Deep learning Model for Driver Behavior Detection in Cyber-Physical System-Based Intelligent Transport Systems (KEY IJP************781)

• Seenu A,Sandeep Kumar G,Ranjith Kumar K,Rajavarman M,Prasanan V

Abstract

In this paper, we present an innovative approach for driver behavior detection within cyber-physical systems (CPS) for intelligent transport systems (ITS) using ESP32 CAM, eye blink sensor and deep learning models. The integration of ESP32 CAM, a compact and versatile camera module, with advanced deep learning techniques enables real-time monitoring and analysis of driver behavior. While an eye blink sensor monitors drowsiness by detecting blink patterns. The collected data is processed using a deep learning model developed in MATLAB, which analyzes visual and sensor inputs to assess driver alertness and behavior. Our system leverages MATLAB for model development and training, employing convolutional neural networks (CNNs) by using VGG-16 classifies various driver states including signs of fatigue and distraction to process and interpret video data captured by the ESP32 CAM. The deep learning model is trained on a diverse dataset of driver actions, including lane deviations, distraction, and fatigue signs. By analyzing the visual input, the system classifies driver behavior and provides actionable insights to enhance road safety and driver assistance. The implementation of this CPS-based approach facilitates timely intervention and adaptive responses, contributing to the overall safety and efficiency of transport systems. When the system detects drowsiness or abnormal behavior, it triggers an alert through a buzzer, ensuring timely notifications to the driver and sends real-time notifications to designated contacts (e.g., car owners, parents) with the drivers live location and status using GPS and GSM technology. At the time, speed of the vehicle can also be controlled. Arduino serves as the central controller, coordinating data collection from the eye blink sensor and managing the alert system.