• 한국통신학회논문지
• /
• 제12권2호
• /
• pp.176-188
• /
• 1987

본 논문은 역필터링을 이용하여 이동에 의해 흐려진 영상을 복원하는 기법에 관한 논문이다. 적절한 역필터를 설계하기 위하여 저하과정(Degradation Process)을 정확히 모델링하여 그의 역특성을 갖는 필터를 구해내야 한다. 고속으로 이동하는 물체의 영상은 Blurring되는데, 이것은 수학적으로 시간에 대한 영상의 적분으로 모델링될 수 있으며, 공간좌표상 방향성 방형 윈도우(Rectangular Window)를 상승(Convolution)하는 것으로 분석된다. 본 논문에서는 주파수 영역에서 이를 분석하여, 이동방향을 고려한 역 Sinc 함수 특성을 갖는 필터를 이용하여 흐려진 영상을 복원하는 기법을 연구한다.

• 산업경영시스템학회지
• /
• 제40권4호
• /
• pp.154-163
• /
• 2017

Collaborative filtering, one of the most widely used techniques to build recommender systems, is based on the idea that users with similar preferences can help one another find useful items. Credit card user behavior analytics show that most customers hold three or less credit cards without duplicates. This behavior is one of the most influential factors to data sparsity. The 'cold-start' problem caused by data sparsity prevents recommender system from providing recommendation properly in the personalized credit card recommendation scenario. We propose a personalized credit card recommender system to address the cold-start problem, using multiple user profiles. The proposed system consists of a training process and an application process using five user profiles. In the training process, the five user profiles are transformed to five user networks based on the cosine similarity, and an integrated user network is derived by weighted sum of each user network. The application process selects k-nearest neighbors (users) from the integrated user network derived in the training process, and recommends three of the most frequently used credit card by the k-nearest neighbors. In order to demonstrate the performance of the proposed system, we conducted experiments with real credit card user data and calculated the F1 Values. The F1 value of the proposed system was compared with that of the existing recommendation techniques. The results show that the proposed system provides better recommendation than the existing techniques. This paper not only contributes to solving the cold start problem that may occur in the personalized credit card recommendation scenario, but also is expected for financial companies to improve customer satisfactions and increase corporate profits by providing recommendation properly.

• 한국정보과학회논문지:데이타베이스
• /
• 제36권6호
• /
• pp.476-485
• /
• 2009

센서 네트워크에서 병합 질의를 효율적으로 처리하기 위한 다양한 인-네트워크 질의 처리 기법들이 제안되었다. 스카이라인 질의는 일반적인 병합 질의와 달리 다차원 데이터에 대한 총괄적인 비교를 요구한다. 따라서, 부분적인 데이터만을 이용할 수 있는 인-네트워크 방식으로 처리하기 어렵다. 스카이라인 질의를 에너지 효율적으로 처리하기 위해서 불필요한 데이터의 전송을 제거하는 것이 중요하다. 기존에 제안된 스카이라인 처리 기법은 전체 네트워크에 필터를 배포함으로써 불필요한 데이터 전송을 차단한다. 하지만 많은 긍정 오류(False Positive) 발생에 따른 불필요한 데이터 전송과 필터 배포시 발생하는 에너지 소모로 인해 네트워크의 수명이 단축된다. 본 논문에서는 필터 배포에 따른 에너지 소모를 줄이기 위한 방법으로 상향식 필터 설정을 통한 스카이라인 질의 처리 기법과 필터링 성능을 향상시키는 PBFiltering 기법을 제안한다. 제안하는 기법은 필터를 미리 배포하지 않고 하위 노드로부터 기지국으로 데이터를 수집하는 과정에서 스카이라인 필터 테이블(SFT)을 만들고 필터링을 수행한다. 그리고 여기서 제안하는 우선순위 맵을 이용한 선 필터링(Pre-filtering) 기법을 통해 필터링 효율을 증가시킨다. 제안하는 알고리즘의 우수성을 보이기 위해 기존에 제안된 MFTAC 기법과의 시뮬레이션을 통해 비교 평가하였다. 그 결과 기존 기법에 비해 다수의 긍정 오류의 발생을 감소시키고, 네트워크 수명이 연장됨을 보였다.

• 한국컴퓨터정보학회논문지
• /
• 제28권7호
• /
• pp.19-27
• /
• 2023

소셜 리뷰를 수집하는 과정에서 주어진 검색어와 상관없는 노이즈 리뷰가 검색 결과에 다수 포함될 수 있으며, 이들을 필터링하기 위해 기계 학습이 이용될 수 있다. 그러나 분석하고자 하는 대상의 리뷰 수가 부족한 경우, 학습 데이터 부족으로 인한 정확도 저하 문제가 발생할 수 있다. 본 논문에서는 리뷰 수가 부족한 플레이스를 대상으로 노이즈 리뷰 필터링의 정확도를 높이기 위한 지도 학습 방법을 소개한다. 제안 방법에서는 개별 플레이스 단위로 학습을 수행하지 않고, 특성이 유사한 여러 플레이스를 그룹으로 묶어 학습을 수행한다. 학습을 통해 얻은 분류기는 그룹에 속한 임의의 플레이스에 공통으로 적용함으로써 학습 데이터 부족 문제를 해결하고자 하였다. 제안 방법의 검증을 위해, LSTM과 BERT를 이용하여 노이즈 리뷰 필터링 모델을 구현하고, 온라인에서 수집된 실제 데이터를 활용한 실험을 통해 필터링 정확도를 체크하였다. 실험 결과, 제안 방법의 정확도는 평균 92.4% 수준이었으며, 리뷰 수가 100개 미만인 플레이스를 대상으로 할 경우 87.5%의 정확도를 제공하였다.

• Fibers and Polymers
• /
• 제2권3호
• /
• pp.164-170
• /
• 2001

Authors studied on the applicability of image analysis technique using a scanner with a CCD (charged coupled deviced) to the evaluation of evenness of nonwovens because it has distinctive features to considerably save time and labor in the analysis compared with other classical methods. As specimens fur the experiment, two different types that are unpatterned and patterned ones were prepared. For the unpatterned specimen, webs were chemically bonded, while for the patterned specimen, webs being thermally calendered with engraved roller. Several webs having various areal densities were prepared and bonded. Coefficient of variation (CV%) was used as a parameter to evaluate the evenness. Scanning conditions could be suitably set up through comparing the total variance to the between-group variance and to the within-group variance, respectively, on the images scanned at the different conditions. The 2D convolution method with smoothing filter kernel was introduced to further filter the noises on the scanned images. After the filtering process, the increase of web areal densities gave an uniform decrease of the CV%. This showed that the scanned image analysis with proper filtering process could be successfully applicable to the evaluation of evenness in nonwovens.

• 동력기계공학회지
• /
• 제13권4호
• /
• pp.43-48
• /
• 2009

The purpose of this study is to show how to eliminate the noises and to obtain the data related with the relationship between AE signal characteristics and mechanical behaviors for the pressure vessel steel. Various kinds of noises are introduced into the AE data in the course of experiments. Accordingly, real AE data have to be obtained after tests. AE test was carried out under four point bending load. Among AE signals, counts and signal strength are used to find out the differences of AE characteristics between the basemetal and weldment. After tests, this paper shows the procedures of filtering the noises against basemetal of the pressure vessel steel to obtain the real data around crack tip. Relationships between plastic zone size and cumulative AE counts are shown also. AE signals were absent within an elastic region, regardless of the specimens. Most of AE signals are produced in the process of plastic deformation. The deformation and fracture modes of basemetal and weldment are quiet different.

• 동력기계공학회지
• /
• 제5권1호
• /
• pp.64-72
• /
• 2001

According to the regulation of IMO, oil discharge from ships is allowed under 15ppm only and an oil filtering equipment is essential. However, for large ships using heavy fuel oil of over S.G 0.98 and viscosity 380cSt and system oil, it has been in difficulty to process with existing filtering type of oily water separator. A parallel plate type oily water separator which is one of gravity type separators can be used as an assistant equipment for the oil filtering system to meet the present IMO standard of 15 ppm, because it is an efficient method in dealing with a large amount of rich oil with high specific gravity. This work is focused on the fundamental investigation of the performance of the plate type oily water separator by visualization method and PIV(Particle Image Velocimetry) measurement to acquire multi-point velocity data simultaneously. The experimental results showed that the space of the plates acts a significant role in separating process and it was found that an important point to minimize a vortex flow is to flow a large amount of fluid in space of the plates in order to promote the efficiency of separation.

• 한국컴퓨터정보학회논문지
• /
• 제21권12호
• /
• pp.35-41
• /
• 2016

In this paper, we propose a new real-time dead pixel detection method based on spatial compare filtering, which are usually used in the small target detection. Actually, the soft dead and the small target are cast in the same mold. Our proposed method detect and remove the dead pixels as applying the spatial compare filtering, into the pixel outputs of a detector after the non-uniformity correction. Therefore, we proposed method can effectively detect and replace the dead pixels regardless of the non-uniformity correction performance. In infrared camera, there are usually many dead detector pixels which produce abnormal output caused by manufactural process or operational environment. There are two kind of dead pixel. one is hard dead pixel which electronically generate abnormal outputs and other is soft dead pixel which changed and generated abnormal outputs by the planning process. Infrared camera have to perform non-uniformity correction because of structural and material properties of infrared detector. The hard dead pixels whose offset values obtained by non-uniformity correction are much larger or smaller than the average can be detected easily as dead pixels. However, some dead pixels(soft dead pixel) can remain, because of the difficulty of uncleared decision whether normal pixel or abnormal pixel.

• 상하수도학회지
• /
• 제13권1호
• /
• pp.113-124
• /
• 1999

We elucidated factors such as membrane pore size, activity of the activated sludge and MLSS. affecting filtering characteristics or filtering resistance in an activated sludge bio-process using tubular type membranes. We used a commercial membrane module with $0.5m^2$ of effective membrane area and a bundle of 7 tubes(13mm of ID and 2m long) in the experiments. As a result, the flux was maintained in the range of $35{\sim}105l/m^2$ hr at 5,000~20,000mg/l of MLSS concentration, and there seemed to be a correlation between the flux and the MLSS. Also the greater the pores were, the higher the flux was in some comparison experiments using UF and MF membranes, but the water permeated through both of the membranes showed almost the same quality. However, the MF membrane showed a great reduction in filtering resistance initially, but went back to a stable stage afterwards. The activity of the activated sludge in the aeration tank was also found to be considered as a factor showing permeation characteristics of the membrane.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 1992년도 한국자동제어학술회의논문집(국제학술편); KOEX, Seoul; 19-21 Oct. 1992
• /
• pp.520-525
• /
• 1992

This paper reports several mathematical properties of the filter vector developed for processing linearly-additive spatially-invariant image sequences. In this filtering of an image sequence into a single filtered image, the information about the image components originally distributed over the entire sequence is compressed into the one new image in a way that the desired component is enhanced and the undesired (interfering) components and noise are suppressed.