YOLOV5 Based A Real Time Automatic Number Plate And Helmet Recognition System

In today’s busy life, many road accidents are seen and it doesn’t take longer to us as a victim. Almost 80% of the road accident fatal occurs due to not wearing the helmet. So, helmet should be worn properly. But from the beginning onwards, a child is trained that it requires a driving force to do some work. For that, it is necessary to enforce the public to wear helmet through fine as per the motor vehicle act of the respective country. To make the fine collection process as automatic one, it is also required to identify the vehicle number plate. In the proposed design, the model is trained with YOLO v5. The dataset collected is annotated and the images along with the labels are fed to YOLO v5 for training and at the end the weight files are created. The proposed model is capable of processing the live video. The video is captured using the camera and split into frames with the help of python using the package OpenCV. The model weight file is loaded in the PyTorch and then the frames from OpenCV are fed in order to start the process of detection. Based on the YOLO weight files from training, the proposed system detects and classify under two classifications called with helmet and without helmet. After the detecting the vehicle rider without helmet, the number plate recognition process takes place for the images which are having rider without the helmet. The number plates are recognized using OCR technique with the help of plate recognizer API and it returns the text extracted from the number plate image. The texts are stored into a csv file. Later, the specific file is used to display the results in the browser using a Flask, a web app framework. The proposed model, confidence score of 84% for rider, 92% for helmet, 90% for no helmet and 82% for number plate have been achieved. Further this can be extended with own OCR engine and better UI Display and security like logins for citizens, officials and admin.