Browsing by Author "Ranasinghe, R. A. J. B."

Filter results by typing the first few letters
Now showing 1 - 1 of 1
  • No Thumbnail Available
    Item
    IoT-enabled intelligent pedestrian crossing signal light system with violation tracking
    (Faculty of Science, University of Kelaniya Sri Lanka, 2024) Rupasinghe, R. A. I. M.; Ranasinghe, R. A. J. B.; Moragoda, Y. G. D.; Navodya, W. D. I.; Premasiri, R. H. M. D.; Chethana, E. J. K. S.; Seneviratne, J. A.; Gunawardana, K. D. B. H.
    The urban pedestrian crossing environment presents numerous challenges in ensuring the safety of pedestrians and maintaining smooth traffic flow. Traditional pedestrian signaling systems operate on fixed timings and have limited capabilities, making it difficult to manage the complexities of modern urban traffic effectively. This research introduces an innovative system for pedestrian crossing signal lights integrated with violation tracking and real-time data analytics to improve pedestrian safety and smooth traffic flow. This encompasses computer vision for pedestrian detection, machine learning (ML) for predictive analysis, adaptive signal light timers, sirens for violation deterrence, and IoT components for seamless real-time operation. The presented methodology combines real-time pedestrian detection, adaptive signal light timing, weather detection, and IoT integration so that all these subsystems work smoothly. The issues resolved include integrating image processing with hardware, selecting an efficient pedestrian detection model, optimizing camera angles for accurate detection, and transitioning from an Arduino to a Raspberry Pi 4 Model B. The Raspberry Pi offered better processing power, enabling faster and more complex data handling. A case study was done at a location proximate to the University of Kelaniya, and the average crossing time taken for the pedestrian crossing was recorded as 18.5 seconds, which can be factored using databases with larger data sets and simple ML models based on the day of the week. The issues that were resolved include integrating image processing with hardware, selecting an appropriate pedestrian detection model such as a Convolutional Neural Network (CNN) that works well within the outdoor environment, setting optimal camera angles for accurate pedestrian detection, and transitioning from an Arduino to a Raspberry Pi 4 Model B for enhanced processing capabilities. Integrating image processing with hardware posed challenges due to the need for real-time data transmission and processing, which required seamless communication between the software and hardware components. The pedestrian detection model was chosen based on its accuracy, speed, and ability to perform well in varying lighting and weather conditions. The transition to the Raspberry Pi 4 Model B, with its superior processing power and memory compared to the Arduino, allowed the system to handle more complex tasks, such as real-time data analysis and multiple input streams, significantly improving performance and efficiency. A custom dataset of overhead views of pedestrians was created, with images acquired from a similar environment within the university, manually labelling the images and achieving an 80% accuracy after training on Google Collab. The real-time data processing system is vital in making dynamic signal timing changes, tracking violations to encourage safe pedestrian behavior, and managing pedestrian and vehicle traffic flow. These findings endorse the broader adoption of intelligent systems, for innovative city projects toward safer and more efficient urban environments.