• 방송공학회논문지
• /
• 제24권5호
• /
• pp.765-774
• /
• 2019

본 논문에서는 다중 RGB-D 카메라의 포인트 클라우드 정합 알고리즘을 제안한다. 일반적으로 컴퓨터 비전 분야에서는 카메라의 위치를 정밀하게 추정하는 문제에 많은 관심을 두고 있다. 기존의 3D 모델 생성 방식들은 많은 카메라 대수나 고가의 3D Camera를 필요로 한다. 또한 2차원 이미지를 통해 카메라 외부 파라미터를 얻는 기존의 방식은 큰 오차를 가지고 있다. 본 논문에서는 저가의 RGB-D 카메라 8대를 사용하여 전방위 3차원 모델을 생성하기 위해 깊이 이미지와 함수 최적화 방식을 이용하여 유효한 범위 내의 오차를 갖는 좌표 변환 파라미터를 구하는 방식을 제안한다.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2005년도 ICCAS
• /
• pp.2301-2306
• /
• 2005

Even if there are lots of researches on localization using 2D range finder in static environment, very few researches have been reported for robust real-time localization of mobile robot in uncertain and dynamic environment. In this paper, we present a new localization method based on ICP(Iterative Closest Point) algorithm for navigation of mobile robot under dynamic or uncertain environment. The ICP method is widely used for geometric alignment of three-dimensional models when an initial estimate of the relative pose is known. We use the method to align global map with 2D scanned data from range finder. The proposed algorithm accelerates the processing time by uniformly sampling the line fitted data from world map of mobile robot. A data filtering method is also used for threshold of occluded data from the range finder sensor. The effectiveness of the proposed method has been demonstrated through computer simulation and experiment in an office environment.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2009년도 정보 및 제어 심포지움 논문집
• /
• pp.396-398
• /
• 2009

Scoliosis is a three-dimensional deformity caused by lateral curvature of the spine. The existing braces used to correct the posture were some drawbacks such as inconvenience, tightness as well as unfitness to wear. In this study, we devised a posture guidance system in order to monitor a posture continuously and lead to pose correctly and a new method fur measuring a Cobb's angle value in third dimension based on two 3-axis accelerometers. As a result, the correlation coefficients between desired and measured angles were and standard error between desired and measured angles were 0.99, 1.32(x-axis), 0.99 and 1.10(y-axis), respectively. The devised system showed good potential for the optimal posture guide and an early detection of scoliosis.

• ETRI Journal
• /
• 제43권4호
• /
• pp.617-629
• /
• 2021

Although the actual visual simultaneous localization and mapping (SLAM) algorithms provide highly accurate tracking and mapping, most algorithms are too heavy to run live on embedded devices. In addition, the maps they produce are often unsuitable for path planning. To mitigate these issues, we propose a completely closed-loop online dense RGB-D SLAM algorithm targeting autonomous indoor mobile robot navigation tasks. The proposed algorithm runs live on an NVIDIA Jetson board embedded on a two-wheel differential-drive robot. It exhibits lightweight three-dimensional mapping, room-scale consistency, accurate pose tracking, and robustness to moving objects. Further, we introduce a navigation strategy based on the proposed algorithm. Experimental results demonstrate the robustness of the proposed SLAM algorithm, its computational efficiency, and its benefits for on-the-fly navigation while mapping.

• 로봇학회논문지
• /
• 제19권1호
• /
• pp.16-22
• /
• 2024

Robotic grasping in unstructured environments poses a significant challenge, demanding precise estimation of gripping positions for diverse and unknown objects. Generative Grasping Convolution Neural Network (GG-CNN) can estimate the position and direction that can be gripped by a robot gripper for an unknown object based on a three-dimensional depth map. Since GG-CNN uses only a depth map as an input, the precision of the depth map is the most critical factor affecting the result. To address the challenge of depth map precision, we integrate the Segment Anything Model renowned for its robust zero-shot performance across various segmentation tasks. We adjust the components corresponding to the segmented areas in the depth map aligned through external calibration. The proposed method was validated on the Cornell dataset and SurgicalKit dataset. Quantitative analysis compared to existing methods showed a 49.8% improvement with the dataset including surgical instruments. The results highlight the practical importance of our approach, especially in scenarios involving thin and metallic objects.

• Structural Engineering and Mechanics
• /
• 제42권6호
• /
• pp.831-847
• /
• 2012

Impact from water-borne debris during tsunami and flood events pose a potential threat to structures. Debris impact forces specified by current codes and standards are based on rigid body dynamics, leading to forces that are dependent on total debris mass. However, shipping containers and other debris are unlikely to be rigid compared to the walls, columns and other structures that they impact. The application of a simple one-dimensional model to obtain impact force magnitude and duration, based on acoustic wave propagation in a flexible projectile, is explored. The focus herein is on in-air impact. Based on small-scale experiments, the applicability of the model to predict actual impact forces is investigated. The tests show that the force and duration are reasonably well represented by the simple model, but they also show how actual impact differs from the ideal model. A more detailed three-dimensional finite element model is also developed to understand more clearly the physical phenomena involved in the experimental tests. The tests and the FE results reveal important characteristics of actual impact, knowledge of which can be used to guide larger scale experiments and detailed modeling. The one-dimensional model is extended to consider water-driven debris as well. When fluid is used to propel the 1-D model, an estimate of the 'added mass' effect is possible. In this extended model the debris impact force depends on the wave propagation in the two media, and the conditions under which the fluid increases the impact force are discussed.

• 전기학회논문지
• /
• 제64권6호
• /
• pp.900-911
• /
• 2015

In this paper, 3D face recognition model is designed by using Polynomial based RBFNN(Radial Basis Function Neural Network) and PNN(Polynomial Neural Network). Also recognition rate is performed by this model. In existing 2D face recognition model, the degradation of recognition rate may occur in external environments such as face features using a brightness of the video. So 3D face recognition is performed by using 3D scanner for improving disadvantage of 2D face recognition. In the preprocessing part, obtained 3D face images for the variation of each pose are changed as front image by using pose compensation. The depth data of face image shape is extracted by using Multiple point signature. And whole area of face depth information is obtained by using the tip of a nose as a reference point. Parameter optimization is carried out with the aid of both ABC(Artificial Bee Colony) and PSO(Particle Swarm Optimization) for effective training and recognition. Experimental data for face recognition is built up by the face images of students and researchers in IC&CI Lab of Suwon University. By using the images of 3D face extracted in IC&CI Lab. the performance of 3D face recognition is evaluated and compared according to two types of models as well as point signature method based on two kinds of depth data information.

• Molecular & Cellular Toxicology
• /
• 제4권4호
• /
• pp.307-310
• /
• 2008

AAD16034 is an alginate lyase from Pseudoalteromonas sp. IAM14594. A very close homologue with known 3D structure exists (marine bacterium Pseudoalteromonas sp. strain no. 272). A three-dimensional structure of AAD16034 was generated based on this template (PDB code: 1J1T) by comparative modeling. The modeled enzyme exhibited a jelly-roll like structure very similar to its template structure. Both enzymes possess the characteristic alginate sequence YFKhG+Y-Q. Since AAD16034 displays enzymatic activity for poly-M alginate, docking of a tri-mannuronate into the modeled structure was performed. Two separate and adjacent binding sites were found. The ligand was accommodated inside each binding site. By considering both binding sites, a plausible binding pose for the poly-M alginate polymer could be deduced. From the modeled docking pose (i.e., the most important factor that attracts alginate polymer into this lyase) the most likely interaction was electrostatic. In accordance with a previous report, the hydroxyl group of Y345 was positioned close to the ${\alpha}$-hydrogen of ${\beta}$-mannuronate, which was suitable to initiate a ${\beta}$-elimination reaction. K347 was also very near to the carboxylatemoiety of the ligand, which might stabilize the dianion intermediate during the ${\beta}$-elimination reaction. This implies that the characteristic alginate sequence is absolutely crucial for the catalysis. These results may be exploited in the design of novel enzymes with desired properties.

• 방송공학회논문지
• /
• 제26권5호
• /
• pp.519-532
• /
• 2021

최근 게임, 영화, 애니메이션 다양한 분야에서 모션 캡처를 이용하여 신체 모델을 구축하고 캐릭터를 생성하여 3차원 공간에 표출하는 콘텐츠가 증가하고 있다. 마커를 부착하여 관절의 위치를 측정하는 방법에서 촬영 장비에 대한 비용과 같은 문제를 보완하기 위해 RGB-D 카메라를 이용하여 애니메이션을 생성하는 연구가 진행되고 있지만, 관절 추정 정확도나 장비 비용의 문제가 여전히 존재한다. 이에 본 논문에서는 애니메이션 생성에 필요한 장비 비용을 줄이고 관절 추정 정확도를 높이기 위해 RGB 이미지를 관절 추정 네트워크에 입력하고, 그 결과를 3차원 데이터로 변환하여 FBX 형식 애니메이션으로 생성하는 시스템을 제안한다. 먼저 RGB 이미지에 대한 2차원 관절을 추정하고, 이 값을 이용하여 관절의 3차원 좌표를 추정한다. 그 결과를 쿼터니언으로 변환하여 회전한 후, FBX 형식의 애니메이션을 생성한다. 제안한 방법의 정확도 측정을 위해 신체에 마커를 부착하여 마커의 3차원 위치를 바탕으로 생성한 애니메이션과 제안된 시스템으로 생성한 애니메이션의 오차를 비교하여 시스템 동작을 입증하였다.

• 한국콘텐츠학회논문지
• /
• 제17권1호
• /
• pp.137-144
• /
• 2017

얼굴 검출은 복잡한 배경 내에서 다양한 얼굴의 자세로 인해 여전히 어려운 문제에 직면하고 있다. 본 논문은 피부색과 깊이 정보를 기반으로 한 한명 또는 여러 명의 얼굴을 검출하는 효과적인 알고리즘을 제안한다. 먼저 우리는 컬러 영상에서 가우시안 혼합 모델을 이용한 피부색 검출 방법에 대해 소개한다. 그리고 Kinect V2의 깊이 센서를 이용하여 획득한 3차원의 깊이 정보는 배경으로부터 사람의 몸을 분할할 때 유용하다. 그리고 레이블링 과정에서 여러 개의 특징을 이용하여 얼굴이 아닌 영역은 성공적으로 제거된다. 실험 결과를 통해 제안한 얼굴 검출 알고리즘은 다양한 조건과 복잡한 배경에서 얼굴이 효과적으로 검출되는 것을 확인할 수 있다.