• Proceedings of the Korea Contents Association Conference
• /
• 2009.05a
• /
• pp.889-894
• /
• 2009

Accompanied by the rapid development of Computer Vision, Visual surveillance has achieved great evolution with more and more complicated processing. However there are still many problems to be resolved for robust and reliable visual surveillance, and the cast shadow occurring in motion detection process is one of them. Shadow pixels are often misclassified as object pixels so that they cause errors in localization, segmentation, tracking and classification of objects. This paper proposes a novel cast shadow removal method. As opposed to previous conventional methods, which considers pixel properties like intensity properties, color distortion, HSV color system, and etc., the proposed method utilizes observations about edge patterns in the shadow region in the current frame and the corresponding region in the background scene, and applies Laplacian edge detector to the blob regions in the current frame and the background scene. Then, the product of the outcomes of application determines whether the blob pixels in the foreground mask comes from object blob regions or shadow regions. The proposed method is simple but turns out practically very effective for Gaussian Mixture Model, which is verified through experiments.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.8 no.10
• /
• pp.3538-3556
• /
• 2014

This paper proposes a novel video stitching method that improves real-time performance and visual quality of a multi-camera video surveillance system. A two-stage seam searching algorithm based on enhanced dynamic programming is proposed. It can obtain satisfactory result and achieve better real-time performance than traditional seam-searching methods. The experiments show that the computing time is reduced by 66.4% using the proposed algorithm compared with enhanced dynamic programming, while the seam-searching accuracy is maintained. A real-time local update scheme reduces the deformation effect caused by moving objects passing through the seam, and a seam-based local color transfer model is constructed and applied to achieve smooth transition in the overlapped area, and overcome the traditional pixel blending methods. The effectiveness of the proposed method is proved in the experiements.

• Journal of Multimedia Information System
• /
• v.7 no.3
• /
• pp.205-214
• /
• 2020

Crime has become one of the major problems in modern society. Even though visual surveillances through closed-circuit television (CCTV) is extensively used for solving crime, the number of crimes has not decreased. This is because there is insufficient workforce for performing 24-hour surveillance. In addition, CCTV surveillance by humans is not efficient for detecting dangerous situations owing to accuracy issues. In this paper, we propose the autonomous detection of dangerous situations in CCTV scenes using a deep learning model with relational inference. The main feature of the proposed method is that it can simultaneously perform object detection and relational inference to determine the danger of the situations captured by CCTV. This enables us to efficiently classify dangerous situations by inferring the relationship between detected objects (i.e., distance and position). Experimental results demonstrate that the proposed method outperforms existing methods in terms of the accuracy of image classification and the false alarm rate even when object detection accuracy is low.

• ETRI Journal
• /
• v.41 no.4
• /
• pp.494-505
• /
• 2019

In this paper, we propose a new framework for detecting the unauthorized dumping of garbage in real-world surveillance camera. Although several action/behavior recognition methods have been investigated, these studies are hardly applicable to real-world scenarios because they are mainly focused on well-refined datasets. Because the dumping actions in the real-world take a variety of forms, building a new method to disclose the actions instead of exploiting previous approaches is a better strategy. We detected the dumping action by the change in relation between a person and the object being held by them. To find the person-held object of indefinite form, we used a background subtraction algorithm and human joint estimation. The person-held object was then tracked and the relation model between the joints and objects was built. Finally, the dumping action was detected through the voting-based decision module. In the experiments, we show the effectiveness of the proposed method by testing on real-world videos containing various dumping actions. In addition, the proposed framework is implemented in a real-time monitoring system through a fast online algorithm.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.6 no.5
• /
• pp.1400-1420
• /
• 2012

Object detection and tracking using visual sensors is a critical component of surveillance systems, which presents many challenges. This paper addresses the enhancement of object detection and tracking via the combination of multiple visual sensors. The enhancement method we introduce compensates for missed object detection based on the partial detection of objects by multiple visual sensors. When one detects an object or more visual sensors, the detected object's local positions transformed into a global object position. Local and global information exchange allows a missed local object's position to recover. However, the exchange of the information may degrade the detection and tracking performance by incorrectly recovering the local object position, which propagated by false object detection. Furthermore, local object positions corresponding to an identical object can transformed into nonequivalent global object positions because of detection uncertainty such as shadows or other artifacts. We improved the performance by preventing the propagation of false object detection. In addition, we present an evaluation method for the final global object position. The proposed method analyzed and evaluated using case studies.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.9 no.4
• /
• pp.69-74
• /
• 2009

Abandoned objects should be treated as possibly dangerous things for public areas until they turn out to be safe because explosive material or chemical substance is intentionally contained in them for public terrors. For large public areas such as airports or train stations, there are limits in man-power for security staffs to check all the monitors for covering the entire area under surveillance. This is the basic motivation of developing the automatic detection system for dangerous abandoned objects based on vision technology. In this research, well-known DBE is applied to stably extract background images and the HOG algorithm is adapted to discriminate between human and stuff for object classification. To show the effectiveness of the proposed system, experiments are carried out in detecting intrusion for a forbidden area and alarming for abandoned objects in a room under surveillance.

• 한국HCI학회:학술대회논문집
• /
• 2006.02a
• /
• pp.653-659
• /
• 2006

다중 객체 추적이란 컴퓨터 비전의 한 분야로, 주어진 비디오 시퀀스 내에서 관심 있는 객체들을 추적하는 것을 말한다. 다중 객체 추적 시스템은 감시 시스템, 사용자 행동 인식, 스포츠 중계, 비디오 회의와 같은 다양한 응용 분야에 핵심 기반 기술로 쓰이고 있어 그 중요성이 매우 크다. 본 논문은 감시 목적의 다중 객체를 추적하는 방법에 대하여 다룬다. 감시 시스템의 특성상, 객체의 외관이나 움직임 등에 대한 가정을 하기가 어렵다. 따라서 본 논문에서는 크기, 색, 형태 같은 객체의 단순하고 직관적인 외관 특성을 이용하면서도, 객체들끼리 부분적으로 혹은 완전히 겹쳐졌을 때에도 객체들의 위치를 적절히 추적할 수 있는 방법을 제안한다. 본 논문에서 제안하는 방법은 객체들의 경로에 대한 정보를 유지하는데 그래프 구조를 이용한다. 그래프를 확장하고, 제거하여 영상에 대한 정보를 추론한다. 크게 보면 객체들을 영역 레벨, 객체 레벨 두 단계에 걸쳐 추적한다. 영역 레벨에서는 각 객체들이 있을 수 있을만한 영역에 대한 가설을 세우고, 객체 레벨에서는 각 가설에 대한 검증을 한다. 제안된 방법은 직관적인 정보만을 이용하여 서로 다른 형태의 객체를 빠르게 추적할 수 있음을 보여준다. 다만 객체의 외관 정보만을 이용하였기 추적하기 때문에, 객체가 다른 객체에 의해 완전히 가려진 채 또다시 다른 객체와 겹쳐지면, 정확한 추적이 되지 않는다. 이를 해결하기 위해서는 객체가 겹쳐졌을 때, 그 관계에 대한 정보를 모아야 하는데 이는 향후 연구를 통해 해결하고자 한다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.4 no.3
• /
• pp.358-379
• /
• 2010

Multiple object association is an important capability in visual surveillance system with multiple cameras. In this paper, we introduce locally initiating line-based object association with the parallel projection camera model, which can be applicable to the situation without the common (ground) plane. The parallel projection camera model supports the camera movement (i.e. panning, tilting and zooming) by using the simple table based compensation for non-ideal camera parameters. We propose the threshold distance based homographic line generation algorithm. This takes account of uncertain parameters such as transformation error, height uncertainty of objects and synchronization issue between cameras. Thus, the proposed algorithm associates multiple objects on demand in the surveillance system where the camera movement dynamically changes. We verify the proposed method with actual image frames. Finally, we discuss the strategy to improve the association performance by using the temporal and spatial redundancy.

• Journal of Korea Multimedia Society
• /
• v.18 no.12
• /
• pp.1432-1438
• /
• 2015

We present a methodology to detect moving shadows in video sequences, which is considered as a challenging and critical problem in the most visual surveillance systems since 1980s. While most previous moving shadow detection methods used hand-crafted features such as chromaticity, physical properties, geometry, or combination thereof, our method can automatically learn features to classify whether image segments are shadow or foreground by using a deep learning architecture. Furthermore, applying Markov Random Field enables our system to refine our shadow detection results to improve its performance. Our algorithm is applied to five different challenging datasets of moving shadow detection, and its performance is comparable to that of state-of-the-art approaches.

• 제어로봇시스템학회:학술대회논문집
• /
• 1996.10a
• /
• pp.145-148
• /
• 1996

In moving object tracking based on the visual sensory feedback, a prerequisite is to determine which feature or which object is to be tracked and then the feature or the object identification precedes the tracking. In this paper, we focus on the object identification not image feature identification. The target identification is realized by finding out corresponding line segments to the hypothesized model segments of the target. The key idea is the combination of the Mahalanobis distance with the geometrica relationship between model segments and extracted line segments. We demonstrate the robustness and feasibility of the proposed target identification algorithm by a moving vehicle identification and tracking in the video traffic surveillance system over images of a road scene.