Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Conditional Generative Models for High-Resolution Range Profiles: Capturing Geometry-Driven Trends in a Large-Scale Maritime Dataset

Edwyn Brient, Santiago Velasco-Forero, Rami Kassab

Abstract

High-resolution range profiles (HRRPs) enable fast onboard processing for radar automatic target recognition, but their strong sensitivity to acquisition conditions limits robustness across operational scenarios. Conditional HRRP generation can mitigate this issue, yet prior studies are constrained by small, highly specific datasets. We study HRRP synthesis on a largescale maritime database representative of coastal surveillance variability. Our analysis indicates that the fundamental scenario drivers are geometric: ship dimensions and the desired aspect angle. Conditioning on these variables, we train generative models and show that the synthesized signatures reproduce the expected line-of-sight geometric trend observed in real data. These results highlight the central role of acquisition geometry for robust HRRP generation.

Conditional Generative Models for High-Resolution Range Profiles: Capturing Geometry-Driven Trends in a Large-Scale Maritime Dataset

Abstract

High-resolution range profiles (HRRPs) enable fast onboard processing for radar automatic target recognition, but their strong sensitivity to acquisition conditions limits robustness across operational scenarios. Conditional HRRP generation can mitigate this issue, yet prior studies are constrained by small, highly specific datasets. We study HRRP synthesis on a largescale maritime database representative of coastal surveillance variability. Our analysis indicates that the fundamental scenario drivers are geometric: ship dimensions and the desired aspect angle. Conditioning on these variables, we train generative models and show that the synthesized signatures reproduce the expected line-of-sight geometric trend observed in real data. These results highlight the central role of acquisition geometry for robust HRRP generation.
Paper Structure (13 sections, 6 equations, 6 figures, 1 table)

This paper contains 13 sections, 6 equations, 6 figures, 1 table.

Table of Contents

  1. Introduction
  2. High-Resolution Range Profile Background
  3. Proposed Method
  4. Overview
  5. Conditioning Generative Models for HRRP
  6. Denoising Diffusion Probabilistic Models (DDPMs)
  7. Conditioning GANs
  8. Experimental Setup and Findings
  9. Dataset
  10. Evaluation Metrics
  11. Results
  12. Discussion
  13. Conclusion

Figures (6)

  • Figure 1: Three different HRRP measurements of the same ship at nearby aspect angles.
  • Figure 2: Illustration of HRRP structure and its dependence on aspect angle: HRRP represents the amplitude of the combined echoes scattered by individual points within each range cell. Low-RCS parts of the ship cause drops in the HRRP, while high-RCS regions appear as peaks. The table highlights how aspect angle affects the coarse-scale structure of the HRRP by providing 3 data of the same ship at different aspect angles.
  • Figure 3: Dataset SNR distribution.
  • Figure 4: Dataset ship dimensions. Blue: train, Red: val/test.
  • Figure 5: LRP profiles from test data and samples generated with the proposed models. Top: test LRPs; gaps in aspect angles reflect missing measurements. Middle: GAN-generated LRPs conditioned on aspect angle, which follow the main trend but do not reproduce outliers. Bottom: LRPs measured on DDPM-generated HRRP conditioned on aspect angle, capturing both the trend and the outliers. Green: theoretical ship projection (TLOP, Eq. \ref{['eq:tlop']}). The different columns correspond to different ships of various dimensions. Overall, LRPs from generated and test data follow the TLOP trend.
  • ...and 1 more figures