The Future of Autonomous Vehicle Sensors: A Shift Toward Advanced Stereo Vision
Why We Believe Stereo Vision Will Become a Critical Component Of All Automated Vehicles
When we started NODAR in 2018, lidar was the golden child of the autonomous vehicle movement and had attracted billions in investment dollars—a revolutionary tool promising precise, realtime 3D mapping and object detection. If you were building a self-driving car, lidar was the sensor to look at.
Even back then, we believed there was a better way forward, one that could offer higher performance, lower costs, and better scalability. That belief led us to stereo vision—and to the development of Hammerhead, our advanced 3D vision platform.
What Is Stereo Vision?
Stereo vision is a technology for perceiving depth using two cameras. Similar to how the human brain estimates depth from images from our two eyes, the cameras capture images from slightly different viewpoints, and by comparing these images software can calculate depth to each pixel using geometry, creating a real-time, high-resolution 3D model of the environment. The result is fast, precise, robust 3D sensing without lasers or moving parts.
It’s the core of our approach at NODAR. We re-invented stereo vision using modern algorithms, processors, and cameras, extending traditional performance by orders of magnitude across metrics - range, speed, resolution, and robustness. We call our advanced stereo vision product Hammerhead, giving a nod to the distance between a Hammerhead shark’s eyes and their unparalleled depth perception. Hammerhead has been designed from the ground up to support safety-critical applications in autonomous vehicles, robotics, and area monitoring.
Why the Market is Moving Away from Lidar
Back in 2005 at the DARPA Grand Challenge, lidar appeared as a viable choice for reliable 3D data for autonomous vehicles. Despite years of investment and promises, lidar has under-delivered. The technology continues to be expensive, power-hungry, offers low resolution and limited range, is difficult to integrate, and performs poorly in harsh weather conditions. Spinning lidars (which is most of them) fail after about 7,000 hours, increasing total cost of ownership.
As a result, the public markets have seen lidar companies lose up to 99% of their stock value, while some have shut down or been acquired for their assets. The need for a more scalable solution has never been clearer.
Why NODAR Advanced Stereo Vision Is Surpassing Expectations
We designed Hammerhead to solve the problems that lidar couldn’t—and the results have been remarkable. Here are the key benefits we’ve seen:
Superior Performance in Challenging Conditions
Our proprietary stereo matcher works in heavy rain, dense fog, and dusty environments with only minor degradation. In fact, in our tests, stereo vision significantly outperformed lidar in all major weather scenarios.
Chart: NODAR stereo vision vs. lidar in dry, rainy, dusty, and foggy conditions. Stereo vision consistently delivers more valid 3D data points—even in extreme weather.
2. Low-Light Dominance
Thanks to NODAR’s patented online calibration and advanced matching algorithms, Hammerhead operates exceptionally well in low light and blurred conditions—better than other camera-based sensors and >2x as well as lidar.
3. Long-Range, High-Resolution Detection
NODAR stereo vision provides 50 million depth measurements per second—10-50x more than high-end lidar. This dramatic increase in resolution and speed enables accurate detection of very small objects at long distances, for example a traffic cone at 200 meters or an adult-sized mannequin lying on its side at 130 meters—even in darkness. For shorter range applications such as automated farming where lower speeds are required but precision is important, NODAR’s wide-baseline stereo vision enables detection of a 5cm object at 50m.
4. Cost-Effective and Scalable
Stereo vision relies on off-the-shelf cameras rather than lasers or moving parts, making it easier to integrate, more affordable, more robust, and more scalable across vehicle platforms.
5. Resilient to Shock and Vibration
Unlike many lidar systems, NODAR’s stereo vision system remains stable and accurate in real-world operating conditions where shock and vibration are a constant.
6. Wide Baseline Advantage
A key enabler of NODAR’s ultra-precise, long-range stereo vision is a wide baseline (distance) between the cameras. The wider the baseline, the farther away we can detect objects.
The Road Ahead
The future of autonomous vehicles depends on sensors that are not only accurate but also cost-effective, scalable, and reliable under real-world conditions. We believe advanced stereo vision, and specifically NODAR’s approach, is the key to unlocking this future.
We’re not just replacing lidar—we’re creating a better solution that performs 24/7, in more weather and lighting conditions, and at a fraction of the cost. We believe this is key to unlocking mainstream autonomy.
If you’re building autonomous vehicles or advanced driver-assistance systems and want to go further, see more, and react faster, we’d love to talk.