Realistic SAR data augmentation using machine learning techniques

Lewis, Benjamin P.; DeGuchy, Omar; Strzalka, Joseph; Kaminski, John E.

doi:10.1117/12.2518452

Cited by 29 publications

(22 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This is motivated by the observation that using synthetic data in its unaltered form yields poor performing models in measured data applications [9]. Some methods in this category rely on training DL-based Generative Adversarial Networks (GANs) and/or Auto-Encoders to learn the transform based on seeing many examples of both synthetic and measured data [10], [11], [9]. Further, Scarnati and Lewis [8] design a pre-processing function for synthetic SAMPLE data which performs a de-specking, quantization, and clutter transfer between (measured, synthetic) pairs.…”

Section: Learning With Synthetic Datamentioning

confidence: 99%

Bridging a Gap in SAR-ATR: Training on Fully Synthetic and Testing on Measured Data

Inkawhich

Davis

et al. 2021

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

Obtaining measured Synthetic Aperture Radar (SAR) data for training Automatic Target Recognition (ATR) models can be too expensive (in terms of time and money) and complex of a process in many situations. In response, researchers have developed methods for creating synthetic SAR data for targets using electromagnetic prediction software, which is then used to enrich an existing measured training dataset. However, this approach relies on the availability of some amount of measured data. In this work, we focus on the case of having 100% synthetic training data, while testing on only measured data. We use the SAMPLE dataset public released by AFRL, and find significant challenges to learning generalizable representations from the synthetic data due to distributional differences between the two modalities and extremely limited training sample quantities. Using deep learning-based ATR models, we propose data augmentation, model construction, loss function choices, and ensembling techniques to enhance the representation learned from the synthetic data, and ultimately achieved over 95% accuracy on the SAMPLE dataset. We then analyze the functionality of our ATR models using saliency and feature-space investigations and find them to learn a more cohesive representation of the measured and synthetic data. Finally, we evaluate the out-oflibrary detection performance of our synthetic-only models and find that they are nearly 10% more effective than baseline methods at identifying measured test samples that do not belong to the training class set. Overall, our techniques and their compositions significantly enhance the feasibility of using ATR models trained exclusively on synthetic data.

show abstract

Section: Learning With Synthetic Datamentioning

confidence: 99%

Bridging a Gap in SAR-ATR: Training on Fully Synthetic and Testing on Measured Data

Inkawhich

Davis

et al. 2021

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

show abstract

“…Lewis et al, in their most recent work, achieved average accuracies of

[ 39 ] and

[ 43 ] by training a DenseNet with the assistance of a Generative Adversarial Network (GAN). Scarnati et al reached an accuracy of

[ 41 ] also by using DenseNet and no more than

[ 45 ] when using Complex-Valued Neural Networks (CVNN).…”

Section: Discussionmentioning

confidence: 99%

“…Feature-based algorithms are those with methods that run offline training supported exclusively by features extracted from the targets of interest. Among the methods employed by feature-based algorithms, we can highlight the following: Template Matching (TM) [ 5 , 6 , 7 , 11 , 30 , 37 ], Hidden Markov Model (HMM) [ 12 , 13 , 22 ], K-Nearest Neighbor (KNN) [ 27 , 28 ], Sparse Representation-based Classification (SRC) [ 8 , 29 ], Convolutional Neural Networks (CNN) [ 17 , 18 , 36 , 38 , 39 , 40 , 41 , 42 , 43 , 44 , 45 , 46 , 47 , 48 ], Support Vectors Machine (SVM) [ 9 ] and Gaussian Mixture Model (GMM) [ 10 ].…”

Section: Introductionmentioning

confidence: 99%

“…Therefore, there is an enormous demand for works that address the SAMPLE dataset [ 51 ]. So far, few works [ 38 , 39 , 40 , 41 , 42 , 43 , 44 , 45 , 46 , 47 , 48 ] have been carried out that focus on the SAMPLE dataset for classifying measured data based exclusively on synthetic data, and they all use a feature-based approach with machine-learning algorithms like CNN or DenseNet.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Non-Cooperative SAR Automatic Target Recognition Based on Scattering Centers Models

Araujo

Machado

Pettersson

2022

Sensors

View full text Add to dashboard Cite

This article proposes an Automatic Target Recognition (ATR) algorithm to classify non-cooperative targets in Synthetic Aperture Radar (SAR) images. The scarcity or nonexistence of measured SAR data demands that classification algorithms rely only on synthetic data for training purposes. Based on a model represented by the set of scattering centers extracted from purely synthetic data, the proposed algorithm generates hypotheses for the set of scattering centers extracted from the target under test belonging to each class. A Goodness of Fit test is considered to verify each hypothesis, where the Likelihood Ratio Test is modified by a scattering center-weighting function common to both the model and target. Some algorithm variations are assessed for scattering center extraction and hypothesis generation and verification. The proposed solution is the first model-based classification algorithm to address the recently released Synthetic and Measured Paired Labeled Experiment (SAMPLE) dataset on a 100% synthetic training data basis. As a result, an accuracy of 91.30% in a 10-target test within a class experiment under Standard Operating Conditions (SOCs) was obtained. The algorithm was also pioneered in testing the SAMPLE dataset in Extend Operating Conditions (EOCs), assuming noise contamination and different target configurations. The proposed algorithm was shown to be robust for SNRs greater than −5 dB.

show abstract

“…Growing awareness of these issues within the radar community has spawned new research routes dedicated to analysing the feasibility of training classifiers on highly augmented [9] and even purely synthetic data [10]. To support this approach, new simulated datasets with corresponding samples from real collections are required.…”

Section: Introductionmentioning

confidence: 99%

SARCASTIC v2.0—High-Performance SAR Simulation for Next-Generation ATR Systems

et al. 2022

View full text Add to dashboard Cite

Synthetic aperture radar has been a mainstay of the remote sensing field for many years, with a wide range of applications across both civilian and military contexts. However, the lack of openly available datasets of comparable size and quality to those available for optical imagery has severely hampered work on open problems such as automatic target recognition, image understanding and inverse modelling. This paper presents a simulation and analysis framework based on the upgraded SARCASTIC v2.0 engine, along with a selection of case studies demonstrating its application to well-known and novel problems. In particular, we demonstrate that SARCASTIC v2.0 is capable of supporting complex phase-dependent processing such as interferometric height extraction whilst maintaining near-realtime performance on complex scenes.

show abstract

Realistic SAR data augmentation using machine learning techniques

Cited by 29 publications

References 10 publications

Bridging a Gap in SAR-ATR: Training on Fully Synthetic and Testing on Measured Data

Bridging a Gap in SAR-ATR: Training on Fully Synthetic and Testing on Measured Data

Non-Cooperative SAR Automatic Target Recognition Based on Scattering Centers Models

SARCASTIC v2.0—High-Performance SAR Simulation for Next-Generation ATR Systems

Contact Info

Product

Resources

About