• 한국게임학회 논문지
• /
• 제18권5호
• /
• pp.123-132
• /
• 2018

실시간 그래픽스 응용에서 연체의 움직임을 효율적으로 생성하기 위해 다양한 방법이 제안되었다. 연체 구성 요소들의 위상을 유지하기 위해서는 서로를 묶는 힘이 존재할 수밖에 없으며, 이는 강직도(stiffness)로서 수치적분의 시간간격의 크기를 제한하고 효율성을 떨어트린다. 이를 해결하기 위해 시간간격을 늘릴 수 있는 암시적 적분이 제안되었지만, 대규모 행렬이 포함된 선형시스템을 풀어야 해서 계산복잡도가 크게 높아진다. 이 문제를 개선한 근사 기법들은 댐핑 효과의 증가와 정확성의 손실을 초래할 수밖에 없다. 본 논문에서는 선형시스템을 풀지 않고도 안정성은 크게 높이기 위해 조화진동에 근거하여 스프링 힘을 적분하고, 이를 근사 암시적 기법과 결합하여 안정성을 극대화 하는 방법을 제안한다. 이 기법은 GPU를 통한 병렬화가 용이하여 거대한 규모를 가진 연체 객체의 움직임을 실시간에 생성할 수 있다.

• 한국전자파학회논문지
• /
• 제26권1호
• /
• pp.88-93
• /
• 2015

대규모 물체의 RCS(Radar Cross Section)값을 임의의 각도에서 예측하기 위해, 미리 계산된 샘플링 지점 외 각도에서는 보간법을 사용한다. 대규모 물체의 경우, RCS 데이터베이스를 구성하기 위해 많은 입사각도에서 RCS값을 계산해야 한다. 이는 수치적으로 시간이 많이 걸려, 실질적으로 필요한 모든 입사각에서 RCS값을 미리 계산하기는 어렵다. 그러므로 가능한 적은 샘플을 이용하여 데이터베이스를 구축하고, 이를 보간하여 RCS값을 예측하는 방법이 필요하다. 본 논문에서는 계산된 RCS를 임의의 각도에서 예측하기 위한 보간법으로 Sinc 함수 및 VSH(Vector Spherical Harmonics) 함수를 이용한 방법을 고려하고, 그 정확성을 시뮬레이션을 통하여 검증한다.

• The Journal of Korean Physical Therapy
• /
• 제6권1호
• /
• pp.121-132
• /
• 1994

This study was carried out to know the change of age-related hand function. 210 adults without neurosurgical and orthopedic disability from 15 to 90 years of age participation this study voluntarily. The results are as followings. 1. The hand function decreased according to age increasing. 2. The dominant and non-dominant hand function was decreased in ever-sixty age groups in the subtests of writing, card turning small common objects, simulated feeding, large light objects and large heavy objects greatly, but decreased between forty and fifty age groups in checkers greatly. 3. The dominant hand function was more excellent than non-dominant hand. 4. The dominant and non-dominant hand function was statistically significant between age groups(p<0.01). 5. The one-way ANOVA of subtests according to age increasing revealed significant statistically(p<0.01). 6. The correlation coefficients between subtests and age increasing revealed significant statistically in the dominant and non-dominant hand(p<0.01).

• Current Optics and Photonics
• /
• 제6권6호
• /
• pp.619-626
• /
• 2022

In this study, we develop a three-dimensional (3D) terahertz time-of-flight (THz-TOF) imaging technique with a large depth range, based on asynchronous optical sampling (ASOPS) methods. THz-TOF imaging with the ASOPS technique enables rapid scanning with a time-delay span of 10 ns. This means that a depth range of 1.5 m is possible in principle, whereas in practice it is limited by the focus depth determined by the optical geometry, such as the focal length of the scan lens. We characterize the spatial resolution of objects at different vertical positions with a focal length of 5 cm. The lateral resolution varies from 0.8-1.8 mm within the vertical range of 50 mm. We obtain THz-TOF images for samples with multiple reflection layers; the horizontal and vertical locations of the objects are successfully determined from the 2D cross-sectional images, or from reconstructed 3D images. For instance, we can identify metallic objects embedded in insulating enclosures having a vertical depth range greater than 30 mm. For feasible practical use, we employ the proposed technique to locate a metallic object within a thick chocolate bar, which is not accessible via conventional transmission geometry.

• 대한임베디드공학회논문지
• /
• 제19권2호
• /
• pp.73-81
• /
• 2024

Recently, mobile manipulators have been utilized in agriculture industry for weed removal and harvest automation. This paper proposes a domain adaptive fruit detection method for harvest automation, by utilizing OWL-ViT model which is an open-vocabulary object detection model. The vision-language model can detect objects based on text prompt, and therefore, it can be extended to detect objects of undefined categories. In the development of deep learning models for real-world problems, constructing a large-scale labeled dataset is a time-consuming task and heavily relies on human effort. To reduce the labor-intensive workload, we utilized a large-scale public dataset as a source domain data and employed a domain adaptation method. Adversarial learning was conducted between a domain discriminator and feature extractor to reduce the gap between the distribution of feature vectors from the source domain and our target domain data. We collected a target domain dataset in a real-like environment and conducted experiments to demonstrate the effectiveness of the proposed method. In experiments, the domain adaptation method improved the AP50 metric from 38.88% to 78.59% for detecting objects within the range of 2m, and we achieved 81.7% of manipulation success rate.

• 한국통신학회논문지
• /
• 제32권4C호
• /
• pp.401-410
• /
• 2007

본 논문에서는 비젼 카메라와 다중 객체 추적 방법을 이용한 실시간 수질 감시 시스템을 제안하였다. 제안된 시스템은 기존의 센서 방식의 감시 시스템과 달리 비젼 카메라를 이용해 객체를 개별적으로 분석한다. 비젼 카메라를 이용한 시스템은 영상에서 개별 객체를 분리해 내는 방법과, 연속하는 두 프레임간의 상관관계에 의해서 다수의 객체를 추적하는 방법으로 구성된다. 실시간 처리를 위해 비모수 예측을 사용하여 배경 영상을 생성하고 이를 이용해 객체를 추출한다. 비모수 예측을 이용하면 연산량을 줄이는 동시에 비교적 정확하게 객체를 추출 할 수 있다. 다중 객체 추적 방법은 개별 객체가 움직이는 방향, 속도 및 가속도를 이용해 다음 움직임을 예측하고 이를 기반으로 추적을 수행하였다. 또한 추적 성공률을 향상시키기 위해 예외처리 알고리즘을 적용하였다. 다양한 환경에서 실험한 결과 제안한 시스템은 처리 시간이 짧고 정확하게 다중 객체를 추적할 수 있어 실시간 수질 감시 시스템에 사용이 가능함을 확인하였다.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 1989년도 하계종합학술대회 논문집
• /
• pp.500-503
• /
• 1989

Our work is an attempt to recognize objects without a rigid ordering of steps and with fuller use of previous results as analysis proceeds. We assume that the difference in brightness between the objects and the background is large enough to detect the background boundaries easily. Lines are mostly proposed insted of found by exhaustive search in the scne, the program is relatively efficient.

• 대한인간공학회:학술대회논문집
• /
• 대한인간공학회 1996년도 추계학술대회논문집
• /
• pp.197-205
• /
• 1996

Unification of neural network with a hierarchical pattern recognition is presented for recognizing large set of objects. A two-step identification procedure is developed for pattern recognition: coarse and fine identification. The coarse identification is designed for finding a class of object while the fine identification procedure is to identify a specific object. During the training phase a course neural network is trained for clustering larger set of reference objects into a number of groups. For training a fine neural network, expert neural network is also trained to identify a specific object within a group. The presented idea can be interpreted as two step identification. Experimental results are given to verify the proposed methodology.

• 천문학회지
• /
• 제38권2호
• /
• pp.173-174
• /
• 2005

On behalf of the IRSF/SIRIUS group, I introduce some recent results from our deep near-infrared surveys (J, Hand Ks bands, limiting magnitude of Ks=17) toward star forming regions in the Milky Way Galaxy (MWG) and Large Magellanic Cloud (LMC) with the near-infrared camera SIRIUS. We discovered a rich population of low-mass young stellar objects associated with the W3 and NGC 7538 regions in the MWG based on the near-infrared colors arid magnitudes. The high sensitivity of our survey enables us to detect intermediate-mass pre-main sequence stars, i.e. HAEBE stars, even in the LMC. We detected many HAEBE candidate stars in the N159/N160 complex star forming region in the LMC with the IRSF 1.4-m telescope. Spatial distributions of the young stellar objects indicate the sequential cluster formation in each star forming region in the complex and large scale (a few ${\times}$ 100 pc) sequential cluster formation over the entire complex.

• 한국정밀공학회지
• /
• 제16권2호통권95호
• /
• pp.68-73
• /
• 1999

Generally, most of machining processes are done according to the dimention of the draft made by CAD. However, there are many cases that a sample is given without the draft because of the simplicity of the shape in the machining of 2D objects. To cut the same shape as the given sample, this paper proposes the method to extract the geometric information about a large sample using the robot vision and to draw the demensional draft for the machining. Because the resolution of one frame in the vision system is too low, it is necessary to set up a camera according to the desired resolution and to capture the image moving along the contour. And the overall outline can be compounded of the sequentially captured images. In the experiment, we compared the product after the cutting with the original sample and found that the size of two objects was coincided within the allowed error bound.