffic shaping of voice, the output cell discarding scheme has been proposed. We analyze the scheme with a MMPP/$D_1$, $D_2$/1/B queueing system to obtain performance measures such as loss probability and waiting time distribution.

• 한국환경과학회지
• /
• 제26권9호
• /
• pp.1073-1080
• /
• 2017

Manufacturing and technology industries produce large amounts of air pollutants. Ulsan Metropolitan City, South Korea, is well-known for its large industrial complexes; in particular, the concentration of $SO_2$ here is the highest in the country. We assessed $SO_2$ monitoring sites based on conditional and joint entropy, because this is a common method for determining an optimal air monitoring network. Monthly $SO_2$ concentrations from 12 air monitoring sites were collected, and the distribution of spatial locations was determined by kriging. Mean absolute error, Root Mean Squared Error (RMSE), bias and correlation coefficients were employed to evaluate the considered algorithms. An optimal air monitoring network for Ulsan was suggested based on the improvement of RMSE.