An intelligent hybrid approach for detection of drowsy driving risk in real environments

Abstract

Up until now, law violation and drowsiness have been major causes of road traffic accident in Thailand. This study presents an approach to detect the risk due to drowsiness and distraction of a driver. In our method, techniques in computer vision are employed to extract facial features of the driver to determine his behaviors. Since the vehicle speed is also an important factor of the accident risk, we use both vehicle speed and driver behavior to analyze and determine the risk level of the driver. The experiment was performed under the real-time illumination condition inside the vehicle. The proposed system consists of three modules. First, the facial image preprocessing adjusts the image quality in order to improve the accuracy rate of feature extraction. Then the facial feature detection module detects the features which are important for risk analysis. Finally, the feature classification and analysis module evaluates the risk and makes a decision on warning. The experimental result has shown that the risk alert yields the 86.30 percent accuracy at any vehicle speed. According to the statistics from the Department of Highways in last three years, the proposed system would be able to reduce the number of accidents by 20,612 accidents in over speed driving and 1,199 accidents from drowsiness if the system was implemented.