Smart Parking Guidance System

With the increasing maturity of deep learning in the field of computer vision, traditional visual methods are gradually being replaced by deep learning algorithms, such as YOLO (You Only Look Once), accelerating the process of computational inference. The parking problem faces limitations in robustness under the constraints of diverse and complex scenarios from a traditional visual perspective. However, in the rapidly evolving modernization process, the development of parking systems has the potential to significantly advance the progress of artificial intelligence applications. This paper presents the design of a parking tracking system utilizing a laptop, D435i camera, STM32, and the YOLOv7 algorithm. By employing a detection-distance measurement-judgment approach, the system achieves target locking and tracking, allowing the vehicle to smoothly follow a specific target. Through experiments, our parking system demonstrates sharp target recognition and prompt responsiveness. The distance measurement error is maintained between 0.28 to 1 meter, and the YOLO inference speed stabilizes at 74 milliseconds until the system responds. This stability enables the parking system to make reliable judgments and respond quickly in real-world scenarios, providing a robust solution for practical applications.