• Review of KIISC
• /
• v.30 no.6
• /
• pp.107-115
• /
• 2020

다양한 응용분야에서 심층신경망 기반의 학습 모델이 앞 다투어 이용됨에 따라 인공지능의 설명 가능한 동작 원리 해석과, 추론이 갖는 불확실성에 관한 분석 또한 심도 있게 연구되고 있다. 이에 심층신경망 기반 기계학습 모델의 취약성이 수면 위로 드러났으며, 이러한 취약성을 이용하여 악의적으로 모델을 공격함으로써 오동작을 유도하고자 하는 시도가 다방면으로 이루어짐에 의해 학습 모델의 강건함 보장은 보안 분야에서의 쟁점으로 부각되고 있다. 모델 추론의 입력으로 이용되는 이미지에 교란값을 추가함으로써 심층신경망의 오분류를 발생시키는 임의의 변형된 이미지를 적대적 사례라 정의하며, 본 논문에서는 최근 인공지능 및 컴퓨터비전 분야에서 이루어지고 있는 이미지 기반 적대적 사례의 생성 기법에 대하여 논한다.

• Journal of the Korean Society of Clothing and Textiles
• /
• v.32 no.3
• /
• pp.505-516
• /
• 2008

Image which appears in fashion illustration on the late twentieth century is not the representative image as an equivalence to the real fashion styles but the virtual image which bears no relation to any reality. The purpose of this study is review the concept of virtuality and analyze in which way virtual image is expressed in fashion illustrations on the background of Jean Baudrillard's simulacre theory. In post-modem paintings the expression methods of image-virtualization were image mixing through photo-image appropriation, image overlapping, and the icons inserted unreasonably, the focus-out effect through scrubbing and the over-painting on the photograph. Image-virtualization in fashion illustration was expressed through image mixing and expression of image uncertainty. Image mixing was made by photo-image appropriation, image overlapping, connection of heterogeneous images and using interface image, and uncertain image was expressed through the expression of visual ambiguity and virtual movement.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2011.06a
• /
• pp.97-98
• /
• 2011

• The Journal of the Korea Contents Association
• /
• v.16 no.1
• /
• pp.595-610
• /
• 2016

This study investigates the parameters that affect the choice of Korean medical services, as well as the relationship between the cultural characteristics and the behavioral intention to use Korean medical services. The cultural characteristics of Chinese and Russian individuals have an impact on the national image and image of Korea's medical services, and that this affects their behavioral intention to use Korean medical services. Nevertheless, there were differences among the cultural characteristics, in terms of specific elements. Among Chinese individuals, the national image is affected by 'power distance', 'individualism-collectivism', and 'long-term orientation'; for Russians, it is affected by 'individualism-collectivism', 'power distance', 'uncertainty avoidance', and 'long-term orientation'. On the other hand, for Chinese, the image of Korean medical services is affected by 'power distance' and 'long-term orientation', while for Russians it is affected by 'power distance' and 'uncertainty avoidance'. These in turn influence their behavioral intention to use Korean medical services. As such, by recognizing these cultural properties, and by designing and offering suitable services with these in mind, Korean medical institutions can nurture among foreign visitors greater satisfaction and a desire to revisit.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2016.04a
• /
• pp.726-727
• /
• 2016

정보이론에서 엔트로피는 불확실성 또는 복잡성을 정량적으로 표현할 수 있는 개념이다. 이 개념을 차용하여 칼라 이미지에서 지역적인 복잡성을 R,G,B에 대해서 각각 엔트로피로 계산하고 R,G,B의 통계적인 특성을 고려하여 가중치를 부여하여 평균값을 구한 후 이 값을 임계치와 비교하여 복잡도가 큰 픽셀을 외곽선의 일부로 결정하는 알고리즘을 제안하였다. 제안 알고리즘을 기존의 소벨과 프레윗 알고리즘과 비교한 결과 다양한 임계치 값에 대해서 기존 알고리즘보다 시각적으로 우수한 외곽선 겸출 효과를 가짐을 확인하였다.

• Journal of The Korean Association of Information Education
• /
• v.19 no.1
• /
• pp.139-148
• /
• 2015

The purpose of image fusion is to combine the relevant information from a set of images into a single image, where the resultant fused image will be more informative and complete than any of the input images. Image fusion techniques can improve the quality and increase the application of these data important applications of the fusion of images include medical imaging, remote sensing, and robotics. In this paper, we suggest a new method to generate a fusion image using the close relation of image features obtained through maximum entropy threshold and mutual information. This method represents a good image registration in case of using a blurring image than other image fusion methods.

• Tunnel and Underground Space
• /
• v.15 no.2 s.55
• /
• pp.145-156
• /
• 2005

MODFLOW, 3-D finite difference code, is widely used to model groundwater flow and has been used to assess the effect of excavations on the groundwater system due to construction of subways and mountain tunnels. The results of numerical analysis depend on boundary conditions, initial conditions, conceptual models and hydrogeological properties. Therefore, its accuracy can only be enhanced using more realistic and field oriented input parameters. In this study, SA(simulated annealing) was used to integrate hydraulic conductivities from a few of injection tests with geophysical reference images. The realizations of hydraulic conductivity random field are obtained and then groundwater flows in each geostatistically equivalent media are analyzed with a numerical simulation. This approach can give probabilistic results of groundwater flow modeling considering the uncertainty of hydrogeological medium. In other words, this approach makes it possible to quantify the propagation of uncertainty of hydraulic conductivities into groundwater flow.

• Journal of Korean Society for Geospatial Information Science
• /
• v.3 no.2 s.6
• /
• pp.91-100
• /
• 1995

The aim of this study is to classify mixed information in a pixel of a remote sensing image data (in the case of SPOT HRV's band $1{\sim}3,\;20m{\times}20m$). First, the loss of information and the uncertainty of mixed pixel are examined. To solve the problems, methods by fuzzy sigmoid function and back-propagation neural network are suggested. Then. the study simulates and comparatively analyzes the two methods.

• The KIPS Transactions:PartD
• /
• v.10D no.6
• /
• pp.873-884
• /
• 2003

So far, significant achievements have been studied on the development of models for spatial and spatiotemporal objects with indeterminate boundaries which are found in many applications for geographic analysis and image understanding. Therefore, in this paper we propose the spatiotemporal data model which is applicable for spatial and spatiotemporal objects with uncertainty. Based on this model, we defined topological relationships among the indeterminate spatiotemporal objects and designed the algorithm for the operations. For compatibility with existing spatial models, the proposed model has been designed by extending the spatiotemporal object model which is based on the open GIS specification. We defined indeterminate spatial objects, such as the objects whose position and the shape change discretely over time, and the objects whose shape changes continuously as well as the position. We defined topological relationships among these objects using the extended 9-IM. The proposed model can be efficiently applied to the management systems of natural resource data, westher information, geographic information. and so on.

• The Journal of the Acoustical Society of Korea
• /
• v.42 no.4
• /
• pp.305-312
• /
• 2023

Physics-Informed Neural Network (PINN) is used to invert bubble size distributions from attenuation losses. By considering a linear system for the bubble population inversion, Adaptive Learned Iterative Shrinkage Thresholding Algorithm (Ada-LISTA), which has been solved linear systems in image processing, is used as a neural network architecture in PINN. Furthermore, a regularization based on the linear system is added to a loss function of PINN and it makes a PINN have better generalization by a solution satisfying the bubble physics. To evaluate an uncertainty of bubble estimation, deep ensemble is adopted. 20 Ada-LISTAs with different initial values are trained using the same training dataset. During test with attenuation losses different from those in the training dataset, the bubble size distribution and corresponding uncertainty are indicated by average and variance of 20 estimations, respectively. Deep ensemble Ada-LISTA demonstrate superior performance in inverting bubble size distributions than the conventional convex optimization solver of CVX.