Amarasooriya, P.M.D.S.Peiris, M.P.P.L.Herath, H.M.D.S.2023-11-132023-11-132023-03-252961-5011https://rda.sliit.lk/handle/123456789/3554In recent years, the number of vehicles in use has shown a steady increase, leading to a clear demand for larger parking areas. However, the traditional methods for detecting occupancy of slots in smart vehicle parking areas are no longer feasible due to the high cost of sensors and the need to monitor larger areas. In response to this challenge, the present study aims to propose a cost-effective, fast, and accurate solution for updating and indicating the real-time number of free parking slots in a parking area. Specifically, the proposed solution utilizes video footage from a camera as the input device and applies the YOLO v3 object detection algorithm for image processing to detect the coordinates of both parking lots and parked vehicles separately. To train and evaluate the model, we used the PKLot database as the dataset and tested the model's performance under different weather conditions. The proposed model achieved an average performance of 88.01%, with the highest performance demonstrated on sunny days and the lowest performance recorded on rainy days.enConvolutional Neural Networkimage processingparking space detectionshortest path algorithmsmart parking systemArticlehttps://doi.org/10.54389/TMIM3086