• Journal of Digital Convergence
• /
• v.14 no.11
• /
• pp.319-324
• /
• 2016

In the case of robot and Argument Reality applications using a camera in environments, a technology to estimate planes is a very important technology. A RGB-D camera can get a three-dimensional measurement data even in a flat which has no information of the texture of the plane;, however, there is an enormous amount of computation in order to process the point-cloud data of the image. Furthermore, it could not know the number of planes that are currently observed as an advance, also, there is an additional operation required to estimate a three dimensional plane. In this paper, we proposed the real-time method that decides the number of planes automatically and estimates the three dimensional plane by using the continuous data of an RGB-D camera. As experimental results, the proposed method showed an improvement of approximately 22 times faster speed compared to processing the entire data.

• The Journal of the Korea Contents Association
• /
• v.22 no.3
• /
• pp.54-61
• /
• 2022

Ship is a huge system including a variety of pipe arrangements. Pipes are installed according to the design layout, however the end poistion of pipes are not well matched owing to its measurement and construction errors. In this situation, the customized pipe fitting is frequently designed to connect with both pipes, the position of which are manually measured. This paper focused that these two coordinates are measured by point cloud from RGBD sensor and the relative transformation induced by positional and orientational differences is calculated by inverse kinematics in robotics theory. Therefore, the result applies for the methodology of the pipe connection design. The pipe coordinate that is estimated by the matching and the probabilistic RANSAC method will be verified by experiments. The kinematic design parameters are computationally calculated by using the minimum degree of freedom that connects both pipe coordinates.

• Journal of Korea Multimedia Society
• /
• v.20 no.8
• /
• pp.1187-1199
• /
• 2017

In recent years, research about Natural User Interface (NUI) has become focused because NUI system can give natural feelings for users in virtual reality. Most important thing in NUI system is how to communicate with the computer system. There are many things to interact with users such as speech, hand gestures, body actions. Among them, hand gesture is suitable for the purpose of NUI because people often use a relatively high frequency in daily life and hand gesture have meaning only by itself. This hand gestures called multi-cultural hand gesture and we proposed the method to recognize this kind of hand gestures. Proposed method is composed of Histogram of Oriented Gradients (HOG) used for hand shape recognition and Histogram of Oriented Displacements (HOD) used for hand center point trajectory recognition.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2012.07a
• /
• pp.311-313
• /
• 2012

본 논문에서는 RGBD(Red Green Blue Depth)센서를 이용하여 얻은 영상의 깊이 정보와 손 이미지의 기하학적 특징을 이용하여 손의 중심을 검출하는 방법을 제안한다. 영상의 깊이 정보와 피부색 정보를 이용하여 손 영역을 검출한다. 검출된 손의 기하학적 정보로 손에 대한 볼록 외피(convex hull)를 형성한다. 볼록 외피의 정점들(vertices)의 위치 정보를 이용하여 손의 중심을 찾는다. 손의 중심은 손의 위치를 추적하거나 손가락 개수를 구하는 것 등에 이용될 수 있다. 이러한 응용은 인간과 컴퓨터의 상호작용(HCI, Human Computer Interface)을 이용한 시스템에 적용될 수 있다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2019.10a
• /
• pp.1101-1104
• /
• 2019

키넥트는 RGBD 카메라로 인체의 뼈대와 관절을 3D 공간에서 스켈레톤 데이터수집을 가능하게 해주었다. 스켈레톤 데이터를 활용한 행동 분류는 RNN, CNN 등 다양한 인공 신경망으로 접근하고 있다. 본 연구는 키넥트를 이용해서 Skeleton Joints를 수집하고, DNN 기반 스켈레톤 모델링 학습으로 행동을 분류한다. Skeleton Joints Processing 과정은 키넥트의 Depth Map 기반의 Skeleton Tracker로 25가지 Skeleton Joints 좌표를 얻고, 학습을 위한 전처리 과정으로 각 좌표를 상대좌표로 변경하고 데이터 수를 제한하며, Joint가 트래킹 되지 않은 부분에 대한 예외 처리를 수행한다. 스켈레톤 모델링 학습 과정에선 3계층의 DNN 신경망을 구축하고, softmax_cross_entropy 함수로 Skeleton Joints를 집는 모션, 내려놓는 모션, 팔짱 낀 모션, 얼굴을 가까이 가져가는 모션 해서 4가지 행동으로 분류한다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2012.11a
• /
• pp.655-657
• /
• 2012

본 논문에서는 심도(Depth) 카메라로부터 실시간 획득한 RGBD 데이터에서 심도 정보 기반의 AAM(Active Appearance Models)과 나이 인식 알고리즘[1]을 통해 4 개의 AG(Age Group)으로 분류하는 실시간 얼굴 나이 인식 시스템(Real-time Facial Age Recognition System)을 설계한다. 기존의 AAM 을 이용한 실시간 얼굴 특징 추출은 평균 약 4.17%의 프레임 손실율을 보였으나, 심도 정보를 활용한 AAM 은 평균 약 0.43%의 프레임 손실율만을 보였다[5]. 본 논문에서는 심도 정보를 활용한 AAM과 병렬 처리 방법인 CUDA 를 결합하여 나이 특징을 추출하고, 실시간 시스템에 적용 가능하도록 나이 인식 알고리즘을 개선하여 실시간 나이 인식 시스템을 설계한다. 설계된 시스템은 1)머리 위치 추적, 2)얼굴 인식 및 특징점 추출, 3)나이 특징 추출, 4) 나이 특징 분석, 5) 나이 분류의 5 가지 단계를 통해 최종적으로 4 개의 AG 로 분류한다.

• Journal of Information Processing Systems
• /
• v.18 no.6
• /
• pp.794-802
• /
• 2022

The recovery of reasonable depth information from different scenes is a popular topic in the field of computer vision. For generating depth maps with better details, we present an efficacious monocular depth prediction framework with coordinate attention and feature fusion. Specifically, the proposed framework contains attention, multi-scale and feature fusion modules. The attention module improves features based on coordinate attention to enhance the predicted effect, whereas the multi-scale module integrates useful low- and high-level contextual features with higher resolution. Moreover, we developed a feature fusion module to combine the heterogeneous features to generate high-quality depth outputs. We also designed a hybrid loss function that measures prediction errors from the perspective of depth and scale-invariant gradients, which contribute to preserving rich details. We conducted the experiments on public RGBD datasets, and the evaluation results show that the proposed scheme can considerably enhance the accuracy of depth prediction, achieving 0.051 for log10 and 0.992 for δ<1.25³ on the NYUv2 dataset.

• Journal of the Korea Computer Graphics Society
• /
• v.21 no.5
• /
• pp.37-44
• /
• 2015

This paper presents a novel 3D model representation, called hybrid model representation, to overcome existing 3D volume-based indoor scene reconstruction mechanism. In indoor 3D scene reconstruction, volume-based model representation can reconstruct detailed 3D model for the narrow scene. However it cannot reconstruct large-scale indoor scene due to its memory consumption. This paper presents a memory efficient plane-hash model representation to enlarge the scalability of the indoor scene reconstruction. Also, the proposed method uses plane-hash model representation to reconstruct large, structural planar objects, and at the same time it uses volume-based model representation to recover small detailed region. Proposed method can be implemented in GPU to accelerate the computation and reconstruct the indoor scene in real-time.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.24 no.4
• /
• pp.607-616
• /
• 2014

Face recognition is a big challenge in surveillance system since different rotation angles of the face make the difficulty to recognize the face of the same person. This paper proposes a novel method to recognize face with different head poses by using 3D information of the face. Firstly, head pose estimation (estimation of different head pose angles) is accomplished by the POSIT algorithm. Then, 3D face image data is constructed by using head pose estimation. After that, 2D image and the constructed 3D face matching is performed. Face verification is accomplished by using commercial face recognition SDK. Performance evaluation of the proposed method indicates that the error range of head pose estimation is below 10 degree and the matching rate is about 95%.