Using Light Field Photography for Robot Vision

Light field photography technology, most known for its use in the consumer Lytro camera, is used for capturing images than can refocused after the photo is taken. A conventional camera takes a photo that records only the intensity of the light entering the lens. A light field camera uses an array of lenses (or multiple cameras) to also record the light’s vector through space. Because the intensity and direction are stored, the image can be post-processed to change the focal point — just like you would usually do before taking the picture.

While this tech is interesting in products like Lytro, it is a fairly niche application. However, a team of engineers at Stanford University and the University of California San Diego have now adapted it for robot vision. Their prototype is the first ever wide view light field camera that uses a single lens. The lens is spherical, and the camera is able to capture a 138° field of view in an image that can be refocused.

The nature of light field photography makes it particularly useful for autonomous robot vision, as the additional depth information can be used to enhance navigation capabilities. The team believes the technology will provide improvements in applications like self-driving cars and drones.

While their system is still just a proof of concept, it seems likely that it will end up in production. The compact size is a major advantage compared to systems commonly in use now, and it will probably be incorporated in a real world robot soon.