SecurePark: Vehicle Intrusion Detection System

A vehicle intrusion detection system includes detection and identification of a vehicle's license plate. This paper describes our web application which uses YOLOv5 (you only look once) as its object detection model for detecting a vehicle's number plate and uses Microsoft Vision API for the OCR (optical character recognition) that identifies the characters from the cropped image of the license plate that is captured during the detection process and sent to the API. Initially different versions of YOLO were tested for object detection and based on the results achieved from the above model training process, YOLOv5 gave the most accurate results among the versions tested. For the identification of the characters on the plate, along with the Microsoft Vision API, the Tesseract OCR engine was used but did not produce satisfactory results as compared to the API. Our main motive was to build an end-to-end integrated web application that makes use of our machine learning model to perform detections using the power of a local computer so that it can be used by any person irrespective of their knowledge about the technologies working under the hood.