Research on the Static Response Mechanism of Event Cameras and Their Applications in Static Scenes

Abstract

Event cameras are novel sensors that have gained significant importance in the field of dynamic vision since the first prototype in 2008. Unlike conventional frame-based cameras, event cameras exhibit independent and asynchronous responses at each pixel to changes in brightness. This unique characteristic allows them to have microsecond time resolution and low latency. Currently, event cameras are widely deployed for dynamic vision tasks such as high-speed motion capture. However, their potential in static scenes remains largely unexplored and not well understood. This research aims to address this gap by delving into the principles and applications of event cameras in static themes. This work explains the phenomenon that event cameras respond differently to static objects of varying brightness. Specifically, a theoretical model is derived, linking event generation to photon absorption rate, and a new concept of static sensitivity is introduced to quantitatively analyze the phenomenon. In addition to the theoretical work, we propose two techniques to unleash the potential of event cameras in static scenes. First, a static calibration method is developed for event cameras, achieving a pattern detection ratio of 0.98 and an average reprojection error of 0.13. Second, we propose to utilize event cameras to detect transparent objects, achieving the successful detection and grasping of a transparent cube. The efficacy and robustness of these techniques are validated through extensive experiments.