High-speed object tracking and localization using event-based extended Kalman filtering

Object tracking algorithms based on traditional cameras exhibit poor performance under conditions of complex backgrounds or significant lighting variations, whereas event cameras, which feature high temporal resolution and wide dynamic range, can overcome these limitations. This paper proposes an efficient and fast tracking and localization algorithm based on event streams. To be specific, a novel event-based fast corner detection algorithm has been designed to perform in a highly efficient and timely manner in high-speed dynamic and low-light conditions. Finally, the extended Kalman filter is chosen as the primary method for estimating the target’s motion state, thereby accomplishing the task of target tracking based on event stream information. Comprehensive evaluation results demonstrate that the proposed algorithm achieves high tracking success rates and low average processing times in various scenarios involving high-speed single and multiple target tracking.