Cold Start Astrometry for High-Precision Airspace and Space Objects Tracking with Neuromorphic Cameras

Supervisors:

Primary supervisor A/Prof. Gregory Cohen

Description:

Objects in the sky – birds, drones, planes, meteorites, and satellites – are typically tracked with radars or motorized telescopes. These heavy instruments are extremely expensive and must be anchored in concrete slabs. It takes years to deploy a new observatory in remote places such as the Australian outback, and the observatory cannot be moved after its construction.

Neuromorphic cameras are silicon sensors whose pixels’ circuits are inspired by biological retinas. They do not generate frames but report changes in luminance in the form of asynchronous events. This sensing strategy breaks the fundamental relationship between frame rate and data rate. As a result, Neuromorphic cameras have an extremely high temporal resolution but generate little data. Their output can be processed in real time with low latency and little power.

Neuromorphic cameras open the door to a new type of sky monitoring device. One that fits in the boot of a car, can be rapidly deployed in remote places, uses little power, is inexpensive, and tracks unknown objects with incredibly high temporal resolution. Such a device – or better yet, a network of such devices – would have a wide range of applications including wildlife monitoring, airspace control, and space debris monitoring.

Outcomes:

During this thesis, the candidate will design novel Neuromorphic computer vision algorithms to tackle the open problems that must be solved to make the device possible. These problems include:

• Use stars to automatically calibrate a motorized Neuromorphic camera deployed in an unknown position (a problem that we dubbed Cold start astrometry).
• Follow an extremely fast object using cues from a Neuromorphic camera, also known as close loop tracking.
• Fuse data from multiple Neuromorphic sensors to reconstruct 3D trajectories.