A Camera That works like the Human Eye
Posted By Vidushi Vij | 04-May-2017
Kingston University London started a project called EPSRC- funded Internet of Silicon Retinas (IOSIRE) with Kings college, London. Their main aim was to develop a machine-to-machine communication system which was motivated by the human eye. Human retina is extremely powerful. It scans the data at any speed. The conventional camera needs to learn a lot from the human retina.
They wanted to develop a communication system that could capture and transmit vision sensors through human retina. This is great for having footage of the distant planets or forests etc.
It took three years to complete this project and was done in collaboration with University college London and Queened Head’s college London. In this Project they focused on the human eye to sharpen the sight of robots.
Recently developed dynamic vision sensors (DVS) behave like a retina. Whenever there is any movement, the image gets updated. On any movement, the light is reflected which is detected by the Andreopoulos sensors. Thus there is very less wastage of memory as compared to conventional cameras. Because of these sensors, the active part of the data is automatically highlighted. These pixels behaves in the same manner as that of eye neuron. Each pixel adjusts on its own and they can handle uneven lightning.
The DVS is built to work with TrueNorth Computer. They created a camera which is good at handling dynamic problems. DVS costs around $2,000, which has been used in many projects. They are even planning to add color sensitivity and increasing retina of the camera. Conventional camera captures everything, whatever it sees. The information is then stored for the later use. This utilises a lot of space and power. In Conventional cameras the video is recorded by creating new images for each frame. This wastes a lot of memory. But when there is some extremely fast scene, the video is not recorded accurately as its processing power is very less.
By using this technique, the speed with which the frames are created has also increased to up to 1000 frames per second to around 30 frames per second for conventional camera. The energy and processing is also reduced thus opening up new possibilities for surveillance from robots and drones to the next generation of retinal implants. These sensors in the forest through a plane. With these sensors we can get the footage without humans help. They are further researching on how the sensors can work together when connected over the internet and operated remotely.