• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.6 no.12
• /
• pp.3315-3337
• /
• 2012

Crowded scenes analysis is a challenging topic in computer vision field. How to detect diverse motion patterns in crowded scenarios from videos is the critical yet hard part of this problem. In this paper, we propose a novel approach to mining motion patterns by utilizing motion information during both long-term period and short interval simultaneously. To capture long-term motions effectively, we introduce Motion History Image (MHI) representation to access to the global perspective about the crowd motion. The combination of MHI and optical flow, which is used to get instant motion information, gives rise to discriminative spatial-temporal motion features. Benefitting from the robustness and efficiency of the novel motion representation, the following motion pattern mining is implemented in a completely unsupervised way. The motion vectors are clustered hierarchically through automatic hierarchical clustering algorithm building on the basis of graphic model. This method overcomes the instability of optical flow in dealing with time continuity in crowded scenes. The results of clustering reveal the situations of motion pattern distribution in current crowded videos. To validate the performance of the proposed approach, we conduct experimental evaluations on some challenging videos including vehicles and pedestrians. The reliable detection results demonstrate the effectiveness of our approach.

• Journal of KIISE:Software and Applications
• /
• v.30 no.11
• /
• pp.1054-1061
• /
• 2003

Q-Learning, a representative algorithm of reinforcement learning, experiences repeatedly until estimation values about all state-action pairs of state space converge and achieve optimal policies. When the state space is high dimensional or continuous, complex reinforcement learning tasks involve very large state space and suffer from storing all individual state values in a single table. We introduce Q-Map that is new function approximation method to get classified policies. As an agent learns on-line, Q-Map groups states of similar situations and adapts to new experiences repeatedly. State-action pairs necessary for fine control are treated in the form of rule. As a result of experiment in maze environment and mountain car problem, we can achieve classified knowledge and extract easily rules from Q-Map

• Journal of KIISE:Computing Practices and Letters
• /
• v.16 no.1
• /
• pp.60-64
• /
• 2010

To discover association rules from nontransactional data, there have been many studies on discretization of attribute values. These studies do not reflect the change of discovered rules' confidence according to the change of the ranges of the discretized attributes, and perform the discretization stage and the rule discovery stage independently. This causes the ranges of attributes not properly discretized, thereby making the rules having high confidence excluded in the result set. To solve this problem, we propose a novel method that performs the discretization and rule discovery stages simultaneously in order to discretize ranges of attributes in such a way that the rules having high confidence are discovered well. To the end, we perform hierarchical clustering on the attributes in the right hand side of rules, then do characterization on every cluster thus obtained. The experimental result demonstrates that our method discovers the rules having high confidence better than existing methods.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2000.04a
• /
• pp.28-35
• /
• 2000

In order to acquire the precise and fast response for an analytical query, proper selection of the views to materialize in data warehouse is very crucial. In traditional algorithms, the whole relation is considered to be selected as materialized views. However, materializing the whole relation rather than a part of relation results in much worse performance in terms of time and space cost. Therefore, we present a new algorithm for selection of views to materialize using clustering method in order to improve the performance of data warehouse including this problem. In the presented algorithm, ASVMR(Algorithm for Selection of Views to Materialize using Reduced table), we first generate reduced tables in data warehouse using automatic clustering based on attribute-values density, then we consider the combination of reduced tables as materialized views instead of the combination of the original base relations. We also show the experimental results in which both time and space cost are approximately 1.8 times better than the conventional algorithms.

• Journal of the Korea Convergence Society
• /
• v.10 no.10
• /
• pp.7-13
• /
• 2019

The natural gas (NG), mostly methane leaks into the air, it is a big problem for the climate. detected NG leaks under U.S. city streets and collected data. In this paper, we introduced a Deep Neural Network (DNN) classification of prediction for a level of NS leak. The proposed method is OrdinalEncoder(OE) based K-means clustering and Multilayer Perceptron(MLP) for predicting NG leak. The 15 features are the input neurons and the using backpropagation. In this paper, we propose the OE method for labeling target data using k-means clustering and compared normalization methods performance for NG leak prediction. There five normalization methods used. We have shown that our proposed OE based MLP method is accuracy 97.7%, F1-score 96.4%, which is relatively higher than the other methods. The system has implemented SPSS and Python, including its performance, is tested on real open data.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.17 no.7
• /
• pp.1873-1893
• /
• 2023

The development of lightweight, low energy and small-sized sensors incorporated with the wireless networks has brought about a phenomenal growth of Wireless Sensor Networks (WSNs) in its different fields of applications. Moreover, the routing of data is crucial in a wide number of critical applications that includes ecosystem monitoring, military and disaster management. However, the time-delay, energy imbalance and minimized network lifetime are considered as the key problems faced during the process of data transmission. Furthermore, only when the functionality of cluster head selection is available in WSNs, it is possible to improve energy and network lifetime. Besides that, the task of cluster head selection is regarded as an NP-hard optimization problem that can be effectively modelled using hybrid metaheuristic approaches. Due to this reason, an Improved Coyote Optimization Algorithm-based Clustering Technique (ICOACT) is proposed for extending the lifetime for making efficient choices for cluster heads while maintaining a consistent balance between exploitation and exploration. The issue of premature convergence and its tendency of being trapped into the local optima in the Improved Coyote Optimization Algorithm (ICOA) through the selection of center solution is used for replacing the best solution in the search space during the clustering functionality. The simulation results of the proposed ICOACT confirmed its efficiency by increasing the number of alive nodes, the total number of clusters formed with the least amount of end-to-end delay and mean packet loss rate.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.37 no.2B
• /
• pp.138-147
• /
• 2012

With the improvement of wireless communication and embedded technology, WSN is used at various fields. Meanwhile, because WSN is resource constrained, it is more vulnerable than other networks. To solve the security problem of WSN, we can use the traditional secure mechanism like as cryptography and authentication. But the traditional secure mechanism is not enough for all security issues that may be happened in WSN, especially attacks caused by the compromised node. So, we need the IDS as the second secure mechanism for WSN. In this paper, we propose the Specification-based Intrusion Detection Mechanism that makes LEACH, which is one of the clustering routing protocol for WSN, more reliable and safety.

• Journal of The Korean Association of Information Education
• /
• v.11 no.1
• /
• pp.91-98
• /
• 2007

Plagiarism is a serious problem in school education due to current technologies such as the internet and word processors. This paper presents how to detect source code plagiarism using similarity based on string comparison methods. The main contribution is to use hierarchical agglomerative clustering technique to classify plagiarism groups, which are then visualized as a dendrogram. Graders can set an empirical threshold to the dendrogram to navigate plagiarism groups. We evaluated the performance of the presented method with a real world data. The result showed the usefulness and applicability of this method.

• Proceedings of the KIEE Conference
• /
• 2006.10d
• /
• pp.34-36
• /
• 2006

In this paper, we will treat the diagnosis problem to accurately determine fault types. The judgement of fault types is accomplished by observing the cluster newly formed with faults and clustering the input current waveforms to intrinsically show the conditions with the dignet that is a clustering algorithm. The types of input current waveforms are, however, constrained during normal operation, though it considers the load character. In case of faults. new clusters are generated outside the clusters. which appear during normal operation, because the input current waveforms of the induction motor are generated by the type which is not observed in case of faults. The diagnosis about the types of faults is essential to building a fault tree about the induction motor, and it removes the causes of the faults using a fuzzy logic. We, first, constitute a fault tree, which connects with the parts and the entire system of the induction motor, and investigate fault modes which can be generated from the fault tree and the relationship of the cause and the effect of each part (of the motor). Also, we distinguish the faults of each part by means of inducing the said of fuzzy relation equations encapsulating the relationship of the fault modes and each part.

• Journal of Korean Society for Quality Management
• /
• v.45 no.1
• /
• pp.25-38
• /
• 2017

Purpose: Hot strip rolling mill consists of a lot of mechanical and electrical units. In condition monitoring and diagnosis phase, various units could be failed with unknown reasons. In this study, we propose an effective method to detect early the units with abnormal status to minimize system downtime. Methods: The early warning problem with various units is defined. K-means and PAM algorithm with Euclidean and Manhattan distances were performed to detect the abnormal status. In addition, an performance of the proposed algorithm is investigated by field data analysis. Results: PAM with Manhattan distance(PAM_ManD) showed better results than K-means algorithm with Euclidean distance(K-means_ED). In addition, we could know from multivariate field data analysis that the system reliability of hot strip rolling mill can be increased by detecting early abnormal status. Conclusion: In this paper, clustering-based monitoring and fault detection algorithm using Manhattan distance is proposed. Experiments are performed to study the benefit of the PAM with Manhattan distance against the K-means with Euclidean distance.