• Proceedings of the Korea Information Processing Society Conference
• /
• 2017.04a
• /
• pp.892-895
• /
• 2017

본 논문에서는 비분할 비디오로부터 이 비디오에 담긴 사람의 행동을 효과적으로 탐지해내기 위한 심층 신경망 모델을 제안한다. 일반적으로 비디오에서 사람의 행동을 탐지해내는 작업은 크게 비디오에서 행동 탐지에 효과적인 특징들을 추출해내는 과정과 이 특징들을 토대로 비디오에 담긴 행동을 탐지해내는 과정을 포함한다. 본 논문에서는 특징 추출 과정과 행동 탐지 과정에 이용할 심층 신경망 모델을 제시한다. 특히 비디오로부터 각 행동별 시간적, 공간적 패턴을 잘 표현할 수 있는 특징들을 추출해내기 위해서는 C3D 및 I-ResNet 합성곱 신경망 모델을 이용하고, 시계열 특징 벡터들로부터 행동을 자동 판별해내기 위해서는 양방향 BI-LSTM 순환 신경망 모델을 이용한다. 대용량의 공개 벤치 마크 데이터 집합인 ActivityNet 비디오 데이터를 이용한 실험을 통해, 본 논문에서 제안하는 심층 신경망 모델의 성능과 효과를 확인할 수 있었다.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2020.01a
• /
• pp.7-9
• /
• 2020

기존 CCTV 비디오에서 딥러닝 기반의 이상 탐지 연구는 객체의 행동 값만을 이용하여 이상을 탐지하였기 때문에, 시간 흐름에 따른 정보가 축소되는 문제점이 있었다. 그러나 CCTV 비디오에서의 이상의 원인은 다양한 요소와 시계열 분석에 따른 정보로 이루어져 있어 시간 정보를 유지하면서 다양한 특징 값을 사용한 모델을 설계할 필요가 있다. 따라서 본 논문에서는 C3D에 광학 흐름을 결합한 새로운 앙상블 모델을 제안한다. 실험 결과 본 논문에서 제안하는 모델이 75.83의 AUC를 얻어 기존에 연구되었던 행동 값만을 사용한 모델보다 높은 정확도를 달성하였다. 또한 이상 탐지 모델 설계 시 객체의 행동에 다양한 측면을 고려할 수 있는 여러 특징 값과 시계열 분석에 따른 정보를 사용하는 것이 적절하다는 결론을 도출하였다.

• Electronics and Telecommunications Trends
• /
• v.35 no.3
• /
• pp.20-33
• /
• 2020

Temporal action detection (TAD) in untrimmed videos is an important but a challenging problem in the field of computer vision and has gathered increasing interest recently. Although most studies on action in videos have addressed action recognition in trimmed videos, TAD methods are required to understand real-world untrimmed videos, including mostly background and some meaningful action instances belonging to multiple action classes. TAD is mainly composed of temporal action localization that generates temporal action proposals, such as single action and action recognition, which classifies action proposals into action classes. However, the task of generating temporal action proposals with accurate temporal boundaries is challenging in TAD. In this paper, we discuss TAD technologies that are considered high performance in terms of representative TAD studies based on deep learning. Further, we investigate evaluation methodologies for TAD, such as benchmark datasets and performance measures, and subsequently compare the performance of the discussed TAD models.

• Electronics and Telecommunications Trends
• /
• v.36 no.2
• /
• pp.75-82
• /
• 2021

Online action detection (OAD) in a streaming video is an attractive research area that has aroused interest lately. Although most studies for action understanding have considered action recognition in well-trimmed videos and offline temporal action detection in untrimmed videos, online action detection methods are required to monitor action occurrences in streaming videos. OAD predicts action probabilities for a current frame or frame sequence using a fixed-sized video segment, including past and current frames. In this article, we discuss deep learning-based OAD models. In addition, we investigated OAD evaluation methodologies, including benchmark datasets and performance measures, and compared the performances of the presented OAD models.

• KIPS Transactions on Software and Data Engineering
• /
• v.2 no.11
• /
• pp.795-800
• /
• 2013

Early detection of oestrus in Korean cows is one of the important issues in maximizing the economic benefit. Although various methods have been proposed, we still need to improve the performance of the oestrus detection system. In this paper, we propose a video surveillance system which can detect unusual behavior of multiple cows including the mounting activity. The unusual behavior detection is to detect the dangerous or abnormal situations of cows in video coming in real time from a surveillance camera promptly and correctly. The prototype system for unusual behavior detection gets an input video from a fixed location camera, and uses the motion vector to represent the motion information of cows in video, and finally selects a SVDD (one of the most well-known types of one-class SVM) as a detector by reinterpreting the unusual behavior into an one class decision problem from the practical points of view. The experimental results with the videos obtained from a farm located in Jinju illustrate the efficiency of the proposed method.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2014.04a
• /
• pp.823-825
• /
• 2014

감시 카메라 환경에서 자동으로 그룹 행동을 인식하는 기술이 최근 많은 관심을 받고 있다. 본 논문에서 제안하는 그룹 해동 인식 시스템은 다른 추가 정보 없이 비디오 프레임만을 인풋으로 받아들여, 자동으로 보행자 탐지, 추적, 행동 인식까지 모두 포괄하는 시스템이다. 시공간 모션 패턴을 만들고 연결 요소들로 모델링 한 뒤 Hidden Markov Model (HMM)을 이용해 그룹 행동을 인식한다. 실험 결과, 기본 논문과 비교하였을 때, 비슷한 인식률을 보이면서 수행 시간을 약 25 배 정도로 획기적으로 단축하였다.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.41 no.9
• /
• pp.1120-1131
• /
• 2016

In this paper, we propose a smoker recognition algorithm, detecting smokers in a video sequence in order to prevent fire accidents. We use description-based method in hierarchical approaches to recognize smoker's activity, the algorithm consists of background subtraction, object detection, event search, event judgement. Background subtraction generates slow-motion and fast-motion foreground image from input image using Gaussian mixture model with two different learning-rate. Then, it extracts object locations in the slow-motion image using chain-rule based contour detection. For each object, face is detected by using Haar-like feature and smoke is detected by reflecting frequency and direction of smoke in fast-motion foreground. Hand movements are detected by motion estimation. The algorithm examines the features in a certain interval and infers that whether the object is a smoker. It robustly can detect a smoker among different objects while achieving real-time performance.

• Journal of Institute of Control, Robotics and Systems
• /
• v.13 no.6
• /
• pp.537-641
• /
• 2007

In this paper, motion analysis algorithm is presented for home surveillance of the aged who lives alone. For the first step, we acquire images from a camera. To enhance the image, we use median filtering and binarize it to reduce processing time. And then morphological operations are performed to remove small blobs and small holes. At the forth step, blobs are analysed to extracts tor foreground region. Then, motions are predicted from these images by using optical tlow technique, and the predicted motion data are refined by comparing our cardboard models so as to judge behavior pattern.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2019.07a
• /
• pp.375-377
• /
• 2019

본 논문에서는 관람객 얼굴 표정 및 제스쳐 인식을 기반으로 인터렉티브 디지털콘텐츠의 문제점 추론 시스템을 제안한다. 관람객이 콘텐츠를 체험하고 다른 장소로 이동하기 전까지의 행동 패턴을 기준으로 삼아 4가지 문제점으로 분류한다. 문제점 분류을 위해 관람객이 콘텐츠 체험과정에서 나타낼 수 있는 얼굴 표정 3가지 종류와 제스쳐 5가지를 구분하였다. 실험에서는 입력된 비디오로부터 얼굴 및 손을 검출하기 위해 Adaboost algorithm을 사용하였고, mobilenet v1을 retraining하여 탐지모델을 생성 후 얼굴 표정 및 제스쳐를 검출했다. 이 연구를 통해 인터렉티브 디지털콘텐츠가 지니고 있는 문제점을 추론하여 향후 콘텐츠 개선과 제작에 사용자 중심 설계가 가능하도록 하고 양질의 콘텐츠 생산을 촉진 시킬 수 있을 것이다.

• Journal of Intelligence and Information Systems
• /
• v.29 no.4
• /
• pp.15-30
• /
• 2023

Recently, various services using artificial intelligence(AI) are emerging in the media field as well However, most of the video editing, which involves finding an editing point and attaching the video, is carried out in a passive manner, requiring a lot of time and human resources. Therefore, this study proposes a methodology that can detect the edit points of video according to whether person in video are spoken by using Video Swin Transformer. First, facial keypoints are detected through face alignment. To this end, the proposed structure first detects facial keypoints through face alignment. Through this process, the temporal and spatial changes of the face are reflected from the input video data. And, through the Video Swin Transformer-based model proposed in this study, the behavior of the person in the video is classified. Specifically, after combining the feature map generated through Video Swin Transformer from video data and the facial keypoints detected through Face Alignment, utterance is classified through convolution layers. In conclusion, the performance of the image editing point detection model using facial keypoints proposed in this paper improved from 87.46% to 89.17% compared to the model without facial keypoints.