• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.26 no.12
• /
• pp.1816-1825
• /
• 2022

In this paper, we propose a new approach for tracking each human in a surveillance camera network with various resolution cameras. When tracking human on multiple non-overlapping cameras, the traditional appearance features are easily affected by various camera viewing conditions. To overcome this limitation, the proposed system utilizes facial information along with appearance information. In general, human images captured by the surveillance camera are often low resolution, so it is necessary to be able to extract useful features even from low-resolution faces to facilitate tracking. In the proposed tracking scheme, texture-based face descriptor is exploited to extract features from detected face after face frontalization. In addition, when the size of the face captured by the surveillance camera is very small, a super-resolution technique that enlarges the face is also exploited. The experimental results on the public benchmark Dana36 dataset show promising performance of the proposed algorithm.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 1995.10a
• /
• pp.1182-1185
• /
• 1995

This paper presents a novel technology for unattented continuous monitoring of radioactive material in hot cell environments. In this monitoring system, the surveillance camera data and NDA data are time synchronized and integrated into the same dimension through data processing. The integrated information is then fed into a neural network to generate diagnostics through data processing. the integrated information of the concept is tested for a spent nuclear fuel transprotation in an operational hot cell at KAERI. The presented integral part of the multi-sensory system and the analytical paradigm may provide an effective technologyical alternative for safeguarding new conceptual hot cell facilities, namely the Dupic facility.

• Journal of Institute of Control, Robotics and Systems
• /
• v.4 no.3
• /
• pp.378-383
• /
• 1998

A FOV, that stands for "Field Of View", refers to the maximum area where a camera could be wholly seen. If a FOV of CCD camera cannot the cover overall inspection area, the overall inspection area should be divided into sub-areas of size FOV. In this paper, we propose a new neural network-based FOV generation method by using a newly modified self-organizing map(SOM) which has multiple structure based on a self-organizing map, and uses new training rule that is composed of the movement, creation and deletion terms. Then, experiment results using real PCB indicate the superiority of the method developed in this study to the existing sequential method.al method.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.61 no.1
• /
• pp.125-129
• /
• 2012

In this paper, we propose an intelligent method to surveil moving objects and to judge an unusual-behavior by using network cameras. To surveil moving objects, the Scale Invariant Feature Transform (SIFT) algorithm is used to characterize the feature information of objects. To judge unusual-behaviors, the virtual human skeleton is used to extract the feature points of a human in input images. In this procedure, the Principal Component Analysis (PCA) improves the accuracy of the feature vector and the fuzzy classifier provides the judgement principle of unusual-behaviors. Finally, the experiment results show the effectiveness and the feasibility of the proposed method.

• Proceedings of the KIEE Conference
• /
• 2006.04a
• /
• pp.258-260
• /
• 2006

In this paper, we designed a embedded system that will perform a primary role of Tangible Space implementation. This hardware includes function of image capture through camera interface, image process and sending off image information by LAN (local area network) or WLAN(wireless local area network). We define this hardware as a network based Visual Agent Platform for Tangible Space

• Proceedings of the KIEE Conference
• /
• 2007.04a
• /
• pp.172-174
• /
• 2007

In this paper, we designed a embedded system that will perform a primary role of Tangible Space implementation. This hardware includes function of image capture through camera interface, image process and sending off image information by LAN(local area network) or WLAN(wireless local area network). We define this hardware as a network based Visual Agent Platform for Tangible Space, This Visual Agent Platform includes the software that is RTLinux and CORBA

• Journal of the Korea Institute of Military Science and Technology
• /
• v.19 no.1
• /
• pp.66-75
• /
• 2016

Stereoscopic image generated by depth image-based rendering(DIBR) for surveillance robot and camera is appropriate in a low bandwidth network. The image is very important data for the decision-making of a commander and thus its integrity has to be guaranteed. One of the methods used to detect manipulation is to check if the stereoscopic image is taken from the original camera. Sensor pattern noise(SPN) used widely for camera identification cannot be directly applied to a stereoscopic image due to the stereo warping in DIBR. To solve this problem, we find out a shifted object in the stereoscopic image and relocate the object to its orignal location in the center image. Then the similarity between SPNs extracted from the stereoscopic image and the original camera is measured only for the object area. Thus we can determine the source of the camera that was used.

• Journal of information and communication convergence engineering
• /
• v.7 no.2
• /
• pp.119-124
• /
• 2009

In this paper, we propose a novel approach to estimating the real-time moving trajectory of an object is proposed in this paper. The object's position is obtained from the image data of a CCD camera, while a state estimator predicts the linear and angular velocities of the moving object. To overcome the uncertainties and noises residing in the input data, a Extended Kalman Filter(EKF) and neural networks are utilized cooperatively. Since the EKF needs to approximate a nonlinear system into a linear model in order to estimate the states, there still exist errors as well as uncertainties. To resolve this problem, in this approach the Kohonen networks, which have a high adaptability to the memory of the input-output relationship, are utilized for the nonlinear region. In addition to this, the Kohonen network, as a sort of neural network, can effectively adapt to the dynamic variations and become robust against noises. This approach is derived from the observation that the Kohonen network is a type of self-organized map and is spatially oriented, which makes it suitable for determining the trajectories of moving objects. The superiority of the proposed algorithm compared with the EKF is demonstrated through real experiments.

• Journal of Korea Multimedia Society
• /
• v.20 no.5
• /
• pp.769-781
• /
• 2017

Deep learning networks like Convolutional Neural Networks (CNNs) show successful performances in many computer vision applications such as image classification, object detection, and so on. For implementation of deep learning networks in embedded system with limited processing power and memory, deep learning network may need to be simplified. However, simplified deep learning network cannot learn every possible scene. One realistic strategy for embedded deep learning network is to construct a simplified deep learning network model optimized for the scene images of the installation place. Then, automatic training will be necessitated for commercialization. In this paper, as an intermediate step toward automatic training under fisheye camera environments, we study more precise human localization in fisheye images, and propose an accurate human localization method, Automatic Ground-Truth Labelling Method (AGTLM). AGTLM first localizes candidate human object bounding boxes by utilizing GoogLeNet-LSTM approach, and after reassurance process by GoogLeNet-based CNN network, finally refines them more correctly and precisely(tightly) by applying saliency object detection technique. The performance improvement of the proposed human localization method, AGTLM with respect to accuracy and tightness is shown through several experiments.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.15 no.8
• /
• pp.3025-3047
• /
• 2021

This paper presents the latest Ethernet standardization of in-vehicle network and the future trends of automotive Ethernet technology. The proposed system provides design and optimization algorithms for automotive networking technology related to AVB (Audio Video Bridge) technology. We present a design of in-vehicle network system as well as the optimization of AVB for automotive. A proposal of Reduced Latency of Machine to Machine (RLMM) plays an outstanding role in reducing the latency among devices. RLMM's approach to real-world experimental cases indicates a reduction in latency of around 41.2%. The setup optimized for the automotive network environment is expected to significantly reduce the time in the development and design process. The results obtained in the study of image transmission latency are trustworthy because average values were collected over a long period of time. It is necessary to analyze a latency between multimedia devices within limited time which will be of considerable benefit to the industry. Furthermore, the proposed reliable camera and video streaming through optimized AVB device settings would provide a high level of support in the real-time comprehension and analysis of images with AI (Artificial Intelligence) algorithms in autonomous driving.