Search papers, labs, and topics across Lattice.
This paper introduces MATA, a Modular Asynchronous Tracking Architecture for UAV-based object tracking, which combines a transformer-based tracker with an Extended Kalman Filter for ego-motion compensation and trajectory modeling. They propose a hardware-independent evaluation protocol and a new metric, Normalized Time to Failure (NT2F), to better assess tracker performance in embedded UAV scenarios. Experiments on UAV benchmarks, including an augmented UAV123 dataset, demonstrate improved Success and NT2F metrics, validated by a ROS 2 implementation on an Nvidia Jetson AGX Orin.
Transformer-based trackers can fly longer and track better on UAVs when fused with Kalman filters and evaluated with a new "Normalized Time to Failure" metric that exposes real-time embedded performance.
Object tracking from Unmanned Aerial Vehicles (UAVs) is challenged by platform dynamics, camera motion, and limited onboard resources. Existing visual trackers either lack robustness in complex scenarios or are too computationally demanding for real-time embedded use. We propose an Modular Asynchronous Tracking Architecture (MATA) that combines a transformer-based tracker with an Extended Kalman Filter, integrating ego-motion compensation from sparse optical flow and an object trajectory model. We further introduce a hardware-independent, embedded oriented evaluation protocol and a new metric called Normalized time to Failure (NT2F) to quantify how long a tracker can sustain a tracking sequence without external help. Experiments on UAV benchmarks, including an augmented UAV123 dataset with synthetic occlusions, show consistent improvements in Success and NT2F metrics across multiple tracking processing frequency. A ROS 2 implementation on a Nvidia Jetson AGX Orin confirms that the evaluation protocol more closely matches real-time performance on embedded systems.
