DMS
Driver fatigue is often a direct cause of traffic accidents.
Introduction to DMS
Driver fatigue is often a direct cause of traffic accidents. There is a clear need for systems that detect a driver's deteriorating psychophysical condition and notify them of it, which could significantly reduce the number of fatigue-related accidents.
Developing such systems is difficult, however, because a driver's fatigue symptoms have to be recognised both quickly and correctly. One of the practical ways to implement driver drowsiness detection is a vision-based approach.
Objective
Active safety systems that significantly reduce the number of accidents are among the top priorities for automotive manufacturers. Over the years, engineers have tried a range of techniques to detect a drowsy driver — one of the most common causes of road accidents. These have included analysing steering wheel operation, measuring a driver's brain waves and heartbeat, and monitoring their response to designed queries. Despite all of them, reliable detection of driver drowsiness remains elusive.
The Tek Labs Driver Drowsiness Detection solution uses an on-board camera to detect eyelid closure. The camera continuously captures the driver's face, and the captured video stream is processed in real time by a proprietary algorithm that identifies drowsiness from the eyelid pattern. As soon as the system detects an impending unsafe condition, it raises a warning so the driver can react and take immediate action.
Concept
The Tek Labs proprietary engine for image processing and drowsiness detection works in three stages.
Image acquisition and processing
After the image is acquired by the video camera, lighting compensation is performed based on the average luminescence of the scene.
The driver's skin is determined from the chrominance information in the image. The image is then converted into binary form and the noise is removed.
Face and eye detection
Using Tek Labs proprietary techniques, the binary image is converted to a labelled matrix and the facial components are determined in order to detect the face.
The region of interest is then reduced to the eye region of the face, as shown below.
Pattern analysis for drowsiness detection
The Tek Labs proprietary algorithm analyses the pattern of the eyes and of eyelid closure to detect driver drowsiness.
Solution highlights
- Drivers with different types of faces, skin colours, eye colours, and facial hair patterns
- Various light conditions in the vehicle, and noise from light reflecting off different objects
- Scenarios such as a tilted driver head
- Dynamic detection while the driver's head moves due to jerks in the vehicle
- High-speed video processing and detection
- Efficient customisation through a model-based architecture
- Scalable and configurable, for ease of customisation or for adding new features such as emotion detection, smile detection, and other dynamic scenarios
- Automatic code generation to the MISRA standard, for a wide range of digital signal processors (DSPs) used in advanced driver assistance systems