• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.22 no.4
• /
• pp.49-57
• /
• 2022

The government is making great efforts to prevent crimes such as missing children by using public CCTVs. However, there is a shortage of operating manpower, weakening of concentration due to long-term concentration, and difficulty in tracking. In addition, applying real-time object search, re-identification, and tracking through a deep learning algorithm showed a phenomenon of increased parameters and insufficient memory for speed reduction due to complex network analysis. In this paper, we designed the network to improve speed and save memory through the application of Yolo v4, which can recognize real-time objects, and the application of Batch and TensorRT technology. In this thesis, based on the research on these advanced algorithms, OSNet re-ranking and K-reciprocal nearest neighbor for re-identification, Jaccard distance dissimilarity measurement algorithm for correlation, etc. are developed and used in the solution of CCTV national safety identification and tracking system. As a result, we propose a solution that can track objects by recognizing and re-identification objects in real-time within situation of a Korean public multi-CCTV environment through a set of algorithm combinations.

• Journal of KIISE:Software and Applications
• /
• v.32 no.5
• /
• pp.449-460
• /
• 2005

This paper proposes a method to track persons robustly in illumination changes and partial occlusions in color video frames acquired from a fixed camera. To solve a problem of changing appearance by illumination change, a time-independent intrinsic image is used to remove noises in an frame and is adaptively updated frame-by-frame. We use a hierarchical human model including body color information in order to track persons in occlusion. The tracked human model is recorded into a persons' list for some duration after the corresponding person's exit and is recovered from the list after her reentering. The proposed method was experimented in several indoor and outdoor scenario. This demonstrated the potential effectiveness of an adaptive model-base method that corrected distorted person's color information by lighting changes, and succeeded tracking of persons which was overlapped in a frame.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.20 no.12
• /
• pp.546-554
• /
• 2019

Typically, closed-circuit television (CCTV) monitoring is mainly used for post-processes (i.e. to provide evidence after an incident has occurred), but by using a streaming video feed, machine-based learning, and advanced image recognition techniques, current technology can be extended to respond to crimes or reports of missing persons in real time. The multi-CCTV cooperation technique developed in this study is a program model that delivers similarity information about a suspect (or moving object) extracted via CCTV at one location and sent to a monitoring agent to track the selected suspect or object when he, she, or it moves out of range to another CCTV camera. To improve the operating efficiency of local government CCTV control centers, we describe here the partial automation of a CCTV control system that currently relies upon monitoring by human agents. We envisage an integrated crime prevention service, which incorporates the cooperative CCTV network suggested in this study and that can easily be experienced by citizens in ways such as determining a precise individual location in real time and providing a crime prevention service linked to smartphones and/or crime prevention/safety information.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.33 no.6
• /
• pp.1099-1110
• /
• 2023

The recognition system of autonomous driving and robot navigation performs vision work such as object recognition, tracking, and lane detection after multi-sensor fusion to improve performance. Currently, research on a deep learning model based on the fusion of a camera and a lidar sensor is being actively conducted. However, deep learning models are vulnerable to adversarial attacks through modulation of input data. Attacks on the existing multi-sensor-based autonomous driving recognition system are focused on inducing obstacle detection by lowering the confidence score of the object recognition model.However, there is a limitation that an attack is possible only in the target model. In the case of attacks on the sensor fusion stage, errors in vision work after fusion can be cascaded, and this risk needs to be considered. In addition, an attack on LIDAR's point cloud data, which is difficult to judge visually, makes it difficult to determine whether it is an attack. In this study, image scaling-based camera-lidar We propose an attack method that reduces the accuracy of LCCNet, a fusion model (camera-LiDAR calibration model). The proposed method is to perform a scaling attack on the point of the input lidar. As a result of conducting an attack performance experiment by size with a scaling algorithm, an average of more than 77% of fusion errors were caused.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.39 no.3
• /
• pp.78-87
• /
• 2002

A macroblock(MB) in MPEG coded domain can have zero, one, or two motion vectors depending on its frame type and prediction direction (forward-, backward-, or hi-directionally). In this paper, we propose a method that converts these motion vectors on MPEG coded domain as a uniform set, independent of the frame type and the direction of prediction, and directly utilizes these re-analyzed motion vectors for understanding video contents. Also, using this frame-type-independent motion vector, we propose novel methods for detecting and tracking moving objects with frame-based detection accuracy on the compressed domain. These algorithms are performed directly from the MPEG bitstreams after VLC decoding with little time consumption. Experimental results show validity and outstanding performance of our methods.

• Journal of the Korea Society of Computer and Information
• /
• v.26 no.1
• /
• pp.155-161
• /
• 2021

Recently, researches for military purposes such as precision tracking and mission completion using UAVs have been actively conducted. In particular, if the posture information of the leading UAV is estimated and the mission UAV uses this information to follow in stealth and complete its mission, the speed of the posture information estimation of the guide UAV must be processed in real time. Until recently, research has been conducted to accurately estimate the posture information of the leading UAV using image processing and Kalman filters, but there has been a problem in processing speed due to the sequential processing of the processing process. Therefore, in this study we propose a way to improve processing speed by applying methods that the image processing area is limited to the ROI area including the object, not the entire area, and the continuous processing is distributed to OpenMP-based multi-threads and processed in parallel with thread synchronization to estimate attitude information. Based on the experimental results, it was confirmed that real-time processing is possible by improving the processing speed by more than 45% compared to the basic processing, and thus the possibility of completing the mission can be increased by improving the tracking and estimating speed of the mission UAV.

• The KIPS Transactions:PartB
• /
• v.14B no.7
• /
• pp.493-502
• /
• 2007

An on-line registration technique is presented to register multi-view range images for the 3D reconstruction of real objects. Using a range camera, we first acquire range images and photometric images continuously. In the range images, we divide object and background regions using a predefined threshold value. For the coarse registration of the range images, the centroid of the images are used. After refining the registration of range images using a projection-based technique, we use a modified KLT(Kanade-Lucas-Tomasi) tracker to match photometric features in the object images. Using the modified KLT tracker, we can track image features fast and accurately. If a range image fails to register, we acquire new range images and try to register them continuously until the registration process resumes. After enough range images are registered, they are integrated into a 3D model in offline step. Experimental results and error analysis show that the proposed method can be used to reconstruct 3D model very fast and accurately.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.8 no.3
• /
• pp.453-458
• /
• 2013

Aside from the issues like energy saving and maximizing network lifetime. WMSN has another issue to deal with: support of quality of service(QoS) which is required especially for handling real-time data such as object tracking and data gathering. This paper proposes a multipath routing algorithm considering the distance to sink node, energy level and link quality of neighbour nodes. Proposed algorithm supports multipath routing path with high quality links. Hence it helps to reduce a power consumption concentration that happens in particular set of nodes along the frequently selected route. It also specifies a service quality pattern and a service quality level depending on traffic pattern. By doing this, the proposed algorithm can realize a differentiated service with QoS guaranteed data transmission.

• Journal of KIISE:Computing Practices and Letters
• /
• v.15 no.6
• /
• pp.459-463
• /
• 2009

Background modeling and motion detection is the one of the most significant real time video processing technique. Until now, many researches are conducted into the topic but it still needs much time for robustness. It is more important when other algorithms are used together such as object tracking, classification or behavior understanding. In this paper, we propose efficient multi-modal background modeling methods which can be understood as simplified learning method of Gaussian mixture model. We present its validity using numerical methods and experimentally show detecting performance.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2019.10a
• /
• pp.972-975
• /
• 2019

건강 보조 식품, 의약품, 화장품 등 현대 제품에는 성분에 대한 제품의 구성정보가 라벨 형태로 상세히 기재 되어있다. 이러한 제품들은 실생활에서 접하기 쉽지만, 비전공자인 일반 사용자들이 이러한 성분들을 모두 기억하고 구분하여 사용하기에는 물질의 종류가 너무 많으며, 각 성분의 역할에 대해 면밀히 조사하기란 사실상 불가능하다. 하지만 제품에 대한 정확한 이해 없이는 제품을 사용 및 섭취함으로써 특정 부작용이 생길 수 있으며, 오용 및 남용할 가능성 또한 다분하다. 따라서, 제품 소비자가 사용하고 있는 제품이 어떠한 성분을 가지고 있는지를 정확히 파악할 필요가 있다. 이를 해결하기 위해, 본 논문에서는 기계 학습을 통한 객체 인식에 사용되는 실시간 객체 추적 기법을 활용하여 제품의 라벨을 1 차적으로 인식하고, 2 차적으로 라벨에 기재되어 있는 제품의 구성성분을 객체 인식하는 기법을 제안하고자 한다. 추가적으로, 해당 기법을 모바일 어플리케이션에 적용하여 건강 보조 식품 관리에 활용할 수 있는 방법에 대해 소개한다.