• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• v.9 no.1
• /
• pp.493-496
• /
• 2005

In this paper, we propose a new approach to detect candidate area of vehicle license plate using compounded color and vertical edge information it's own. Also, we propose a verification course, to compressed image generated by Fast DCT, using SVM to increase accuracy of extracted vechicle license plate area. Proposed method is consider that vehicle's position, become a object of it's license plate recognition, has various angle, scale and include enough environment informations. As a experimental results, proposed method shows a superior performance compared with the case that not includes verification course using SVM.

• Journal of Broadcast Engineering
• /
• v.21 no.6
• /
• pp.913-919
• /
• 2016

In this paper, we propose the visual tracking algorithm that takes advantage of multiple random walkers. We first show the tracking method based on support vector machine as [1] and suggest a method that suppresses feature vectors extracted from backgrounds while preserve features vectors from foregrounds. We also show how to discriminate between foregrounds and backgrounds. Learned by reducing influences of backgrounds, support vector machine can clearly distinguish foregrounds and backgrounds from the image whose target objects are similar to backgrounds and occluded by another object. Thus, the algorithm can track target objects well. Furthermore, we introduce a simple method improving tracking speed. Finally, experiments validate that proposed algorithm yield better performance than the state-of-the-art trackers on the widely-used benchmark dataset with high speed.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.15 no.2
• /
• pp.364-370
• /
• 2011

This paper propose a novel method for tracking moving object based on covariance matrix and Riemannian Manifolds. With image backgrounds continuously changed, we use the covariance matrices to extract features for tracking nonrigid object undergoing transformation and deformation. The covariance matrix can make fusion of different types of features and has its small dimension, therefore we enable to handle the spatial and statistical properties as well as the component correlation. The proposed method can estimate the position of the moving object by employing the covariance matrix of object region as a feature vector and comparing the candidate regions. Rimannian Geometry is efficiently adapted to object deformation and change of shape and improve the accuracy by using geodesic distance to predict the estimated position with the minimum distance. The experimental results have shown that the proposed method correctly tracked the moving object.

• Proceedings of the Korean Institute of IIIuminating and Electrical Installation Engineers Conference
• /
• 2004.05a
• /
• pp.479-485
• /
• 2004

The purpose of this study is to develop a practical image inspection system that could recognize it correctly, endowing flexibility to the productive field, although the same object for work will be changed in the size and rotated. In this experiment, it selected a fighter, rotating the direction from $30^{\circ}\;to\;45^{\circ}$ simultaneously while changing the size from 1/4 to 1/16, as an object inspection without using another hardware for exclusive image processing. The invariant moments, Hu has suggested, was used as feature vector moment descriptor. As a result of the experiment the image inspection system developed from this research was operated in real-time regardless of the chance of size and rotation for the object inspection, and it maintained the correspondent rates steadily above from 94% to 96%. Accordingly, it is considered as the flexibility can be considerably endowed to the factory automation when the image inspection system developed from this research is applied to the productive field.

• Proceedings of the KSRS Conference
• /
• 1999.11a
• /
• pp.75-79
• /
• 1999

Geo-features play a key role in object-oriented or feature-based geo-processing system. So the strategy for how-to-model and how-to-manage the geo-features builds the main architecture of the entire system and also supports the efficiency and functionality of the system. Unlike the conventional 2D geo-processing system, geo-features in 3B GIS have lots to be considered to model regarding the efficient manipulation and analysis and visualization. When the system is running on the Web, it should also be considered that how to leverage the level of detail and the level of automation of modeling in addition to the support for client side data interoperability. We built a set of 3D geo-features, and each geo-feature contains a set of aspatial data and 3D geo-primitives. The 3D geo-primitives contain the fundamental modeling data such as the height of building and the burial depth of gas pipeline. We separated the additional modeling data on the geometry and appearance of the model from the fundamental modeling data to make the table in database more concise and to allow the users more freedom to represent the geo-object. To get the users to build and exchange their own data, we devised a file format called VGFF 2.0 which stands for Virtual GIS File Format. It is to describe the three dimensional geo-information in XML(eXtensible Markup Language). The DTD(Document Type Definition) of VGFF 2.0 is parsed using the DOM(Document Object Model). We also developed the authoring tools for. users can make their own 3D geo-features and model and save the data to VGFF 2.0 format. We are now expecting the VGFF 2.0 evolve to the 3D version of SVG(Scalable Vector Graphics) especially for 3D GIS on the Web.

• The Journal of the Acoustical Society of Korea
• /
• v.16 no.5
• /
• pp.26-35
• /
• 1997

This paper deals with establishing correspondent points of feature pints and sampling period when we estimate object motion parameters from image information of freely moving objects in space of gravity-free state. Replacing the inertial coordinate system with the camera coordinate system which is equipped within a space robot, it is investigated to be able to analyze a problem of correspond points from image information, and to obtain sequence of angular velocity $\omega$ which determine a motion of object by means of computer simulation. And if a sampling period ${\Delta}t$ is shortened, the relative errors of angular velocity are increased because the relative errors against moving distance of feature points are increased by quantization. In reverse, if a sampling period ${\Delta}t$ is lengthened too much, the relative error are likewise increased because a sampling period is long for angular velocity to be approximated, and we confirmed the precision that grows according to ascending of resolution.

• Proceedings of the Korean Association of Geographic Inforamtion Studies Conference
• /
• 1999.12a
• /
• pp.47-51
• /
• 1999

Geo-features play a key role in object-oriented or feature-based geo-processing system. So the strategy for how-to-model and how-to-manage the geo-features builds the main architecture of the entire system and also supports the efficiency and functionality of the system. Unlike the conventional 2D geo-processing system, geo-features in 3D GIS have lots to be considered to model regarding the efficient manipulation and analysis and visualization. When the system is running on the Web, it should also be considered that how to leverage the level of detail and the level of automation of modeling in addition to the support for client side data interoperability. We built a set of 3D geo-features, and each geo-feature contains a set of aspatial data and 3D geo-primitives. The 3D geo-primitives contain the fundamental modeling data such as the height of building and the burial depth of gas pipeline. We separated the additional modeling data on the geometry and appearance of the model from the fundamental modeling data to make the table in database more concise and to allow the users more freedom to represent the geo-object. To get the users to build and exchange their own data, we devised a fie format called VGFF 2.0 which stands for Virtual GIS File Format. It is to describe the three dimensional geo-information in XML(extensible Markup Language). The DTD(Document Type Definition) of VGFF 2.0 is parsed using the DOM(Document Object Model). We also developed the authoring tools for users can make their own 3D geo-features and model and save the data to VGFF 2.0 format. We are now expecting the VGFF 2.0 evolve to the 3D version of SVG(Scalable Vector Graphics) especially for 3D GIS on the Web.

• Journal of Digital Convergence
• /
• v.10 no.6
• /
• pp.297-302
• /
• 2012

In this paper, we present a object detection and tracking method based on image context analysis. It is robust from the image variations such as complicated background, dynamic movement of the object. Image context analysis is carried out using the hybrid network of k-means and RBF. The proposed object detection employs context-driven adaptive Bayesian framework to relive the effect due to uneven object images. The proposed method used feature vector generator using 2D Haar wavelet transform and the Bayesian discriminant method in order to enhance the speed of learning. The system took less time to learn, and learning in a wide variety of data showed consistent results. After we developed the proposed method was applied to real-world environment. As a result, in the case of the object to detect pass outside expected area or other changes in the uncertain reaction showed that stable. The experimental results show that the proposed approach can achieve superior performance using various data sets to previously methods.

• Journal of Intelligence and Information Systems
• /
• v.28 no.1
• /
• pp.89-106
• /
• 2022

Recently, the amount of video data collected from smartphones, CCTVs, black boxes, and high-definition cameras has increased rapidly. According to the increasing video data, the requirements for analysis and utilization are increasing. Due to the lack of skilled manpower to analyze videos in many industries, machine learning and artificial intelligence are actively used to assist manpower. In this situation, the demand for various computer vision technologies such as object detection and tracking, action detection, emotion detection, and Re-ID also increased rapidly. However, the object detection and tracking technology has many difficulties that degrade performance, such as re-appearance after the object's departure from the video recording location, and occlusion. Accordingly, action and emotion detection models based on object detection and tracking models also have difficulties in extracting data for each object. In addition, deep learning architectures consist of various models suffer from performance degradation due to bottlenects and lack of optimization. In this study, we propose an video analysis system consists of YOLOv5 based DeepSORT object tracking model, SlowFast based action recognition model, Torchreid based Re-ID model, and AWS Rekognition which is emotion recognition service. Proposed model uses single-linkage hierarchical clustering based Re-ID and some processing method which maximize hardware throughput. It has higher accuracy than the performance of the re-identification model using simple metrics, near real-time processing performance, and prevents tracking failure due to object departure and re-emergence, occlusion, etc. By continuously linking the action and facial emotion detection results of each object to the same object, it is possible to efficiently analyze videos. The re-identification model extracts a feature vector from the bounding box of object image detected by the object tracking model for each frame, and applies the single-linkage hierarchical clustering from the past frame using the extracted feature vectors to identify the same object that failed to track. Through the above process, it is possible to re-track the same object that has failed to tracking in the case of re-appearance or occlusion after leaving the video location. As a result, action and facial emotion detection results of the newly recognized object due to the tracking fails can be linked to those of the object that appeared in the past. On the other hand, as a way to improve processing performance, we introduce Bounding Box Queue by Object and Feature Queue method that can reduce RAM memory requirements while maximizing GPU memory throughput. Also we introduce the IoF(Intersection over Face) algorithm that allows facial emotion recognized through AWS Rekognition to be linked with object tracking information. The academic significance of this study is that the two-stage re-identification model can have real-time performance even in a high-cost environment that performs action and facial emotion detection according to processing techniques without reducing the accuracy by using simple metrics to achieve real-time performance. The practical implication of this study is that in various industrial fields that require action and facial emotion detection but have many difficulties due to the fails in object tracking can analyze videos effectively through proposed model. Proposed model which has high accuracy of retrace and processing performance can be used in various fields such as intelligent monitoring, observation services and behavioral or psychological analysis services where the integration of tracking information and extracted metadata creates greate industrial and business value. In the future, in order to measure the object tracking performance more precisely, there is a need to conduct an experiment using the MOT Challenge dataset, which is data used by many international conferences. We will investigate the problem that the IoF algorithm cannot solve to develop an additional complementary algorithm. In addition, we plan to conduct additional research to apply this model to various fields' dataset related to intelligent video analysis.

• The Journal of Society for e-Business Studies
• /
• v.17 no.4
• /
• pp.209-219
• /
• 2012

Objects such as products, product reviews, and user profiles are important in e-commerce domain. Vector is one of the most widely used object representation scheme. Information of e-commerce objects may be modeled by vectors in which the featured values are assigned to various dimensions. E-commerce objects are in general quantitatively large while some are similar or even same in reality. It Plays, therefore, an important role to measure the similarity between objects. In this paper, we survey the state-of-the -art vector similarity measures. Similarity measures are analyzed to feature the algebraic characteristics and relationship of those, and upon which we classify the related measures accordingly. We then present such features that standard vector similarity measures should convey.