A machine learning approach for image classification in synthetic aperture RADAR

2025-08-06

AI Summary

This study addresses geometric shape and sea-ice-type classification for ground targets in synthetic aperture radar (SAR) imagery. We propose an end-to-end convolutional neural network (CNN)-based recognition framework. Methodologically, we innovatively integrate a single-scattering approximation physical model to synthesize SAR data, and systematically compare classification performance across three input modalities: simulated SAR data, reconstructed SAR images, and real Sentinel-1 acquisitions. To our knowledge, this work is the first to quantitatively analyze the impact of antenna height variation on SAR image discriminability. Experimental results demonstrate that the CNN achieves >75% accuracy on both fine-grained classification tasks, validating the effectiveness of physics-informed data synthesis in enhancing deep learning generalizability for SAR imagery. The proposed approach establishes a reproducible benchmark framework for low-sample, high-noise SAR interpretation.

Technology Category

• Computer Vision: Low Level & Physics-based Vision
• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

We consider the problem in Synthetic Aperture RADAR (SAR) of identifying and classifying objects located on the ground by means of Convolutional Neural Networks (CNNs). Specifically, we adopt a single scattering approximation to classify the shape of the object using both simulated SAR data and reconstructed images from this data, and we compare the success of these approaches. We then identify ice types in real SAR imagery from the satellite Sentinel-1. In both experiments we achieve a promising high classification accuracy ($geq$75%). Our results demonstrate the effectiveness of CNNs in using SAR data for both geometric and environmental classification tasks. Our investigation also explores the effect of SAR data acquisition at different antenna heights on our ability to classify objects successfully.

Problem

• Classifying ground objects in SAR images using CNNs
• Comparing simulated and reconstructed SAR data for shape classification
• Assessing antenna height impact on SAR object classification accuracy

Innovation

• Uses Convolutional Neural Networks for SAR image classification
• Applies single scattering approximation for shape classification
• Tests classification accuracy with varying antenna heights