New Lidar/Camera Hybrid Could Be The Key to Self Driving Car Success

lidar ouster.pngTo make a car self-driving, you need information. The car needs to see the road and the obstacles just like a human does. But regular cameras don't provide a real sense of depth. In order to achieve a human level of accuracy car manufacturers have to use a combination of cameras, radar, and lidar. 

Lidar is a portmanteau of light and radar. In its simplest forms, a lidar device shoots out pulsed lasers and then measures the return wavelength which it uses to create a 3D representation of an object. Lidar has only been around for a little over a decade and was invented by Velodyne and has been the primary reason that self-driving cars exist and operate as well as they do. Now there are dozens of companies trying to get in on the lira action.  

New silicon valley Lidar company OUSTER was founded by ANGUS PACALA who had previously founded Quanergy Systems... another company that makes liar units. This guy is serious about lidar, and his new company may have found a clever new way to use it.

Ouster has figured out how to use the sensors on its OS-1's lidar unit as a camera, creating panoramic two-dimensional images of the surrounding area. 

Angus Pacala explains:

"The OS-1's optical system has a larger aperture than most DSLRs, and the photon counting ASIC we developed has extreme low-light sensitivity... So we're able to collect ambient imagery even in low-light conditions."

We've covered Lidar before, but this new system is unique. Check out the video below of what kind images they can get out of their ouster OS-1-64.

That footage looks like it came straight out of an 80s science fiction movie. The car is creating a 3D world around it, mapping where everything is. The difference with Ousters way of doing things is not just trying to combine information from a lidar, radar, and camera into one thing, but to use the lidar as you would a camera. That creates images that can be fed into the deep learning algorithms that were made for cameras. The algorithms can then interpret the images around them as trees, cars, people etc. 

It's truly fascinating and just like the lidar is being used in a way that wasn't conceived before, I bet the actual product itself will have uses outside of self-driving cars. 

Source: OUSTER