• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.28 no.4C
• /
• pp.406-412
• /
• 2003

A fast multiple object motion detection method using periodogram has been proposed. Each frame is projected to vertical and horizontal directions, and then temporal FFT is applied. A line in two dimensional frequency domain indicates velocity. To estimate the velocity of an object, we intergrate along a line which passes through the origin in frequency domain. If a frame contains multiple moving objects, multiple peaks are detected corresponding to the velocity of each object. After pairing these peaks, we can determine the velocities of an object. In the proposal method we can easily pair horizontal and vertical velocity components efficiently with simple computation by combining projections in 45$^{\circ}$ and 135 $^{\circ}$ directions in addition to the vertical and horizontal direction.

• Proceedings of the Korean Institute of Navigation and Port Research Conference
• /
• 2012.10a
• /
• pp.261-262
• /
• 2012

In this research, intelligent protection system for laser boat is introduced. The function of system is to measure the distance and velocity of object from our boat and generate control signals to avoid collision with moving targets. A novel approach to estimate object's position from our ship is tackled on this paper. To do this laser sensors are used to measure distance from ship to targets. The ship position and velocity is estimated by th Kalman filter algorithm. In the real phase, the filtering method will be applied to process signal gathered by laser sensors. Simulation to estimate ship's position and velocity under noise are executed and the results are introduced to show the effectiveness of the algorithm.

• Broadcasting and Media Magazine
• /
• v.22 no.4
• /
• pp.41-49
• /
• 2017

본 논문에서는 최근 각광받고 있는 Convolutional Neural Network(CNN)과 같은 딥러닝 기반의 얼굴 인식 연구 동향을 살펴 보고자 한다. 얼굴 인식은 입력 영상이 들어왔을 때 자동으로 누구인지 알아내는 알고리즘으로 크게 얼굴 검출, 얼굴 랜드마크 검출 및 얼굴 특징 추출로 나누어진다. 본 논문에서는 얼굴 검출, 랜드마크 검출 및 얼굴 특징 추출에 특화된 딥러닝 알고리즘을 하나씩 살펴보고 이들이 어떻게 발전해 왔는지를 확인하고자 한다. 특히, 딥러닝 기반 얼굴 인식 알고리즘들은 딥러닝 기반 물체 인식의 발전 방향과 유사하게 진행되어 오다가 최근에는 얼굴 인식에 특화된 딥러닝 아키텍처 형태로 발전하고 있다. 어떤 방향이 얼굴 인식에 더 도움이 될지에 대해서도 확인하고 실제로 어떤 문제를 해결하고 있는지 확인하고자 한다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2007.10c
• /
• pp.398-403
• /
• 2007

본 논문에서는 스테레오 비전을 이용하여 지하철 선로에 존재하는 장애물을 검출하는 기법을 제안한다. 본 논문에서는 두 대의 아날로그 카메라를 이용하여 좌우 영상을 획득하며, 영역 기반 정합을 이용하여 영상의 불균형을 계산한다. 불균형 영상의 배경을 매 프레임에서 갱신하고 현재 프레임의 불균형과의 차이에 의해 선로상의 장애물을 검출한다. 실내 역사에서 임의로 투척한 신문지, 상자, 인형과 임의로 진압한 사람을 모두 검출하였으며, 한 대의 카메라를 이용하는 기법에서 해결하기 어려운 배경과 비슷한 색상을 가지는 물체도 모두 검출하였다. 따라서 제안하는 기법은 선로상의 위험물을 검출하는 핵심 기술로 이용될 수 있다.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2013.05a
• /
• pp.886-888
• /
• 2013

An edge includes information of objects such as magnitude, orientation, and location. Conventional edge detection methods to detect those edge are methods using Sobel, Prewitt, Roberts, Laplacian operator. Existing methods use fixed weighted mask to detect edges, and their edge detection characteristics are insufficient. Therefore, to remedy weakness of conventional methods, in this paper, an edge detection algorithm using weight with standard deviation and thresholds is proposed.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2017.11a
• /
• pp.957-960
• /
• 2017

디지털 비디오의 발전이 가속됨에 따라, 비디오 샷 경계 검출은 비디오 분석 및 카타로깅 등 여러 분야에 있어 필수적인 요소가 되었다. 기존 샷 경계 검출 방법들은 잠음이나 카메라 혹은 물체의 이동, 그리고 색상의 급격한 변화 등에 민감한 성능을 보인다. 본 논문에서는 개선된 지역 방향 패턴 기반(eLDP) 검출 방법을 제안한다. 제안하는 방법은 RGB 색상의 일부와 eLDP의 특징을 결합해 더욱 강인한 샷 경계 검출 성능을 보였다. 또한, 셋톱박스 품질검사 시 필요한 채널 간 동기화의 신뢰성을 높였고, 실시간으로 검사하면서도 안정적인 샷 경계 검출이 가능함을 입증하였다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2012.06c
• /
• pp.353-355
• /
• 2012

본 논문에서는 지능형 감시 시스템을 위한 3가지 이상행위 검출 방법을 제안한다. 단순히 직접 감시나 센서에 의존한 문제점 검출이 아닌 비전 기반 기술을 적용하여 특정지역 및 모든 감시구역에 대하여 객체의 이상 행동을 감지하는 방법들을 소개한다. 제안하는 이상행위의 분류는 배회, 도주, 특정 감시 지역 침입 3가지로 정의한다. 휘도 기반의 평균 배경 모델링 방법을 통하여 움직임 물체를 검출하고, 검출된 객체를 분석(위치, 크기, 방향, 속도) 및 정의한다. 이때 이상행위의 판단에 따라 정의된 시나리오 환경으로 구성하고 분석하였다. 제안하는 방법은 실험에 사용된 3가지 이상행위에 대해 1초 안에 검출되는 것을 보였다.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.53 no.7
• /
• pp.138-143
• /
• 2016

The designed and fabricated millimeter-wave security screening system receives radiation energy from an object and a human body. The imaging system consist of sixteen array antennas, sixteen four-stage LNAs, sixteen detectors, an infrared camera, a CCD camera, reflector, and a focusing lens. This system requires high sensitivity and wide bandwidth to detect the input thermal noise. The LNA module of the system has been measured to have 65.8 dB in average linear gain and 82 GHz~102 GHz in bandwidth to enhance the sensitivity for thermal noise, and to receive it over a wide bandwidth. The detector is used for direct current (DC) output translation of millimeter-wave signals with a zero bias Schottky diode. The lens and front-end of the millimeter-wave sensor are important in the system to detect the input thermal noise signal. The frequency range in the receiving sensitivity of the detectors was 350 to 400 mV/mW at 0 dBm (1 mW) input power. The developed W-band imaging system is effective for detecting and identifying concealed objects such as metal or plastic.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.43 no.5 s.311
• /
• pp.38-47
• /
• 2006

Conventional classification method uses a single classifier based on shape or motion feature. However this method exhibits a weakness if naively used since the classification performance is highly sensitive to the accuracy of moving region to be detected. The detection accuracy, in turn, depends on the condition of the image background. In this paper, we propose to resolve the drawback and thus strengthen the classification reliability by employing a Bayesian decision fusion and by optimally combining the decisions of three classifiers. The first classifier is based on shape information obtained from Fourier descriptors while the second is based on the shape information obtained from image gradients. The third classifier uses motion information. Our experimental results on the classification Performance of human and vehicle with a static camera in various directions confirm a significant improvement and indicate the superiority of the proposed decision fusion method compared to the conventional Majority Voting and Weight Average Score approaches.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.29 no.11C
• /
• pp.1527-1540
• /
• 2004

In this paper, velocity measurement of fast moving object for traffic information acquisition using line sampling of image is proposed. Velocity measurement for traffic information acquisition of moving object is that the first sample line and second sample line on the road is set, then car is detected by using difference image method between time-variance hue data of image when car is passing two sample lines and hue data of the reference image, and velocity of the car is measured by using frame number of video which is occupied by two sample lines. Identification of the car is performed by hue of the detected car between the first sample line and second sample line, respectively To examine the propriety of the proposed algorithm, identification and velocity measurement for driving car is evaluated. The evaluated results is that it is identified by hue data of car passing two sample lines, and the velocity measurement for driving car is less than 3% comparing with X-band speed gun.