• 방송공학회논문지
• /
• 제27권4호
• /
• pp.477-486
• /
• 2022

본 논문에서는 컨볼루션 신경망 네트워크를 이용하여 VVC 화면 내 예측으로 얻은 예측 블록을 개선하여 잔차 신호를 보다 줄이는 화면 내 예측 방법을 제안한다. 기존의 화면 내 예측 방법은 일부 고정 규칙을 기반으로 주변의 재구성된 참조 샘플로부터 예측 블록을 생성하므로 복잡한 콘텐츠의 예측 블록을 생성하기 어렵다는 한계가 있다. 또한, 참조 샘플로 이용할 수 있는 정보의 양이 시간적 주변 정보에 비해 적기 때문에 화면 간 예측보다 낮은 부호화 성능을 가진다. 본 연구에서는 앞서 언급한 문제를 해결하기 위해 기존의 비디오 부호화 과정의 화면 내 예측을 통해 생성되는 예측 블록에 CNN을 적용하여 원본 블록과 예측 블록의 차분 신호를 줄이는 화면 내 예측 방법을 제안한다. 부호기에서는 제안 알고리즘의 활성 여부를 나타내는 플래그가 함께 부호화된다. 제안하는 화면 내 예측 방법은 최신 비디오 압축 표준인 Versatile Video Coding의 참조 모델인 VTM version 10.0 대비 휘도 성분에 대하여 향상된 압축 성능을 제공한다.

• 한국철도학회:학술대회논문집
• /
• 한국철도학회 2011년도 정기총회 및 추계학술대회 논문집
• /
• pp.3201-3207
• /
• 2011

This paper reviewed wave propagation of train vibration based on the study of high speed railway soft ground section with pile slab construction. In a filed of railway, concrete track has been adapted in a railway construction. And in order to maintain its track, soil improving method was required to control residual settlement. Within many soft ground settlement prevention techniques, pile slab method has an effect of minimizing residual settlement of soft ground. This is possible using support embankment load method by construct pile slab or cap the upper soft ground. This paper reviewed vibration wave characteristic of soft ground section with pile slab using numerical analysis application through finite element analysis. Pile slab method is established between high stiffened soft ground and embankment this creates a possibility of vibration block or slab amplification. Thus analyzed of wave propagation was done with roadbed and structure property to confirm application performance of pile slab method of high speed railway structure.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 1998년도 추계학술대회 논문집 학회본부 C
• /
• pp.907-909
• /
• 1998

This paper describes the effect of impulse current on degradation of ZnO blocks. In this study, an impulse current generator which can produce 8/20 [${\mu}s$], 3 [kA] and 4/10 [${\mu}s$], 5 [kA] waveform is designed and fabricated to simulate the lightning impulse current. The residual voltage, reference voltage, and leakage current flowing to the ZnO blocks are observed. The experimental results show that the leakage current increases continuously with the number of applied impulse current, but no significant changes in residual voltage and in operating voltage are observed until the ZnO block is destroyed. Also, it is confirmed that the main factor on degradation of ZnO blocks is rather the total energy applied to ZnO blocks than the peak value of the impulse current.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 2004년도 하계학술대회 논문집 D
• /
• pp.2227-2229
• /
• 2004

Fault Detection and Isolation(FDI) schemes using unknown input functional observers with very low order are presented. These schemes resolve the major practical difficulties with all FDI systems employing multiple observers for residual generation and can be implemented by the use of microprocessors that are normally used in commercial processes mainly due to the simplicity of the residual generation block. Various design objectives including detection, isolation, estimation and compensation of instrument fault/or process fault are achievable with these schemes. The proposed FDI scheme is applied to an inverted pendulum control system for instrument fault detection.

• 한국진공학회:학술대회논문집
• /
• 한국진공학회 2013년도 제45회 하계 정기학술대회 초록집
• /
• pp.104.2-104.2
• /
• 2013

A residual gas analyzer (RGA) system has been developed in this laboratory. Characteristics of the RGA system parts such as ion source, quadrupole mass filter and sensitivity are introduced. Some efforts have been made to improve performance of the two types of ion sources, open ion source (OIS) and closed ion source (CIS). A metal mesh was placed onto the electron beam entrance of the CIS anode tube to block the filament field penetration. Sensitivity of the CIS ion sources with and without the mesh was compared by mass spectra of SF6 gas (97% He base) introduced into the CIS anode through a needle valve. About ten-times improvement in the RGA sensitivity was observed for the CIS with the mesh in the electron entrance. Computer simulation showed an axi-symmetric anode potential distribution and improved focusing of the electron beam inside the anode tube with the mesh.

• ETRI Journal
• /
• 제43권3호
• /
• pp.538-548
• /
• 2021

This paper proposes an approach to improve the performance of no-reference video quality assessment for sports videos with dynamic motion scenes using an efficient spatiotemporal model. In the proposed method, we divide the video sequences into video blocks and apply a 3D shearlet transform that can efficiently extract primary spatiotemporal features to capture dynamic natural motion scene statistics from the incoming video blocks. The concatenation of a deep residual bidirectional gated recurrent neural network and logistic regression is used to learn the spatiotemporal correlation more robustly and predict the perceptual quality score. In addition, conditional video block-wise constraints are incorporated into the objective function to improve quality estimation performance for the entire video. The experimental results show that the proposed method extracts spatiotemporal motion information more effectively and predicts the video quality with higher accuracy than the conventional no-reference video quality assessment methods.

• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2022년도 추계학술발표대회
• /
• pp.300-302
• /
• 2022

Palm vein identification has attracted attention due to its distinct characteristics and excellent recognition accuracy. However, many contactless palm vein identification systems suffer from the issue of having low-quality palm images, resulting in degradation of recognition accuracy. This paper proposes the use of U-Net architecture to correctly segment the vascular blood vessel from palm images. Attention gate mechanism and residual block are also utilized to effectively learn the crucial features of a specific segmentation task. The experiments were conducted on CASIA dataset. Hessian-based Jerman filtering method is applied to label the palm vein patterns from the original images, then the network is trained to segment the palm vein features from the background noise. The proposed method has obtained 96.24 IoU coefficient and 98.09 dice coefficient.

• 한국자동차공학회논문집
• /
• 제12권4호
• /
• pp.173-180
• /
• 2004

The tread pattern of a tire has an important effect on tire performances such as handling, wear, noise, hydroplaning and so on. However, a finite element analysis of a patterned tire with detailed tread blocks has been limited owing to the complexity of making meshes for tread blocks and the huge computation time. The computation time has been shortened due to the advance in the computer technology. The modeling of tread blocks usually requires creating a solid model using a CAD software. Therefore it is a very complicated and time-consuming job to generate meshes of a patterned tire using a CAD model. A new efficient and convenient method for generating meshes of a patterned tire has been developed. In this method, 3-D meshes of tread pattern are created by mapping 2-D meshes of tread geometry onto 3-D tread surfaces and extruding them through tread depth. Then, the tread pattern meshes are assembled with the tire body meshes by the tie contact constraint. Residual aligning torque and frictional energy are calculated by using a patterned tire model and compared to the experimental results. It is shown that the calculated results of a patterned tire model are in a good agreement with the experimental ones.

• 대한조선학회논문집
• /
• 제44권5호
• /
• pp.496-503
• /
• 2007

Although weld-induced deformation is inevitable in shipbuilding, it is important to reduce it as low as possible during fabrication for a more efficient production of ships' blocks. The weld-induced deformation is more serious in thin plates than in thick plates because heat affect zone of thin plates is wider than that of thick plates, and in addition internal and external constraints much more influence upon weld-induced deformation of thin plates. This paper deals with the application of the mechanical tensioning method to butt weld of thin plates to reduce the transverse and longitudinal deformation. in order to investigate the quantitative effect of tensioning method upon the reduction of angular deformation and shrinkage in longitudinal and transverse direction of weld line, butt welding test have been carried out for several thin plate specimens with varying plate thickness and magnitude of tensile load. Numerical simulation has been also carried out to compare the weld-induced deformation and residual stress. From the present study, it has been found that the tensioning method is very effective on reduction of weld-induced residual stress as well as weld-induced deformation.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제16권3호
• /
• pp.757-770
• /
• 2022

Palmprint recognition has drawn increasingly attentions in the past decade due to its uniqueness and reliability. Traditional palmprint recognition methods usually use high-resolution images as the identification basis so that they can achieve relatively high precision. However, high-resolution images mean more computation cost in the recognition process, which usually cannot be guaranteed in mobile computing. Therefore, this paper proposes an improved low-resolution palmprint image recognition method based on residual networks. The main contributions include: 1) We introduce a channel attention mechanism to refactor the extracted feature maps, which can pay more attention to the informative feature maps and suppress the useless ones. 2) The ResStage group structure proposed by us divides the original residual block into three stages, and we stabilize the signal characteristics before each stage by means of BN normalization operation to enhance the feature channel. Comparison experiments are conducted on a public dataset provided by the Hong Kong Polytechnic University. Experimental results show that the proposed method achieve a rank-1 accuracy of 98.17% when tested on low-resolution images with the size of 12dpi, which outperforms all the compared methods obviously.