Ouster's new REV8 Color LiDAR chip is being delivered now, and captures color and depth simultaneously without relying on a traditional pixel grid.
The company's CEO, Angus Pacala, was not shy about the possibilities of the tech, either. "The goal is to obviate cameras," he told Tech Crunch. It’s early-stage tech aimed at robotics and self-driving systems, but the implications for imaging are enormous.
The chip is being supplied in small quantities to major partners, including Google and drone firm Skydio. Instead of collecting a flat image – a grid of pixels – as we are used to in photography, it collects a 'point cloud' – a pattern of points which represent distance from the sensor.
In practice, it might be a little while before the tech is ready to replace cameras as we know them, and the point-cloud concept is nothing new in photography. It is a technology used by a lot of companies to build up an understanding of the area around the sensor.
Apple's iPhones, DJI's drones, and plenty more have LiDAR sensors near their cameras, which point in the same direction and collect distance information.
Apple uses this data as part of its computational photography, so it can accurately apply depth-of-field simulations because it knows roughly what areas of the overall image were near to the camera and which were far.
The 'computation' is the making sense of the data from a CMOS chip and a LiDAR camera, which imposes a lot of work that could be side-stepped by collecting it all in one go, which is exactly what is new here.
What is new, here, is that the sensor chip is the same one – the same lens and the same chip and lens is receiving the distance and color information.
Each point in the cloud recorded – and that's up to 10.4 million points per second in the new OS1 Max flagship variant – has color data as well as location.
Technology like this has uses outside traditional photography – indeed its development is more directly associated with robotic taxis.
The other day I was walking outside our London office and snapped a Waymo car passing by me. It was a bit of a surprise as – while fleets are already operational in San Francisco and LA – they're new and out learning in the UK.
These vehicles need a massive array of LiDAR cameras on the top to feed data to the vehicle's AI driver. It currently needs separate cameras to detect things like the color of brake lights and traffic lights – this kind of tech is what the Ouster Rev8 is initially looking to change.
Smaller LiDAR systems open up the possibility of more autonomous systems, like drones, which have less need to process data on board.
It doesn't take a lot of imagination to see how steadily increasing resolution of color LiDAR on combined chips could revolutionise the camera, and wholly merge spatial and traditional image data in a way that we've only seen tweaked at around the edges so far.
Chinese company Hesai has already announced something similar with its '6D' platform in April this year, meaning it is also already a competitive space and, yes, that is also a company with a focus on "hands-off self-driving cars."
LiDAR started at a measuring tech with a single 'point' in the sixties. The sensor chips chips hit 100,000 points or more in the 2010s, and are now in color and over the million mark, so progress is rapid.
A point cloud of data could be immensely desirable for some photographers – like a 'RAW' file but so much more – so, as soon as it is practical, why wouldn't it supplant CMOS?
Sure, the tech needs to keep getting smaller, and the frame rate will be a challenge, but all these things can likely be overcome.
Remember CCD chips? No, me neither 😉
