• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.11a
• /
• pp.215-216
• /
• 2020

Optical Flow는 컴퓨터 비전 분야의 많은 응용기술에 사용된다. 객체 탐지, 추적, 연속 영상 보간, 3D Reconstruction과 같은 최근에 활발히 연구되는 여러 분야에서 사용되는 기반 기술이다. 최근 딥러닝을 기반으로 한 다양한 연구가 활발히 진행되어 왔으며 높은 정확도를 보이고 있다. 이런 분야들은 많은 경우에 실시간 시스템에 적용되어 이미지로부터 정보를 연산한다. 본 논문은 MaskFlownet, SelFlow, LiteFlowNet2 등과 같은 높은 정확도를 가진 신경망 네트워크로 추정된 Optical Flow를 살펴본다. 각 신경망 네트워크로 얻어진 정확도를 비교하고 디스플레이 기술과 이미지 센서 기술의 발전으로 사용 수요가 많아진 고화질의 이미지를 실시간으로 처리하는 경우, 적용 가능한 Optical Flow의 성능을 분석하였다.

• Electronics and Telecommunications Trends
• /
• v.29 no.6
• /
• pp.71-81
• /
• 2014

5G 무선통신시스템은 동일한 영역에서 스펙트럼 사용 효율성을 개선하기 위해 매크로셀과 소형셀이 공존하는 이종 네트워크(HetNet: Heterogeneous Network) 형태로 진화하고 있으며, 급증하는 모바일 트래픽을 효율적으로 처리하기 위해 소형셀들을 고밀도 네트워크(High dense network)로 구축하는 방안이 연구되고 있다. 매크로셀과 고밀도 소형셀들이 중첩되어 구축되는 HetNet 기반 셀룰러 네트워크에서 소형셀 시스템의 구성과 파라미터 최적화를 통한 성능 유지를 운영자가 수동으로 조정하는 것은 한계가 있으므로 네트워크 환경변화에 따라 시스템에서 자율적으로 파라미터를 조정하여 시스템 성능을 유지하는 기술이 요구되고 있다. 본고에서는 시스템 운용 중 자율적인 최적화를 통해 시스템 성능을 최적으로 유지하고 유지비용을 최소화하는 3GPP 자율적 네트워크 최적화 기술을 소개한다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.10 no.11
• /
• pp.5400-5418
• /
• 2016

Femtocells are envisioned as a key solution to embrace the ever-increasing high data rate and thus are extensively deployed. However, the dense and random deployments of femtocell access points (FAPs) induce severe intercell inference that in turn may degrade the performance of spectral efficiency. Hence, unrestrained proliferation of FAPs may not acquire a net throughput gain. Besides, given that numerous FAPs deployed in ultra-dense networks (UDNs) lead to significant energy consumption, the amount of FAPs deployed is worthy of more considerations. Nevertheless, little existing works present an analytical result regarding the optimal FAP density for a given User Equipment (UE) density. This paper explores the realistic scenario of randomly distributed FAPs in UDN and derives the coverage probability via Stochastic Geometry. From the analytical results, coverage probability is strictly increasing as the FAP-to-UE ratio increases, yet the growing rate of coverage probability decreases as the ratio grows. Therefore, we can consider a specific FAP-to-UE ratio as the point where further increasing the ratio is not cost-effective with regards to the requirements of communication systems. To reach the optimal FAP density, we can deploy FAPs in line with peak traffic and randomly switch off FAPs to keep the optimal ratio during off-peak hours. Furthermore, considering the unbalanced nature of traffic demands in the temporal and spatial domain, dynamically and carefully choosing the locations of active FAPs would provide advantages over randomization. Besides, with a huge FAP density in UDN, we have more potential choices for the locations of active FAPs and this adds to the demand for a strategic sleeping policy.

• Journal of radiological science and technology
• /
• v.43 no.5
• /
• pp.397-404
• /
• 2020

Coronavirus disease(COVID-19) is highly infectious disease that directly affects the lungs. To observe the clinical findings from these lungs, the Chest Radiography(CXR) can be used in a fast manner. However, the diagnostic performance via CXR needs to be improved, since the identifying these findings are highly time-consuming and prone to human error. Therefore, Artificial Intelligence(AI) based tool may be useful to aid the diagnosis of COVID-19 via CXR. In this study, we explored various Deep learning(DL) approach to classify COVID-19, other viral pneumonia and normal. For the original dataset and lung-segmented dataset, the pre-trained AlexNet, SqueezeNet, ResNet18, DenseNet201 were transfer-trained and validated for 3 class - COVID-19, viral pneumonia, normal. In the results, AlexNet showed the highest mean accuracy of 99.15±2.69% and fastest training time of 1.61±0.56 min among 4 pre-trained neural networks. In this study, we demonstrated the performance of 4 pre-trained neural networks in COVID-19 diagnosis with CXR images. Further, we plotted the class activation map(CAM) of each network and demonstrated that the lung-segmentation pre-processing improve the performance of COVID-19 classifier with CXR images by excluding background features.

• Journal of Astronomy and Space Sciences
• /
• v.36 no.1
• /
• pp.35-43
• /
• 2019

The Optical Wide-field patroL-Network (OWL-Net) is a global optical network for Space Situational Awareness in Korea. The primary operational goal of the OWL-Net is to track Low Earth Orbit (LEO) satellites operated by Korea and to monitor the Geostationary Earth Orbit (GEO) region near the Korean peninsula. To obtain dense measurements on LEO tracking, the chopper system was adopted in the OWL-Net's back-end system. Dozens of angle-only measurements can be obtained for a single shot with the observation mode for LEO tracking. In previous work, the reduction process of the LEO tracking data was presented, along with the mechanical specification of the back-end system of the OWL-Net. In this research, we describe an integrity assessment method of time-position matching and verification of results from real observations of LEO satellites. The change rate of the angle of each streak in the shot was checked to assess the results of the matching process. The time error due to the chopper rotation motion was corrected after re-matching of time and position. The corrected measurements were compared with the simulated observation data, which were taken from the Consolidated Prediction File from the International Laser Ranging Service. The comparison results are presented in the In-track and Cross-track frame.

• Applied Microscopy
• /
• v.30 no.3
• /
• pp.265-272
• /
• 2000

In this study, the brachial ganglion of Octopus minor was investigated with light microscope and electron microscope,andthefollowingresultswereobtained. The brachial ganglions of the octopus, round in shapes , are located under each of suckers. Their sizes are proportional to those of the suckers. A brachial ganglion of round shape consists of cortex and medulla. In cortex, nerve cells exist collectively while neuropiles in medulla. Three kinds of nerve cells (large, middle, and small neurons) are found in the cluster of nerve cells. The small one is a round cell of about $0.9{\mu}m$ in diameter while the middle and large ones are an elliptical cell of $1.6\times1.3{\mu}m$ and an ovoid cell of $2.8{\mu}m$ in diameter, respectively. All of those cells look light due to their low electron densities , in which cell organelle are not well developed. It was also observed that the middle neurons are surrounded by median electron-dense neuroglial cells of pyramidal shapes and about $0.6\times0.4{\mu}m$ in sizes. In the neuropiles of medulla, dendrites and axons of various sizes make a complex net. They contain four kinds of chemical synaptic vesicles-electron-dense synaptic vesicle of 100 nm in diameter, median electron-dense synaptic vesicle of 90 nm in diameter, electron-dense cored synaptic vesicle of 90 nm in diameter, and electron-lucent synaptic vesicle of 50 nm in diameter.

• Journal of Multimedia Information System
• /
• v.8 no.3
• /
• pp.175-182
• /
• 2021

Oral cancer is ranked second most diagnosed cancer among Indian population and ranked sixth all around the world. Oral cancer is one of the deadliest cancers with high mortality rate and very less 5-year survival rates even after treatment. It becomes necessary to detect oral malignancies as early as possible so that timely treatment may be given to patient and increase the survival chances. In recent years deep learning based frameworks have been proposed by many researchers that can detect malignancies from medical images. In this paper we have proposed a deep learning-based framework which detects oral cancer from histopathology images very efficiently. We have designed our model to split the color channels and extract deep features from these individual channels rather than single combined channel with the help of Efficient NET B3. These features from different channels are fused by using feature fusion module designed as a layer and placed before dense layers of Efficient NET. The experiments were performed on our own dataset collected from hospitals. We also performed experiments of BreakHis, and ICML datasets to evaluate our model. The results produced by our model are very good as compared to previously reported results.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.21 no.3
• /
• pp.159-168
• /
• 2021

As the information age enters, the use of biometrics using the body is gradually increasing because it is very important to accurately recognize and authenticate each individual's identity for information protection. Among them, finger vein authentication technology is receiving a lot of attention because it is difficult to forge and demodulate, so it has high security, high precision, and easy user acceptance. However, the accuracy may be degraded depending on the algorithm for identification or the surrounding light environment. In this paper, we designed and manufactured a side-type finger vein recognizer that is highly versatile among finger vein measuring devices, and authenticated using the deep learning model of DenseNet-201 for high accuracy and recognition rate. The performance of finger vein authentication technology according to the influence of the infrared light source used and the surrounding visible light was analyzed through simulation. The simulations used data from MMCBNU_6000 of Jeonbuk National University and finger vein images taken directly were used, and the performance were compared and analyzed using the EER.

• Korean Journal of Remote Sensing
• /
• v.38 no.6_1
• /
• pp.1007-1014
• /
• 2022

Compact Advanced Satellite 500 (CAS500) can be used for various purposes, including vegetation, forestry, and agriculture fields. It is expected that it will be possible to acquire satellite images of various areas quickly. In order to use satellite images acquired through CAS500 in the agricultural field, it is necessary to develop a satellite image-based extraction technique for crop-cultivated areas.In particular, as research in the field of deep learning has become active in recent years, research on developing a deep learning model for extracting crop cultivation areas and generating training data is necessary. This manuscript classified the onion and garlic cultivation areas in Hapcheon-gun using PlanetScope satellite images and farm maps. In particular, for effective model learning, the model performance was analyzed according to the proportion of crop-cultivated areas. For the deep learning model used in the experiment, Fully Convolutional Densely Connected Convolutional Network (FC-DenseNet) was reconstructed to fit the purpose of crop cultivation area classification and utilized. As a result of the experiment, the ratio of crop cultivation areas in the training data affected the performance of the deep learning model.

• The Bulletin of The Korean Astronomical Society
• /
• v.42 no.1
• /
• pp.38.1-38.1
• /
• 2017

We report the results of three-dimensional radiation hydrodynamic simulations of star cluster formation in turbulent molecular clouds, with primary attention to how stellar radiation feedback controls the lifetime and net star formation efficiency (SFE) of their natal clouds. We examine the combined effects of photoionization and radiation pressure for a wide range of cloud masses (10^4 - 10^6 Msun) and radii (2 - 80 pc). In all simulations, stars form in densest regions of filaments until feedback becomes strong enough to clear the remaining gas out of the system. We find that the SFE is primarily a function of the initial cloud surface density, Sigma, (SFE increasing from ~7% to ~50% as Sigma increases from ~30 Msun/pc^2 to ~10^3 Msun/pc^2), with weak dependence on the initial cloud mass. Control runs with the same initial conditions but without either radiation pressure or photoionization show that photoionization is the dominant feedback mechanism for clouds typical in normal disk galaxies, while they are equally important for more dense, compact clouds. For low-Sigma clouds, more than 80% of the initial cloud mass is lost by photoevaporation flows off the surface of dense clumps. The cloud becomes unbound within ~0.5-2.5 initial free-fall times after the first star-formation event, implying that cloud dispersal is rapid once massive star formation takes place. We briefly discuss implications and limitations of our work in relation to observations.