Motion Compensation of Event-Streams for Multitarget Tracking in Sensor-Based Sorting

Event-based cameras excel in representing dynamic changes efficiently but can be overwhelmed in scenarios with strong global motion, such as conveyor belt systems. We propose a motion compensation method that filters events to focus on relative motion deviations. Our method significantly reduces the event-stream size while maintaining or even improving tracking accuracy. Our approach incorporates prior knowledge of global motion and demonstrates its efficacy in sensor-based sorting tasks.