Latest from colin Barnden - CES15 Jan 2025 11:40
Https://www.linkedin.com/posts/colin-barnden-1081376_mde-tof-algorithms-activity-7285253272423026688-foDL/?utm_source=share&utm_medium=member_ios
One of the most exciting demos I saw at CES was monocular depth estimation hashtag#MDE, shown here from Seeing Machines. Top left is the standard 2D image taken from the RGB-IR image sensor, here sited in the mirror location. Bottom left is the depth map, with red showing estimates of nearest to the sensor, and green further away. Shown on the right is 3D depth inferenced from 2D data. The software enables 360 degree rotation of the 2D data. 3D sensing has been seen in various guises for a decade, typically using structured light, time-of-flight hashtag#ToF or stereo vision. MDE has been around a similar time, and various hashtag#algorithms exist to create 3D from 2D data, such as Midas, ZoeDepth, Marigold, and PatchFusion. Some MDE tech is now finding its way into the automotive domain.
How does MDE relate to the automotive DMS market? The NCAP29 framework is likely to introduce scoring for fusing data from the hashtag#DMS hashtag#OMS sensor to understand body position and body pose, enabling suppression of airbag deployment in unsafe scenarios. Various technologies exist today to perform occupant classification to meet, for example, FMVSS 208 in the U.S. What we are seeing is ways the guidelines are adapting to take advantage of the safety benefits of DMS/OMS image sensors, both in Europe and the U.S. OEMs are now exploring ways to meet the advanced hashtag#safety roadmap entirely in software, using hardware which is already present in the vehicle. DMS is being introduced to understand driver state for hashtag#drowsiness, hashtag#distraction, hashtag#impairment, hashtag#attention, and hashtag#engagement, hands-on wheel detection, seat-belt use, phone and smoking detection etc. Vital signs monitoring is enabled using the same IR optical path, and the next evolution is to take the existing hardware and 2D data and perform 3D depth sensing using MDE.
What we can stitch together from CES is that OEMs are rapidly moving away from seeing DMS/OMS as a "tick-the-box" exercise for drowsiness, distraction, and automated driving, and starting to understand how the hardware can be leveraged for advanced safety, convenience and comfort features in the cabin. There are as many possible strategies that can be developed here as there are OEMs, so the ways that this tech will be integrated will be under discussion now. CES was all about the software-defined vehicle hashtag#SDV and hashtag#GenAI. Monocular depth estimation for DMS/OMS leverages both of these trends to enhance vehicle safety and the user experience.
The 2025 edition of Semicast's market report "ADAS & Automotive Vision-based Driver Monitoring Systems" is now available for purchase. Significant segmentation of the DMS/OMS market is provided:
- by movement detection into "Owl-only" and "Owl & Lizard" detection
- by feature set
- by proces
Sign In
Follow the stocks
