• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제12권8호
• /
• pp.3790-3803
• /
• 2018

Content-based image retrieval is an approach used to query images based on their semantics. Semantic based retrieval has its application in all fields including medicine, space, computing etc. Semantically generated binary hash codes can improve content-based image retrieval. These semantic labels / binary hash codes can be generated from unlabeled data using convolutional autoencoders. Proposed approach uses semi-supervised deep hashing with semantic learning and binary code generation by minimizing the objective function. Convolutional autoencoders are basis to extract semantic features due to its property of image generation from low level semantic representations. These representations of images are more effective than simple feature extraction and can preserve better semantic information. Proposed activation and loss functions helped to minimize classification error and produce better hash codes. Most widely used datasets have been used for verification of this approach that outperforms the existing methods.

• 한국콘텐츠학회:학술대회논문집
• /
• 한국콘텐츠학회 2016년도 춘계 종합학술대회 논문집
• /
• pp.257-258
• /
• 2016

In my paper, an evaluation method on naturalness and urbanness characterizations of animation image is proposed. In contrast to the prior works, we intend to give the exact degree of naturalness and urbanness by combining the low-level visual feature-based schemes with the biological visual schemes. In our method, four descriptors are extracted and pseudoinverse is adopted. By this method, we can experiment on the exact degree of naturalness and urbanness with satisfactory results.

• 조명전기설비학회논문지
• /
• 제30권1호
• /
• pp.79-86
• /
• 2016

In this paper, we propose a current-driven synchronous rectifier driver circuit for LLC resonant half-bridge converters. The proposed driver circuit detects a relatively low current in the primary side of the transformer although a large current is flowing in the secondary side. Due to this feature, the driver circuit has a simple circuit structure and stabilizes the switching operation with a logic-level switching voltages for the synchronous rectifier. The operation and performance of the proposed driver circuit are confirmed with a prototype of 1kW class LLC resonant half-bridge converter. The experimental results proved that the proposed synchronous rectifier driver method improves the power conversion efficiency by around 1% and reduces the internal power loss by 17W.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 2003년도 학술회의 논문집 정보 및 제어부문 B
• /
• pp.392-395
• /
• 2003

This paper describes a new iris segmentation and recognition method, which is robust to noises. Combining statistical classification and elastic boundary fitting, the iris is first segmented. Then, the localized iris image is smoothed by a convolution with a Gaussian function, down-sampled by a factor of filtered with a Laplacian operator, and quantized using the Lloyd-Max method. Since the quantized output is sensitive to a small shift of the full-resolution iris image, the outputs of the Laplacian operator are computed for all space shifts. The quantized output with maximum entropy is selected as the final feature representation. An appropriate formulation of similarity measure is defined for the classification of the quantized output. Experimentally we showed that the proposed method produces superb performance in iris segmentation and recognition.

• 한국정밀공학회지
• /
• 제30권1호
• /
• pp.53-58
• /
• 2013

In this paper, we present a framework for collision avoidance of UGVs by vision-based control. On the image plane which is created by perspective camera rigidly attached to UAV hovering stationarily, image features of UGVs are to be controlled by our control framework so that they proceed to desired locations while avoiding collision. UGVs are assumed as unicycle wheeled mobile robots with nonholonomic constraint and they follow the image feature's movement on the ground plane with low-level controller. We used potential function method to guarantee collision prevention, and showed its stability. Simulation results are presented to validate capability and stability of the proposed framework.

• 교육녹색환경연구
• /
• 제12권3호
• /
• pp.58-67
• /
• 2013

This study is to analyze the satisfaction and the image evaluation of landscape elements in outdoor space by types of the university campus. The results are as follows. 1) Out of outdoor elements at university campus, planting area, resting area, access road, and water feature are recognized as major landscape elements. Among them, planting area and access roads are evaluated low in terms of satisfaction levels, therefore, improvement on these elements are required. 2) In outdoor space image evaluation, university campus has image such as 'simple', 'clear', and 'safe'. By scale of universities, both 'A' university, which is the biggest in terms of size of campus, and 'B' university, which has a medium sized campus, have a positive image. However, 'C' university, which is the smallest in terms of size of campus, has a passive and negative image. 3) 6 factors are extracted through Factor Analysis for image evaluation. All of the universities show positive image in the categories of 'clarity' and 'familiarity', however, 'B' university and 'C' university show negative image in the category of 'scale'. 4) In Correlation Analysis between landscape elements satisfaction level and image evaluation, it is showed that the group of landscape facility becomes a relation factor of overall image evaluation. As a result, the higher satisfaction level goes, the better image evaluation of overall outdoor space at university campus is.

• Journal of international Conference on Electrical Machines and Systems
• /
• 제1권4호
• /
• pp.457-463
• /
• 2012

Cascaded H-bridge multilevel inverters shows a useful circuit configuration to increase the number of output voltage levels to obtain high quality output voltage. By applying the concept of the power of three to dc voltage sources, it can increase the number of output voltage levels effectively. To realize this concept, two approaches may be considered. One is to use independent dc voltage sources pre-scaled in the power of three, and the other is to use instantaneous dc voltage sources generated from a cascaded transformer, which has the secondary turn-ratios scaled in the power of three in sequence. A common feature in both approaches is to use the concept of the power of three for dc voltage sources, and a point of difference is whether it adopts a low frequency transformer or not, and where the transformer is located. According to the difference, application areas are limited and show different characteristics on THD of output voltages. We compare and analyze both approaches for their circuit configurations, voltage level generating method, THD characteristics of output voltage, efficiency, application areas, limitations, and other characteristics by experiments using 500 [W] prototypes when they generate a 27-level output voltage.

• 한국정보통신학회논문지
• /
• 제15권7호
• /
• pp.1517-1524
• /
• 2011

본 논문은 의미특징과 워드넷 기반의 의사연관피드백을 이용하여 사용자의 질의에 관련 있는 의미 있는 문장을 추출하여 문서요약을 하는 새로운 방법을 제안한다. 제안된 방법은 비음수 행렬 분해로부터 유도된 의미특정이 문서의 잠재의미를 잘 나타나기 때문에 문서요약의 질을 향상할 수 있다. 또한 의미특정과 워드넷기반의 의사연관피드백을 이용하여서 사용자의 요구사항과 제안방법의 요약결과 사이의 의미적 차이를 감소시킨다. 실험결과 제안방법이 유사도, 비음수행렬분해를 이용한 방법들에 비하여 좋은 성능을 보인다.

• Wind and Structures
• /
• 제26권1호
• /
• pp.45-56
• /
• 2018

Wind damage of urban trees arises to be a serious issue especially in the typhoon-prone areas. As a family of tree species widely-planted in Southeast China, the structural behaviors of Plane tree is investigated. In order to accommodate the complexities of tree morphology, a fractal theory based finite element modeling method is proposed. On-site measurement of Plane trees is performed for physical definition of structural parameters. It is revealed that modal frequencies of Plane trees distribute in a manner of grouped dense-frequencies; bending is the main mode of structural failure. In conjunction with the probability density evolution method, the fragility assessment of urban trees subjected to wind excitations is then proceeded. Numerical results indicate that small-size segments such as secondary branches feature a relatively higher failure risk in a low wind level, and a relatively lower failure risk in a high wind level owing to windward shrinks. Besides, the trunk of Plane tree is the segment most likely to be damaged than other segments in case of high winds. The failure position tends to occur at the connection between trunk and primary branches, where the logical protections and reinforcement measures can be implemented for mitigating the wind damage.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제10권3호
• /
• pp.1229-1248
• /
• 2016

With the feature of convenience and low cost, remote healthcare monitoring (RHM) has been extensively used in modern disease management to improve the quality of life. Due to the privacy of health data, it is of great importance to implement RHM based on a secure and dependable network. However, the network connectivity of existing RHM systems is unreliable in disaster area because of the unforeseeable damage to the communication infrastructure. To design a secure RHM system in disaster area, this paper presents a Secure VANET-Assisted Remote Healthcare Monitoring System (SVC) by utilizing the unique "store-carry-forward" transmission mode of vehicular ad hoc network (VANET). To improve the network performance, the VANET in SVC is designed to be a two-level network consisting of two kinds of vehicles. Specially, an innovative two-level key management model by mixing certificate-based cryptography and ID-based cryptography is customized to manage the trust of vehicles. In addition, the strong privacy of the health information including context privacy is taken into account in our scheme by combining searchable public-key encryption and broadcast techniques. Finally, comprehensive security and performance analysis demonstrate the scheme is secure and efficient.