Back to Search
Start Over
Drone Virtual Fence Using a Neuromorphic Camera
- Source :
- ICONS
- Publication Year :
- 2021
- Publisher :
- ACM, 2021.
-
Abstract
- Neuromorphic cameras are well suited to detect the motion of propellers (blades) on Unmanned Aerial Systems (UAS), or drones. In this paper, we introduce the concept of a virtual fence which is a low-cost networked situational awareness device to quickly alert that a drone has entered the zone. Neuromorphic cameras significantly reduce the amount of data that must be processed as opposed to conventional cameras. Processing is required only when events are generated. Those events can be generated by a drone, by another low altitude airborne object (projectiles or birds), or by variations in the background. We propose two complementary algorithms that allow us to differentiate the signature of propeller blades from other events. Those algorithms exploit the periodic nature of propellers’ signal and the presence of sub-harmonics in the detected signal. Those sub-harmonics are introduced in the signal when a camera pixel misses some high-frequency events. We also show how to adjust the optics of the camera so as to reduce the contrast of background events, simplifying the categorization task. A prototype of a system consuming, during normal operations, 5.14 W with a battery autonomy of to 27 hours is presented. This prototype can detect drones up to an altitude of 9 m using a DAVIS 346 from IniVation with a field of view of about 70 degrees. Based on the actual improvement in resolution of current and next generation neuromorphic cameras, it is expected that the range of detection will increase and the virtual fence concept could be deployed operationally in the next few years.<br />ICONS 2021: International Conference on Neuromorphic Systems 2021, July 27-29, 2021, Knoxville, Tennessee [Virtual Event]<br />Series: International Conference Proceeding Series (ICPS)
Details
- Database :
- OpenAIRE
- Journal :
- International Conference on Neuromorphic Systems 2021
- Accession number :
- edsair.doi.dedup.....80908a8ff71fe48c386236cfebd653c2