Automatic Vacant Parking Places Management System Using Multicamera Vehicle Detection

Managing vacant parking spaces efficiently is crucial for smart city applications. This paper tackles the challenge by introducing a deep neural network approach for detecting parking occupancy. For large parking areas, using a single high-mounted camera offers an effective way to monitor the entire space. To achieve this, we leveraged the well-known YOLO object detection model, which is recognized for its high precision and real-time speed. We propose a modified, lightweight version of the YOLO-v5 architecture tailored for parking management. Our model is designed to detect vehicles of various sizes, from large trucks to small cars, using a multi-scale learning mechanism. This approach enables the model to learn detailed, discriminative features across different scales. By optimizing the architecture, we reduced the number of trainable parameters from 7.28 million in YOLO-v5S to 7.26 million in our model while significantly boosting precision. Our model achieves a detection speed of 30 frames per second (fps) and outperforms the larger YOLO-v5-L and YOLO-v5-X versions, particularly excelling in detecting small vehicles, with a 33% improvement compared to YOLO-v5-X. Experimental results demonstrate that our approach is both efficient and accurate, making it a promising solution for managing parking spaces autonomously in smart cities.