Browsing by Author "Wimaladharma, S.T.C.I."
Now showing 1 - 14 of 14
Results Per Page
Sort Options
Item Artificial Neural Network based Signature Recognition and Verification System(Uva Wellassa University of Sri Lanka, 2013) Prabodith, N.P.C.; Wimaladharma, S.T.C.I.The signature is an important biometric attribute of a person which can be used to authenticate the personal identity because of its uniqueness for each person. Now a day the personal signature has a significant value in day to day works. Because of its uniqueness there is a certain pattern which can be identified by extracting certain unique features. Though in present days signatures are using as the valid authentication mechanism, some peoples are trying to imitate another person’s signature to achieve some narrow goals. It is also common sight in Sri Lanka place like banks, government organizations, universities etc. Therefore, it is essential to introduce a high accuracy validation mechanism for personal authentication. The proposed system uses an efficient image processing and feature extraction methods as well as well-trained neural network system. Methodology Implementation of the system mainly based on two phases, Training phase and recognition phase. Several technologies, Programming languages and Libraries are used to design and implement the system. In the training phase there are several steps to be completed by the system before the training get started. In pre-processing activities, system is mainly focusing on background elimination, cropping (crop signature image according to the signature bounds), thresholding, thinning, and image width transformation (Abikoye et al., 2011). Feature extraction is one of the important parts of the system because powerful features directly affect to the accuracy of the final output. In here firstly, system will collect some global features such as pixel density, width to height ratio, maximum horizontal count and maximum vertical count. Then system will extract some unique points based on vertical splitting and horizontal splitting (Ashwiniet al., 2012). Those feature points are related with the image geometric centre point. After the feature extraction process is completed and then system normalizes all the features and added them to the input vector. Main purpose of the normalization is to convert values into an acceptable range for the neural network (range between 0 and 1). Then, the normalize data are used to train the Neural Network. In the Recognition phase, all the steps up to feature normalization are accomplished and those features are compared with trained Neural Network. Finally, the system will generate a unique value and which reveals the acceptance and rejection of the relevant signature.Item An Artificial Neural Network Model for Rainfall Prediction on the Basis of Agro-ecological Regions (AER) in Sri Lanka(Uva Wellassa University of Sri Lanka, 2016) Surin, A.A.D.; Wimaladharma, S.T.C.I.There are many studies done for testing the capability of rainfall forecasting using artificial neural networks. In Sri Lankan context, this study presents another model that uses Agra Ecological Regions as the basis instead of currently popular administrative districts. An Agra Ecological Region represents a particular combination of climate, soil and relief, so this study tries to find out the advantage and accuracy of giving a weather report per each Agra Ecological Region. For the purpose, two locations from WL4 region, Bandaranayaka International Airport (UM) and Galle weather station (WM043495) were considered. The Artificial Neural Network was trained using 10 years of daily data from each selected station using split-sample method. The accuracy of predicting rainfall probability, rain or non-rain status is tested. Selected Artificial Neural Network is a pattern recognition, feedforward based neural network, which uses hyperbolic tangent function as the transferring function and back propagation method as the training algorithm. Mean temperature, mean dew point, mean humidity, mean sea level pressure, mean visibility and mean wind speed were selected as inputs for the neural network. According to the binary classification strategy, the status of each day was defined as rain or non-rain. Mean value of each weather parameter was compared with each locations, and there were no significant difference between the actual values and the predicted values. The estimated error rate of making predictions using the proposed model was less than 35%. According to the results obtained, this model tends to be giving more precise results compare to the district based weather forecasts. Keeping several classification parameters as targets and using more observatory points are the recommendations to improve the accuracy of the results of this forecasting model. Keywords: Artificial Neural Network, binary classification, rainfall forecasting, climate, pattern recognitionItem A Conceptual Framework for Flood Early-Warning System for the Lower Flood Plains of Kalu Ganga Using Twitter Crowd-sourcing and Internet of Things(Uva Wellassa University of Sri Lanka, 2018) Ranasingha, T.B.B.M.; Chathuranga, D.W.R.; Gunasekara, K.M.; Wimaladharma, S.T.C.I.Flood is one of the common natural disasters in all over the world. Sri Lanka has two major monsoons: Southwest (May to September) and Northeast monsoon (December to February) causing for floods along the one third of low lands. The objective of this study is to develop a framework that is relatively credulous community based flood early-warning system for the populous areas near by the riverbanks of Kalu Ganga in Kalutara district in Sri Lanka. The study focuses on two major affected areas that are Palindanuwara and Agalawatta. There are six major tributaries joined to the river between Kalutara and Ratnapura making Ratnapura as a considerable catchment area for the river. Therefore, the system collects real-time bulletins, associated with predefined keywords and posted by the Twitter crowdsourcing living in Ratnapura and surroundings, using Twitter stream API. It uses hashtags to filter locations and performs the text analysis. While the percentage of likelihood of flooding is estimated based on the number of positive twitters, the possibility of a flood is verified using the incline or decline trend of the water levels collected from Ground Control Units located in flood risk areas. If the Ground Control Unit confirms that there is a possibility of a flood, the system generates a flood-positive alert that can be used to warn people living in those areas. The proof of the concept was successfully tested by simulating the flood situation using the Ground Control Units. Thus, it can be concluded that the Twitter crowdsourcing can be effectively used to warn the community about upcoming flooding situations beforehand.Item Enrich the Awareness of Road Rules in Sri Lanka for the Tourists Using a Game-Based Driving Learning System(Uva Wellassa University of Sri Lanka, 2020) Sapumohotti, C.H.V.; Sampath, A.G.A.; Wimaladharma, S.T.C.I.; Sampath, J.D.B.Tourism is one of the major economic sectors which affects the country’s Gross Domestic Product (GDP) and Gross National Product (GNP). Traffic safety within the cities which are highly attracted by tourists is becoming an important problem in the country. Based on investigations it is identified that unawareness of rules and lack of practice to hazardous incidents are the main reason for that. Although most of the tourists have a valid driving license, they may be having very low knowledge about the road rules in Sri Lanka. So, when they are driving in the country, they need to have a proper understanding or an awareness about the road rules in the country. The primary objective of the system is to provide the tourist a platform to allow practicing to overcome the hazardous challenges and to have a self-evaluation about their knowledge about the road rules in the country through a point-based method defined upon the rules, road conditions and driving ethics established in the country. To provide a realistic environment that is similar to the country, virtual environments are modeled based on different criteria. Through the usage of Artificial Intelligence techniques like non-player characters and objects, the reality of the environment was enhanced. It can be concluded that training the learners in a virtual environment that similar to the real environment with a proper assessment of their awareness of the rules and road signs, and driving ethics will solve most of the problems we face today. Keywords: Driving game, Tourism, Simulator, Rules, Traffic safetyItem A Game-Based Driving Learning System for Sri Lankan Driving Learners to Enrich the Awareness of Road Rules(Uva Wellassa University of Sri Lanka, 2019-02) Sapumohotti, C.H.V.; Sampath, A.G.A.; Sampath, J.D.B.; Wimaladharma, S.T.C.I.Traffic safety is becoming an important problem in most of the countries. Based on investigations it has been identified that the unawareness of road rules, lack of practice of sudden reactions in hazardous situations are the major causes for accidents. Though there are many driving simulators available, most of them have not addressed the road rules and hazardous incidences that a driver must be aware. Also they are lacking of a proper evaluation of the driving skills and awareness of the driver. Primary objective of the system is to provide a driving learning platform for the learners, trainers as well as evaluators to overcome the existing challenges, which has mainly focused on creating a virtual environment to facilitate the training and testing process in the local context and main areas of violating road rules and regulations by drivers are taken into account. In order to provide a realistic road environment, virtual environments are modeled based on different criteria. Artificial Intelligence techniques like non-player characters and objects, are employed. Through that, the responsiveness and intelligent behavior of the simulator has been improved. One of the major components of the simulator is the driver evaluation: a point based method defined upon the rules, road conditions and driving ethics established in the country. Further, the virtual environment provides all the road conditions available, countryside as well as the urban traffic conditions with different weather conditions. The effectiveness of the developed simulator is measured by allowing a selected group of learners to use the simulator for a specific period and assess their driving skills in a real driving environment. It can be concluded that training the learners in a virtual environment that similar to the real environment with a proper assessment of their driving skills, awareness of the rules and road signs, and the driving ethics will solve most of the problems we face today.Item Identification of Anomalous Clients’ Request by Analyzing Server Log File using Apache Hadoop Framework and Tableau(Uva Wellassa University of Sri Lanka, 2019-02) Bavathuja, V.; Raahini, S.; Ramashini, M.; Wimaladharma, S.T.C.I.Information systems provide information about its state and operation in the form of log records. These records are composed of log entries containing information related to a specific event, which can be related to security. Potential security breaches can be revealed by analyzing log files and looking for anomalies that occurred at a certain time during the device operation. Log files from proxy server of Uva Wellassa University of Sri Lanka will be analyzed using Hadoop Framework and Apache Pig in order to identify anomalous clients’ Request. Anomalous clients’ request identification refers to the problem of finding pattern in data that do not conform to expected behavior. These nonconforming patterns are often referred to as anomalies, outliers or exceptions in different application domains. Log files of a proxy server are created and maintained by the server itself and analyzing theses files will offer a valuable insight into server usage while they can be used in various applications, such as detecting intrusions on the web. The log files will be stored in Hadoop Distributed File System. Data preprocessing and analyzation will be done using Apache Pig: a platform for analyzing large data sets. The analyzed data will be reported through Tableau dashboard. According to the research study, the total number of records after cleaning is 817,426 and 856 unique IP addresses have accessed the proxy server from the period of Thursday, 26 April 2018 01:14:48.138 to the period of Friday, 27 April 2018 10:31:23.834. Several findings including the total visits and bandwidth were found and displayed using graph and charts. This information along with other findings can be applied to find solutions for many legitimate problems such as, user/customer behavior analysis, etc.Item An Intelligent Postal Mail Sorter: Sinhala Hand Written Address Recognition Method Using Geometric Feature Extraction Technique and Artificial Neural Network(Uva Wellassa University of Sri Lanka, 2018) Sri Darshana, B.P.S.R.; Attanayake, A.M.U.L.; Perera, A.A.L.A.C.; Wimaladharma, S.T.C.I.The main objective of this study is to develop a methodology to recognize Sinhala handwritten characters that can be used in postal mail sorting. The Department of Posts, Sri Lanka uses the manual sorting mechanism, while most of the developed countries are using automated sorting machines. The main reason for not having such types of machinery in local postal collecting and distribution centers is the initial cost of implementation. The machines have to be tailor-made due to the localized language. The proposed methodology is based on the geometric feature, such as Corner detection, Curve fitting and Edge detection, extraction technique and Artificial Neural Network backpropagation technique. The benchmarking of the classification system is carried out using 34 Sinhala characters that are mostly related to the district names. The neural network consists of three layers, where the input layer with 108 input nodes, the output layer with 34 nodes and a hidden layer of 78 nodes. The training and testing are performed by 850 characters and 510 characters, respectively. The accuracy of the system is around 78% of giving a correct answer. The resultant set of characters then be extracted and used to control the sorting machine. In order to prove the concept, an embedded system is developed using Arduino microprocessor. The sorting mechanism is simulated by using a servomotor that indicates the relevant mail bucket using a rotating arm.Item Location Based Exploratory Decision Support Approach for Midwifery and Grama Niladhari Divisions in Hunuwala-North: Ratnapura District, Sri Lanka(Uva Wellassa University of Sri Lanka, 2018) Wickramasinghe, W.B.P.L.; Ariyarathna, K.L.G.A.M.; Wimaladharma, S.T.C.I.Spatial information is being used as a supportive component for the process of decision making in various disciplines and applications. Generally, the governing activities, facts of citizens and properties, and natural and man-made phenomena are associated with locations. In Sri Lanka, the smallest administration unit is the Grama Niladhari division, whereas the midwife is the closest health care person that the community encounters. Most of the time two divisions are overlapped with each other resulting that they collect common facts about citizens. All the data about villages or citizens collected by Grama Niladharies and midwives become the data sources to make decisions by the top-level officers. The main objective of this study is to develop a location based (spatial) decision support approach for multiple criteria decision model with geo-visualization for decision making officers in various government sectors such as divisional and district secretaries, top-level officers of healthcare sector and their upper administrative levels: Its architecture consists of three major components namely spatial layer, attribute layer and the criteria layer. Each attribute record is associated with at least one spatial record resulting to a geospatial database, which has citizen level data, with predefined rules and criteria compiled according to the administrative policies and healthcare rules and regulations of the government. A proof of concept is developed and tested with the actual data. Therefore, it is proven that the introduced approach has a significant effect for the decision makers to make cognitive decisions rather than emotional decisions.Item Recommending a Usability Practices for Websites Developers and Designers(Uva Wellassa University of Sri Lanka, 2016) Madushani, T.D.; Wimaladharma, S.T.C.I.Websites are fundamentally designed to work for all people. Using these websites, people meet different kind of goals. When the web meets this goal, it is accessible to people with a diverse range of hearing, movement, sight and cognitive ability. Website creating which is compatible with accessibility standards is kind of best practices. These standards enable to extend group of information users such as people with disabilities, people who encounter incompatibility and technology problems. Considering the problems, World Wide Web Consortium (W3C) proposed worth standards called Web Content Accessibility Guidelines (WCAG). WCAG developed through the W3C process in cooperation with individuals and organizations around the world with a goal of proving a single shared standard for web content accessibility that meets the needs of individuals, organizations and governments internationally. A-Tester and 508-checker version 1.4 are the tools used for evaluation and they follow the guidelines from WCAG. When website is 508 compliant, they are accessible to all users. Using 508 checker can quickly check web page for 508 compliance and learn more about how to become 508 compliant across the entire organization. A-Tester checks the pre-enhanced version of a web page designed with progressive enhancement against Evaluera's "WCAG 2.0 Level — AA conformance statements for HTML5 foundation mark-up". A selected set of websites from all over the world are evaluated using above tools and found the most common accessibility issues. Finally, an accessibility standards checklist is developed in order to minimize the mistakes made by the developers. Keywords: Web usability, Accessibility, Disable people, Web accessibility standardsItem Smart Food Safety Management Framework for Small Scale Restaurants(Uva Wellassa University of Sri Lanka, 2019) Sandipani, H.A.D.C.L.; Dharmarathna, E.K.G.P.U.; Wimaladharma, S.T.C.I.; Abeyrathne, E.D.N.S.Food safety inspection is a crucial factor in small scale restaurants, to prevent food borne illnesses among the consumers. HACCP is the international tool to manage food safety effectively which can be used as a unique protocol to assure the food safety in any food company. However due to the limited time as a local public health inspector, food safety is a minor concerned subject in their field of working. Objective of this study is to enhance the existing Sri Lankan food inspection process to Badulla area through a developed risk-based food inspection system, which analyze, diagnose and implement main principles of food safety. Interviews with health professionals and pilot survey for small scale restaurants (30) were conducted to understand the existing food inspection programme. Based on that pilot survey, among the targeted group of food handlers most were lack of knowledge on food safety and sanitation. Hence, assessment model based on HACCP for food safety inspection in small scale restaurants was developed. In order to facilitate the end-users to use this developed model, an Android food safety application which consists of optimized user interfaces and offline database was developed. Prototypical development was achieved the user satisfaction in the field level due to its effectiveness and accessibility. As a result, it assures the food safety in small scale restaurants and has potential to improve the food safety practices in food services in the areas covered by the national hygienic and sanitary regulations. Furthermore, “big data” collection through this mobile application can be used for further data analysis, which creates multiple research opportunities.Item Smart Reply Generation for SMS Using Natural Language Processing(Uva Wellassa University of Sri Lanka, 2020) Kaumada, M.W.S.; Sumanasekara, S.S.; Jayasekara, N.E.C.; Wimaladharma, S.T.C.I.The use of Short Message Service (SMS) is increasing due to the rapid increase in mobile phone usage and the simplicity in sending SMS messages. With the increasing complexity of human lives, people are seeking more efficient activities to save time. This research proposes an end-to-end method that automatically generates short responses known as Smart Replies by identifying the content of an SMS using natural language processing. There are a few pieces of research done on the topic of Smart Reply. Most of them are carried out for the emails. And the efficiency and the size of those existing models cannot be used in an offline mobile device. The application will use Natural Language Processing to process an incoming message and then uses a neural network to predict the most likely responses which will allow us to send it directly or edit it before sending it to the recipient. The Ubuntu Corpus dataset was used for training and testing the model by analysing its properties. It is identified that there are three main approaches: TF-IDF, Recurrent Neural networks (RNN), and Long Short-Term Memory (LSTM) that can be used in the model. After a performance test, identified the most suitable approach is LSTM. Accordingly built a Sequential Neural Network with a Dense with sigmoid activation using LSTM. Finally, extract the highest three responses from the trained model to show in the SMS application. This proposed model achieved around 92% percent of accurate results and it can be used offline and also it is a lightweight file that can be easily handled in a mobile device. Keywords: Smart reply, SMS, Natural language processing, Long short-term memory, Sequential neural networkItem Smart SMS Classification for Android Operating System Using Natural Language Processing(Uva Wellassa University of Sri Lanka, 2020) Sumanasekara, S.S.; Kaumada, M.W.S.; Jayasekara, N.E.C.; Wimaladharma, S.T.C.I.The use of Short Message Service (SMS) is increasing as more people exchange SMS messages very frequently due to the rapid increase of mobile phone usage and the simplicity in sending SMS messages. However, this has led to an increase in mobile device attacks using SMS Spam. The two main categories of SMS Messages are spam messages and ham (legitimate) messages. Up to now, several kinds of research were done on SMS classification but all of them are on spam filtering techniques by using various algorithms and machine learning techniques. In this paper, we present a novel approach that can detect and filter both spam and ham messages into a better organization under six different predefined categories named as Primary for legitimate messages, Bank and Finance, Social and Web, Promotions, Service Provider Messages, and Spam Messages by using Natural Language Processing for Android Operating System. A smart messaging application that can properly organize SMS into categories will help to identify the SMS easily as they are classified under different tabs. Even though SMS can be identified and categorized manually with little or no effort by people, it remains difficult for mobile phones. A dataset is created according to the Sri Lankan context and various experiments are performed to evaluate the performance of the SMS Classification. Initially, the features were selected based on the behavior of messages and extracted the features from the dataset to get the feature vectors. Naive Bayes and Support Vector Machines algorithms were used to select the best classification algorithm. With the highest accuracy rate, the Support Vector Machines algorithm is selected to train the model while k-Fold cross-validation is used to perform the validation. Our proposed approach achieved a 93% accuracy rate and the model is deployed in the Android environment and its performance is confirmed using a proof of concept. Keywords: SMS classification, Natural language processing, Support vector machines, Naive bayes algorithm, AndroidItem Smart Tour Planner for Sri Lanka(Uva Wellassa University of Sri Lanka, 2019-02) Kumar, P.; Deemantha, S.P.S.; Lakmal, E.K.H.; Ariyadasa, H.M.S.N.; Wimaladharma, S.T.C.I.Tourism industry is an asset to Sri Lankan economy. It contributes around 5% in national revenue. The development of tourism industry is significantly slow as compared to the other tourist countries because we are not using new technologies in this field. In recent years, several applications are developed for the tourism industry. However, those applications are only used to provide the information about the places. The most important thing when it comes to going for a tour is to manage the time. Moreover, managing time at an unknown place is very difficult. Climatic changes also affect the tourists to visit some places. To overcome these issues, we have proposed an Android based mobile application that can help tourists to plan their tour to Sri Lanka before arriving here. The application covers 5 major areas with the methodology. 1) Finding the shortest route using “Nearest Neighbor algorithm” that covers the as many as possible places to visit with respect to the tourist’s budget, time and tour type. 2) An intelligent system that records the time spend at a place, then by using machine learning algorithms using Google’s “TensorFlow” that can predict the time needed to visit that particular place for new tourists. 3) A schedule for tourists that they have to follow during their tour that helps to manage time. 4) Alerting the tourists for any emergency situations (flood, tsunami, Land sliding) using crowdsourcing. 5) Comment summarization and sentiment analysis that can give a brief idea about the places those tourists are planning to visit. At last, user gets a scheduled tour that he can follow up during visit to Sri Lanka. The evaluation of this application depends upon the time and money saved due to scheduling of the tour, and that saved time is use to visit some more places. This application helps the tourists to plan their tour easily that significantly increase the tourism in the country and will positively affects the country’s economy.Item Sri Lankan Vein Graphite Classification Using Image Processing and Neural Network(Uva Wellassa University of Sri Lanka, 2016) Wickramasinghe, E.M.C.G.; Cooray, J.T.; Wimaladharma, S.T.C.I.Image classification is an essential task in pattern recognition applications. Rock and mineral images are some of the typical examples for natural images, and their analysis is more important in rock and mineral industry. Ore mineral classification is based on specific visual descriptors extracted from the images. These textures are used to identify their visual similarity and categorise them accordingly. This research primarily addresses the problem of automatic measurement of graphite ore textures by image analysis in a way that it is relevant to mineral processing in Kahatagaha Graphite Lanka Limited, Sri Lanka. Specifically, it addresses three major hypotheses: Automatic separation of graphite ore by image analysis provides a feasible alternative to manual curing by mineralogists and labourers, Image analysis can quantify process mineralogy by physical parameters and Image analysis provides potential benefits to process mineralogy and better retains the information of manual logging. Traditionally, minerals are visually recognized and manually outlined prior to the digitizing and subsequent analysis. The preciseness of the outcomes is affected by the conventional methods. This limitation can be overcome by using multichannel methods of classification with Artificial Neural Network, in which the minerals in multichannel digital images are accurately recognized based on their unique spectral or elemental signatures, established by a training stage prior to classification. The technique is applied here for model analysis of images, which are digitized using a standard digital camera. In all case studies of the analysis of graphite lumps, the resulting mineral modes are sufficiently precise to identify significant compositional heterogeneities between groups of samples. This model can be readily applied to automated vein graphite ore classification in mineral processing industry. Keywords: Mineral classification, Image processing, Neural network, Vein graphite classification