The Chronocam, a camera that only sees what moves, ushers in a huge inventive step in the domain of smart sensors used for machine vision. This technology, developed as a spinoff from France’s CNRS Vision Institute, is termed game-changing and evidence of a paradigm shift by the inventors.
|Copyright © Chronocam|
Up to the Chronocam, all the pixels in a picture captured a tiny part of a scene, which image synthesizers organize into an image. Images taken in succession, for example 24 frames per second, as in motion picture, each contain all of the information in the scene. However, this sort of image acquisition also creates a huge amount of redundant information, from one frarme to the almost identical subsequent frame, which video compression encoding cannot completely delete. In addition to generating lots of redundancy this prior mode of image acquisition also loses a lot of information. The information lost is all that happens between the 24 frames per second, or all the information that has moved or changed much faster.
In other words, what if you could actually capture what happens at the speed of 100,000 frames per second? What if you captured all the information lost to 24 frames per second, without generating a massive amount of data, and a gigantic storage problem?
In a nutshell, this is exactly what the Chronocam does. Instead of each pixel capturing a tiny part of a scene in each of 24 frames per second, the pixels in Chronocam image acquisition only capture what moves, and what moves as fast as 100,000 frames per second, or less. Since not everything moves, this eliminates the prior redundancy. And since the acquisition works much faster, at the speed of 100,000 frames per second, it is also far more sensitive to all the events happening in the scene, especially those phenomena that happen very fast.
Consider for example being able to capture the real-time microcirculation of blood, on a cellular level. What radically changes in this mode of image acquisition is that, since the pixels only capture what moves in a scene, the acquisition is actually no longer clock-dependent, it depends on what's moving in the scene.
The Chronocam has applications wherever machines require artificial vision, in particular for autonomous cars, since the motion dependent image acquisition technology is unaffected by light conditions, such as glare. Applications also exist in industrial settings, for inspection and monitoring of production and installations, for always-on visual input applications, and for scientific microscopy. Applications also exist for drones and Remotely Piloted aerial systems (RPAS), plus more.
The various aspects of Chronocam technology are patented in at least 17 different patent families. Members from each separate patent family are hyperlinked in the list appearing below.
- US2018024343 (A1) ― 2018-01-25 - Imaging device and method
- AU2016248758 (A1) ― 2017-10-26 - Pixel cell circuit and implant
- US2017111619 (A1) ― 2017-04-20 - Device for displaying an image sequence and system for displaying a scene
- US2017110045 (A1) ― 2017-04-20 - Display control method and device for implementing said method
- US2017053407 (A1) ― 2017-02-23 - Method of tracking shape in a scene observed by an asynchronous light sensor
- BR112013028742 (A2) ― 2017-01-24 - Method and device for controlling a device for aiding vision
- WO2017060590 (A1) ― 2017-04-13 - Method of optimizing decomposition of an asynchronous signal
- WO2017009543 (A1) ― 2017-01-19 - Data-processing device with representation of values by time intervals between events
- WO2017013065 (A1) ― 2017-01-26 - Method for downsampling a signal outputted by an asynchronous sensor
- EP3271869 (A1) ― 2018-01-24 - Method for processing an asynchronous signal
- EP3272119 (A1) ― 2018-01-24 - Method for the 3d reconstruction of a scene
- US2016086344 (A1) ― 2016-03-24 - Visual tracking of an object
- KR20140130106 (A) ― 2014-11-07 - 발명의 명칭 비동기식 광센서의 베이스에 대한 광흐름 평가 방법
- AU2011247114 (B2) ― 2014-11-27 - Implant having three-dimensional shape for electrically stimulating a nerve structure
- WO2013083848 (A1) ― 2013-06-13 - Method of 3d reconstruction of a scene calling upon asynchronous sensors
- FR2922074 (A1) ― 2009-04-10 - Procédé de synchronisation de flux vidéo
- US2007008405 (A1) ― 2007-01-11 - Method for calibrating at least two video cameras relatively to each other for stereoscopic filming and device therefor
The video below shows Chronocam image acquisition tracking a person.
CNRS - Institut de la vision