A Noval Event-IMU Odometry Based on Quadtree-Accelerated Asynchronous Event Corner and Graph Optimization.

Event cameras, a type of neuromorphic vision sensor, capture pixel-level intensity changes to generate asynchronous event streams with precise timestamps. They offered several advantages, including a high dynamic range, high temporal resolution (low latency), and freedom from motion blur. Leveraging event information and corresponding visual algorithms holded significant promise for the development of robotic autonomous systems. In this paper, we proposed an event-inertial odometry method based on asynchronous event corner detection and tracking. We introduced a novel approach for fast event corner detection using a quadtree-based method, ensuring a spatiotemporally uniform distribution of corners. Additionally, we presented a high-precision state estimation framework that combines IMU measurements and event corner observations through graph optimization. To assess the performance of our method, we conducted experiments using a publicly available dataset and compared our results with state-of-the-art (SOTA) methods. Our findings demonstrated that our approach excels in terms of both accuracy and speed.