Smart Vehicle Identification and Tracking Using YOLO-Based License Plate Recognition

Manual verification of vehicle registration plates can be time consuming and error-prone. This paper details a functional prototype which enables license plate detection, reading and results presentation. Once images or videos have been uploaded the system uses Python to detect license plates within them, and then extracts and displays the plate information through a simple web interface. Detection is done by training a YOLO model using the Ultralytics library. The license plate is cropped using OpenCV, and then the characters are read using EasyOCR. If a video is used, ByteTrack will link detections between frames using temporary tracking IDs. The system has been made using Gradio to create a dashboard for the user. A detection section allows users to upload an image and detect license plates. There is also a video tracking section, detection history section and CSV export. On one example test image the detection had an accuracy of around 90.1%. The OCR stage found the plate text "DL1CA6207 IND" with an accuracy of around 99.43%. These are only for one example and do not give an indication of the overall accuracy. This project contributes by providing a functional, open source solution to license plate detection, recognition, tracking and recording. The limitations are well discussed and include OCR errors, additional printing on the plates, temporary tracking IDs and there needs to be more testing on a larger scale. A clear plan is provided to properly evaluate the project