Low-Latency Motion Extraction from Event Cameras on FPGA

Abstract

Event-based vision sensors provide asynchronous, high-temporal-resolution measurements that are attractive for low-latency robotic perception. This project investigates a lightweight, hardware-friendly method for detecting motion from event-camera data in real time by testing a small set of motion hypotheses rather than computing full optical flow for each pixel. The goal is to validate the approach and assess how efficiently it can be mapped to FPGA hardware.

Method Overview

Incoming events are discretized into short time bins and processed independently along the horizontal and vertical axes using shift-register- based logic and windowed accumulation.

Resources

Contact

Arianna Alonso Bizzi at alari@upenn.edu
University of Pennsylvania