E2-VINS: An Event-Enhanced Visual–Inertial SLAM Scheme for Dynamic Environments

• DOI: 10.3390/app15031314
• Publication Date: 2025-01-27T09:59:10Z
• Authors: Jiafeng Huang; Shengjie Zhao; Lin Zhang
• Keywords: Technology; QH301-705.5; T; Physics; QC1-999; event camera; Engineering (General). Civil engineering (General); robot vision; Chemistry; dynamic SLAM; TA1-2040; Biology (General); QD1-999; robust bundle adjustment

Simultaneous Localization and Mapping (SLAM) technology has garnered significant interest in the robotic vision community over past few decades. The rapid development of SLAM resulted its widespread application across various fields, including autonomous driving, robot navigation, virtual reality. Although SLAM, especially Visual–Inertial (VI-SLAM), made substantial progress, most classic algorithms this field are designed based on assumption that observed scene is static. In complex real-world environments, presence dynamic objects such as pedestrians vehicles can seriously affect robustness accuracy systems. Event cameras, which use recently introduced motion-sensitive biomimetic sensors, efficiently capture changes (referred to “events”) with high temporal resolution, offering new opportunities enhance VI-SLAM performance environments. Integrating kind innovative sensor, we propose first event-enhanced framework specifically for termed E2-VINS. Specifically, system uses visual–inertial alignment strategy estimate IMU biases correct measurements. calibrated measurements used assist motion compensation, achieving spatiotemporal events. event-based dynamicity metrics, measure each pixel, then generated these aligned Based visual residual terms different pixels adaptively assigned weights, namely, weights. Subsequently, E2-VINS jointly alternately optimizes state (camera poses map points) effectively filtering out features through a soft-threshold mechanism. Our scheme enhances against features, significantly resulting an average improvement 1.884% mean position error compared state-of-the-art methods. superior validated both qualitative quantitative experimental results. To ensure our results fully reproducible, all relevant data codes have been released.