This video shows our proof of concept counter-unmanned aerial system (CUAS) drone tracking at various distances using the NeuroSight™ event camera platform with the Prophesee IMX636 sensor. The video was processed using a generic, untrained detection and tracking algorithm designed for event data ingest. NeuroSight was tracking a DJI Inspire 2 drone at various distances using a 75 mm lens to understand basic system limitations.
Video recorded August 12, 2025
This video shows the accumulated event data displayed as a frame imaging the layers of plumes and explosions from conventional fireworks using the Prophesee IMX636 sensor with no sensor bias tuning, in daylight, at a rate of 33 ms accumulation per frame at 30 fps. The false color denotes the event polarity (white is an ON event, or increase in relative brightness and blue is an OFF event, or decrease in relative brightness). Dark blue denotes areas where no events were captured within the sensor parameter bounds.
An event sensor differs from conventional frame sensors in that each pixel activates asynchronously when a relative change in brightness occurs beyond a user settable threshold (on the order of 1 μs). If an event is seen, the pixel returns the sparse XYPT (X,Y position in the pixel array, Polarity of the event, microsecond resolution Timestamp of when the event occurred), and returns to a ready state to receive more events. Traditional frame sensors rely on global or rolling electronic shutters with a given exposure time to accumulate the desired data where event sensors have no general exposure time and provide a variable event data stream based on the motion in the scene instead of individual frames of data. This general exposure time makes it difficult to use frame sensors for high-speed imaging applications without the need for expensive sensors and added light.
Video recorded May 19, 2025