MFN Decomposition and Related Metrics for High-Resolution Range Profiles Generative Models

2025-10-04

International Radar Conference

AI Summary

This work addresses the lack of interpretable, multi-level evaluation methods for high-resolution range profile (HRRP) generation models, which often rely on black-box classifiers and thus hinder physical-level quality analysis. The authors propose the first interpretable decomposition framework tailored for generated HRRP data, decoupling each HRRP into three physically meaningful components: Mask, Features, and Noise (MFN decomposition). By integrating radar target physics, they design evaluation metrics with clear physical interpretations, moving beyond conventional black-box assessment paradigms. The framework is validated on real-world, high-cost HRRP datasets and demonstrates its ability to accurately discriminate the output quality of different generative models, thereby establishing a new paradigm for both HRRP generation and its evaluation.

Technology Category

• Natural Language Processing: Generation
• Computer Vision: Low Level & Physics-based Vision
• Machine Learning: Multimodal Learning

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

High-resolution range profile (HRRP) data are in vogue in radar automatic target recognition (RATR). With the interest in classifying models using HRRP, filling gaps in datasets using generative models has recently received promising contributions. Evaluating generated data is a challenging topic, even for explicit data like face images. However, the evaluation methods used in the state-of-the-art of HRRP generation rely on classification models. Such models, called “black-box”, do not allow either explainability on generated data or multilevel evaluation. This work focuses on decomposing HRRP data into three components: the mask, the features, and the noise. Using this decomposition, we propose two metrics based on the physical interpretation of those data. We take profit from an expensive dataset to evaluate our metrics on a challenging task and demonstrate the discriminative ability of those.

Problem

• HRRP
• generative models
• evaluation metrics
• radar automatic target recognition
• data decomposition

Innovation

• MFN decomposition
• HRRP generative models
• explainable evaluation
• physical-based metrics
• radar target recognition