I built an end-to-end Automatic Number Plate Recognition (ANPR) system to automate vehicle identification from images and video streams. The system detects license plates using YOLO, extracts the plate region, performs OCR to recognize the registration number, and automatically blurs the detected plate to preserve privacy when required.

To improve recognition accuracy, I implemented multiple image preprocessing techniques such as grayscale conversion, adaptive thresholding, denoising, and contrast enhancement before passing the cropped plate to the OCR engine. These preprocessing steps significantly improved text extraction under varying lighting conditions and image quality.

The project follows a modular inference pipeline where object detection, OCR, preprocessing, and post-processing are separated into independent components, making the system easy to extend and deploy. The architecture supports both image and video inputs and can be adapted for applications such as parking management, toll collection, traffic monitoring, and surveillance.