Tiny, 3D-Printed "Eagle Eyes" for Robots
Researchers from the University of Stuttgart in Germany have developed a highly miniaturized camera that mimics the natural vision of predators. They 3D-printed different multi-lens objectives directly onto a complementary metal-oxide semiconductor (CMOS) image sensor. Each lens is made from plastic and about the size of a grain of salt. The lenses mimic the action of the 'fovea,' a key physiological feature of the eyes of humans and eagles, which allows for faster image processing. The image processing system collects the most information from the center of the field of view. Capturing scenes at different resolutions saves processing time and can be used in systems where quick decisions are key, like self-driving cars. The approach allows for fast design iterations and can lead to many different miniaturized multi-aperture imaging systems with applications in fields like endoscopy, optical sensing, surveillance drones, or security.
Transcript
00:00:01 This imaging system is designed to have vision that works like our eyes do These little lenses are used to create images with higher resolution in the center than at the edges Our eyes work the same way thanks to a region called the fovea, which is rich in light receptors Foveated electronic sensors can see a wide, low-resolution view and a narrower, high-resolution view at the same time, instead of one or the other
00:00:30 Which means that this camera can capture a scene at low and high resolution simultaneously Having a wide view with high resolution in the center preserves detail where it’s needed without sacrificing processing time In the camera, a set of four compound lenses is 3D printed directly onto an image sensing chip Some of these lenses see from a wide angle, while others take a narrower view with higher resolution
00:01:00 The information from each lens is digitally combined to create a foveated image with the most information in the center A foveated camera can do things like helping self-driving cars make quick judgments based on visual information At 300 square microns in size with lenses only a few tens of microns across, this camera is unique in its small size and image quality There’s still room to improve the manufacturing time
00:01:28 and optical performance of this device But it could be a boon to engineers who need a device with quick, sharp eagle eyes