• Journal of the Korean Society of Systems Engineering
• /
• v.15 no.1
• /
• pp.16-24
• /
• 2019

In this study, an improved method of setting the RAM goals for surveillance system using OMS/MP and field operations data of similar systems is suggested and a case study of surveillance system A is given. Newly suggested methods for setting the RAM goals consist of the procedures such as On/Off time analysis of configuration equipment based on OMS/MP, field operations data analysis of similar systems, adjustment of preventive maintenance time, the number of corrective maintenances, and TALDT in wartime, introducing new subsystem to reliability structure of surveillance system, and reflecting expert's opinion. A case study of surveillance system A shows that newly suggested method in this study for setting the RAM goals is reasonably acceptable and practically applicable to other similar systems.

• Convergence Security Journal
• /
• v.23 no.5
• /
• pp.57-64
• /
• 2023

In this study, we analyzed how to steal the target model without training data. Input data is generated using the generative model, and a similar model is created by defining a loss function so that the predicted values of the target model and the similar model are close to each other. At this time, the target model has a process of learning so that the similar model is similar to it by gradient descent using the logit (logic) value of each class for the input data. The tensorflow machine learning library was used as an experimental environment, and CIFAR10 and SVHN were used as datasets. A similar model was created using the ResNet model as a target model. As a result of the experiment, it was found that the model stealing method generated a similar model with an accuracy of 86.18% for CIFAR10 and 96.02% for SVHN, producing similar predicted values to the target model. In addition, considerations on the model stealing method, military use, and limitations were also analyzed.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.22 no.2
• /
• pp.181-186
• /
• 2012

Recently there has been many active researches on searching similar sequences from data generated with the passage of time. Those data are classified as time series data or sequence data and have different semantics from scalar data of traditional databases. In this paper similar sequence search retrieves sequences that have a similar trend of value changes. At first we have transformed the original sequences by applying DFT. The converted data are more suitable for trend analysis and they require less number of attributes for sequence comparisons. In addition we have developed a region-based query and we applied bitmap indexes which could show better performance in data warehouse. We have built bitmap indexes with varying number of attributes and we have found the least cost query plans for efficient similar sequence searches.

• The KIPS Transactions:PartC
• /
• v.11C no.5
• /
• pp.621-632
• /
• 2004

Recent measurement studies of real teletraffic data in modern telecommunication networks have shown that self-similar (or fractal) processes may provide better models of teletraffic in modern telecommunication networks than Poisson processes. If this is not taken into account, it can lead to inaccurate conclusions about performance of telecommunication networks. Thus, an important requirement for conducting simulation studies of telecommunication networks is the ability to generate long synthetic stochastic self-similar sequences. A new generator of pseu-do-random self-similar sequences, based on the fractional Gaussian nois and a wavelet transform, is proposed and analysed in this paper. Specifically, this generator uses Daubechies wavelets. The motivation behind this selection of wavelets is that Daubechies wavelets lead to more accurate results by better matching the self-similar structure of long range dependent processes, than other types of wavelets. The statistical accuracy and time required to produce sequences of a given (long) length are experimentally studied. This generator shows a high level of accuracy of the output data (in the sense of the Hurst parameter) and is fast. Its theoretical algorithmic complexity is 0(n).

• The Journal of the Korea Contents Association
• /
• v.6 no.1
• /
• pp.114-125
• /
• 2006

Since wireless Internet services and small mobile communication devices come into wide use as well as the use of GPS is rapidly growing, researches on moving object, whose location information shifts sequently in accordance with time interval, are being carried out actively. Especially, the researches on vehicle moving object are applied to Advanced traveler information system, vehicle tracking system, and distribution transport system. These systems are very useful in searching previous positions, predicted future positions, the optimum course, and the shortest course of a vehicle by managing historical data of the vehicle movement. In addition, vehicle historical data are used for distribution transport plan and vehicle allocation. Vehicle historical data are stored at regular intervals, which can have a pattern. For example, a vehicle going repeatedly around a specific section follows a route very similar to another. If historical data of the vehicle with a repeated route course are stored at regular intervals, many redundant data occur, which result in much waste of storage. Therefore this thesis suggest a vehicle historical data store scheme for vehicles with a repeated route course using similar trajectory which efficiently store vehicle historical data.

• Journal of KIISE:Databases
• /
• v.30 no.1
• /
• pp.14-26
• /
• 2003

Moving objects' trajectories play an important role in indexing video data on their content and semantics for content-based video retrieval. In this paper, we propose new similar sub-trajectory retrieval schemes based on k-warping algorithm for efficient retrieval on moving objects' trajectories in video data. The proposed schemes are fixed-replication similar sub-trajectory retrieval(FRSR) and variable-replication similar sub-trajectory retrieval(VRSR). The former can replicate motions with a fixed number for all motions being composed of the trajectory. The latter can replicate motions with a variable number. Our schemes support multiple properties including direction, distance, and time interval as well as a single property of direction, which is mainly used for modeling moving objects' trajectories. Finally, we show from our experiment that our schemes outperform Li's scheme(no-warping) and Shan's scheme(infinite-warping) in terns of precision and recall measures.

• Journal of the Korea Institute of Military Science and Technology
• /
• v.25 no.3
• /
• pp.300-310
• /
• 2022

A vast amount of labeled data is required for deep neural network training. A typical strategy to improve the performance of a neural network given a training data set is to use data augmentation technique. The goal of this work is to offer a novel image augmentation method for improving object detection accuracy. An object in an image is removed, and a similar object from the training data set is placed in its area. An in-painting algorithm fills the space that is eliminated but not filled by a similar object. Our technique shows at most 2.32 percent improvements on mAP in our testing on a military vehicle dataset using the YOLOv4 object detector.

• Journal of Intelligence and Information Systems
• /
• v.3 no.2
• /
• pp.33-55
• /
• 1997

A data model is a diagram that describes the properties of different categories of data and the associations among them within a business or information system. In spite of its importance and usefulness, data modeling activity requires not only a lot of time and effort but also extensive experience and expertise. The data models for similar business areas are analogous to one another. Therefore, it is reasonable to reuse the already-developed data models if the target business area is similar to what we have already analyzed before. In this research, we develop a case-based reasoning system for data model reuse, which we shall call CB-DM Reuser (Case-Based Data Model Reuser). CB-DM Reuse consists of four subsystems : the graphic user interface to interact with end user, the data model management system to build new data model, the case base to store the past data models, and the knowledge base to store data modeling and data model reusing knowledge. We present the functionality of CB-DM Reuser and show how it works on real-life a, pp.ication.

• The KIPS Transactions:PartD
• /
• v.15D no.1
• /
• pp.31-40
• /
• 2008

We generally use lower-dimensional transformations to convert high-dimensional sequences into low-dimensional points in similar sequence matching. These traditional transformations, however, show different characteristics in indexing performance by the type of time-series data. It means that the selection of lower-dimensional transformations makes a significant influence on the indexing performance in similar sequence matching. To solve this problem, in this paper we propose a hybrid approach that integrates multiple transformations and uses them in a single multidimensional index. We first propose a new notion of hybrid lower-dimensional transformation that exploits different lower-dimensional transformations for a sequence. We next define the hybrid distance to compute the distance between the transformed sequences. We then formally prove that the hybrid approach performs the similar sequence matching correctly. We also present the index building and the similar sequence matching algorithms that use the hybrid approach. Experimental results for various time-series data sets show that our hybrid approach outperforms the single transformation-based approach. These results indicate that the hybrid approach can be widely used for various time-series data with different characteristics.

• Journal of Korea Multimedia Society
• /
• v.8 no.8
• /
• pp.1134-1141
• /
• 2005

With ever more popularity of video web-publishing, popular content is being compressed, reformatted and modified, resulting in excessive content duplication. Such overlapped data can cause problem of search speed and rate of searching. However, duplicated data on other site can provide alternatives while specific site cause problem. This paper proposes the efficient method, for retrieving. similar video data in large database. In this research we have used the method to compare summarized video image instead of the raw video data, and detected similar videos through clustering in that dimension feature vector through PCA(principle component analysis). We show that our proposed method is efficient and accurate through our experiment.