The Blue Oval is working on its own version of Tesla's camera-only self-driving systems.
Ford apparently believes that Tesla's decision to replace LiDAR sensors with cameras is the right one. A new patent filed with the German Patent and Trade Mark Office (DPMA), discovered by CarBuzz, reveals that the Blue Oval is currently developing a camera-based means of determining a vehicle's distance from other objects. More than just a typical patent that may never come to light, we believe this technology will, as Ford also trademarked a name for a “Digital camera systems for use in vehicles,” with the EUIPO. The name? F-Sight.
The disclosure suggests that a new method, using monocular image data (image data from a single camera), would be able to extract necessary details about the environment from a picture and thereafter generate a depth map. If it works as intended, this would reduce the number of components needed to produce a car with advanced driver assistance features like semi-autonomous driving.
The camera would be able to detect an object in the image. Then, a “bounding box” surrounding the object would be applied to the depth map. The depth map is then masked, and the distance to the object can be determined.
The details are somewhat complex, but the essence is that the Ford system would be able to capture an image, identify an object or obstacle, and then analyze the data in that image, essentially counting pixels, to determine how far the car is from the obstruction.
To determine what type of object it is, the system may rely on a machine learning program (artificial intelligence). AI has already reaped rewards in Ford's bid to improve semi-autonomous driving technology in BlueCruise-equipped vehicles like the Mustang Mach-E and has made hitching a trailer a hands-free operation too.
We have our concerns about this. LiDAR is not infallible, but there is no doubt that it can detect objects from further away, and while using cameras may make it cheaper and simpler to produce a semi-autonomous car, industry experts believe that LiDAR is still more reliable and is the only way to achieve truly autonomous self-driving.
Tesla CEO Elon Musk has flip-flopped on this concept several times. He's argued that cameras can be just as good as ultrasonic sensors. Tesla then removed sensors from many cars, only for Musk to later concede that Vision (the company's camera-based Autopilot system) “with high-res radar would be better than pure Vision.”
In closing, while we can applaud efforts to reduce costs and democratize advanced driver assist systems, the primary focus of these inventions should be safety. Once the world knows how to make a safe self-driving car, only then should it focus on making the technology cheaper. Any other approach gambles with human life.
Keyword: Ford's New F-Sight Tech Will Make Self-Driving Cars Cheaper And Better Than Tesla