• Proceedings of the Korea Information Processing Society Conference
• /
• 2004.11a
• /
• pp.911-914
• /
• 2004

본 논문에서는 주어진 영상으로부터 보다 빠르고 효율적인 의미정보 추출을 위하여 블록분할 및 영역구성에 의한 기본영역 및 확장영역을 제안하며, 각 영역들을 구성하는 블록들의 구성관계에 의한 블록탐색 기법도 제안하고 있다. 기본영역은 영상의 중심을 기반으로 구성되는 중심영역과 이웃영역으로 구성되며, 확장영역은 기본영역들의 결합에 의해 생성된다. 블록탐색은 영역을 구성하는 블록간의 구성관계를 기반으로 블록들이 가질 수 있는 특징들의 유사도와 영역정보에 따라 탐색할 수 있는 방법이다. 얼굴추출은 분할된 블록들로부터 피부색상 존재여부를 판별하여 피부색이 존재하는 블록들로부터 얼굴 후보영역들을 획득한 후, 추출된 후보영역들로부터 얼굴을 구성하는 지역적 특성을 비교평가하여 얼굴을 추출할 수 있다. 또한 추출된 얼굴 영역정보는 연속적인 영상이 주어졌을 때, 해당영역들의 블록들에 대한 정합을 통하여 이동경로와 얼굴영역을 탐색할 수 있다.

• Animal Systematics, Evolution and Diversity
• /
• v.8 no.2
• /
• pp.189-200
• /
• 1992

In the present investigation, some cluster analyses were done for each population of the native honeybee workers(Apis cerana), which were selected for 15 different localities in spring and 16 in summer from Korea. In this analysis, the seasonal segregation was perfectly revealed by both Ward's and the average linkage between groups methods, whereas, it did not necessarily revealed any systematic relationship between geographical groups.

• Annual Conference on Human and Language Technology
• /
• 2022.10a
• /
• pp.105-108
• /
• 2022

오픈 도메인 질의 응답은 사전학습 언어모델의 파라미터에 저장되는 정보만을 사용하여 답하는 질의 응답 방식과 달리 대량의 문서 등에서 질의에 대한 정답을 찾는 문제이다. 최근 등장한 Dense Retrieval은 BERT 등의 모델을 사용해 질의와 문서들의 벡터 연산으로 질의와 문서간의 유사도를 판별하여 문서를 검색한다. 이러한 Dense Retrieval을 활용하는 방안 중 RAG는 Dense Retrieval을 이용한 외부 지식과 인코더-디코더 모델에 내재된 지식을 결합하여 성능을 향상시킨다. 본 논문에서는 RAG를 한국어 오픈 도메인 질의 응답 데이터에 적용하여 베이스라인에 비해 일부 향상된 성능을 보임을 확인하였다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2009.04a
• /
• pp.171-172
• /
• 2009

질감 분류 기술은 패턴인식과 컴퓨터 비전 분야에서 널리 사용되는 기술로서, 최근 들어서는 감시 카메라 시스템에서의 정확한 객체 인식을 위한 회전 변화에 강인한 질감 분류 연구가 진행되고 있다. 본 논문에서는 순환 가보 웨이블렛 필터를 이용한 회전 변환에 강인한 질감 분류 방법을 제안한다. 제안하는 방법은 순환 가보 웨이블렛 필터링된 영상에서 전역 및 지역 특징 벡터를 계산하고 특징 벡터의 차이를 이용한 유사도 측정 판별식으로 질감 분류를 수행한다. Brodatz 질감 앨범을 이용한 실험에서 기존의 방법들보다 2~6% 향상된 질감 분류 비율을 확인할 수 있었다. 제안하는 방법은 질감 기반 객체 인식에 관련된 응용 분야에서 유용하게 사용될 수 있다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2008.05a
• /
• pp.285-288
• /
• 2008

사업에 미치는 환경적 요인들이 점점 다양해짐에 따라 사업평가에 있어서 사업 자체의 성과뿐만 아니라 기업의 환경과 특색까지도 고려할 수 있는 에이전트의 필요성이 부각되고 있다. 본 연구에서는 사업의 타당성 검토와 진행패턴 분석 시 기업의 특수성이 반영된 평가요인과 사업형태의 분류에 따라 에이전트가 각기 다른 방식으로 동작하도록 설계하였다. 본 에이전트는 사업유사성 분석방법으로 판별 분석과 획득가치 접근법을 사용함으로써 유사한 기존사업의 진행패턴과 성과패턴 및 정보를 경영자나 사업담당자에게 보다 직관적으로 제공할 수 있다. 이러한 방법은 경영자로 하여금 기존 시스템 하의 일관된 방식에 의한 평가의 오류를 줄이는 데 도움이 될 것으로 기대된다.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.44 no.5
• /
• pp.719-731
• /
• 2024

For efficient construction management, linking BIM with schedule and cost is essential, but there are limits to the application of 5D BIM due to the difficulty in disassembling thousands of WBS and CBS. To solve this problem, a standardized WBS-CBS set is configured in advance, and when a new construction project occurs, the CBS in the BOQ is automatically linked to the WBS when a text most similar to it is found among the standard CBS (Public Procurement Service standard construction code) of the already linked set. A method was used to compare the text similarity of CBS more efficiently using artificial intelligence natural language processing techniques. Firstly, we created a civil term dictionary (CTD) that organized the words used in civil projects and assigned numerical values, tokenized the text of all CBS into words defined in the dictionary, converted them into TF-IDF vectors, and determined them by cosine similarity. Additionally, the search success rate increased to nearly 70 % by considering CBS' hierarchical structure and changing keywords. The threshold value for judging similarity was 0.62 (1: perfect match, 0: no match).

• Korean journal of applied entomology
• /
• v.47 no.3
• /
• pp.303-308
• /
• 2008

An unidentified moth was captured in sex pheromone traps of the oriental fruit moth, Grapholita molesta, especially at spring season in apple orchards and their vicinity. Though the captured males were similar in appearance to G. molesta males, they were easily distinguished by a matted difference in body size. Their occurrence pattern was also similar to that of overwintering G. molesta population from April to May, at which more males were captured in the pheromone traps installed in the vicinity of apple orchards than within apple orchards. After May, they were no longer captured in the pheromone traps. To investigate any larval damage due to this unidentified moth, molecular markers needed to be developed. Four PCR-RFLP markers originated from cytochrome b region of mitochondrial DNA could distinguish this unidentified moth from G. molesta.

• Journal of Korea Society of Industrial Information Systems
• /
• v.17 no.6
• /
• pp.51-58
• /
• 2012

Scirtothrips Dorsalis (Thysanoptera: Thripidae) recently has been recognized as a major source of the pest damage in the citrus fruit orchards. So its arrival has been predicted periodically but it is difficult to identify adults of the pest with the naked eyes because of their size smaller than the 0.8mm. In this paper, we propose a method to detect candidate areas for automatic identification of Scirtothrips Dorsalis on forecasting traps. The proposed method uses a histogram-based template matching where the composite image synthesized with the gray-scale image and the gradient image is used. In our experiments, images are acquired by the optical microscopy with 50 magnifications. To show the usefulness of the proposed method, it is compared with the method we previously suggested. Also, the performances when the proposed method is applied to noise-reduced images and gradient images are examined. The experimental results show that the proposed method is approximately 14.42% better than our previous method, 41.63% higher than the case that the noise-reduced image is used, and 21.17% higher than the case that the gradient image is used.

• Journal of KIISE:Computing Practices and Letters
• /
• v.15 no.8
• /
• pp.611-615
• /
• 2009

Due to the popular use of digital camera, a great number of photos are taken at every usage of camera. It is essential to reveal relationship between photos to manage digital photos efficiently. We propose a method that tessellates image into unit blocks and applies 2D alignment to extend content-based similar region from seed block pair having high similarity. Through an alignment, we can get a block region scoring best matching value on whole image. The method can distinguish whether photos are sharing the same object or background. Our result is less sensitive to transition or pause change of objects. In experiment, we show how our alignment method is applied to real photo and necessities for further research like photo clustering and massive photo management.

• The Journal of the Convergence on Culture Technology
• /
• v.8 no.6
• /
• pp.867-871
• /
• 2022

Amid the recent rapid trend change, the change in design has a great impact on the sales of fashion companies, so it is inevitable to be careful in choosing new designs. With the recent development of the artificial intelligence field, various machine learning is being used a lot in the fashion market to increase consumers' preferences. To contribute to increasing reliability in the development of new products by quantifying abstract concepts such as preferences, we generate new images that do not exist through three adversarial generative neural networks (GANs) and numerically compare abstract concepts of preferences using pre-trained convolution neural networks (CNNs). Deep convolutional generative adversarial networks (DCGAN), Progressive growing adversarial networks (PGGAN), and Dual Discriminator generative adversarial networks (DANs), which were trained to produce comparative, high-level, and high-level images. The degree of similarity measured was considered as a preference, and the experimental results showed that D2GAN showed a relatively high similarity compared to DCGAN and PGGAN.