Smart Street Parking Management Using Real-Time Computer Vision and YOLOv8

The current paper introduces a smart parking system with privacy concerns, which can be applied to edge computing to detect the presence of vehicles in real time. The main goal is to obtain lowlatency and exact detection of computationally limited edge devices. The principles of the proposed methodology lie in the fact that the optimized YOLOv8 detection model will be implemented, which will be supplemented with geometric overlap checks and temporal consistency checks to reduce false detections that may occur due to changes in the illumination and the presence of an object. Real-life urban applications have shown that the system can run at 28 frames per second and only uses low power requirements with F1-scores of 90.9 percent when used in daytime and 79.9 percent when used at night. The offered vision-based system has competitive detection capability compared to traditional sensor-based parking systems: it consumes much lower deployment and maintenance expenses. The system design in general is based on privacy-by-design philosophy and is capable of being deployed across a wide range of urban environments, which is why it is appropriate in the practical application of smart city parking systems in various cities around the world.