• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2007.06a
• /
• pp.615-618
• /
• 2007

본 실험을 통하여 연속된 프레임을 갖는 이미지 영상을 히스토그램의 절대값을 이용하여 영상의 모션에너지를 분석하는 방법을 제안한다. 입력되어지는 영상은 그 내용의 흐름에 따라 각각의 프레임마다 다른 모션에너지를 발생하고 모션에너지의 피크값을 검출 할 임계값과 영상의 분할을 통하여 영상내 객체의 이동방향이나 움직임의 강도 등을 파악할 수 있다. 이 모션에너지는 여러 가지로 활용할 수 있으며 실험을 통하여 차량의 주행속도를 자동으로 검출해 낼 수 있었다.

• Journal of Korea Society of Industrial Information Systems
• /
• v.19 no.5
• /
• pp.9-14
• /
• 2014

Recently, various contents using human motion are developed in computer vision and game industries. If we try to apply human motion to application programs and contents, users can experience a sense of immersion getting into it so that the users feel a high level of satisfaction from the contents. In this research, we analyze human head motion using images captured from an webcam and then we apply the result of motion recognition to a game without special devices as an interface. The proposed method, first, segments human head region using an image composed of MHI(Motion History Image) and the result of skin color detection, and then we calculate the direction and distance by the MHI sequence. In experiments, the proposed method for human head motion recognition was tested for controlling a game player. From the experimental results we proved that the proposed method can make a gammer feel more immersed into the game. Furthermore, we expect the proposed method can be an interface of a serious game for medical or rehabilitation purposes.

• KIPS Transactions on Software and Data Engineering
• /
• v.11 no.10
• /
• pp.419-426
• /
• 2022

In this paper, we propose a method to improve the accuracy of 3D human pose estimation model in various move motions. Existing human pose estimation models have some problems of jitter, inversion, swap, miss that cause miss coordinates when estimating human poses. These problems cause low accuracy of pose estimation models to detect exact coordinates of human poses. We propose a method that consists of detection and correction methods to handle with these problems. Deep learning-based outlier detection method detects outlier of human pose coordinates in move motion effectively and rule-based correction method corrects the outlier according to a simple rule. We have shown that the proposed method is effective in various motions with the experiments using 2D golf swing motion data and have shown the possibility of expansion from 2D to 3D coordinates.

• Journal of KIISE:Software and Applications
• /
• v.30 no.5_6
• /
• pp.497-502
• /
• 2003

Content-based video copy detection is a complementary approach to watermarking. As opposed to watermarking, which relies on inserting a distinct pattern into the video stream, video copy detection techniques match content-based signatures to detect copies of video. Existing typical content-based copy detection schemes have relied on image matching which is based on key frame detection. This paper proposes a motion directional histogram, which is quantized and accumulated the direction of motion, for video copy detection. The video clip is represented by a motion directional histogram as a 1-dimensional graph. This method is suitable for real time indexing and counting the TV CF verification that is high motion video clips.

• The Journal of the Korea Contents Association
• /
• v.10 no.3
• /
• pp.1-9
• /
• 2010

In this paper, we propose a new method of improving accuracy of motion estimation in mobile environments, compared with Rosten's algorithm. The present method selects corners as feature points. The Rosten's algorithm uses simple addition and subtraction to detect the corners. Although it has the advantage of faster processing speed, Rosten's algorithm has a drawback of low performance in motion estimation. We use the NCC(Normalized Cross Correlation) coefficients to match the corners, and remove in two steps the outliers of inaccurate matching corners. We compare the proposed algorithm with Rosten's algorithm by applying both to the real images. We find that the proposed method shows better performance than Rosten's algorithm in motion estimation. In addition, we implement the present method on mobile devices and confirm that it works in mobile environments in real time.

• Proceedings of the Korean Information Science Society Conference
• /
• 2006.06b
• /
• pp.325-327
• /
• 2006

본 논문은 모션 그래디언트 히스토그램(Motion Gradient Histogram : 이하 'MGH')을 적용하여 동영상에서 나타나는 다수 객체들의 동작 검출 및 인식을 실시간으로 구현하는 방법을 제안한다. 인식하고자 하는 대상에 대한 기본적인 템플릿 동영상들의 MGH와 일정 프레임 간격마다 동영상의 MGH를 비교하여 검출 및 인식이 이루어진다. 동시에 다수의 동작이 있는 경우 동작이 발생하는 영역을 모션 에너지 영상(Motion Energy Image : MEI) 기법으로 추출하여 해당 영역별 MGH를 구함으로써 다수 동작을 인식할 수 있도록 한다.

• Annual Conference of KIPS
• /
• 2021.11a
• /
• pp.787-790
• /
• 2021

본 논문에서는 최신 모션 인식 기술을 활용하여 골프 스윙 비디오에서 사람의 자세를 추정한 후 다양한 원인으로 오검출된 좌표들을 보정하여 자세 추정의 정확도를 높이는 방법을 제안한다. 기존의 사람 자세 추정 모델은 골프 스윙 데이터에서 오검출, 반전, 불안정성, 미검출의 문제를 보여 정확한 자세 추정을 어렵게 했다. 이를 해결하기 위하여 본 연구에서는 자세 추정시 발생하는 이상치 데이터들을 Bi-LSTM 으로 학습하고 골프 스윙의 특징을 고려한 간단한 규칙을 통하여 이상치 데이터를 효과적으로 검출하고 이를 보정하는 방법을 제안한다. 또한 다양한 실험과 분석을 통하여 제안하는 방법이 골프 스윙 모션에서 사람의 자세를 정확히 추정할 수 있음을 보인다.

• Journal of KIISE:Software and Applications
• /
• v.34 no.12
• /
• pp.1075-1082
• /
• 2007

In this paper, we propose the method of multiple human action recognition on video clip. For being invariant to the change of speed or size of actions, Spatial-Temporal Pyramid method is applied. Proposed method can minimize the complexity of the procedures owing to select Motion Gradient Histogram (MGH) based on statistical approach for action representation feature. For multiple action detection, Motion Energy Image (MEI) of binary frame difference accumulations is adapted and then we detect each action of which area is represented by MGH. The action MGH should be compared with pre-learning MGH having pyramid method. As a result, recognition can be done by the analyze between action MGH and pre-learning MGH. Ten video clips are used for evaluating the proposed method. We have various experiments such as mono action, multiple action, speed and site scale-changes, comparison with previous method. As a result, we can see that proposed method is simple and efficient to recognize multiple human action with stale variations.

• Proceedings of the Korea Contents Association Conference
• /
• 2009.05a
• /
• pp.672-676
• /
• 2009

Recently, intelligent surveillance camera systems are needed popularly. However, current researches are focussed on improvement of a single module rather than implementation of an integrated system. In this paper, we implemented a wireless surveillance camera system which is composed of face detection, and using motion sensor. In our implementation, we used a camera module from SHARP, a pair of wireless video transmission module from ECOM, a pair of ZigBee RF wireless transmission module from ROBOBLOCK, and a motion sensor module (AMN14111) from PANASONIC. We used OpenCV library for face dection and MFC for implement software. We identified real-time operations of face detection, PTT control, and motion sensor detecton. Thus, the implemented system will be useful for the applications of remote control, human detection, and using motion sensor.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 1999.11b
• /
• pp.123-128
• /
• 1999

본 논문에서는 기존에는 샷 경계 검출에 초점을 맞춘 것과는 달리 본 논문에서는 샷 보다 상위레벨인 비디오 씬 추출에 초점을 맞추어 디지털 비디오를 구조화하는 시스템을 제안한다. 샷간의 유사도를 측정하기 위해서 칼라와 모션 특징을 이용하였으며, 비디오 내의 동적 또는 정적 특성을 반영하기 위해서 적응적 가중치를 적용하였다. 칼라 특징을 추출하기 위해서 각 샷의 내부에서 대표 프레임을 추출하였고, 각 샷 내부의 모션 정보는 MPEG 비디오 모션 벡터를 이응해서 추출하였다. 또한, 비디오 씬 분할 시 연산 시간을 줄이기 위한 기법을 제시하였다. 마지막으로 추출된 비디오 씬에 대해서 성능평가를 하였다.