Driver Information System based on Computer Vision

In the last decade, rapid advancements in computer technology have significantly impacted various industries, especially the automotive sector. This paper presents a software solution for a system for displaying information about the profile of the driver of the vehicle in front of the ego-vehicle, based on license plate detection (LPD) and recognition. The system consists of two applications, one for detecting vehicle license plates (VLPs) and recognizing text on them, and a web application for displaying driver profile data and reviewing drivers. Various algorithms for LPD and text recognition were explored within this research, whereas YOLOv5 small and PaddleOCR are selected as the most suitable for these tasks, respectively. The experimental results show that the selected YOLOv5 small achieved a precision of 0.9905, recall of 0.9811, and F1 score of 0.9858 for the LPD task while processing 2.69 frames per second (FPS) when tested on an Intel(R) Core(TM) i7-8550U CPU @ 1.80 GHz 1.99 GHz processor. Regarding PaddleOCR, the system’s maximum recognition distance for VLPs was determined to be when the vehicle in front of the ego-vehicle is distanced at approximately 8 meters for images with resolution of $1920 \times 1080$ pixels, where PaddleOCR achieved an accuracy of 77.08% after testing. A Django web application was developed to display driver profile data and enable driver reviews based on license plates and the experimental results confirmed its functionality.