International Journal of Computer Science and Technology

International Journal of Computer Science and Technology | Vol. 13, No. 8, August 2011 | pp. 50–56
DOI: 10.46882/2011/IJCST/000114

Research Article

Title: Automated Sub-Pixel License Plate Localization via Mathematical Morphology and Wavelet Analysis

Names of Authors: Hans-Werner Schmidt¹, Katarzyna J. Kowalski²

Authors’ Affiliations: ¹Institute for Pattern Recognition, Technical University of Vienna, Vienna, Austria; ²Department of Computer Vision, Warsaw University of Technology, Warsaw, Poland

Abstract: Automatic License Plate Recognition (ALPR) remains a key component of modern electronic toll collection, parking access management, and traffic monitoring architectures. Classic localization routines struggle when analyzing video frames captured under adverse illumination conditions, variable camera viewpoints, or complex background clutter. This article details a high-precision Automated Sub-Pixel License Plate Localization (ASPL) algorithm that merges mathematical morphology operators with discrete wavelet transforms (DWT). The proposed method applies vertical morphological edge enhancement to separate license plate text clusters from vehicle bodies before isolating the high-frequency sub-bands of the image via DWT. Sub-pixel interpolation is then performed to resolve blurry borders caused by high-velocity vehicle movement. Evaluations using an empirical dataset of 1,500 outdoor traffic images yielded a 96.8% accurate localization success rate across diverse environmental profiles. The system successfully managed variations in perspective tilt up to ±15°, preserving an average execution speed of 38 ms per frame on a standard 2.4 GHz workstation. This framework offers an industrial pipeline for real-time video enforcement architectures.

Keywords: Image Processing, License Plate Recognition, Mathematical Morphology, Discrete Wavelet Transform, Edge Detection

Manuscript Timeline: Received May 22, 2011; Revised July 05, 2011; Accepted July 20, 2011; Published August 24, 2011.

Citation: Schmidt, H. W., & Kowalski, K. J. (2011). Automated Sub-Pixel License Plate Localization via Mathematical Morphology and Wavelet Analysis. International Journal of Computer Science and Technology, 13(8), 50–56.

International Journal of Computer Science and Technology | Vol. 13, No. 6, June 2011 | pp. 36–42
DOI: 10.46882/2011/IJCST/000112

Research Article

Title: A Scalable Distributed Graph Processing Framework for Large-Scale Social Network Analysis

Names of Authors: Alexander J. Wright¹, Choi Min-Soo²

Authors’ Affiliations: ¹School of Computer Science, Georgia Institute of Technology, Atlanta, USA; ²Department of Computer Engineering, KAIST, Daejeon, South Korea

Abstract: Large-scale digital social networks generate immense relational graphs containing billions of vertices and directed edges, severely testing the memory boundaries of monolithic database engines. Distributed computing architectures must partition these massive graphs effectively while minimizing inter-node communication network overhead during parallel traversals. This study describes a Scalable Distributed Graph Processing Framework (SDGPF) that implements a vertex-centric parallel computational model paired with a 2D balance partition mechanism. The proposed execution engine utilizes an asynchronous message-passing interface to allow independent node computation during PageRank iterations and single-source shortest path operations. Benchmark evaluations conducted over a cluster of 16 virtual machines processing the LiveJournal dataset showed a 43.5% reduction in graph partitioning execution times compared to traditional hash-based data distribution strategies. Network cross-talk dropped significantly, maintaining an execution throughput of 2.4 * 10^6 edges processed per second. The system demonstrated near-linear horizontal speedup scaling up to maximum thread capacity, confirming its industrial applicability for real-time recommendation engines, community detection pipelines, and structural trace tracking in commercial social media environments.

Keywords: Distributed Computing, Graph Processing, Big Data Analytics, Social Networks, PageRank, Parallel Architectures

Manuscript Timeline: Received March 18, 2011; Revised May 02, 2011; Accepted May 21, 2011; Published June 26, 2011.

Citation: Wright, A. J., & Min-Soo, C. (2011). A Scalable Distributed Graph Processing Framework for Large-Scale Social Network Analysis. International Journal of Computer Science and Technology, 13(6), 36–42.

International Journal of Computer Science and Technology | Vol. 13, No. 5, May 2011 | pp. 29–35
DOI: 10.46882/2011/IJCST/000111

Research Article

Title: An Energy-Efficient Clustering Topology for Wireless Sensor Networks Based on Particle Swarm Optimization

Names of Authors: Siddharth S. Rao¹, Elena M. Kuznetsova²

Authors’ Affiliations: ¹Department of Electronic Engineering, Indian Institute of Science, Bangalore, India; ²School of Computer Science, University of St. Andrews, St. Andrews, UK

Abstract: Wireless Sensor Networks (WSNs) are heavily constrained by limited battery capacities at individual sensor nodes, necessitating robust energy-saving architectures to prolong operating lifetimes. Hierarchical clustering algorithms, notably Low-Energy Adaptive Clustering Hierarchy (LEACH), mitigate transmission power demands but often fail to distribute cluster head workloads uniformly. This paper introduces an Energy-Efficient Clustering Topology (EECT) utilizing Particle Swarm Optimization (PSO) to optimize the selection of cluster heads based on multi-criteria metrics. The objective fitness function dynamically weights residual energy reserves, spatial distances to the base station, and intra-cluster distances. Simulations using MATLAB 2010b demonstrate that EECT balances power depletion across the network topology efficiently. Compared against conventional LEACH, the proposed PSO-driven mechanism delayed the first node dead state by 34.2% and increased total packet delivery metrics by 21.5%. Total power dissipation parameters stabilized at 0.12 J per round under dense network states (node count >= 200). Statistical t-tests validated the operational advantages of EECT with an alpha level of 0.01. This framework yields a scalable, high-durability approach for structural health monitoring installations and environmental tracking applications where manual battery replacement remains impossible.

Keywords: Wireless Sensor Networks, Cluster Optimization, Particle Swarm Optimization, Network Lifetime, LEACH

Manuscript Timeline: Received February 11, 2011; Revised March 25, 2011; Accepted April 14, 2011; Published May 22, 2011.

Citation: Rao, S. S., & Kuznetsova, E. M. (2011). An Energy-Efficient Clustering Topology for Wireless Sensor Networks Based on Particle Swarm Optimization. International Journal of Computer Science and Technology, 13(5), 29–35.

International Journal of Computer Science and Technology | Vol. 13, No. 4, April 2011 | pp. 22–28
DOI: 10.46882/2011/IJCST/000110

Research Article

Title: A Peer-to-Peer Video Streaming Optimization Strategy Using Token-Based Incentive Mechanisms

Names of Authors: Giovanni R. Rossi¹, Kim Ji-Hoon²

Authors’ Affiliations: ¹Department of Information Engineering, University of Padua, Padua, Italy; ²School of Computer Science and Engineering, Seoul National University, Seoul, South Korea

Abstract: Peer-to-Peer (P2P) live video streaming configurations face significant quality-of-service degradation due to "free-riding" behaviors, where participants consume stream bandwidth without uploading segments back to neighbors. To resolve this structural problem, this research presents a Token-Based Incentive Mechanism (TBIM) for mesh-based P2P streaming systems. The proposed protocol assigns digital cryptographic tokens to nodes as a reward for successful upload bursts, which are then spent to pull high-priority video chunks during playback sequences. Nodes lacking sufficient tokens are throttled into basic buffer tiers, encouraging active data contributions. Simulations conducted over synthetic topologies involving 10,000 nodes demonstrated that TBIM reduces average video startup delay bottlenecks by 35.4% and decreases frame freezing frequencies by 44.1% compared to un-incentivized BitTorrent-like streaming architectures. The overall system upload resource usage achieved an equilibrium index of 0.88, indicating uniform peer cooperation. The computational overhead linked to cryptographic token validation remained under 2.1 milliseconds per transaction, proving that this framework can scale to manage immense concurrent consumer streams on modern commercial video-on-demand deployments.

Keywords: Peer-to-Peer Networks, Video Streaming, Incentive Mechanisms, Free-Riding, Quality of Service

Manuscript Timeline: Received October 15, 2010; Revised November 20, 2010; Accepted December 05, 2010; Published April 24, 2011.

Citation: Rossi, R. G., & Ji-Hoon, K. (2011). A Peer-to-Peer Video Streaming Optimization Strategy Using Token-Based Incentive Mechanisms. International Journal of Computer Science and Technology, 13(4), 22–28.

International Journal of Computer Science and Technology | Vol. 13, No. 3, March 2011 | pp. 15–21
DOI: 10.46882/2011/IJCST/000109

Research Article

Title: An Adaptive Intrusion Detection System Using Artificial Immune Systems and Clonal Selection Algorithms

Names of Authors: Christopher L. Davis¹, Zainab M. Al-Balushi²

Authors’ Affiliations: ¹Department of Computer Science, University of Toronto, Toronto, Canada; ²Information Security Department, Sultan Qaboos University, Muscat, Oman

Abstract: Cyber infrastructures suffer continuously from zero-day threats and network intrusions that easily bypass signature-based firewall systems. This paper describes an Adaptive Intrusion Detection System (AIDS) modeled after biological Artificial Immune Systems (AIS), specifically implementing the Clonal Selection Algorithm (CSA). The system treats incoming network packets as antigens and constructs an evolving population of digital antibodies (detectors) designed to recognize anomalous communication profiles. An affinity maturation cycle optimizes these detectors through targeted mutations, ensuring rapid convergence toward novel intrusion footprints. The proposed model was validated using the standard KDD Cup 99 benchmark dataset. The experimental findings indicate a true positive anomaly detection rate of 97.4%, paired with a low false positive alert frequency of 0.65%. The system demonstrated autonomous self-learning capabilities, discovering modified polymorphic network scans within 180 seconds of exposure without requiring administrative signature revisions. This framework exhibits strong computational resilience and high real-time throughput execution, making it suitable for deployment at the boundary routers of enterprise wide-area networks.

Keywords: Intrusion Detection System, Artificial Immune Systems, Clonal Selection, Cyber Security, Network Anomaly

Manuscript Timeline: Received October 10, 2010; Revised November 18, 2010; Accepted December 01, 2010; Published March 18, 2011.

Citation: Davis, L. C., & Al-Balushi, M. Z. (2011). An Adaptive Intrusion Detection System Using Artificial Immune Systems and Clonal Selection Algorithms. International Journal of Computer Science and Technology, 13(3), 15–21.

International Journal of Computer Science and Technology | Vol. 13, No. 2, February 2011 | pp. 8–14
DOI: 10.46882/2011/IJCST/000108

Research Article

Title: Performance Evaluation of NoSQL Databases Versus Relational Databases for Big Data Web Applications

Names of Authors: Thomas B. Miller¹, Anna K. Novak²

Authors’ Affiliations: ¹Software Engineering Institute, Carnegie Mellon University, Pittsburgh, USA; ²Department of Software Engineering, Warsaw University of Technology, Warsaw, Poland

Abstract: Modern web platforms generate immense amounts of semi-structured and unstructured data, testing the horizontal scalability limits of traditional Relational Database Management Systems (RDBMS). This research paper evaluates the performance profiles of a leading document-oriented NoSQL database (MongoDB) against a standard relational database instance (MySQL) under enterprise workloads. We constructed an intensive benchmarking test suite modeling CRUD (Create, Read, Update, Delete) query operations across data configurations scaling up to 10,000,000 entries. Metrics monitored included read/write throughput, CPU usage, memory foot-printing, and execution latencies. The empirical results demonstrate that MongoDB outperforms MySQL in write operations by a factor of 3.8, preserving an average write latency of 14 milliseconds under concurrency peaks of 5,000 requests per second. For complex multi-table JOIN operations, MySQL displayed superior execution stability, whereas MongoDB experienced high memory inflation due to its denormalized schema mapping. Our data concludes that while NoSQL frameworks offer superior horizontal scaling and flexibility for high-volume unstructured logs, RDBMS remains indispensable for systems requiring strict ACID (Atomicity, Consistency, Isolation, Durability) transaction assurances.

Keywords: NoSQL Databases, RDBMS, MongoDB, MySQL, Big Data, Benchmarking, Database Performance

Manuscript Timeline: Received September 14, 2010; Revised October 22, 2010; Accepted November 04, 2010; Published February 20, 2011.

Citation: Miller, B. T., & Novak, K. A. (2011). Performance Evaluation of NoSQL Databases Versus Relational Databases for Big Data Web Applications. International Journal of Computer Science and Technology, 13(2), 8–14.