• 한국도로학회논문집
• /
• 제18권6호
• /
• pp.61-67
• /
• 2016

OBJECTIVES : The objective of this study is to detect road cavities using multi-channel 3D ground penetrating radar (GPR) tests owned by the Seoul Metropolitan Government. METHODS : Ground-penetrating radar tests were conducted on 204 road-cavity test sections, and the GPR signal patterns were analyzed to classify signal shape, amplitude, and phase change. RESULTS : The shapes of the GPR signals of road-cavity sections were circular or ellipsoidal in the plane image of the 3D GPR results. However, in the longitudinal or transverse direction, the signals showed mostly unsymmetrical (or symmetrical in some cases) parabolic shapes. The amplitude of the GPR signals reflected from road cavities was stronger than that from other media. No particular pattern of the amplitude was found because of nonuniform medium and utilities nearby. In many cases where road cavities extended to the bottom of the asphalt concrete layer, the signal phase was reversed. However, no reversed signal was found in subbase, subgrade, or deeper locations. CONCLUSIONS : For detecting road cavities, the results of the GPR signal-pattern analysis can be applied. In general, GPR signals on road cavity-sections had unsymmetrical hyperbolic shape, relatively stronger amplitude, and reversed phase. Owing to the uncertainties of underground materials, utilities, and road cavities, GPR signal interpretation was difficult. To perform quantitative analysis for road cavity detection, additional GPR tests and signal pattern analysis need to be conducted.

• 한국정보통신학회:학술대회논문집
• /
• 한국정보통신학회 2015년도 춘계학술대회
• /
• pp.571-575
• /
• 2015

실시간으로 입력되는 영상으로부터 이동객체의 움직임 특징을 분석하는 시간적 히스토그램 기반의 행동패턴분석 알고리즘을 제안한다. 이동객체의 추적 및 분석을 위해 배경과 이동객체를 분리하는 배경학습을 행한다. 배경학습으로 추출된 이동객체는 무게중심 및 좌표연관성을 이용하여 객체를 식별한 후 객체별 추적을 행한다. 추적된 각 객체의 시작프레임, 종료프레임, 좌표정보, 크기정보를 연결리스트에 저장하여 관리한다. 시간적 히스토그램은 x, y좌표와 시간을 이용해 움직임 특징 패턴을 정의한 것으로 각 객체의 좌표정보와 비교하여 움직임특징 및 행동패턴을 파악한다. 시간적 히스토그램 기반 행동패턴분석시스템은 자체 수집한 데모영상에 대한 실험을 통해 초당 45~50 fps의 높은 처리속도를 유지하며 95%이상의 높은 추적율을 확인하였다.

• The International Journal of Costume Culture
• /
• 제2권2호
• /
• pp.60-70
• /
• 1999

The purposes of this study were to investigate preference and images of Korean traditional motifs, and to identify the relationship between them. The subjects consisted of 369 male and 356 female undergraduate students. The experimental materials used in this study were 48 stimuli and a questionnaire composed of 7-point semantic differential scales of 26 adjectives. Twelve motifs selected from 3 groups of Korean motifs were used as pattern design stimuli. Twelve repeated patterns were constructed from them to be applied on a CAD-simulated dress. The data were analyzed by ANOVA, Duncan's multiple range test, and regression. The major findings were as follows : 1. Category, composition type, and application object had a significant effect on the preference. Interpretation type has no significant effects on the preference independently, but it had interaction effects when combined with composition type, and category. Especially the composition type had a greater effect than the other variables on the preference. Cloud motif and its abstract and decorative type were found to be more related to the preference than the other category and interpretation type. On the basis of the analysis results, image charts and preference charts were developed. By combining information from the image chart and preference chart, motifs and images preferred by consumers may be selected and developed into new valuable designs. 2. The preference was affected mainly by 'quality'image followed by 'simplicity', 'interest', and 'modernity'image. The preference on pattern design was affected by 'quality', 'simplicity', 'interest', and 'interest', 'simplicity', and 'modernity'image in the order. The relationship between the preference and sensibility images has been represented by equations.

• 한국정보통신학회:학술대회논문집
• /
• 한국정보통신학회 2018년도 추계학술대회
• /
• pp.178-179
• /
• 2018

최근 자율주행 및 음성인식 등 인공지능 분야에서 기계학습을 이용한 방법이 활발히 연구되고 있다. 디지털 영상에서 특정 사물이나 영역을 인식하기 위해 고전적인 경계검출 및 패턴인식 등의 고전적인 영상처리 방법으로는 많은 한계를 가지고 있으나 deep-learning 등 기계학습 방법을 이용하면 사람의 인지수준에 근접한 결과를 얻을 수 있다. 하지만 기본적으로 deep-learning 등 기계학습은 방대한 양의 학습데이터가 확보되어야 한다. 따라서 환경 분석을 위한 항공사진처럼 데이터의 양이 매우 적은 경우 영역 구분을 위해 기계학습을 적용하기 어렵다. 본 연구에서는 입력영상의 dataset 크기가 적고 입력 영상의 형태가 training dataset의 category에 포함되지 않는 경우 사용할 수 있는 transfer-learning 기법을 적용하며 이를 이용하여 영상 내에서 특정 영역 검출을 수행한다.

• 한국의류산업학회지
• /
• 제10권1호
• /
• pp.68-75
• /
• 2008

Demands and attempts to express Korean traditional beauty are increasing. Some schools choose the uniform in designs expressing the image of Hanbok, but recently there is a few increment of wearing the uniforms. For the improvement of the uniforms, first of all, the shape and the clothing materials of the uniforms should be considered. We have compared and analyzed the shape and the clothing materials of the uniforms reflecting the image of Hanbok, according to seasons and clothing items. The uniforms reflecting the image of Hanbok were worn in eight schools and all of them are Saenghwal Hanbok style. The shape of Saenghwal Hanbok school uniforms showed both Korean style and western style characteristics. Korean style characteristics were expressed through the appearance-related parts while western style characteristics showed through the performance-related parts such as open/closure method, armhole pattern and straight sleeve line. It was shown that girls' uniforms are more diverse and similar to the western style uniforms than boys' uniforms. It was found that most of the clothing material were synthetic fiber or mixture of natural/synthetic fiber, and polyester was shown to be used most.

• 한국구조물진단유지관리공학회 논문집
• /
• 제16권1호
• /
• pp.64-77
• /
• 2012

본 연구에서는 디지털 이미지 분석기술을 이용하여 콘크리트 구조물의 대표적인 열화인자인 콘크리트의 균열을 검지할 수 있는 기술을 개발하기 위해 형태수리학에 근거한 모폴로지 기법이 도입된 이미지 분석기법을 적용한 균열검출 시스템을 개발하였다. 개발된 시스템에는 분석결과의 효율적인 관리를 고려하여 촬영된 다수의 이미지를 하나의 전체 이미지로 재구성하는 이미지 조합기술이 추가로 적용되었다. 그리고 개발된 시스템의 적용성 및 신뢰도 검증은 균열이 발생된 콘크리트 암거를 이용한 현장검증 실험을 통해 이루어졌으며, 실험에 사용된 촬영장비로는 40m까지 0.2mm의 균열을 검지할 수 있는 성능을 확인하였다. 이미지 분석을 통해 산정된 균열폭의 경우 실측 균열폭과의 차이가 최대 0.08mm로 나타나 일정 수준 이상의 정확도를 구명하였고, 이미지 조합의 경우 추상 패턴 이미지 면에 대한 분할 촬영 이미지를 조합한 결과 실제 전체촬영 원본 이미지와 육안으로 차이를 확인할 수 없을 정도로 우수한 조합결과를 도출하였다.

• 한국정밀공학회:학술대회논문집
• /
• 한국정밀공학회 2002년도 춘계학술대회 논문집
• /
• pp.263-266
• /
• 2002

An optical method for surface roughness estimation based on statistical analysis of the light intensity of a scattered laser beam pattern. The method is very simple but has a disadvantage that no more information than the averaged roughness is estimated. In this study a new try was conducted to derive more advanced surface information from the details of the light intensity distribution. Some periodic ripples among the light intensity distribution being assumed to relate with scratch left on the machined surface, a corresponding surface profile is estimated from the ripples using FFT and IFFT algorithm. IFFT technique is used to extract some dominant signal components among the intensity distribution. Compared to the measured profiles by a stylus type surf-tester, the profiles obtained through the proposed method are probably acceptable in a sense of the profile shape. Calibration of the amplitude needs more works in the future.

• 한국멀티미디어학회논문지
• /
• 제2권2호
• /
• pp.137-144
• /
• 1999

오차확산법은 계조화상을 이진화상으로 재현하는 것이 우수하지만 이진화상에 상관패턴이 생긴다. 본 논문에서는 오차확산계수의 주파수 분석을 통해서 상관패턴을 제거하고 경계를 강조하는 새로운 오차확산계수를 제안한다. 주목화소의 앞줄은 경계를 강조하도록 확산계수를 정하고 주목화소의 앞화소의 확산계수를 계수들이 대칭이 되도록 정한다. 그리고 제안하는 오차확산계수는 1,2로 구성되어 있기 때문에 계산량이 작다. 실험을 통해서 제안하는 확산계수를 이용한 이진화상의 화질이 기존의 확산계수를 이용한 이진화상보다 우수한 것을 보인다.

• 대한의용생체공학회:학술대회논문집
• /
• 대한의용생체공학회 1997년도 춘계학술대회
• /
• pp.68-71
• /
• 1997

The early detection of brest cancer is clearly a key ingredient for any strategy designed to reduce breast cancer mortality. Microcalcification(MCC) is one of the primary signatures to discriminate between normal and cancerous tissue. The detection and locating procedures can be automated by digital image processing, however, MCCs have various sizes, shapes, and intensity levels in film images, so it is difficult to find accurate locations and sizes. Firstly, we made quantitative analysis for many characteristic features of mammograms that can be used to segment MCCs from normal tissues. Secondly, we developed algorithms proper to segmentation like pattern matching. The performance was evaluated with TP and FP rates.

• 한국정밀공학회:학술대회논문집
• /
• 한국정밀공학회 1997년도 춘계학술대회 논문집
• /
• pp.91-95
• /
• 1997

ESPI(Electronic Speckle Pattern Interferometry) is new optical measuring method to be able to measure the surface deformations of engineering components and materials in industrial areas. Conventional measuring method of surface deformation such as the strain gauge have many demerits because it is contact and point-to-point measuring one. But ESPI that is non-contact, whole field measuring method can overcome previous disadvantages. The speckle pattern to be formed with interference phenomena of scattering light from rough surfaces illuminated by laser light have phase information of surface In this study we used this interference phenomena and the phase shifting method to measure the in- plane deformation, together with the use of digital equipment to process the information contained in the speckle pattern and to display consequent inter ferograms. Finally we obtained good agreement between the experimenta results and those of FEM..