Deep Learning and Vision - Based Smart Parking Vehicle System Using YOLO11n and Edge AI

Abstract

Vehicle ownership has A lot grown in cities, leading to serious problems like difficulties in parking management, traffic congestion, high fuel consumption, and driver discomfort. Conventional parking management systems mostly depend on sensor-based infrastructures that are quite expensive to install and maintain. To overcome these drawbacks, this paper suggests a Deep Learning and Vision-Based Smart Parking Vehicle System using YOLO11n and Edge AI for realtime parking occupancy detection. The system is based on the PKLot dataset that contains 12,416 parking lot images taken in different weather conditions like sunny, cloudy, and rainy. A compact YOLO11n detection model is utilized to ascertain the occupancy status of parking spaces and to label the spaces as either occupied or vacant. Besides, the Intersection over Union (IoU)-based occupancy detection technique is adopted to increase the parking-space classification accuracy. The model has been trained so that it can run on NVIDIA Jetson edge devices; this local processing reduces reliance on the network, lowers latency, and also enhances data privacy. Test results show a very high level of detection accuracy, with the system achieving major Precision Recall F1-Score, and mean Average Precision (mAP) values. The method suggested here offers a scalable, less costly, and intelligent parking management system that can be applied in smart city and intelligent transportation scenarios.