Autonomous Transport System Embedded with Alcohol Detection and Ignition Lock, Driver Anti-snooze System and Passenger Counting

Abstract

Every year, the number of road accidents continues to rise and kill a lot of people. Approximately, 80% of these accidents occur on public transportation and are caused by poor driving habits such as drunk driving and exhaustion. Machine learning-based approaches can learn and improve over time. This paper introduces an autonomous transport system based on a machine learning approach. The system is embedded with 3 key features, driver alcohol detection with the ignition lock, Driver anti-snooze system, and passenger counting. The alcohol detection system automatically locks the ignition and prevents the driver from starting the vehicle if he/she is classified as being drunk. Our model is based on logistic regression and achieves an accuracy score of 85% based on its predictions. The anti-snooze system alerts the driver and his passenger if fatigue is detected. This is accomplished by using the Intel RealSense 3D camera and SDK to perform driver face tracking and detection. Based on the driver’s facial expressions and movements, this method detects fatigue patterns with 97.2% accuracy. Passenger counting is the final embedded system introduced in this paper. To count passengers, this system combines Open CV and TensorFlow object detection with object tracking algorithms. This method reduces the computational load required to complete this task. This results in an accuracy score of 87.9%. This paper brings the capabilities of humans to understand data and be able to make predictions based on previous observations. It contributes to the ongoing work of trying to make roads safe for all drivers.