• 한국멀티미디어학회논문지
• /
• 제24권7호
• /
• pp.890-902
• /
• 2021

Although the object detection accuracy with still images has been significantly improved with the advance of deep learning techniques, the object detection problem with video data remains as a challenging problem due to the real-time requirement and accuracy drop with occlusion. In this research, we propose a method in pig detection for video monitoring environment. First, we determine a motion, from a video data obtained from a tilted-down-view camera, based on the average size of each pig at each location with the training data, and extract key frames based on the motion information. For each key frame, we then apply YOLO, which is known to have a superior trade-off between accuracy and execution speed among many deep learning-based object detectors, in order to get pig's bounding boxes. Finally, we merge the bounding boxes between consecutive key frames in order to reduce false positive and negative cases. Based on the experiment results with a video data set obtained from a pig farm, we confirmed that the pigs could be detected with an accuracy of 97% at a processing speed of 37fps.

• 한국멀티미디어학회논문지
• /
• 제20권7호
• /
• pp.986-993
• /
• 2017

Object Tracking is a technique for tracking moving objects over time in a video image. Using object tracking technique, many research are conducted such a detecting dangerous situation and recognizing the movement of nearby objects in a smart car. However, it still remains a challenging task such as occlusion, deformation, background clutter, illumination variation, etc. In this paper, we propose a novel deep visual object tracking method that can be operated in robust to many challenging task. For the robust visual object tracking, we proposed a Convolutional Neural Network(CNN) which shares weight of the convolutional layers. Input of the CNN is a three; first frame object image, object image in a previous frame, and current search frame containing the object movement. Also we propose a method to consider the motion of the object when determining the current search area to search for the location of the object. Extensive experimental results on a authorized resource database showed that the proposed method outperformed than the conventional methods.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2008년도 하계종합학술대회
• /
• pp.1211-1212
• /
• 2008

In this paper, we propose image process modeling as a part of location based surveillance system for unauthorized target recognition and tracking in harbor, airport, military zone. For this, we compress and store background image in lower resolution and perform object extraction and motion tracking by using sobel edge detection and difference picture method between real images and a background image. In addition to, we use Independent Component Analysis Neural Network for moving target recognition. Experiments are performed for object extraction and tracking of moving targets on road by using static camera in 20m height building and it shows the robust results.

• 한국정보과학회:학술대회논문집
• /
• 한국정보과학회 2008년도 한국컴퓨터종합학술대회논문집 Vol.35 No.1 (D)
• /
• pp.444-448
• /
• 2008

유비쿼터스 스마트 홈 (ubiquitous smart home) 은 인간과 홈의 컨텍스트(context) 정보를 이용하여 인간에게 자동적인 홈 서비스 (Home service)를 제공해줄 수 있는 미래의 환경이다. 인간의 위치와 모션은 유비쿼터스 스마트 홀에서 굉장히 중요한 컨텍스트이다. 본 논문은 유비쿼터스 스마트 홀에서 인간의 위치와 모션을 예측할 수 있는 실시간 휴먼 트랙커 (tracker)를 연구하였다. 실시간 휴먼 트랙커를 위해 우리는 4개의 네트워크 카메라를 사용하였다. 본 논문에서는 실시간 휴먼 트랙커의 구조를 설명하고, 인간의 위치와 모션을 자동적으로 예측 및 판단하는 알고리즘을 제안하였다. 인간 위치를 위해서 3개의 배경이미지를 이용하였다 (이미지1 : 빈 방, 이미지2: 가구 및 가전, 이미지3: 이미지 2 와 거주자를 포함). 실시간 휴먼 트랙커는 3개의 이미지를 비교하여 각 이미지로부터 추출되는 특징 값을 결정하고, 이들 특징 값을 SVM (Support Vector Machine)을 이용하여 각각의 모션을 예측하였다. 3 개의 배경 이미지를 이용한 인간 위치 인식실험은 평균 0.037 초가 소요되었다. SVM을 이용한 모션 인식 요소에서, 우리는 각 동작에 대하여 1000번씩 측정했고, 모든 모션의 정확도 평균은 86.5% 의 정확도를 보였다.

• 정보처리학회논문지A
• /
• 제16A권3호
• /
• pp.209-216
• /
• 2009

스마트 홈(smart home)은 인간과 홈의 컨텍스트(context) 정보를 이용하여 인간에게 자동적인 홈 서비스(Home service)를 제공해줄 수 있는 미래의 환경이다. 인간의 위치와 모션은 스마트 홈에서 굉장히 중요한 컨텍스트이다. 본 논문은 스마트 홈에서 인간의 위치와 모션을 예측할 수 있는 실시간 휴먼 트랙커(tracker)를 연구하였다. 실시간 휴먼 트랙커를 위해 4개의 네트워크 카메라를 사용하였다. 본 논문에서는 실시간 휴먼 트랙커의 구조를 설명하고, 인간의 위치와 모션을 자동적으로 예측 및 판단하는 알고리즘을 제안하였다. 인간 위치를 위해서 3개의 배경 이미지를 이용하였다(이미지1: 빈 방 이미지, 이미지2: 거주자가 제외 된 가구 및 가전 이미지, 이미지3: 전체 이미지). 실시간 휴먼 트랙커는 3개의 이미지를 비교하여 각 이미지로부터 추출되는 특징 값을 결정하고, 이들 특징 값을 SVM(Support Vector Machine)을 이용하여 각각의 모션을 예측하였다. 3개의 배경 이미지를 이용한 인간 위치 인식실험은 평균 0.037 초가 소요 되었다. SVM을 이용한 모션 인식 요소에서, 각 동작에 대하여 1000번씩 측정했고, 모든 모션의 정확도 평균은 86.5% 의 정확도를 보였다.

• 대한방사선기술학회지:방사선기술과학
• /
• 제16권2호
• /
• pp.19-26
• /
• 1993

Magnetic Resonance Image represents three-dimensional diagnostic imaging technique using both nuclear magnetic resonance phenomenon and computer. Compared with computed tomography (CT), MRI have advantages harmless to patient's body, three-dimensional image with high resolution and disadvantages long data acquisition time because of long T1 relaxation time, relatively low signal to noise ratio, high cost of setting, also. As physiologic motion of tissue results in motion ghost in MRI, high 2.0Tesla make improve low signal to noise ratio. This study have aim to improve image quality with controling motion ghost of tissue. Supposing a moving pixel in constant frequency, one pixel make two ghosts which are same size and different anti-phase. So, this study will show adjust parameter on locational control of motion ghost. Author made moving phantom replaced by respiratory movement of human, researched change of motion frequency, FOV by location shift, and them decided optimal FOV (field of view). The results are as follows: 1. The frequency content of the motion determines how far the image always appear in phase-encoding direction, the morphology of the ghost image is characteristic of the direction of the motion and its amplitude. 2. Double FOV of fixed signal object for locational control of motion ghost is recommended. Decreasement of spatial resolution by increasing FOV can compensate on increasing of matrix in spite of scan time increasement.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2005년도 ICCAS
• /
• pp.610-615
• /
• 2005

Visual servoing is the fusion of results from many elemental areas including high-speed image processing, kinematics, dynamics, control theory, and real-time computing. It has much in common with research into active vision and structure from motion, but is quite different from the often described use of vision in hierarchical task-level robot control systems. We present a new approach to visual feedback control using image-based visual servoing with the stereo vision in this paper. In order to control the position and orientation of a robot with respect to an object, a new technique is proposed using a binocular stereo vision. The stereo vision enables us to calculate an exact image Jacobian not only at around a desired location but also at the other locations. The suggested technique can guide a robot manipulator to the desired location without giving such priori knowledge as the relative distance to the desired location or the model of an object even if the initial positioning error is large. This paper describes a model of stereo vision and how to generate feedback commands. The performance of the proposed visual servoing system is illustrated by the simulation and experimental results and compared with the case of conventional method for dual-arm robot made in Samsung Electronics Co., Ltd.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 1999년도 제14차 학술회의논문집
• /
• pp.13-16
• /
• 1999

An active contour model, Snake, was developed as a useful segmenting and tracking tool lot rigid or non-rigid (i.e. deformable) objects by Kass in 1987 In this research, Snake is newly designed to cover this large moving case. Image flow energy is proposed to give Snake the motion information of the target object. By this image flow energy Snake's nodes can move uniformly along the direction of the target motion in spite of the existences of local minima. Furthermore, when the motion is too large to apply image flow energy to tracking, a jump mode is proposed for solving the problem. The vector used to make Snake's nodes jump to the new location can be obtained by processing the image flow. The effectiveness of the proposed Snake is confirmed by some simulations.

• 한국항행학회논문지
• /
• 제12권2호
• /
• pp.164-169
• /
• 2008

본 논문에서는 영상분할에 의한 동작 추적 기법을 이용하여 침입자를 감시하고 관련 정보를 모바일 기반으로 확인하도록 하였다. 먼저, 탐지하고자 하는 일정한 영역을 촬영한 동영상에서 프레임을 추출하고, 인접한 두 프레임 사이의 이미지 차를 사용하여, 고정된 배경과 움직이는 대상을 분할한다. 분할된 전경 물체에서 에지를 검출하여 지정된 위치별로 추출된 에지의 중간 값을 추정하여 동작을 분석함으로써 침입자를 감시할 수 있도록 하였다. 동작이 검출되면, 영상은 WAP 풀 기반 영상 전송 방법을 사용하여 모바일 클라이언트로 전송한다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제9권1호
• /
• pp.169-189
• /
• 2015

Detection of anomalous events from video streams is a challenging problem in many video surveillance applications. One such application that has received significant attention from the computer vision community is traffic video surveillance. In this paper, a Lossy Count based Sequential Temporal Pattern mining approach (LC-STP) is proposed for detecting spatio-temporal abnormal events (such as a traffic violation at junction) from sequences of video streams. The proposed approach relies mainly on spatial abstractions of each object, mining frequent temporal patterns in a sequence of video frames to form a regular temporal pattern. In order to detect each object in every frame, the input video is first pre-processed by applying Gaussian Mixture Models. After the detection of foreground objects, the tracking is carried out using block motion estimation by the three-step search method. The primitive events of the object are represented by assigning spatial and temporal symbols corresponding to their location and time information. These primitive events are analyzed to form a temporal pattern in a sequence of video frames, representing temporal relation between various object's primitive events. This is repeated for each window of sequences, and the support for temporal sequence is obtained based on LC-STP to discover regular patterns of normal events. Events deviating from these patterns are identified as anomalies. Unlike the traditional frequent item set mining methods, the proposed method generates maximal frequent patterns without candidate generation. Furthermore, experimental results show that the proposed method performs well and can detect video anomalies in real traffic video data.