New Camera Technology Can Spot Drunk Drivers as Soon as They Get in Their Car

A new advancement in camera technology aimed at identifying alcohol-impaired drivers is currently under development at Edith Cowan University (ECU).

This system utilizes data from standard RGB (red, green, and blue) camera footage to detect signs of alcohol intoxication in drivers before they hit the road.

New Camera Technology Aims to Identify Intoxicated Drivers

Researchers at ECU collaborated with Mix by Powerfleet to conduct experiments capturing videos of drivers across varying levels of alcohol intoxication using a driving simulator setup.

They aim to collect data in a controlled yet realistic environment to train their computer tracking technology. The core of their innovation lies in a machine learning system that analyzes drivers' facial features, gaze direction, and head position from the captured videos.

Presented at the IEEE/CVF Winter Conference on Applications of Computer Vision, their research indicates that this system can classify alcohol intoxication levels with an overall accuracy of 75% across the three defined levels.

According to Ensiyeh Keshtkaran, an ECU PhD student involved in the project, this not only benefits vehicles with driver monitoring systems and eye-tracking technologies but also has the potential for smartphone integration, making alcohol intoxication detection more effective.

Dr. Syed Zulqarnain Gilani, Senior Lecturer at ECU, noted that their approach is distinct as it relies solely on a standard RGB camera. This makes it feasible for broader applications, such as surveillance cameras on roadsides and could enable law enforcement agencies to prevent drunk driving incidents proactively before they occur.

The next phase of their research involves defining the optimal image resolution required for the algorithm to function effectively. Should low-resolution videos prove sufficient, it could pave the way for widespread adoption of this technology across various vehicle types and infrastructure without requiring specialized in-cabin installations.

Integrating Computer Vision Into Road Cameras

Gilani highlighted the potential of integrating their computer vision-based approach into existing road cameras, similar to systems that currently monitor seatbelt usage or detect mobile phone activity while driving.

This technology also has a comprehensive dataset comprising 3D and infrared videos of drivers' faces, rearview RGB videos capturing posture and steering interactions of the driver, logs of driving simulation events, and screen recordings of driving behavior.

This dataset not only supports ongoing research efforts at ECU but will also be a valuable resource for the wider scientific community, which is interested in advancing similar studies.

"This research confirms that it is possible to detect intoxication levels using just a simple camera. The next step in our research is to define the image resolution needed to employ this algorithm," Gilani said in a press release. "If low resolution videos are proven sufficient, this technology can be employed by surveillance cameras installed on roadside, and law enforcement agencies can use this to prevent drink driving."