INTELLIGENT COMPUTER VISION-BASED DRIVER AND ROAD SAFETY MONITORING SYSTEM USING DEEP LEARNING

Abstract

Driver and road safety require continuous monitoring of both the driver and the surrounding road environment. This work presents an Intelligent Computer Vision-Based Driver and Road Safety Monitoring System Using Deep Learning that combines different safety-monitoring functions into a single framework. The system uses MediaPipe Face Mesh, facial landmark analysis, deep learning models, YOLOP, YOLO11, and image processing techniques to monitor the driver and road conditions. The driver monitoring module analyzes facial features to identify Normal, Drowsy, Impaired, and Yawning states, along with driver absence and abnormal head movements using eye, mouth, and headpose analysis. The system also detects seatbelt usage and potholes, while YOLOP and YOLO11 are used for road-scene and road-object detection. CLAHE is used to improve image quality under different lighting conditions. The outputs from these modules are combined using a Unified Processor and Safety Decision Engine to determine the overall safety state and generate an audible alert when an unsafe condition is detected. The system produces annotated output frames and buzzer-synchronized H.264 video, while safety events and session details are stored in a SQLite database. A Streamlit-based analytics dashboard is provided to visualize safety events, session records, detection results, and system outputs. The proposed system provides an integrated approach to driver and road safety monitoring by combining computer vision, deep learning, safety decision-making, real-time alerts, data logging, and analytics in a single framework.