15th November 2022 http://ir.lib.seu.ac.lk/handle/123456789/6282 2026-05-07T00:42:11Z 2026-05-07T00:42:11Z http://ir.lib.seu.ac.lk/handle/123456789/6361 2023-04-18T23:16:04Z 2022-11-15T00:00:00Z

Preliminaries

2022-11-15T00:00:00Z Akmal Jahan, M. A. C http://ir.lib.seu.ac.lk/handle/123456789/6360 2023-04-18T23:16:04Z 2022-11-15T00:00:00Z

Image quality estimation of contactless infant foot prints using enhancement filters Akmal Jahan, M. A. C Biometric systems have been using physiological & behavioral traits of humans for the identification or verification of an individual. Most biometric systems have been developed for adults in several applications particularly, in civilian and forensic domains. There is a lack of well-defined systems for infant identification or verification, and newborn recognition has got attention in recent years. There are several applications that have a requirement to use of infant recognition particularly, infant tracking, identifying a missing child, child swapping, etc. It is observed that image acquisition for infant biometric systems does not follow the same procedures as for adults. Since infants have different laying positions, acquiring face, fingers and eye-related biometric is difficult. However, footprints can be easily collected using some mobile-based devices even if the infants are in sleeping positions. When dealing with such images, applying enhancement filters without affecting the image quality is a crucial step. In this work, the quality of acquired images is comparatively evaluated. A set of enhancement filters have experimented with original and enhanced images, and the quality of images is measured using image quality metrics. From the analysis, the Jerman enhancement filter and unsharp masking show better-quality preservation and slight improvement in performance with infant footprint biometric system.

2022-11-15T00:00:00Z Haneesa, A. L. Manikandan, S. http://ir.lib.seu.ac.lk/handle/123456789/6359 2023-04-18T23:16:03Z 2022-11-15T00:00:00Z

Virtualized cloud optimizer for predicting cloudlets Haneesa, A. L.; Manikandan, S. Virtualization is one of the important factors in cloud computing to select the cloud service and model. Different cloud applications are providing services at the platform level and it varies depending upon application services. Load balancing is a major issue while accessing the resources. Cloudlet is the device or machine or computing tool to access the resources when it is required. An optimizer is required to predict the user profile based on active users and cloud services. In this research, the proposed method model identifies a solution for efficient load balancing by considering factors such as processing time, and response time to reduce carbon footprint in the cloud computing environment The Proposed algorithm is based on the genetic algorithm Ant colony optimization which uses path cost and threshold. The major components are the User Base, Datacenter selector, Virtual Machine (VM) selector and allocator and Efficiency analyzer. In this work, we provide virtualized Cloud Optimizer for selecting the cloud services and ranking the Cloudlets. The experiments are done by using CloudSim and the dataset is selected from the UCI repository.

2022-11-15T00:00:00Z Akmal Jahan, M. A. C. Vithusa, B. http://ir.lib.seu.ac.lk/handle/123456789/6358 2023-04-18T23:16:03Z 2022-11-15T00:00:00Z

Depression analysis on users of social network using machine learning algorithms Akmal Jahan, M. A. C.; Vithusa, B. Depression is a serious conditioned mental disorder that has significant effects on the quality of life of a person. Internet sources state that the number of people suffering from depression is getting increased day by day and it affects teenagers more than adults. Our project in this work is to find the status of a user's posts or comments which show depression mood or not, using different types of machine learning classification algorithms. The dataset is collected from users who share their day-to-day status on social networks. The dataset is preprocessed and tokenized to make it compatible to feed into different types of algorithms such as Naïve Bayes, Random Forest, Linear Regression, and Support Vector Machine. During the process, the accuracy level of each algorithm is compared and the algorithm with the highest accuracy has chosen as suitable to process further prediction.

2022-11-15T00:00:00Z