Monocular Event-Inertial Odometry with Adaptive decay-based Time Surface and Polarity-aware Tracking

TL;DR

Comparative analysis with visual-inertial and event-inertial odometry methods shows that the proposed monocular event-inertial odometry incorporating an adaptive decay-based time surface outperforms state-of-the-art techniques, with competitive results across various datasets.

Abstract

Event cameras have garnered considerable attention due to their advantages over traditional cameras in low power consumption, high dynamic range, and no motion blur. This paper proposes a monocular event-inertial odometry incorporating an adaptive decay kernel-based time surface with polarity-aware tracking. We utilize an adaptive decay-based Time Surface to extract texture information from asynchronous events, which adapts to the dynamic characteristics of the event stream and enhances the representation of environmental textures. However, polarity-weighted time surfaces suffer from event polarity shifts during changes in motion direction. To mitigate its adverse effects on feature tracking, we optimize the feature tracking by incorporating an additional polarity-inverted time surface to enhance the robustness. Comparative analysis with visual-inertial and event-inertial odometry methods shows that our approach outperforms state-of-the-art techniques, with competitive results across various datasets.

Figures (6)

• Figure 1: Event cameras have potential advantages over traditional cameras in high dynamic range and high-speed motion scenes. Time surfaces with an exponential decay kernel (the third column) are unable to adjust their internal parameter (time constant $\eta$) in response to the event stream's dynamic characteristics, resulting in bold edges and redundant events. Conversely, adaptive decay-based time surfaces (the fourth column) offer clear details and less noise.
• Figure 2: System overview. The system includes the Time Surface Map Generation Module, the Adaptive Tracking Module and the State Estimator. The Adaptive Tracking Module fuses the tracking results of Polarity-weighted and Polarity-inverted time surfaces to alleviate the tracking loss that occurs when the direction of motion or illumination changes.
• Figure 3: The impact of event activity on $t_{init}$. The increased value of event activity shortens the temporal scope of active events, thereby mitigating interference from past events.
• Figure 4: The direction of motion influences the polarity of events. Abrupt changes in motion direction can cause a reversal in event polarity, which may lead to tracking loss. Lines are only used to highlight polarity changes.
• Figure 5: Trajectory Estimates Comparison for $hdr\_boxes$ (top) and $hdr\_poster$ (down) in the DAVIS 240C Dataset mueggler2017event