• Proceedings of the Korean Information Science Society Conference
• /
• 2010.06c
• /
• pp.507-511
• /
• 2010

휴먼 행동 인식은 크게 3D 모델 기반 방법과 템플릿 기반 방법으로 나눌 수 있다. 3D 모델 기반 방법은 휴먼의 포즈를 3D로 재구성한 뒤 특징을 추출하는 것으로 인식 정확도는 높으나 연산량이 많아 매우 비효율적이다. 반면 템플릿 기반의 방법은 간단하고 수행 시간이 빠르기 때문에 여러 논문들에서 채택되고 있다. 그러나 템플릿을 이용한다는 특성 때문에 시점, 행동 스타일의 변화 등에 따라 실루엣의 변화가 심해 인식 성능에 한계점을 가진다. 본 논문에서는 핵심-포즈들의 히스토그램으로 표현되는 핵심-포즈 분포와 광류의 변화를 이용하여 다중 시점에서의 휴먼 행동 인식 방법을 제안한다. 제안하는 방법은 IXMAS 데이터 셋을 이용한 실험에서 적은 수의 템플릿을 이용하면서도 평균 87.9%의 높은 인식률을 보였다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2012.11a
• /
• pp.355-358
• /
• 2012

본 논문은 사용자의 포즈를 인터페이스로 사용하는 가상 비행 체험 콘텐츠에 대하여 기술한다. 사용자의 포즈를 인식하기 위해서 제스처를 구성하는 상반신의 포즈를 식별하여야 한다. 본 논문에서 기술한 콘텐츠는 한정된 공간에서 사용자의 움직임을 인식하고 가상공간에 아바타를 이용하여 표현하고 있다. 그러므로 사용자는 가상공간에서 정의된 포즈를 사용하여 가상 비행을 체험할 수 있고 인식된 포즈는 OS-Value 이벤트를 이용하여 가상 비행 체험 콘텐츠에서 인터페이스로 활용이 가능하다.

• International Journal of Advanced Culture Technology
• /
• v.9 no.4
• /
• pp.302-306
• /
• 2021

Indoor localization is one of the basic elements of Location-Based Service, such as indoor navigation, location-based precision marketing, spatial recognition of robotics, augmented reality, and mixed reality. We propose a Voxel Labeling-based visual positioning system using object simultaneous localization and mapping (SLAM). Our method is a method of determining a location through single image 3D cuboid object detection and object SLAM for indoor navigation, then mapping to create an indoor map, addressing it with voxels, and matching with a defined space. First, high-quality cuboids are created from sampling 2D bounding boxes and vanishing points for single image object detection. And after jointly optimizing the poses of cameras, objects, and points, it is a Visual Positioning System (VPS) through matching with the pose information of the object in the voxel database. Our method provided the spatial information needed to the user with improved location accuracy and direction estimation.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.11a
• /
• pp.678-680
• /
• 2023

본 연구는 2022년 유네스코 인류무형유산 대표목록에 등재된 탈춤 동작을 디지털화하여 후속 세대에게 정보를 제공하는 것을 목적으로 한다. 데이터 수집은 국가무형문화제로 지정된 탈춤 단체 13개, 시도무형문화재 단체 5개에 소속된 무형문화재, 전승자 39명이 관성식 모션 캡처 장비를 착용하고, 8대의 카메라를 이용하여 수집하였다. 데이터 가공은 바운딩박스를 수행하였고, 탈춤동작 추정은 YOLO v8을 사용하였고 탈춤 동작 분류는 YOLO v8에 CNN모델을 결합하여 130개의 탈춤을 분류하였다. 연구결과, mAP-50은 0.953, mAP50-95는 0.596, Accuracy 70%를 달성하였다. 향후 학습용 데이터셋 구축량이 늘어나고, 데이터 품질이 개선된다면 탈춤 분류 성능은 더욱 개선될 것이라 기대한다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2008.05a
• /
• pp.11-12
• /
• 2008

본 논문은 이차원 영상에 투영된 삼차원 인체의 포즈를 인식하기 위하여 이차원 영상에 투영된 인체의 실루엣 정보를 이용하였다. 인체는 삼차원 공간에서 움직이므로 이차원 영상으로 모든 정보를 알아내기에는 부족한 면이 있다. 따라서 본 논문에서는 인체 포즈의 주시 방향을 결정한 후 인체의 실루엣 영상 Convex-hull 특징점 정보를 이용하여 인체의 삼차원 포즈를 인식하였다. 인체의 포즈는 PCA로 차원을 축소하였으며 Diffusion Distance로 데이터베이스의 포즈모델 중 가장 가까운 모델을 선택하였다.

• 한국HCI학회:학술대회논문집
• /
• 2008.02a
• /
• pp.342-347
• /
• 2008

Face tracking and recognition are difficult problems because the face is a non-rigid object. If the system tries to track or recognize the unknown face continuously, it can be more hard problems. In this paper, we propose the method to track and to recognize the face of the unknown person on video sequences using linear combination of nonlinear manifold models that is constructed in the system. The arbitrary input face has different similarities with different persons in system according to its shape or pose. Do we can approximate the new nonlinear manifold model for the input face by estimating the similarities with other faces statistically. The approximated model is updated at each frame for the input face. Our experimental results show that the proposed method is efficient to track and recognize for the arbitrary person.

• The KIPS Transactions:PartB
• /
• v.16B no.5
• /
• pp.347-358
• /
• 2009

In this thesis, we develop a vision based hand motion recognition system using a camera with two rotational motors. Existing systems were implemented using a range camera or multiple cameras and have a limited working area. In contrast, we use an uncalibrated camera and get more wide working area by pan-tilt motion. Given an image sequence provided by the pan-tilt camera, color and pattern information are integrated into a tracking system in order to find the 2D position and direction of the hand. With these pose information, we estimate 3D motion plane on which the gesture motion trajectory from approximately forms. The 3D trajectory of the moving finger tip is projected into the motion plane, so that the resolving power of the linear gesture patterns is enhanced. We have tested the proposed approach in terms of the accuracy of trace angle and the dimension of the working volume.

• 제어로봇시스템학회:학술대회논문집
• /
• 2005.06a
• /
• pp.1202-1205
• /
• 2005

This paper describes a method for vision-based person identification that can detect, track, and recognize person from video using multiple cues: height and dressing colors. The method does not require constrained target's pose or fully frontal face image to identify the person. First, the system, which is connected to a pan-tilt-zoom camera, detects target using motion detection and human cardboard model. The system keeps tracking the moving target while it is trying to identify whether it is a human and identify who it is among the registered persons in the database. To segment the moving target from the background scene, we employ a version of background subtraction technique and some spatial filtering. Once the target is segmented, we then align the target with the generic human cardboard model to verify whether the detected target is a human. If the target is identified as a human, the card board model is also used to segment the body parts to obtain some salient features such as head, torso, and legs. The whole body silhouette is also analyzed to obtain the target's shape information such as height and slimness. We then use these multiple cues (at present, we uses shirt color, trousers color, and body height) to recognize the target using a supervised self-organization process. We preliminary tested the system on a set of 5 subjects with multiple clothes. The recognition rate is 100% if the person is wearing the clothes that were learned before. In case a person wears new dresses the system fail to identify. This means height is not enough to classify persons. We plan to extend the work by adding more cues such as skin color, and face recognition by utilizing the zoom capability of the camera to obtain high resolution view of face; then, evaluate the system with more subjects.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.50 no.10
• /
• pp.171-180
• /
• 2013

Generally, when we apply zernike moments value for matching, we can use those moments value obtained from projecting image information under circumscribed circle to zernike basis function. However, the problem is that the power of discrimination can be reduced because hand images include lots of overlapped information due to its special characteristic. On the other hand, when distinguishing hand poses, information in specific area of image information except for overlapped information can increase the power of discrimination. In this paper, in order to solve problems like those, we design R3 ring mask by combining image obtained from R2 ring mask, which can weight information of the power of discrimination and image obtained from R1 ring mask, which eliminate the overlapped information. The moments which are obtained by R3 ring mask decrease operational time by reducing dimension through principle component analysis. In order to confirm the superiority of the suggested method, we conducted some experiments by comparing our method to other method using seven different hand poses.

• KSCI Review
• /
• v.14 no.1
• /
• pp.241-248
• /
• 2006

The recognition of a relatively big and rarely movable object. such as refrigerator and air conditioner, etc. is necessary because these objects can be crucial global stable features of Simultaneous Localization and Map building(SLAM) in the indoor environment. In this paper. we propose a novel method to recognize these big objects using a sequence of 3D scenes. The particles representing an object to be recognized are scattered to the environment and then the probability of each particles is calculated by the matching test with 3D lines of the environment. Based on the probability and degree of convergence of particles, we can recognize the object in the environment and the pose of object is also estimated. The experimental results show the feasibility of incremental object recognition based on particle filtering and the application to SLAM