Maritime object classification with SAR imagery using quantum kernel methods

Illegal, unreported, and unregulated (IUU) fishing causes global economic losses of $10-25 billion annually and undermines marine sustainability and governance. Synthetic Aperture Radar (SAR) provides reliable maritime surveillance under all weather and lighting conditions, but classifying small maritime objects in SAR imagery remains challenging. We investigate quantum machine learning for this task, focusing on Quantum Kernel Methods (QKMs) applied to real and complex SAR chips extracted from the SARFish dataset. We tackle two binary classification problems, the first for distinguishing vessels from non-vessels, and the second for distinguishing fishing vessels from other types of vessels. We compare QKMs applied to real and complex SAR chips against classical Laplacian, RBF, and linear kernels applied to real SAR chips. Using noiseless numerical simulations of the quantum kernels, we find that QKMs are capable of obtaining equal or better performance than the classical kernel on these tasks in the best case, but do not demonstrate a clear advantage for the complex SAR data. This work presents the first application of QKMs to maritime classification in SAR imagery and offers insight into the potential and current limitations of quantum-enhanced learning for maritime surveillance.

💡 Research Summary

This paper presents a pioneering investigation into the application of Quantum Kernel Methods (QKMs) for maritime object classification using Synthetic Aperture Radar (SAR) imagery. The primary motivation is to enhance surveillance capabilities against Illegal, Unreported, and Unregulated (IUU) fishing, a global problem causing significant economic and environmental damage. While SAR provides all-weather, day-and-night imaging capability, classifying small vessels within these images remains technically challenging.

The authors explore whether quantum machine learning, specifically QKMs, can offer advantages for this task. QKMs are kernel methods that use a quantum computer to evaluate a kernel function, potentially enabling access to high-dimensional feature spaces that are computationally intractable classically. A particular hypothesis is that quantum computers, which operate natively in complex Hilbert spaces, might be naturally suited to process the complex-valued data inherent in raw SAR signals (Single Look Complex - SLC), as opposed to the commonly used real-valued detected products (Ground Range Detected - GRD).

The study utilizes the SARFish dataset, which contains coincident real (GRD) and complex (SLC) SAR image chips. Two binary classification tasks are defined: 1) distinguishing vessels from non-vessels (e.g., offshore platforms), and 2) distinguishing fishing vessels from other vessel types. The performance of QKMs applied to both real and complex data is benchmarked against classical Support Vector Machines (SVMs) employing standard Laplacian, Radial Basis Function (RBF), and linear kernels on real data. Crucially, the quantum kernels are evaluated using noiseless numerical simulations to assess their theoretical potential without current hardware limitations.

The results demonstrate that, in their best configuration, QKMs can achieve performance equal to or better than the best classical kernel on these tasks. This finding is significant as it provides a proof-of-concept that QKMs can be effectively applied to a real-world, non-trivial image classification problem in remote sensing. However, the study did not observe a clear, consistent advantage of using QKMs with complex-valued SAR data over real-valued data. This indicates that the specific quantum feature maps and encoding strategies used may not have fully leveraged the phase information in the complex data, or that such information may not be decisively beneficial for these particular classification tasks within the SARFish dataset.

In conclusion, this work establishes the first application of Quantum Kernel Methods to maritime classification in SAR imagery. It confirms the feasibility and potential of QKMs for this domain while offering a nuanced insight: realizing a practical quantum advantage requires careful co-design of the quantum algorithm (feature map/encoding) with the specific characteristics of the problem domain, and theoretical suitability (like handling complex numbers) does not automatically translate into superior performance. The study lays groundwork for future research to explore more specialized quantum embeddings tailored for SAR signal processing.