http://dlib.iit.ac.lk/xmlui/handle/123456789/181 2026-04-08T18:12:25Z 2026-04-08T18:12:25Z Pathirana, Hiruni Amanda http://dlib.iit.ac.lk/xmlui/handle/123456789/188 2019-03-05T07:14:30Z 2018-01-01T00:00:00Z

An application to identify Diabetic Retinopathy using digital color fundus photography Pathirana, Hiruni Amanda Diabetic patient population is rising against to the world population. All the diabetic patients are at a risk of developing Diabetic Retinopathy. As author highlights along in this research this circumstance could damage the retina, back of the eye and might lead to complete blindness. Therefore, early detection of this health matter and the treatments are required to avoid this serious complication. Therefore, author has proposed a screening system in order to identify the earliest Diabetic Retinopathy pathology Microaneurysm, during Non-Proliferative Retinopathy which is known as the first stage of Diabetic Retinopathy. The proposed system utilized an image processing approach to preprocess image, Segment and extract features which are appropriate for detection of Microaneurysm regions. Furthermore, machine learning approach will be use to classify the selected regions. Number of feature vectors were identified for the classification. a novel method of segmentation vessels using thresholding approach was attempted.

2018-01-01T00:00:00Z Perera, Jason Archana http://dlib.iit.ac.lk/xmlui/handle/123456789/187 2019-03-05T07:09:18Z 2018-01-01T00:00:00Z

Collision Detection Algorithm Improvement Perera, Jason Archana Collision detection is the process of identifying if two or more objects have made contact with each other. It is a technology that has applications in the fields of gaming, robotics and even medical science. A performance gap has emerged due to rapid developments in the related field of graphic rendering. This gap has been made apparent through reduced efficiency and/or accuracy in existing collision detection algorithms. This project aims to modify the GJK algorithm in order to improve its performance in terms of efficiency, as per the requirements gathered from stakeholders. The resultant algorithm shows minor improvements in efficiency, while experiencing a slightly higher drop in accuracy. While achieving the goal of the research, there is further scope in improving the algorithm’s ability to handle non-convex objects.

2018-01-01T00:00:00Z Jayakody, J. A. Danura Ishara http://dlib.iit.ac.lk/xmlui/handle/123456789/186 2019-03-05T07:06:12Z 2018-01-01T00:00:00Z

A Hybrid Solution for Stock Market Symbol prediction Jayakody, J. A. Danura Ishara When it comes to prediction, it's always a complicated and challenging process. Sometimes it might not helpful to use traditional methods of prediction with your requirement. Stock market prediction has been an interesting area due to its impotency. This report is containing the information based on a research conduct in stock market prediction and suggesting an alternative hybrid system did by using KNN (K Nearest Neighbor) algorithm (unsupervised) and a supervised algorithm which is rich with a higher accuracy level. The system is getting an accuracy of 65% to 75% when using only the KNN algorithm and to improve the accuracy of the system another algorithm has been written and sorting the results coming from the 1st algorithm. When this step was done the accuracy was climb up to 85% to 90% and its different from symbol to symbol. However, the overall accuracy is in between 80% to 90%. All the statistics and graphs are included in the report under the relevant chapters. The proposed system has been evaluated and tested and all the test results, design, implementation and documentations are expressed in an efficient manner.

2018-01-01T00:00:00Z Hisham, Shadir http://dlib.iit.ac.lk/xmlui/handle/123456789/185 2019-03-05T06:11:59Z 2018-01-01T00:00:00Z

“CropRec” - Crop Recommendation System Hisham, Shadir One of Sri Lanka’s biggest livelihoods and a major economic contributor is agriculture. Rich with a tropical atmosphere, Sri Lanka is an ideal island to contribute our agricultural resources to a global scale. With all of these factors in mind, Sri Lanka still has a threat to food security on major crops like Rice. Since the past few years, predicting weather for cultivation for planning has been challenging due to changes of weather patterns as effects of global warming. Although there has considerable amount of research in the domain of agriculture and weather prediction, in Sri Lanka those two haven’t been merged together effectively to provide predictions to the attached stakeholders like farmers. The already available literature is only serving a specific scenario. CropRec is a system that uses large amounts of weather data over 15 years to find a pattern and predict the future weather patterns. The outcome of this process is taken to recommend a suitable crop. This recommendation is sent through a SMS alert to the farmer within a time period helping them to plan their cultivation based on the changing weather patterns. The system was implemented and tested using qualitative and quantitative approaches, evaluation was done on the overall system and completed effectively.

2018-01-01T00:00:00Z