• Journal of Korea Multimedia Society
• /
• v.19 no.9
• /
• pp.1659-1668
• /
• 2016

Understanding 3D structure of scenes is of a great interest in various vision-related tasks. In this paper, we present a unified approach for estimating depth from a single monocular image. The key idea of our approach is to take advantages both of parametric learning and non-parametric sampling method. Using a parametric convolutional network, our approach learns the relation of various monocular cues, which make a coarse global prediction. We also leverage the local prediction to refine the global prediction. It is practically estimated in a non-parametric framework. The integration of local and global predictions is accomplished by concatenating the feature maps of the global prediction with those from local ones. Experimental results demonstrate that the proposed method outperforms state-of-the-art methods both qualitatively and quantitatively.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2019.10a
• /
• pp.1002-1003
• /
• 2019

Face recognition is an extensively researched area of computer vision. Visible, infrared, thermal, and 3D modalities have been used against various challenges of face recognition such as illumination, pose, expression, partial information, and disguise. In this paper we present a multi-modal approach to face recognition using convolutional neural networks. We use visible and thermal face images as two separate inputs to a multi-input deep learning network for face recognition. The experiments are performed on IRIS visible and thermal face database and high face verification rates are achieved.

• Korean Journal of Remote Sensing
• /
• v.23 no.5
• /
• pp.439-445
• /
• 2007

Stereoscopic viewing is an efficient technique for not only computer vision but also remote sensing applications. Generally, stereo pair obtained at the same time is necessary for 3D viewing, but it is possible to synthesize a stereomate suitable for stereo view with a single image and disparity-map. There have been researches concerning the generation of the synthetic stereomate from remote sensing imagery. However it is hard to find researches concerning the restoration of occlusion in stereomate. In this paper, we generated synthetic stereomates from remote sensing images, focused on the occlusion restoration. In order to figure out proper restoration methods depending on the spatial resolution of remote sensing imagery, we tested several methods including general interpolation and inpainting technique, then evaluated the results.

• International Journal of Internet, Broadcasting and Communication
• /
• v.13 no.4
• /
• pp.90-96
• /
• 2021

The logistics market is continuously growing due to the development of technology and the growth of the online market. In addition, the social atmosphere that emphasizes non-face-to-face due to the pandemic situation is accelerating the growth of logistics. Delivery of goods ordered online requires delivery process through courier worker. In order for the courier worker to ship the product, the work of loading the product on the truck must be preceded. The accident caused by such delivery and loading work is increasing and it is emerging as a social problem. This study proposes a robot-based automated loading system to efficiently handle the increasing volume of courier service and to construct a more efficient and safe working environment by replacing the physical labor that was overloaded to courier workers. The proposed system replaces the loading of the courier worker and proposes the optimal loading function through the automation system.

• Journal of the Korean Society of Mechanical Technology
• /
• v.20 no.6
• /
• pp.790-795
• /
• 2018

In this study, we analyzed the structural safety and vibration characteristics of rotational drive in 3D CT scan equipment using finite element analysis. The analysis results showed a safety factor of 9.2 and a left and right vertical deflectional deviation of 0.24mm from the maximum equivalent stress. After applying weight compensation of 27.7kgf, the structural analysis reduced the safety factor to 7.6, but the deflectional deviation of the left and right structure was reduced to 0mm. Also, we presented the optimum design of rotational drive through the vibration analysis.

• The Journal of Korea Robotics Society
• /
• v.16 no.1
• /
• pp.29-34
• /
• 2021

Traditional actuation system such as electric and pneumatic actuator has obvious advantages and disadvantages. To combine advantages and compensate disadvantages of the traditional actuation, a pneumatic actuation system with an internal air pressure source is noteworthy approach. In this paper, a soft pneumatic actuation system based on origami pump is described for haptic feedback glove. To improve wearability, an origami pump is introduced because the origami pump is much lighter than air compressor. The miniaturized electric actuation system is also designed with 3D printed planetary gear in order to reduce the volume of the system. To figure out the performance of the system, shrinkage distance of origami pump was measured with vision camera. The pressure in the origami pump was also estimated to understand the performance of the system.

• International journal of advanced smart convergence
• /
• v.11 no.1
• /
• pp.1-10
• /
• 2022

Single-view reconstruction (SVR) is a fundamental method in computer vision. Often used for reconstructing human-made environments, the Manhattan world assumption presumes that planes in the real world exist in mutually orthogonal directions. Accordingly, this paper addresses an automatic SVR algorithm for Manhattan worlds. A method for estimating the directions of planes using graph-cut optimization is proposed. After segmenting an image from extracted line segments, the data cost function and smoothness cost function for graph-cut optimization are defined by considering the directions of the line segments and neighborhood segments. Furthermore, segments with the same depths are grouped during a depth-estimation step using a minimum spanning tree algorithm with the proposed weights. Experimental results demonstrate that, unlike previous methods, the proposed method can identify complex Manhattan structures of indoor and outdoor scenes and provide the exact boundaries and intersections of planes.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2020.11a
• /
• pp.992-995
• /
• 2020

시각-언어 이동 작업은 에이전트가 주어진 지시를 따라 특정 실내 공간 내에서 목적 위치로 이동하는 작업이다. 시각-언어 이동 작업의 특성상 자연어 지시 속에 등장하는 랜드마크인 장소 정보를 인지하는 것은 작업을 수행하는 데 큰 도움이 된다. 본 논문에서는 환경을 구성하는 주요 장소 정보를 저장하기 위한 장소 미리보기 메모리를 제안한다. 에이전트는 장소 미리보기 메모리에 저장된 장소 정보를 고려하여 작업을 수행하게 된다. 본 논문에서는 Matterport3D 시뮬레이션 환경에서의 실험을 통해 R2R 벤치마크 데이터 집합에서 가장 높은 성능을 보였다.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.07a
• /
• pp.269-271
• /
• 2020

In this letter, we propose a novel approach for stitching stereoscopic panoramas. When stitching stereoscopic panoramas, the amount of depth retrieved is the most important factor to pay attention for. Also, it is very crucial to deliver the two left and right panoramas with the right depth information to deliver good 3D perception. However, when stitching the two panoramas independently using the state-of-the-art algorithms and methods, we do still have some inconsistencies with the disparity map retrieved from the panoramas. To overcome this problem, we propose a method that modifies the latest conventional algorithm by making the two panoramas dependent of one another. This brings two panoramas with a much more consistent disparity map that lets users fully immerse into a comfortable stereoscopic vision.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2022.06a
• /
• pp.146-149
• /
• 2022

기존 RDO(Rate Distortion Optimization) 기반 압축 방식은 압축 성능에 초점을 두기 때문에 영상 내 인지 특성이 무시될 수 있다. 따라서 RoI(Region of Interest)을 기반으로 압축률을 조절하는 연구가 고안[1, 2, 3, 4] 되었으며, HVS(Human Visual System) 관점에서 영상 내 중요한 부분에 대해 더 높은 품질로 영상을 압축하는 연구가 대부분이다. 최근 인공지능 기술이 발전함에 따라 지능형 영상 분석에 대한 수요가 증가하고 있으며, 이에 따라 머신 비전을 위한 영상 부호화 및 효율적인 전송에 대한 필요성이 대두되고 있다. 본 논문에서는 VVC(Versatile Video Coding)의 dQP(delta Quantization Parameter)를 활용하여 RoI(Region of Interest) 기반압축 방법을 제안하고, 두가지의 RoI 추출 방식을 소개한다. Detectron2 Faster R-CNN X101-FPN [5]의 첫번째 탐지기를 통해 후보 영역 기반 RoI 을 추출하고, 두번째 탐지기를 통해 객체 기반 RoI 을 추출하여, 영상 내 객체 부분과 비객체 부분으로 나누어 서로 다른 압축률로 압축을 수행하였으며, 이에 따른 성능을 비교하고자 한다.