Automotive safety has been transformed in the past 10 years through the adoption of active safety systems, which have leveraged external facing, active sensors—particularly cameras—to build an understanding of the exterior environment around the vehicle, identifying dangerous situations and inteverning on the driver’s behalf. This trend is now being extended into the vehicle’s interior environment, with camera-based Driver Monitoring Systems (DMS) and Occupant Monitoring Systems (OMS) set to deliver improved safety and richer digital cockpit experiences.
Driver distraction remains a persistent factor in road accidents, and the slow rollout of autonomous driving means that human drivers are set to play a central role in the driving process for the foreseeable future. Therefore, most applications for in-vehicle perception will focus on the driver, detecting signs of faitigue, as well as visual, manual, and cognitive distraction, with typical responses including Human-Machine Interface (HMI) prompts to bring the driver back into the loop, or the adjustment of other Advanced Driver-Assistance Systems (ADAS) to accommodate for the impaired driver. At the same time, a better assessment of the state of front and rear passengers can yield safety advantages for all vehicle occupants in the longer term.
As the adoption of DMS and OMS is set to grow rapidly, driven by regulation and the activities of safety ratings agencies, Original Equipment Manufacturers (OEMs) are already pursuing valueadded applications delivered over the enabling hardware of the DMS.
Overall, delivering on the triple requirement of robust fatigue distraction, rapid integration, and value-added differentiation means OEMs need to make the right choices when specifying DMS, especially the integration point of the primary camera sensor. A key candidate for DMS integration in the future is the rearview mirror, with this location offering excellent coverage of the whole vehicle environment and the form factor of the device enabling rapid and seamless integration into multiple models.
