• Proceedings of the Korea Contents Association Conference
• /
• 2015.05a
• /
• pp.451-452
• /
• 2015

다양한 인공 지능이 탑재된 가전 제품의 스마트화 속도가 빨라지면서, 음성과 영상 등 멀티미디어의 기능을 이용한 '스마트 냉장고'가 출시되고 있는 추세이다. 스마트 냉장고의 출현으로 인해 기존의 냉장고를 사용할 때 냉장고 안에 어떤 내용물이 얼마나 남아 있는지 자세히 파악할 수 없던 점, 유통기한이 지나 버려지는 식재료가 발생하는 점 등 기존의 냉장고가 가진 문제들이 해결되는 듯 보였지만, 600만 원에 달하는 고가의 가격 때문에 보편화되지 못하여 아직까지도 상당수의 문제가 난제로 남아 있다. 이에 대한 해결 방안으로 우리는 현재 출시된 스마트 냉장고를 재설계해 IoT 냉장고 용기를 개발하고, 스마트폰의 애플리케이션과 통신하여 용기 안의 식재료 정보를 주고받고, 이에 따른 레시피를 추천해 주는 시스템을 설계하고자 한다. 이 IoT 냉장고를 활용한 레시피 추천 시스템은 스마트 냉장고뿐만 아니라 일반 냉장고에도 적용시킬 수 있을 것이라고 기대된다.

• Journal of Hydrogen and New Energy
• /
• v.15 no.3
• /
• pp.250-256
• /
• 2004

Magnesium and its alloys have the high potential as hydrogen storage materials because of their highest hydrogen storage capacity, low density and abundant resources. But poor kinetic properties of hydriding and dehydriding and high working temperature have limited their practical applications. In this study, the Mg-Ni binary alloys with different amount of Ni were produced by gravity casting and characterized in order to investigate the relationship between the microstructures and hydriding properties. The maximum hydrogen absorption capacity decreased, but the absorption kinetics increased with Ni content. The difference in the absorption kinetics was resulted from the differences in the sort and shape of primary solid phases and eutectic microstructure.

• Journal of KIISE:Computing Practices and Letters
• /
• v.15 no.6
• /
• pp.459-463
• /
• 2009

Background modeling and motion detection is the one of the most significant real time video processing technique. Until now, many researches are conducted into the topic but it still needs much time for robustness. It is more important when other algorithms are used together such as object tracking, classification or behavior understanding. In this paper, we propose efficient multi-modal background modeling methods which can be understood as simplified learning method of Gaussian mixture model. We present its validity using numerical methods and experimentally show detecting performance.

• Journal of KIISE:Software and Applications
• /
• v.28 no.4
• /
• pp.369-379
• /
• 2001

본 논문은 부영역 분할과 코드워드 인텍스의 캐시 개념을 이용하여 벡터 양자화를 위한 고속코드북 생성 및 부호화 방법을 제안한다. 제안한 방법은 인접한 입력 벡터는 대개 코드북내 특정 코드워드에 의해 나타내어지는 국부성에 바탕을 두고 있다. 초기에 모든 학습 벡터가 거리에 기반한 근접성을 이용하여 정해진 수의 부영역으로 분할된다. 각 부영역에 하나의 코드워드 인덱스 캐시가 할당되는데 이 캐시는 학습 초기에는 전체 코드북 크기에 대응하는 코드워드 인덱스를 갖는다. 학습이 진행되면서 입력 벡터가 갖는 국부성 때문에 각 부영역내 캐시중 사용되지 않는 코드워드 인덱스가 점차 발생하게 되므로 이들은 LRU(Least Recently Used) 삭제 알고리즘에 의해 제거된다. 학습이 진행됨에 따라 부영역 캐시에는 주어진 입력 벡터에 의해 참조되는 코드워드 인덱스만이 남게 되므로 한 학습 주기 동한 필요한 학습 시간이 점차 짧아지게 되어 전체적으로 코드북 생성 시간을 크게 줄일 수 있게 된다. 제안한 방법은 매 학습주기마다, 코드워드 인덱스 삭제 후보 중 주어진 부영역 중심으로부터 거리에 의해 멀리 떨어진 것부터 반만을 제거함에 따라. 복원된 영상의 화질 열화가 거의 없다. 시뮬레이션 결과 제안한 방법은 기존의 LBG 방법에 비해 화질 열화는 거의 없지만 코드북 생성 (또는 부호화) 속도를 2.6-5.4배 (또는3.7-18.8배) 향상시킨다.

• Journal of IKEEE
• /
• v.21 no.3
• /
• pp.280-283
• /
• 2017

HEVC (high efficiency video coding) exploits CABAC (context-based adaptive binary arithmetic coding) for entropy coding, where a context model estimates the probability for each syntax element. In this paper, a context modeler was designed and implemented for CABAC decoding. lookup table was used to reduce computation and to increase speed. 12 simulations for HEVC standard test sequences and encoder configurations were performed, and the context modeler was verified to perform correction operations. The designed context modeler was synthesized in 0.18um technology. Maximum frequency, maximum throughput, and gate count are 200 MHz, 200 Mbin/s, and 29,268 gates, respectively.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2014.11a
• /
• pp.142-145
• /
• 2014

동영상에서의 객체 추적 알고리즘에 대한 활발한 연구가 진행되고 있음에도 불구하고 실시간 객체추적을 위해서는 여전히 정확도, 복잡도 등에서의 성능향상이 필요하다. 압축영역 기반 방식에서는 전역 움직임 보상(GMC : Global Motion Compensation)과정을 거쳐 추적하려는 객체와 배경을 구분한다. 전역 움직임 보상방법은 프레임 전 영역을 대상으로 하는 연산으로 전체 추적 시스템에서 차지하는 복잡도가 높다. 본 논문은 관심영역(ROI : Region Of Interest) 기반 전역 움직임 보상방법을 이용한 ST-MRF(Spatio-Temporal Markov Random Field)기반 추적기 고속화 방법을 제안한다. 관심영역을 기반으로 전역 움직임 보상을 적용함으로써 객체와 배경을 분리할 뿐만 아니라 알고리즘의 복잡도를 효과적으로 줄일 수 있다. 제안하는 방법의 추적성능은 평균 precision 87.29%, recall 82.58%, F-measure 83.78%로 기존방법과 비교하여 약 1%의 차이를 유지하였으며 전체 시스템의 수행시간은 평균 29.95ms로 기존방법과 비교하여 1.74배의 속도향상을 보였다.

• Journal of Korean Tunnelling and Underground Space Association
• /
• v.24 no.3
• /
• pp.247-262
• /
• 2022

In domestic tunnels, it is mandatory to install CCTVs in tunnels longer than 200 m which are also recommended by installation of a CCTV-based automatic accident detection system. In general, the CCTVs in the tunnel are installed at a low height as well as near by the moving vehicles due to the spatial limitation of tunnel structure, so a severe perspective effect takes place in the distance of installed CCTV and moving vehicles. Because of this effect, conventional CCTV-based accident detection systems in tunnel are known in general to be very hard to achieve the performance in detection of unexpected accidents such as stop or reversely moving vehicles, person on the road and fires, especially far from 100 m. Therefore, in this study, the region of interest is set up and a new concept of inverse perspective transformation technique is introduced. Since moving vehicles in the transformed image is enlarged proportionally to the distance from CCTV, it is possible to achieve consistency in object detection and identification of actual speed of moving vehicles in distance. To show this aspect, two datasets in the same conditions are composed with the original and the transformed images of CCTV in tunnel, respectively. A comparison of variation of appearance speed and size of moving vehicles in distance are made. Then, the performances of the object detection in distance are compared with respect to the both trained deep-learning models. As a result, the model case with the transformed images are able to achieve consistent performance in object and accident detections in distance even by 200 m.

• Geophysics and Geophysical Exploration
• /
• v.16 no.4
• /
• pp.211-216
• /
• 2013

In general, smoothing filters regularize functions by reducing differences between adjacent values. The smoothing filters, therefore, can regularize inverse solutions and produce more accurate subsurface structure when we apply it to full waveform inversion. If we apply a smoothing filter with a constant coefficient to subsurface image or velocity model, it will make layer interfaces and fault structures vague because it does not consider any information of geologic structures and variations of velocity. In this study, we develop a selective smoothing regularization technique, which adapts smoothing coefficients according to inversion iteration, to solve the weakness of smoothing regularization with a constant coefficient. First, we determine appropriate frequencies and analyze the corresponding wavenumber coverage. Then, we define effective maximum wavenumber as 99 percentile of wavenumber spectrum in order to choose smoothing coefficients which can effectively limit the wavenumber coverage. By adapting the chosen smoothing coefficients according to the iteration, we can implement multi-scale full waveform inversion while inverting multi-frequency components simultaneously. Through the successful inversion example on a salt model with high-contrast velocity structures, we can note that our method effectively regularizes the inverse solution. We also verify that our scheme is applicable to field data through the numerical example to the synthetic data containing random noise.

• Investigative Magnetic Resonance Imaging
• /
• v.6 no.1
• /
• pp.41-46
• /
• 2002

The Fourier transform arteriography (FTA) exploits the periodic variation of arterial flow velociety of arterial flow velocity in stnchronized with cardiac cycles. This technique is intrinsically unique compared to other modern techniques. This technique separates the arteries from the veins using the pulsatile arterial flow without using the presaturation RF pulses. Therefore, it has less RF deposition and is free from the dark band artifacts that can arise from retrograde flow and curved arteries. Furthermore, it is free from the artifacts induced by eddy currents. However, there are some drawbacks such as a single projection view and the saturation of arteries at the end of an imaging slab. These drawbacks are circumvented by applying recently developed techniques. The fast gradient switching capability of modern MRI systems enabled us to incorporate dual projection views into the conventional FTA sequence without increasing the repetition time. In addition, signals from the distal arteries were enhanced by use of a ramped RF pulse and therefore the distal arteries were less saturated. By use of the FTA sequence with dual projection views and the ramped RF pulse, we acquired the sagittal and coronal projection views of femoral arteriograms simultaneously with more enhanced signals of distal arteries than the conventional FTA.

• Journal of the Korean Association of Geographic Information Studies
• /
• v.10 no.1
• /
• pp.35-46
• /
• 2007

In the study digital imagery was used to examine asphalt concrete pavements. With digitally mastered-image information that was filmed with a video camera fixed on a car travelling on road at a consistent speed, a road surface management system that can gain road surface information (Crack, Rutting, IRI) was developed using an object-oriented language "Delphi". This system was designed to improve visualized effects by animations and graphs. After analyzing the accuracy of 3-D coordinates of road surfaces that were decided using multiple image orientation and bundle adjustment method, the average of standard errors turned out to be 0.0427m in the X direction, 0.0527m in the Y direction and 0.1539m in the Z direction. As a result, it was found to be good enough to be put to practical use for maps drawn on scales below 1/1000, which are currently producted and used in our country, and GIS data. According to the analysis of the accuracy in crack width on 12 spots using a digital video camera, the standard error was found to be ${\pm}0.256mm$, which is considered as high precision. In order to get information on rutting, the physically measured cross sections of 4 spots were compared with cross sections generated from digital images. Even though a maximum error turned out to be 10.88mm, its practicality is found in work efficiency.