Explainability of Deep SAR ATR Through Feature Analysis

Belloni, Carole; Balleri, Alessio; Aouf, Nabil; Caillec, Jean-Marc Le; Merlet, Thomas

doi:10.1109/taes.2020.3031435

Cited by 29 publications

(17 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To know a target or scene for analysis, detection, or classification, it is desirable to have its SAR image acquired from different positions [26], [27]. Different relative viewing angles (resulting from changes of flight direction or target movement in different revisits) results in a kind of target rotation in SAR image.…”

Section: A Basic Sar Principlesmentioning

confidence: 99%

Analytical Interpretation of Latent Codes in InfoGAN with SAR Images

Feng¹,

Daković²,

Ji³

et al. 2022

Preprint

View full text Add to dashboard Cite

Generative Adversarial Networks (GANs) can synthesize abundant photo-realistic synthetic aperture radar (SAR) images. Some recent GANs (e.g., InfoGAN), are even able to edit specific properties of the synthesized images by introducing latent codes. It is crucial for SAR image synthesis since the targets in real SAR images are with different properties due to the imaging mechanism. Despite the success of InfoGAN in manipulating properties, there still lacks a clear explanation of how these latent codes affect synthesized properties, thus editing specific properties usually relies on empirical trials, unreliable and time-consuming. In this paper, we show that latent codes are disentangled to affect the properties of SAR images in a non-linear manner. By introducing some property estimators for latent codes, we are able to provide a completely analytical nonlinear model to decompose the entangled causality between latent codes and different properties. The qualitative and quantitative experimental results further reveal that the properties can be calculated by latent codes, inversely, the satisfying latent codes can be estimated given desired properties. In this case, properties can be manipulated by latent codes as we expect.

show abstract

Section: A Basic Sar Principlesmentioning

confidence: 99%

Analytical Interpretation of Latent Codes in InfoGAN with SAR Images

Feng¹,

Daković²,

Ji³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…In this part, we use the visualization method to analyze the key parts that affect the target classification. In Belloni et al [53], researchers use a black square to occlude target images and the percentage of correctly classified images is used as the new intensity of the pixels located in the center of the black square in the classification map. Different from it, we occlude target parts instead of pixel squares.…”

Section: E Analysis On Key Parts With Part Occlusionmentioning

confidence: 99%

SAR Target Classification Based on Integration of ASC Parts Model and Deep Learning Algorithm

Feng¹,

Ji²,

Zhang³

et al. 2021

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

View full text Add to dashboard Cite

“…The learned correlation between these weights and each class's image features is critical to ensure that the CNN correctly identifies new images being analyzed [2]. Even the slightest of perturbations in pixel intensities in test image features can deter accurate image classification [3]. In order to accurately label any new image to a learned dataset class, the set of weights must represent a sufficiently diverse set of feature representations belonging to each image class [2].…”

Section: Introductionmentioning

confidence: 99%

“…This limitation ultimately leads to the need to train CNNs with a large number of examples containing all potential representations of the same image class. For imaging techniques involving electromagnetic (EM) probing, this often means parameterizing all applied incidence angles, phases, and instances of capture intensities and frequencies under different environments and backgrounds [3]. Similarly, for any other image platform, this leads to parameterizing every factor contributing to how each class image is developed or represented post processing [4].…”

Section: Introductionmentioning

confidence: 99%

Application of multidomain sensor image fusion and training data augmentation for enhanced CNN image classification

Arnous

Narayanan

2022

J. Electron. Imag.

View full text Add to dashboard Cite

Convolutional neural networks (CNNs) provide the sensing and detection community with a discriminative approach for classifying images. However, one of the largest limitations for deep CNN image classifiers is the need for extensive training datasets containing a variety of image representations. While current methods such as GAN data augmentation, additions of noise, rotations, and translations can allow CNNs to better associate new images and their feature representations to ones of a learned image class, many fail to provide new contexts of ground truth feature information. To expand the association of critical class features within CNN image training datasets, an image pairing and training dataset augmentation paradigm via a multi-sensor domain image data fusion algorithm is proposed. This algorithm uses a mutual information and merit-based feature selection subroutine to pair highly correlated cross domain images from multiple sensor domain image datasets. It then re-augments the corresponding cross domain image pairs into the opposite sensor domain's feature set via a highest mutual information, cross sensor domain, image concatenation function. This augmented image set then acts to retrain the CNN to recognize greater generalizations of image class features via cross-domain, mixed representations. Experimental results indicated an increased ability of CNNs to generalize and discriminate between image classes during testing of class images from SAR vehicle, solar cell device reliability screening and lung cancer detection image datasets.

show abstract

Explainability of Deep SAR ATR Through Feature Analysis

Cited by 29 publications

References 42 publications

Analytical Interpretation of Latent Codes in InfoGAN with SAR Images

Analytical Interpretation of Latent Codes in InfoGAN with SAR Images

SAR Target Classification Based on Integration of ASC Parts Model and Deep Learning Algorithm

Application of multidomain sensor image fusion and training data augmentation for enhanced CNN image classification

Contact Info

Product

Resources

About