• 한국산업정보학회논문지
• /
• 제19권2호
• /
• pp.85-91
• /
• 2014

본 연구는 MLS 데이터로부터 자동으로 철도 시설물들을 인식하여 시설물 간의 기하학적인 공간정보를 추출하는데 기여 하고자 한다. 본 연구에서 제안된 방법은 9개 주요 철도 시설물(노반, 레일, 철로, 수목, 플렛폼, 방음벽, 철주, 절연체, 고압선)들의 분류를 목적으로 하고 있다. 이를 위해 제안된 방법은 크게 두 단계로 나뉘어 진행된다. 첫 번째 단계에서는 포인트, 라인, 체적과 수직 프로파일 레벨에서 데이터의 맥락 특징(contextual feature)들이 추출된다. 두 번째 단계에서는 CRF(Conditional Random Field)가 맥락 분류자(contextual classifier)로 사용되어 각 데이터 포인트에 객체 정보가 할당되고 철도 시설물들이 분류된다. 사용된 CRF 모델은 다른 맥락 분류자 와는 달리 로컬지역에서 데이터들의 분류정보가 일관성을 유지하게 하는 장점이 있다. 제안된 방법의 성능은 commission과 omission 오류분석을 통해 입증되었다.

• 대한교통학회지
• /
• 제5권2호
• /
• pp.97-112
• /
• 1987

The objective of this study is to construct not only trip production and attraction in Pusan but also to study and examine appropriateness of the model positively. Depending on the estimation models of trip production and attraction of each zone that have been constructed in this study, it has been proved that the formula of multiple regression by the explanation variables like the indices of total employees, total students, floor spaces of residentials and floor spaces of educational and cultural areas within the study areas have very high explanatory capacity and appropriateness. It si considered that a study of method on new division, integration or omission etc. of the existing zones preceeding for reduction of calculation quantity and a study of estimation error have to be done for future study, if these models are used actually.

• 대한음성학회지:말소리
• /
• 제46호
• /
• pp.117-126
• /
• 2003

This paper presents our language model adaptation for Korean spontaneous speech recognition. Korean spontaneous speech is observed various characteristics of content and style such as filled pauses, word omission, and contraction as compared with the written text corpus. Our approaches focus on improving the estimation of domain-dependent n-gram models by relevance weighting out-of-domain text data, where style is represented by n-gram based tf/sup */idf similarity. In addition to relevance weighting, we use disfluencies as Predictor to the neighboring words. The best result reduces 9.7% word error rate relatively and shows that n-gram based relevance weighting reflects style difference greatly and disfluencies are good predictor also.

• Wind and Structures
• /
• 제9권5호
• /
• pp.397-412
• /
• 2006

This paper presents the wind excited acceleration responses of a 50 m guyed mast under the action of Typhoon Dujuan. The response of the structure is reconstructed from using a full finite element model and an equivalent beam-column model. The wind load is modelled based on the measured wind speed and recommendations for high-rise structures. The nonlinear time response analysis is conducted using the Newton Raphson iteration procedure. Comparative studies on the measured and computed frequencies and acceleration responses show that the torsional vibration of the structure is significant particularly in the higher vibration modes after the first few bending modes. The equivalent model, in general, gives less accurate amplitude predictions than the full model because of the omission of torsional stiffness of the mast in the vibration analysis, but the root-mean-square value is close to the measured value in general with an error of less than 10%.

• ETRI Journal
• /
• 제33권4호
• /
• pp.547-557
• /
• 2011

This study presents an approach for extracting boundaries of various buildings, which have concave boundaries, inner yards, non-right-angled corners, and nonlinear edges. The approach comprises four steps: building point segmentation, boundary tracing, boundary grouping, and regularization. In the second and third steps, conventional algorithms are improved for more accurate boundary extraction, and in the final step, a new algorithm is presented to extract nonlinear edges. The unique characteristics of airborne light detection and ranging (LIDAR) data are considered in some steps. The performance and practicality of the presented algorithm were evaluated for buildings of various shapes, and the average omission and commission error of building polygon areas were 0.038 and 0.033, respectively.

• 대한음성학회:학술대회논문집
• /
• 대한음성학회 2003년도 5월 학술대회지
• /
• pp.83-86
• /
• 2003

This paper presents our style-based language model adaptation for Korean conversational speech recognition. Korean conversational speech is observed various characteristics of content and style such as filled pauses, word omission, and contraction as compared with the written text corpora. For style-based language model adaptation, we report two approaches. Our approaches focus on improving the estimation of domain-dependent n-gram models by relevance weighting out-of-domain text data, where style is represented by n-gram based tf*idf similarity. In addition to relevance weighting, we use disfluencies as predictor to the neighboring words. The best result reduces 6.5％ word error rate absolutely and shows that n-gram based relevance weighting reflects style difference greatly and disfluencies are good predictor.

• International Journal of Advanced Culture Technology
• /
• 제6권4호
• /
• pp.165-171
• /
• 2018

As the artificial intelligence (AI) plays a crucial role in machine translation (MT) which has loomed large as a new translation paradigm, concerns have also arisen if MT can produce a quality product as human translation (HT) can. In fact, several MT experimental studies report cases in which the MT product called post-editing (PE) as equally as HT or often superior ([1],[2],[6]). As motivated from those studies on translation quality between HT and PE, this study set up an experimental situation in which Korean literature was translated into English, comparatively, by 3 translators and 3 post-editors. Afterwards, a group of 3 other Koreans checked for accuracy of HT and PE; a group of 3 English native speakers scored for fluency of HT and PE. The findings are (1) HT took the translation time, at least, twice longer than PE. (2) Both HT and PE produced similar error types, and Mistranslation and Omission were the major errors for accuracy and Grammar for fluency. (3) HT turned to be inferior to PE for both accuracy and fluency.

• 한국산학기술학회논문지
• /
• 제10권12호
• /
• pp.3863-3869
• /
• 2009

최근 자동차산업에서의 시장의 세계화와 증가하는 고객의 까다로운 욕구는 다양한 제품의 증가와 제조업체가 공급해야 하는 생산라인의 다양한 방식의 증가로 이어지게 되었다. 또한, 자동차 생산시스템의 모듈화는 자동차부품 공급업체들의 생산시스템에도 많은 영향을 미치고 있다. 이러한 변화는 혼합형 모델 조립라인이라 불리는 비용 효율적이고 유연한 생산시스템의 구현을 요구하고 있다. 본 논문에서는 이러한 혼합형 모델 조립에서 발생하는 여러 문제점들과 이러한 조립오류를 방지할 수 있는 기술에 대해 살펴보고, 혼합형 모델 자동차부품 조립라인에서 RFID를 이용한 오류방지 모니터링시스템을 개발하였다. 특히, 자동차부품 조립현장에서 일어나는 부품의 잘못된 설치와 누락과 같은 문제점들을 최소화하기 위하여 RFID를 이용한 모니터링시스템을 개발함으로써 문제점을 해결하고 조립오류를 예방하여 기존의 시스템을 보완하고자 하였다.

• 한국보건간호학회지
• /
• 제16권1호
• /
• pp.176-189
• /
• 2002

The purpose of this study were to illustrate the various medication error types and causes and identified to related drugs to provide basic data for preventing nurses' medication error by analysing 73 cases of AJN 'medication Error' column(1993, Oct -2000, Nov). Nurses' types of medication error were classified into 7 types. The most frequent error types are wrong medication$(21.9\%)$ and the wrong dose$(21.9\%)$ together. The others are wrong $time(4.1\%)$, $omission(2.7\%)$, mechanical $error(2.7\%)$, incorrect IV $rate(1.4\%)$. wrong route $administration(1.4\%)$ in order. Nurses' causes of medication error were 9 kinds. The most frequent type is confusing between similar drug shape, color, size, name, injection devices and patient's $name(43.9\%)$ and the others are lack of knowledge about $drugs(26.8\%),\; slips(7.3\%),\; miscalculating\;dose(4.9\%)$, incorrect adjusts $devices(4.9\%)$, difficulty to read or illegible decimal $point(4.9\%),$ $abbreviation(2.4\%)$, fatigue with $overwork(2.4\%)$ and no communication with $patient(2.4\%)$ in order. Related drugs with medication error are as follows. - dose unit(IU. minims. mcg/min. mEq) : Heparin. insulin. synthetic calcitonin, some enzymes and hormones, vitamins, some antibiotics, tuberculin injection. MgSO4 injection. nitroglycerin - similar size, color and shape drug : $0.9\%$ N/S and acetic acid $0.25\%$ for irrigation. premixed 2mg lidocaine sol. and $0.9\%$ N/S, gentamycin 20mg/2mL for children and 80mg/2mL for adult, dextroamphetamine 5mg and 10mg capsule. sedatives chloral hydrate 250mg/5mL and 500mg/5mL - similar name :Aredia(pamidronate disodium) and Adriamycin(doxorubicin), Lamictal (lamotrigine) and Lamisil 250mg. Elderpryl and enalapril, cefotaxime and cefoxitin, carboplatin and cisplatin, sumatriptan and zolmitriptan, Celebrex and Celexa, Humulin and Humalog, Percodan and Percocet, Diabeta and Diabinese, Epivir and Retrovir, Xanax(alprazolam) and Zantac(ranitidine) - decimal point : low molecular weight warfarin, methotrexate - unfamiliar drug uses of familiar drug ; methotrexate. droperidol, imipramine, propranolol - number of drug name(misleading chemical name) : 6-thioguanine, 6-mercaptopurine, 5-fluorouracil - type of administration route : Oxycodone(OxyContin). - administration time : acarbose(Precose). - injection way (Z-track method): hydroxyzine - epidural cathether : LMWHs(enoxaparin, dalteparin), - ADD Vantage self contained delivery system : ceftriaxone(Rocephin)

• 대한원격탐사학회지
• /
• 제20권4호
• /
• pp.261-274
• /
• 2004

위성영상 정보를 이용하여 지형지물을 인식하거나 추출하는 것은 영상정보의 실제적인 응용을 위한 기본 연구로 간주되고 있다. 또한 고 해상도 영상정보가 다양한 민간 분야에서 활발하게 이용됨에 따라 정확한 지형지물의 추출에 대한 필요성이 더욱 강조되고 있다. 지금까지 중-저 해상도 영상정보를 대상으로 하여 도로 경계를 포함하는 선형 지형지물을 자동으로 추출하기 위한 여러 가지 방법들이 개발되고 있어 왔으나 이를 고 해상도 영상정보에 적용하거나 적용 결과를 분석한 사례는 많지 않다. 본 연구에서는 항공사진 정보니 고 해상도 영상정보를 대상으로 하여 선형 지형지물의 추출에 적합한 기법으로 최근에 제안된 구배 방향 프로파일 알고리즘(GDPA)과 Hough 변환기법에 따른 적용 결과를 동일한 자료에 적용하고 이러한 결과를 비교하고자 하였다. 각 방법에 대한 적용 결과를 정량적으로 비교하기 위해 수치지도의 도로 중심선과 도로 경계선 레이어를 기준으로 하는 Commission, Omission 오차를 이용한 Ranking기법을 적용하여 수행하였다. 연구 결과, Hough 변환기법이 추출 결과 영상에서는 평균적으로 20%정도의 높은 정확도를 보이며, 처리속도는 GDPA의 경우가 Hough 변환 기법에 비하여 빠른 실행 속도를 보이는 것으로 나타났다 그러나 GDPA 알고리즘에 잡음제거를 시행한 경우에는 Hough 변환에 의한 결과와 유사한 정확도를 보인다. 결론적으로 위성 영상정보로부터 지형지물을 추출하는 어플리케이션에서는 GDPA 알고리즘이 Hough 변환 기법에 비하여 적용성이 좋은 것으로 생각된다.