Training Deep Learning Models with Hybrid Datasets for Robust Automatic Target Detection on real SAR images

TL;DR

This paper introduces a deep learning method for automatic target detection in SAR images using hybrid datasets of synthetic and real data, achieving high accuracy without real target examples in training.

Contribution

It presents a novel hybrid dataset creation pipeline and training strategy that effectively bridges the domain gap between synthetic and real SAR images for target detection.

Findings

Achieves up to 90% Average Precision on real data.

Hybrid datasets eliminate image overlay bias.

Synthetic training data can replace real targets effectively.

Abstract

In this work, we propose to tackle several challenges hindering the development of Automatic Target Detection (ATD) algorithms for ground targets in SAR images. To address the lack of representative training data, we propose a Deep Learning approach to train ATD models with synthetic target signatures produced with the MOCEM simulator. We define an incrustation pipeline to incorporate synthetic targets into real backgrounds. Using this hybrid dataset, we train ATD models specifically tailored to bridge the domain gap between synthetic and real data. Our approach notably relies on massive physics-based data augmentation techniques and Adversarial Training of two deep-learning detection architectures. We then test these models on several datasets, including (1) patchworks of real SAR images, (2) images with the incrustation of real targets in real backgrounds, and (3) images with the incrustation of synthetic background objects in real backgrounds. Results show that the produced hybrid datasets are exempt from image overlay bias. Our approach can reach up to 90% of Average Precision on real data while exclusively using synthetic targets for training.