Data rate estimation based on scene complexity for dynamic vision sensors on unmanned vehicles

Abstract

Event-based cameras are very promising sensors for high speed mobile robots, self-driving cars and drones. Event-based Vision Sensors output a sequence of events data rather than a video bitstream as for classical cameras. The data rate output by such sensors depends upon the type of scene and camera motion speed. Estimating the data rate associated to such sensors is needed for the appropriate design of transmission systems. This paper suggests considering information on the scene content to support such estimation. The considered content parameter shows a high correlation with the event rate, appearing as a promising parameter for event rate and hence data rate estimation.