• Annual Conference on Human and Language Technology
• /
• 2021.10a
• /
• pp.203-208
• /
• 2021

본 논문에서는 맥락에 따라 개체명의 범주가 달라지는 어휘를 중심으로 교차 태깅된 개체명의 성능을 레이블과 스팬 정답률, 문장 성분과 문장 위치에 따른 정답률로 나누어 살펴 보았다. 레이블의 정확도는 KoGPT2, mBERT, KLUE-RoBERTa 순으로 정답률이 높아지는 양상을 보였다. 스팬 정답률에서는 mBERT가 KLUE-RoBERTa보다 근소하게 성능이 높았고 KoGPT2는 매우 낮은 정확도를 보였다. 다만, KoGPT2는 개체명이 문장의 끝에 위치할 때는 다른 모델과 비슷한 정도로 성능이 개선되는 결과를 보였다. 문장 종결 위치에서 인식기의 성능이 좋은 것은 실험에 사용된 말뭉치의 문장 성분이 서술어일 때 명사의 중첩이 적고 구문이 패턴화되어 있다는 특징과 KoGPT2가 decoder기반의 모델이기 때문으로 여겨지나 이에 대해서는 후속 연구가 필요하다.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2012.07a
• /
• pp.43-46
• /
• 2012

컴퓨터 비전을 이용한 다양한 응용 분야에 있어서, 특징점을 이용한 응용 분야가 많이 이루어지고 있다. 그 중에 Global feature는 표현의 위험성과 부정확성으로 인해서 많이 사용되고 있지 않으며, Local feature를 이용한 연구가 주로 이루고 있다. 그 중에 SURF(Speeded-Up Robust Features) 알고리즘은 다수의 영상에서 같은 물리적 위치에 있는 동일한 특징점을 찾아서 매칭하는 방법으로 널리 알려진 특징점 매칭 알고리즘이다. 하지만 SURF 알고리즘을 이용하여 특징점을 매칭하여 정합 쌍을 구하였을 때 매칭되는 특징점들의 정확도가 떨어지는 단점이 있다. 본 논문에서는 특징점 매칭 알고리즘인 SURF를 사용하여 대응되는 특징점들을 들로네 삼각형의 기하학적 특징을 이용하여 정확도가 높은 특징점을 분류하여 SURF 알고리즘의 매칭되는 대응점들의 정확도를 높이는 방법을 제안한다.

• Journal of the Institute of Electronics Engineers of Korea TC
• /
• v.45 no.12
• /
• pp.97-104
• /
• 2008

Location-based services provide customized information or services according to the user's location. The existing localization schemes for outdoor environment are not applicable to the indoor localization system which requires higher accuracy of location estimation than that of the outdoor localization system. In this paper, we employ the received signal strength(RSS) to approximate the distance between a moving target and a reference point and use the triangulation method to estimate the location of the moving target for the indoor localization system in IEEE 802.15.4 wireless PAN(personal area network). For the indoor localization system, we propose a scheme which selects the best reference points to enhance the localization accuracy and adaptively reflects the changes in propagation environments of a moving target to the distance approximation. Through the implementation of the localization system, we have verified the performance of the proposed scheme in terms of the estimation accuracy.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.13 no.5
• /
• pp.55-63
• /
• 2013

In cognitive radio networks, acquirement of position information of primary user is very important to secondary network since localization information of primary users can be utilized for improving the spectrum efficiency of secondary network and for avoiding harmful interference to primary users by using proper power control. Among various location methods, Received Signal Strength (RSS)-based localization has been widely used for distance measurements in the location detection process despite its inherent inaccuracy because it can be easily implemented without any additional hardware cost. In the RSS-based localization, the distance is measured by the received signal strength, and distance error can be caused by many factors such as fading, shadowing and obstacle between two nodes. In the paper, therefore we propose a localization scheme based on estimation of path-loss exponent to localize the location of primary users more accurately. Through simulations, it is shown that the proposed scheme can provide less localization error and interference rate to primary users than other schemes.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.9 no.1
• /
• pp.83-96
• /
• 1991

The use of SPOT Imagery is a growing trend in the field of small and middle scale mapping, as well as in establishing topographic database. This study is about 3-D positioning using the SPOT Imagery, where the accuracy and the gemetric characteristics of SPOT photos are analysed according to the preprocessing level (level 1AP,1B). As a result of this study the following could be determined, i. e 1) the geometric characteristics of SPOT Imagery according to the preprocessing level, 2) the optimal polynomial type for exterior orientations of each preprocessing level, and 3) the type of significant additional parameters. It was found that both the geometric precision and accuracy of level 1AP is higher than those of level 1B, which implies that level 1AP is more suitable for precise 3-D positioning and map production.

• Journal of Korean Society of Transportation
• /
• v.17 no.1
• /
• pp.105-112
• /
• 1999

Image Processing Technique has been used as an efficient method to collect traffic information on the road such as vehicle counts, speed, queues, congestion and incidents. Most of the current methods which have been used to detect vehicles by the image processing are based on point processing, dealing with the local gray level of each pixel in the small window. However, these methods have some drawbacks. Firstly, detection is restricted by image quality. Secondly, they can not deal with occlusion and perspective projection problems, In this research, a new method which possibly deals with occlusion and perspective problems will be proposed. It extracts spatial information such as the position, the relationship of vehicles in 3-dimensional space, as well as vehicle detection in the image. The main algorithm used in this research is based on an extension of the Hough Transform. The Hough Transform which is proposed to estimates parameters of vertices and directed edges analytically on the Hough Space, is a valuable method for the 3-dimensional analysis of static scenes, motion detection and the estimation of viewing parameters.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.51 no.5
• /
• pp.88-96
• /
• 2014

In modern electronic warfare systems, the necessity of a more accurate estimation method based on non-AOA (arrival of angle) measurement, such as TDOA and FDOA, have been increased. The previous researches using single TDOA have been carried out in terms of not only the development of emitter location algorithms but also the enhancement of measurement accuracy. Recently, however, the combined TDOA/FDOA method is of considerable interest because it is able to estimate the velocity vector of a moving emitter and acquire a pair of TDOA and FDOA measurements from a single sensor pair. In this circumstance, it is needed to derive the required FDOA measurement accuracy in order that the TDOA/FDOA combined localization system outperforms the previous single TDOA localization systems. Therefore, we analyze the contribution of FDOA measurement accuracy to emitter location, then propose the criterion based on CRLB (Cramer-Rao lower bound). Simulations are included to examine the validity of the proposed criterion by using the Gauss-Newton algorithm.

• Korean Journal of Remote Sensing
• /
• v.31 no.5
• /
• pp.421-432
• /
• 2015

There are significant differences in geometric property and stereo model accuracy between single-sensor stereo that uses two images taken by stereo acquisition mechanism within identical sensor and dual-sensor stereo that randomly combines two images taken from two different sensors. This paper compares the two types of stereo pairs thoroughly. For experiment, two single-sensor stereo pairs and four dual-sensor stereo pairs were constituted using SPOT-5 stereo and KOMPSAT-2 stereo covering same area. While the two single-sensor stereos have stable geometry, the dual-sensor stereos produced two stable and two unstable geometries. In particular, the unstable geometry led to a decrease in stereo model accuracy of the dual-sensor stereos. The two types of stereo pairs were also compared under the stable geometry. Overall, single-sensor stereos performed better than dual-sensor stereos for vertical mapping, but dual-sensor stereos was more accurate for horizontal mapping. This paper has revealed the differences of two types of stereos with their geometric properties and positioning accuracies, suggesting important considerations for handling satellite stereo images, particularly for dual-satellite stereo images.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.27 no.5
• /
• pp.529-534
• /
• 2009

In this study, we investigate the feasibility of modeling entire image strips that has been acquired from the same orbital segments. We tested sensor models based on satellite orbit and attitude with different sets(Type1 ~ Type4) of unknowns. We checked the accuracy of orbit modeling by establishing sensor models of one scene using control points extracted from the scene and by applying the models to adjacent scenes within the same orbital segments. Results indicated that modeling of individual scenes with 1st or 2nd order unknowns was recommended. We tested the accuracy of around control points, digital map using the HIST-DPW (Hanjin Information Systems & Telecommunication Digital Photogrammetric Workstation) As a result, we showed that the orbit-based sensor model is a suitable sensor model for making 1/25,000 digital map.

• Proceedings of the Korean Information Science Society Conference
• /
• 2005.11a
• /
• pp.376-378
• /
• 2005

유비쿼터스 컴퓨팅을 구현하려면 위치인식 기술과 이를 기반으로 하는 다양한 위치 인식시스템이 필수적이다. 그러나 복잡한 실내 환경에서는 Global Positioning System (GPS)의 이용이 불가능하며 실내 위치 측정을 위한 전문적인 시스템이 필요하다. RFID 기술은 적은 시설투자와 낮은 비용으로 위치 측정 시스템을 구축할 수 있는 가능성을 제공한다. RFID기술을 이용한 실내 위치 측정 시스템 LANDMARC는 기타 시스템과 달리 참고태그라는 개념을 도입하여 RFID리더기의 배치 수량을 줄이는 동시에 위치 측정의 정확성을 보장함으로써 시스템을 도입하는데 소요되는 비용을 절감하였으며 현실에서의 높은 사용 가능성을 제공한다. 본 논문에서는 기존의 LANDMARC시스템을 분석하여 문제점들을 해결하고 실내에서의 위치측정의 정확도를 진일보 향상시킬 수 있는 보다 효율적인 방안을 제안했다. 실험 분석 결과 동일한 실내 환경하에서 본 논문에서 제안하는 방법이 기존의 LANDMARC 시스템에 비하여 더 높은 정확성을 보장함을 입증하였다.