Motion Segmentation and Egomotion Estimation from Event-Based Normal Flow

2025-07-19

AI Summary

Existing motion segmentation and ego-motion estimation methods for neuromorphic vision sensors typically rely on optical flow or explicit depth estimation, limiting their efficiency and robustness in dynamic, low-texture, or depth-ambiguous scenes. Method: This paper proposes a lightweight, joint optimization framework that leverages event-driven normal flow—implicitly encoding scene geometry—augmented by inertial measurements and geometric constraints, without requiring optical flow or dense depth. Contribution/Results: Key innovations include event-based superpixel segmentation, residual-driven iterative optimization, and hierarchical clustering integrating motion similarity and temporal consistency—significantly improving segmentation accuracy at object boundaries and translational ego-motion estimation. Evaluated on the EVIMO2v2 dataset, the method demonstrates robustness, real-time performance (≥1 kHz event-rate scalability), and temporal resolution fidelity under high-speed motion. It establishes a new paradigm for event-camera-enabled robotic navigation, offering efficient, geometry-aware perception with minimal computational overhead.

Technology Category

• Computer Vision: Motion & Tracking
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Search and Optimization: Learning to Search

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

This paper introduces a robust framework for motion segmentation and egomotion estimation using event-based normal flow, tailored specifically for neuromorphic vision sensors. In contrast to traditional methods that rely heavily on optical flow or explicit depth estimation, our approach exploits the sparse, high-temporal-resolution event data and incorporates geometric constraints between normal flow, scene structure, and inertial measurements. The proposed optimization-based pipeline iteratively performs event over-segmentation, isolates independently moving objects via residual analysis, and refines segmentations using hierarchical clustering informed by motion similarity and temporal consistency. Experimental results on the EVIMO2v2 dataset validate that our method achieves accurate segmentation and translational motion estimation without requiring full optical flow computation. This approach demonstrates significant advantages at object boundaries and offers considerable potential for scalable, real-time robotic and navigation applications.

Problem

• Robust motion segmentation using event-based normal flow
• Egomotion estimation without full optical flow computation
• Handling sparse event data for real-time robotic applications

Innovation

• Uses event-based normal flow for motion segmentation
• Incorporates geometric constraints with inertial data
• Optimization pipeline with hierarchical clustering refinement