Phenomics

Exploring the Potential of Quantum Enhanced Multi-Object Tracking and Re-Identification

---

Abstract

Quantum computing is an emerging technology that leverages quantum bits (qubits) capable of existing in superposition and entanglement to enable fundamentally new approaches to computation. Unlike classical systems that process bits through logical gates, quantum computers do not simply accelerate tasks but require specialized algorithms to exploit their unique properties for solving complex problems efficiently. While applications in agricultural science remain largely unexplored, this work demonstrates a potential use case by addressing a computer vision challenge: multi-object tracking. Specifically, we introduce a novel re-identification module formulated as a network flow optimization problem and implemented using quantum techniques to improve tracking consistency across frames.The goal of multi-object tracking is to track detected objects of interest across frames. In our case, we used YOLOv8 together with BoT-SORT to detect and track cows. State-of-the-art frameworks like BoT-SORT use the shape and expected trajectory to derive track sequences. Here, we propose a new implementation of a framework for the pair-wise comparison of tracked objects at consecutive timepoints based on the overlap between bounding boxes. Tracklets are subsequently obtained by solving an optimization problem using the Intersection over Union as the primary cost metric, this avoids relying on appearance-based features that are either not present or obscured by camera view, changing light conditions, or dust and dirt. The developed approach is available as an extension for all object trackers and is implemented for both classical and quantum hardware, as a Mixed-Integer Linear Programming (MILP) and a Quadratic Unconstrained Binary Optimization (QUBO) algorithm, respectively. On a test dataset, including 16 cows in a single barn over several video snippets of approximately three minutes, our approach outperformed BoT-SORT with default ReID on key performance metrics for detection performance and identity consistency, improving HOTA from 89.66 to 90.81, MOTA from 95.44 to 96.27, IDF1 from 95.95 to 97.40, and reducing the number of ID switches from 9 to 5. Although the quantum solution at this stage is slower than its classical counterpart, we expect long-term promises as quantum computing as a technology is still in its infancy and quantum hardware is expected to become more reliable and computationally performant in the future. This work provides a first proof-of-concept on the use of quantum computing for computer vision applications that can complement classical methods in future precision livestock monitoring systems.

Keywords: 2026

How to Cite:

Aguilera, E., Boschero, J., Cramer, S., Doornweerd, J., Leijnse, K., Pook, T., Scavuzzo Montana, L., van Vliet, M. & Vandenplas, J., (2026) “Exploring the Potential of Quantum Enhanced Multi-Object Tracking and Re-Identification”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2252487. doi: https://doi.org/10.31274/wcgalp.23386

Rights: 1