High-Speed Object Tracking with Dynamic Vision Sensor

Abstract

High-speed object tracking is still a great challenge for video processing. Traditional cameras can hardly capture the motion trajectory of the high-speed moving object. With differential logarithmic photodetector and nanosecond response latency to fast stimuli, dynamic vision sensor (DVS) is extremely sensitive to the moving object (especially for the object with high speed). However, existing object tracking algorithms, which are limited by their frame-by-frame processing mode, are no longer suitable for DVS. In this work, we introduce a novel event coherence detection algorithm for high-speed objective tracking. The moving target is determined by judging the coherence of the event according to the event distribution at a certain moment. Experimental results demonstrate that the proposed algorithm can accurately track the small objects with high speed. Meanwhile, the proposed algorithm performs efficiently, which can run in real time.