Industry Voices: Driver Monitoring Systems Rev Up

 Sponsored Content

Thomas Falck is senior product marketing manager at Osram Opto Semiconductors and Ann Russell is senior applications engineer at Osram Opto Semiconductors.

The driver monitoring systems (DMS) market is expected to grow to more than $2.3Bn, according to Emergan Research, owing to demands for increased road safety and new government regulations.

For example, in July 2020 the US House of Representatives passed the Moving Forward Act with safety measures for automobiles, including in-cabin monitoring of distracted and intoxicated drivers. The European Union Council of Ministers passed a general safety regulation mandating that all newly manufactured vehicles be equipped with advanced safety systems, such as DMS, beginning 2022 with full implementation in 2026. China also has legislated the mandatory installation of DMSs for commercial vehicles.

DMS is an essential requirement for advanced driving assistance systems (ADAS) and autonomous driving systems, which use both visible and infrared light sources for illumination. Basic driver monitoring features include head tracking, gaze tracking, eye state analysis (blink rate), blink duration, eye open or closed duration, gaze direction, yawning and head movements. All can be used in driver safety applications to detect driver distraction and drowsiness.

DMS solutions traditionally have been developed using computer vision with image processing. This approach needs fine-tuning to address different conditions such as lighting, skin color or unusual facial characteristics. It is extremely difficult to anticipate such issues and meet all conditions. For example, bright or low light may lead to unusable saturated images. In these extreme cases, even the most advanced algorithms will not work. While driving, illumination on the driver’s face can change dramatically when passing under trees or bridges. As the car moves, this direct illumination can change in both magnitude and spatial extent.

Active in-car illumination in a narrow spectral band, in which sunlight and streetlamp brightness are minimal, is the most effective means to combat this challenge. The 940-nm wavelength is particularly interesting due to its absorption of interfering light and sun by ambient water droplets. This provides a window for artificial light to operate with reduced harmful environmental factors. In addition, red glow at 850 nm is visible to the driver, and therefore 940 nm is preferred because it cannot be detected by the human eye.

Another principal challenge for camera-based monitoring is the low ratio of signal-to-background/noise (SNR). Using a 40-nm bandwidth near infrared (NIR) LED improves SNR. This is especially true if a filter is in front of the camera. Light from the LED is transmitted and the visible part of the spectrum is removed with special interest at 940 nm because of reduced solar intensity via atmospheric water absorption. Thus, LEDs also enable in-cabin monitoring at night when using emitters with a 940-nm emission wavelength, which is invisible to the human eye and does not disturb the driver or passenger.

Lighting variations that may occur at any frequency in the NIR spectrum have negligible impact on the SNR if a filter centered around the correct wavelength is used. Also, off-the-shelf secondary optics are available to control the direction of light, thus preventing hotspots and unwanted illumination areas. This makes the image classification algorithm more dependable.

There also are alternative ways to handle lighting variations. System lighting using a ~2-nm bandwidth vertical cavity surface emitting laser (VCSEL) array can be used to further improve ambient light rejection with a corresponding 10-nm notch filter. VCSELs emit coherent light and can create nanosecond-level pulses easily, enabling both infrared camera monitoring and time-of-flight applications. VCSELs also exhibit very little variation with temperature, thus enabling such narrow notch filter usage.

Another advantage of VCSELs is the inherent efficiency for light coupling. VCSELs allow for very precise tuning of the field of view to the area of observation. The single-mode emitters can be accurately expanded to the camera field of view and remain collimated more easily. VCSELs have a narrow field of view and typically require expansion versus other forms of illumination with a high divergence angle, which can result in etendue mismatch.

Also, different wavelengths can be used simultaneously with different cameras and corresponding notch filters. However, the benefits of using laser illumination are not without tradeoffs. VCSELs may produce speckle, as the highly coherent light reflects off the human face and results in constructive and destructive interference. This produces a dark and bright spotted effect, introducing noise into the operation. To reduce speckle effects, Osram has carefully engineered diffuser structures, which dictate the field of view to address these complications.

Osram has tested the feasibility of using both narrow and broadband 940-nm NIR illumination to measure high infrared efficiency signals. We have found that it is possible to accurately illuminate at 940 nm indoors in both controlled and varying ambient light. The SNR ratio is significantly increased in the NIR range compared to visible cameras in controlled and varying lighting conditions.

As camera makers continue to increase the quantum efficiency of automotive qualified camera systems, we see many opportunities to enhance safety. Carefully selected lighting provides images that can be captured in all lighting conditions, which leads to improved driver monitoring and safer roads for everyone.