• 디지털산업정보학회논문지
• /
• 제16권1호
• /
• pp.55-65
• /
• 2020

With the development of display hardware, image interpolation techniques have been used in various fields such as image zooming and medical imaging. Traditional image interpolation methods, such as bi-linear interpolation, bi-cubic interpolation and edge direction-based interpolation, perform interpolation in the spatial domain. Recently, interpolation techniques in the discrete cosine transform or wavelet domain are also proposed. Using these various existing interpolation methods and machine learning, we propose decision tree classification-based image interpolation methods. In other words, this paper is about the method of adaptively applying various existing interpolation methods, not the interpolation method itself. To obtain the decision model, we used Weka's J48 library with the C4.5 decision tree algorithm. The proposed method first constructs attribute set and select classes that means interpolation methods for classification model. And after training, interpolation is performed using different interpolation methods according to attributes characteristics. Simulation results show that the proposed method yields reasonable performance.

• 대한음성학회지:말소리
• /
• 제56호
• /
• pp.207-223
• /
• 2005

The goal of language model adaptation is to improve the background language model with a relatively small adaptation corpus. This study presents a language model adaptation technique where additional text data for the adaptation do not exist. We propose the information retrieval (IR) technique with N-gram language modeling to collect the adaptation corpus from baseline text data. We also propose to use a dynamic language model interpolation coefficient to combine the background language model and the adapted language model. The interpolation coefficient is estimated from the word hypotheses obtained by segmenting the input speech data reserved for held-out validation data. This allows the final adapted model to improve the performance of the background model consistently The proposed approach reduces the word error rate by $13.6\%$ relative to baseline 4-gram for two-hour broadcast news speech recognition.

• Journal of information and communication convergence engineering
• /
• 제21권1호
• /
• pp.17-23
• /
• 2023

Recently, several high-performance instance segmentation models have used the Mask R-CNN model as a baseline, which reached a historical peak in instance segmentation in 2017. There are numerous derived models using the Mask R-CNN model, and if the performance of Mask R-CNN is improved, the performance of the derived models is also anticipated to improve. The Mask R-CNN uses interpolation to adjust the image size, and the input differs depending on the interpolation method. Therefore, in this study, the performance change of Mask R-CNN was compared when various interpolation methods were applied to the transform layer to improve the performance of Mask R-CNN. To train and evaluate the models, this study utilized the PennFudan and Balloon datasets and the AP metric was used to evaluate model performance. As a result of the experiment, the derived Mask R-CNN model showed the best performance when bicubic interpolation was used in the transform layer.

• 한국측량학회지
• /
• 제23권3호
• /
• pp.251-260
• /
• 2005

CCD의 개발은 사진측량에 많은 발전을 이루었다. 본 연구는 근접수치사진측량으로 수리모형 해석시 지형 보간최적분석기법에 관한 연구이다. 먼저 근접수치사진측량과 삼각수준측량으로 수리모형의 상사성을 분석하기 위해 통계분석의 일종인 T-검정을 실시하였다 그리고 수치사진측량으로 분석한 수리모형해석을 컴퓨터로 실제지형과 비슷한 형상을 가지는 수치표고모형을 만들기 위해 역거리, 크리깅, 최근린, 불규칙삼각망 보간법 중에서 수리모형 적용에 적합한 최적 보간법을 제시하였다. 수리모형에 대한 기하학적 상사분석을 위해 보간법을 적용 결과, 크리깅 보간법과 불규칙삼각망 보간법이 효율적인 것으로 판단되었다

• 한국항해항만학회지
• /
• 제36권9호
• /
• pp.699-706
• /
• 2012

This paper describes Declination interpolation method in the application of Digital Magnetic Compass(DMC) with small capacity memory for relatively low-priced system. The purpose of this study is to find accurate declination values using Declination Lookup Table(DLT) and Secular Variation Table(SVT) with the Bilinear interpolation method having four points data only. World Magnetic Model with Gauss coefficients produced by NOAA is used in the calculation of DLT and SVT at desired years. To verify the applicability of the proposed method, the simulation tests and the error analysis are carried out in this work. As results from tests, the interpolation error is within 0.01 degree that is much enough to implement high accurate DMC comparing with commercial DMCs on the market. In summary, the declination interpolation method, proposed in this study, can be useful in the application of DMC.

• 지구물리와물리탐사
• /
• 제20권2호
• /
• pp.78-87
• /
• 2017

MWNI (Minimum Weighted Norm Interpolation)를 이용한 내삽 방법은 고차원으로 확장이 용이하고 상대적으로 계산 속도가 빠르다는 장점을 가지고 있으나 알리아스 효과가 존재하는 자료의 내삽에 취약하다. 이런 문제의 개선을 위해 제안된 방법이 모델제약(model-constrained) MWNI이다. 이 논문에서는 MWNI를 이용한 방법과 모델제약 MWNI 방법의 두가지 모듈을 개발한 후 알리아스 효과가 존재하는 자료의 내삽 결과를 비교하였다. 시공간 영역(t-x domain)과 주파수-파수 영역(f-k domain)의 결과 그림을 통해서 모델제약 MWNI를 적용했을 때의 결과가 더 효과적임을 확인할 수 있었다. 동해 울릉분지의 가스 하이드레이트 부존 지역의 현장 자료에 내삽을 적용한 결과, 가스침니 구간 전후로 진폭이 급격하게 변하는 자료에서도 내삽이 가능함을 확인할 수 있었다. 또한 매우 불규칙하고 넓은 구간에서 누락된 인공지진파 자료의 정규화를 통해 신호의 연결성 향상이 가능함을 보일 수 있었다. 결과적으로 이 논문에서 개발된 모듈은 현장의 다양한 여건에 의해 불규칙하거나 넓은 간격으로 얻어진 탄성파 자료의 정규화나 내삽에 효율적으로 활용될 수 있음을 확인하였다.

• 환경영향평가
• /
• 제21권1호
• /
• pp.219-227
• /
• 2012

Daily solar radiation is essential for water resources planning and environmental impact assessment. However, radiation data is not commonly available in Korea other than in big cities, and there has been no direct measurement for rural areas where water resources planning and environmental impact assessment is usually most needed. In general, missing radiation data is estimated from nearby regional stations within a certain distance, and this study compared two dominant methods (modified Angstrom equation and transmittance interpolation method) at six stations in Nakdong River watershed area. Two methods shows a similar level of accuracy but the transmittance interpolation method is likely to be superior in that there is no need for any measurement element since the modified Angstrom equation require the sunshine hour measurement. This study will contribute to improve water resource and water quality management in Nakdong River watershed.

• 대한공간정보학회지
• /
• 제14권4호통권38호
• /
• pp.27-36
• /
• 2006

현재 위성영상에서 정확한 위치정보를 얻고자 할 때 일반적으로 위성영상에 대응되는 지상에서의 위치 정보, 즉 지상기준점이 필요하다. 이 논문에서는 지상기준점 없이 영상의 위치 정보를 얻기 위하여 동일궤도에서 연속적으로 촬영한 위성영상들 중에서 한 위성영상의 획득한 기준점으로 모델을 수립한 뒤, 수립된 센서모델을 동일궤도상의 다른 영상에 적용할 때의 센서모델의 정확도를 분석하고자 한다. 분석에 사용한 센서모델은 궤도기반센서모델을 사용하며, 여러 위성의 위치 자세 내삽법 및 미지수조합을 시험하여 궤도모델링에 적합한 내삽법(interpolation)과 최적의 미지수 조합을 알아보고자 했다. 실험은 총 420Km의 길이에 해당하는 SPOT-3의 영상 7장과 GPS수신기에서 취득한 기준점을 사용하였다. 실험결과 단일영상에서는 내삽법과 미지수 조합에 따른 결과의 차이는 크게 나타나지 않았으나 궤도모델링 시에는 미지수 조합과 자세와 위치의 내삽법에 따라 다양한 결과가 나타남을 알 수 있다. 또한 적절한 미지수조합과 내삽법을 사용하면 한 영상에서 추출한 기준점만으로 총 420Km의 궤도를 정확하게 모델 할 수 있음을 알 수 있었다.

• Korea-Australia Rheology Journal
• /
• 제15권3호
• /
• pp.131-150
• /
• 2003

A new numerical algorithm of finite element methods is presented to solve high Deborah number flow problems with geometric singularities. The steady inertialess planar 4 : 1 contraction flow is chosen for its test. As a viscoelastic constitutive equation, we have applied the globally stable (dissipative and Hadamard stable) Leonov model that can also properly accommodate important nonlinear viscoelastic phenomena. The streamline upwinding method with discrete elastic-viscous stress splitting is incorporated. New interpolation functions classified as rational interpolation, an alternative formalism to enhance numerical convergence at high Deborah number, are implemented not for the whole set of finite elements but for a few elements attached to the entrance comer, where stress singularity seems to exist. The rational interpolation scheme contains one arbitrary parameter b that controls the singular behavior of the rational functions, and its value is specified to yield the best stabilization effect. The new interpolation method raises the limit of Deborah number by 2∼5 times. Therefore on average, we can obtain convergent solution up to the Deborah number of 200 for which the comer vortex size reaches 1.6 times of the half width of the upstream reservoir. Examining spatial violation of the positive definiteness of the elastic strain tensor, we conjecture that the stabilization effect results from the peculiar behavior of rational functions identified as steep gradient on one domain boundary and linear slope on the other. Whereas the rational interpolation of both elastic strain and velocity distorts solutions significantly, it is shown that the variation of solutions incurred by rational interpolation only of the elastic strain is almost negligible. It is also verified that the rational interpolation deteriorates speed of convergence with respect to mesh refinement.

• 치과방사선
• /
• 제29권1호
• /
• pp.65-85
• /
• 1999

Objectives: To determine the effect of interpolation functions when processing the digital periapical images. Material and Methods: The digital images were obtained by Digora and CDR system on the dry skull and human subject. 3 oral radiologists evaluated the 3 portions of each processed image using 7 interpolation methods, and ROC curves were obtained by trapezoidal methods. Results: The heighest Az value(0.96) was obtained with cubic spline method and the lowest Az value(0.03) was obtained with facet model method in Digora system. The heighest Az value (0.79) was obtained with gray segment expansion method and the lowest Az value(0.07) was obtained with facet model method in CDR system. There was significant difference of Az value in original image between Digora and CDR system at a=0.05 level. There were significant differences of Az values between Digora and CDR images with cubic spline method, facet model method, linear interpolation method and non-linear interpolation method at α=0.1 level.