Browsing by Author "Vithanage, D.S."

Now showing 1 - 3 of 3
  • No Thumbnail Available
    Item
    Automated Traffic Violation Detection
    (Uva Wellassa University of Sri Lanka, 2021) Ruwan, A.D.; Vithanage, D.S.
    One of the most serious health risks has been and will continue to be road accidents. The number of deaths and injuries caused by traffic accidents has been proven statistically. Road accident is a most unwanted thing to happen, especially on the pedestrian crossings to a road user, though they happen quite often. Some reasons for the accidents and crashes are due to human errors such as drunk driving, high speed, red light jumping and overtaking on the pedestrian crossing, etc. Among these reasons, especially an overtaking on the pedestrian crossing is one of the most common traffic rules violations in Sri Lanka, and the accidents associate with this violation cause a huge loss to life and property. Although automated techniques for detecting some traffic offenses exist, such as detection of the speed limit and drunk drivers, currently there is no automatic mechanism for the detection of the vehicles which are overtaking on pedestrian crossings. Manual identification of overtaking vehicles on the pedestrian crossing is more critical than anything else because detection of moving vehicles, then tracking and classifying them in real-time in a complicated environment, is extremely tough. Therefore, an accurate and efficient automatic method for detecting traffic violations is a very useful tool for road safety. This paper describes an automatic detection of traffic violation offender on pedestrian crossings. This paper proposed an improved dynamic background-updating approach and a feature-based tracking method to detect overtaking vehicles on the pedestrian crossing. This can fill the gap of manual detection with automatic detection and no labour costs. Thus, it is beneficial in various ways such as the confirmation of road safety. The application is proposed as a mobile application. A complete traffic violation detection system is realized in C++ with Open CV libraries. The accuracy of the system was found as above 73% after the train and validate the model. In conclusion, the developed method can help to detect vehicles that have violated the traffic rules on the pedestrian crossing accurately. Keywords: Traffic Violation; Road Safety; Mobile Application; Manual Detection;
  • No Thumbnail Available
    Item
    Classification of Monkeys for the Automatic Monkey Repellenter using Transfer Learning
    (Uva Wellassa University of Sri Lanka, 2021) Liyanage, K.L.K.S.; Vithanage, D.S.
    Monkeys take over the cultivations looking for foods, and it makes the fields wasted and it has become one of the major problems that farmers have to face in many areas. Furthermore, monkeys attack the normal households, and cities have become a common problem in the present society. The typical method to warding off monkeys is through human involvement such as shouting and lighting torches, air riffles, and fire crackers to repel the monkeys that come to their crops. Moreover, these methods are not easy, and some of them are harmful to monkeys as well as human. Additionally, these methods are familiar to the monkeys. Therefore, even if they leave at those times, they are used to coming back again. This is a lot of time wasting for the farmers, and it is difficult to protect the cultivations whole day. In some scenarios, monkeys attack farmers when they attempt to repel them from cultivations. Due to the damages that made by monkeys to the cultivations in Sri Lanka, there is a huge lose to economy. Therefore, accurate, efficient automatic method for chase different classes of monkeys is very useful tool. This paper describes classification of three classes of monkeys using transfer learning and electronic monkey repellenter to prevent this issue. This can help to protect the cultivation for the farmers whole the day without any labor cost. This can fill the gap of experts with cheap labors and computers or mobiles. The developed application consists of four main steps namely image pre-processing, data augmentation, train the model and visualize the results. Furthermore, this model can identify the three classes of monkeys and repel them from the cultivations by emitting a frequency wave. The accuracy of the system was calculated after visualized the results with the help of the prediction of the labels of the test images, and found as 98%. In the conclusion, the developed method can help farmers to recognize three classes of monkeys and warding off them early and cheaply using the Automatic Repellenter. Keywords: Monkey; Transfer learning; Data augmentation; Cultivations
  • No Thumbnail Available
    Item
    Detect Appropriate Period to Apply Fertilizer for the Tea Plantation Using Image Processing
    (Uva Wellassa University of Sri Lanka, 2021) Yasanthi, H.I.G.E.; Vithanage, D.S.
    Tea cultivation is one of the main income sources of the export field in Sri Lanka and the main sources of employment in the country, employing more than one million workers. Tea inflation is a national challenge. Tea revenue drops several times a year due to the low yield of tea and it is the main threat to the tea industry. The yields can be reduced due to the poor nutrition of the plant. Fertilizer should be applied to the plant on time to maintain the nutrition of the plant and can get a proper harvest. When the plants do not have fertilizer, the fibers in the leaves change, and gradually the color of the leaf change to yellow. Experts' eye observation is the ordinary method to recognize the time to fertilize. Especially, it is hard to recognize the exact period of the fertilizer application to tea plants by eye observation with the help of color changes on leaves for novel tea planters. Hence, they may require to grab the assistance of expertise, which is more expensive. Early recognition of the period of the fertilizer application is the key to avert losses in the quality and quantity of tea products. Therefore, an accurate, efficient automatic method for recognizing fertilizing period is a very useful tool for novel planters among tea cultivators. This research describes the automatic recognition of the period of fertilizer application for the 'TRI 4049' type of Tea plants using image processing technique. This can fill the gap of experts with cheap labor and computers. It is beneficial to get a nutritional harvest for a large estate. The developed application consists of four main steps namely image acquisition, image pre-processing, segmentation, and feature extraction and Classification. The green pixels percentage was calculated to find out the suitable period for the fertilizer application of the Tea plants. The accuracy of the system was found as above73% after the train and validate the model. In the conclusion, the developed method can help novice planters to recognize the most suitable period of the fertilizer application for tea plants early and cheaply. Keywords: Image Processing; Feature Extraction; Classification; Fertilizer Applications