Intel Corp. is jumping into the lidar market.

The chipmaker today took the covers off a new RealSense camera that it describes as the smallest and most power-efficient entry into the high-resolution lidar sensor category. The device, the RealSense LiDAR Camera L515, is a compact disk the size of a tennis ball (pictured) that comes in at about 3.6 ounces.

Lidar is a popular surveying method employed by autonomous machines such as self-driving cars to navigate the world. The technique uses invisible laser light to generate a three-dimensional model of the environment. Lidar systems fire off countless laser pulses into the air, catch beams reflected by nearby objects and convert those beams into pixels in a map that’s continuously updated with new information.

Intel said that the L515 camera is built for indoor applications such as warehouse robots and automated shelf-scanning systems like the ones Wal-Mart Stores Inc. is piloting in some stores. The flat, compact form factor allows the sensor to be attached to the end of a robotic arm or a worker’s mobile device. It’s capable of seeing objects located as close as 10 inches to the lens or as far as 30 feet away, with a 1024-pixel-by-768-pixel maximum depth resolution. 

The L515 is based on what’s known as solid-state lidar technology. Whereas traditional lidar sensors spin on an axle so they can see in different directions, Intel’s camera achieves the same effect using a micro-electro-mechanical mirror that requires far less space and electricity to scatter light around. Intel claims it has developed a new proprietary type of micro-electro-mechanical mirror technology for the L515 that is even more efficient.

Besides the lidar unit, the L515 has a camera, a gyroscope and an accelerometer that gather additional data. The information from the sensors is fed to an internal “vision processor” that Intel said handles tasks such as clearing up motion blur when an object is moving. The result is that image quality goes up while the device to which the L515 is attached has to perform less computational heavy lifting to gain a high-fidelity view of the environment.  

Image: Intel