• ETRI Journal
• /
• 제44권1호
• /
• pp.62-72
• /
• 2022

In this study, we propose a multi-GPU-based 8KVR stitching system that operates in real time on both local and cloud machine environments. The proposed system first obtains multiple 4 K video inputs, decodes them, and generates a stitched 8KVR video stream in real time. The generated 8KVR video stream can be downloaded and rendered omnidirectionally in player apps on smartphones, tablets, and head-mounted displays. To speed up processing, we adopt group-of-pictures-based distributed decoding/encoding and buffering with the NV12 format, along with multi-GPU-based parallel processing. Furthermore, we develop several algorithms such as equirectangular projection-based color correction, real-time CG overlay, and object motion-based seam estimation and correction, to improve the stitching quality. From experiments in both local and cloud machine environments, we confirm the feasibility of the proposed 8KVR stitching system with stitching speed of up to 83.7 fps for six-channel and 62.7 fps for eight-channel inputs. In addition, in an 8KVR live streaming test on the 5G MEC/cloud, the proposed system achieves stable performances with 8 K@30 fps in both indoor and outdoor environments, even during motion.

• 한국철도학회:학술대회논문집
• /
• 한국철도학회 2010년도 춘계학술대회 논문집
• /
• pp.533-539
• /
• 2010

2차원 카메라와 라인 레이저를 이용한 비접촉식 레일 마모도 측정 시스템은 레일의 정확한 마모도 측정을 위하여 레일에 조사된 레이저 영역을 왜곡 없이 획득하고 이를 레일 단면과 정확하게 일치시켜야 한다. 그러나 일반적인 비접촉식 레일 마모도 측정 시스템의 카메라는 레일의 측면에서 비스듬하게 레일을 촬영하여 사선방향(Oblique Viewpoint)의 원시영상을 획득하는 방식을 취하기 때문에 카메라 모델(Camera Model)과 투시 변환(Perspective Transform)을 이용하여 카메라로부터 획득한 원시영상이나 영상으로부터 추출된 레이저 영역의 좌표 데이터를 레일의 단면과 일치시키기 위한 영상 후처리 작업을 수행하게 된다. 이때 라인 레이저가 조사된 지점에서 획득한 레일 프로파일 데이터는 전체 측정 구간에 대하여 일정한 간격으로 측정된 불연속 데이터의 특성을 가지므로 측정된 각 레일 프로파일 데이터 사이의 영역에서는 정확한 레일의 마모도를 측정할 수 없는 한계를 가진다. 본 논문에서는 레일의 양방향에서 획득한 영상에 대하여 3차원 레일 프로파일 생성을 위한 변형된 카메라 모델과 투시 변환을 적용함으로서 레일 마모도의 측정 정밀도를 향상하기 위한 3차원 레일 프로파일 재구성 기법을 제안하고자 한다. 제안된 3차원 레일 프로파일 재구성 기법을 통하여 전체 측정 구간내에 포함된 레일 마모도 측정 데이터의 불연속성을 효과적으로 제거하여 연속적인 측정면(Measurement Surface)을 생성함으로서 측정하고자 하는 전체 레일에 대한 향상된 측정 정밀도를 제공하고자 한다.

• 제어로봇시스템학회논문지
• /
• 제5권5호
• /
• pp.550-557
• /
• 1999

This paper describes a road-following controller using the proposed neural network for autonomous vehicle. Road-following with visual sensor like camera requires intelligent control algorithm because analysis of relation from road image to steering control is complex. The proposed neural network, relative similarity modular network(RSMN), is composed of some learning networks and a partitioniing network. The partitioning network divides input space into multiple sections by similarity of input data. Because divided section has simlar input patterns, RSMN can learn nonlinear relation such as road-following with visual control easily. Visual control uses two criteria on road image from camera; one is position of vanishing point of road, the other is slope of vanishing line of road. The controller using neural network has input of two criteria and output of steering angle. To confirm performance of the proposed neural network controller, a software is developed to simulate vehicle dynamics, camera image generation, visual control, and road-following. Also, prototype autonomous electric vehicle is developed, and usefulness of the controller is verified by physical driving test.

• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송공학회 2009년도 IWAIT
• /
• pp.732-736
• /
• 2009

In this paper we focus on EPI (Epipolar-Plane Image), the horizontal cross section of Ray-Space, and we propose a novel method that chooses objects we want and edits scenes by using multi-view images. On the EPI acquired by camera arrays uniformly distributed along a line, all the objects are represented as straight lines, and the slope of straight lines are decided by the distance between objects and camera plane. Detecting a straight line of a specific slope and removing it mean that an object in a specific depth has been detected and removed. So we propose a scheme to make a layer of a specific slope compete with other layers instead of extracting layers sequentially from front to back. This enables an effective removal of obstacles, object manipulation and a clearer 3D scene with what we want to see will be made.

• 한국자동차공학회논문집
• /
• 제22권6호
• /
• pp.14-22
• /
• 2014

An autonomous vehicle requires a technology of generating maps by recognizing surrounding environment. The recognition of the vehicle's environment can be achieved by using distance information from a 2D laser scanner and color information from a camera. Such sensor information is used to generate 2D or 3D maps. A 2D map is used mostly for generating routs, because it contains information only about a section. In contrast, a 3D map involves height values also, and therefore can be used not only for generating routs but also for finding out vehicle accessible space. Nevertheless, an autonomous vehicle using 3D maps has difficulty in recognizing environment in real time. Accordingly, this paper proposes the technology for generating 2D maps that guarantee real-time recognition. The proposed technology uses only the color information obtained by removing height values from 3D maps generated based on the fusion of 2D laser scanner and camera data.

• 한국정밀공학회지
• /
• 제33권8호
• /
• pp.641-646
• /
• 2016

The best method of measuring wood diameter is a contact-type device: however, obtaining an accurate result can be problematic under certain circumstances. In this study, we used a laser beam and a CCD camera that did not require contact with wood. The wood is illuminated by the laser beam, and the CCD camera captures this illumination. The measurement results were determined by processing of the captured image sequences. This paper explains the use of image processing and laser systems for measurement of wood under circumstances in which physical contact is impossible.

• 한국공작기계학회:학술대회논문집
• /
• 한국공작기계학회 2000년도 춘계학술대회논문집 - 한국공작기계학회
• /
• pp.115-120
• /
• 2000

This study is on the application the image processing algorithm for inspection of the aluminium mill strip surface defect. The image of surface defect data was obtained using the CCD camera with the digital signal board. The edge was found from the difference of pixel intensity between the normal image and defect image. Two step were taken to find the edge in the image processing algorithm. First, noise was removed by using the median filter in the image. Second, the edge was sharpened in detail by using the sharpening convolution filter in the image. Canny algorithm was used to defect the exact edge. The defect section was separated from the original image is to find the coordination point p1 and p2 which include the defect image

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2003년도 ICCAS
• /
• pp.772-777
• /
• 2003

A vision-based tool position estimation system for the inspection and maintenance manipulator working inside the steam generator bowl of nuclear power plants can help human operators ensure that the inspection probe or plug are inserted to the targeted tube. Some previous research proposed a simplified tube position verification system that counts the tubes passed through during the motion and displays only the position of the tool. In this paper, by using a general camera calibration approach, tool orientation is also estimated. In order to reduce the computation time and avoid the parameter bias problem in an ellipse fitting, a small number of edge points are collected around the large section of the ellipse boundary. Experiment results show that the camera calibration parameters, detected ellipses, and estimated tool position are appropriate.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2003년도 컴퓨터소사이어티 추계학술대회논문집
• /
• pp.107-110
• /
• 2003

In this paper, we propose sharing 3D media between multisite using enhanced Access Grid (e-AG) which is a composition of 3D display and Access Grld (AG) Conventional AG and other collaborative systems have a limitation to share immersive 3D media Thus, proposed system supports sharing 3D media contents in a AG meeting section. Real object can be shared by acquiring stereo image with pre-calibrated stereo camera and by delivering, and virtual object can be shared by transmitting state information after downloading 3D model. And also, real video scene acquired by stereo camera and virtual object from 3D model can be displayed on the 3D display system of each node adaptively. The characteristics of proposed sharing method are sharing 3D media, displaying 3D media on a system adaptively, supporting real-time interaction. The proposed sharing method will be used remote lecture, remote collaboration with 3D media.

• 대한교통학회지
• /
• 제33권2호
• /
• pp.182-191
• /
• 2015

무인과속단속시스템은 속도위반 차량을 효과적으로 단속하기 위해 도입되었으나, 기존의 지점 과속단속시스템의 경우 차량 내비게이션, 스마트폰 등을 통해 제공되는 위치정보 서비스로 인해 단속 효과가 현격히 낮아지게 되었다. 구간 과속단속시스템은 지점 과속단속카메라 설치 지점에서 나타나는 카메라 전후에서 급감속 및 재가속을 반복하는 회피 거동으로 인한 여러 가지 부정적인 문제점을 보완하기 위해 도입되어 설치를 확대해 나가고 있다. 이에 본 연구에서는, 국내 고속도로 상에 설치된 구간 과속단속시스템에서 나타나는 교통류 특성에 대해 대조 구간을 설정하여 비교 분석하였다. 이를 위해 구간의 특성 등을 고려하여, 중부내륙선, 대전-통영선, 영동선 내의 구간 과속단속 시행구간에 대해 VDS 검지기 데이터를 수집하였으며, 데이터필터링을 통해 비교분석에 용이한 조건을 설정하고 자유속도 상태에서의 교통류 특성을 분석하였다. 분석결과, 구간 과속단속구간 내에서는 대조 구간에 비해 속도 평균 및 속도 분산이 낮아지는 경향을 보였으며, 통계적 검정 방법을 이용하여 이를 확인하였다.