Self-Supervised Learning Based Anomaly Detection in Synthetic Aperture Radar Imaging

Muzeau, Max; Ren, Chengfang; Angelliaume, Sébastien; Datcu, Mihai; Ovarlez, Jean‐Philippe

doi:10.1109/ojsp.2022.3229618

Cited by 4 publications

(5 citation statements)

References 74 publications

(108 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A deep-learning self-supervised algorithm developed explicitly for SAR anomaly detection [8] is used to test the proposed RX self-supervision method. The standard L 1 loss function is replaced with the newly defined loss.…”

Section: Proposed Methodsmentioning

confidence: 99%

“…An ablation study is realized to test the apport of the new loss. The architecture is the same as the one described in [8]…”

Section: Methodsmentioning

confidence: 99%

“…One of the most used algorithms for anomaly detection is the autoencoder or his modified version, the adversarial autoencoder (AAE) [5]. Because SAR data is corrupted by a strong noise called speckle [6], the use of these algorithms is inefficient and a despeckling pre-processing step has to be applied [7,8]. Furthermore, the goal of an autoencoder is to reconstruct an image as best as possible, which is not the goal of anomaly detection.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Self-Supervised SAR Anomaly Detection Guided with RX Detector

Muzeau,

Ren,

Angelliaume

et al. 2023

IGARSS 2023 - 2023 IEEE International Geoscience and Remote Sensing Symposium

View full text Add to dashboard Cite

HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L'archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d'enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

show abstract

Section: Proposed Methodsmentioning

confidence: 99%

“…An ablation study is realized to test the apport of the new loss. The architecture is the same as the one described in [8]…”

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Self-Supervised SAR Anomaly Detection Guided with RX Detector

Muzeau,

Ren,

Angelliaume

et al. 2023

IGARSS 2023 - 2023 IEEE International Geoscience and Remote Sensing Symposium

View full text Add to dashboard Cite

show abstract

“…Letting these kernels be trainable could enhance the performance of classification and segmentation. Furthermore, most despeckling techniques such as the one presented on Reference [49] can not be used as it only works for real-valued images.…”

Section: Coherency Vs Paulimentioning

confidence: 99%

Impact of PolSAR Pre-Processing and Balancing Methods on Complex-Valued Neural Networks Segmentation Tasks

Barrachina

Ren

Morisseau

et al. 2023

IEEE Open J. Signal Process.

View full text Add to dashboard Cite

In this paper, we investigated the semantic segmentation of Polarimetric Synthetic Aperture Radar (PolSAR) using Complex-Valued Neural Network (CVNN). Although the coherency matrix is more widely used as the input of CVNN, the Pauli vector has recently been shown to be a valid alternative. We exhaustively compare both methods for six model architectures, three complex-valued, and their respective real-equivalent models. We are comparing, therefore, not only the input representation impact but also the complex-against the real-valued models. We then argue that the dataset splitting produces a high correlation between training and validation sets, saturating the task and thus achieving very high performance. We, therefore, use a different data pre-processing technique designed to reduce this effect and reproduce the results with the same configurations as before (input representation and model architectures). After seeing that the performance per class is highly different according to class occurrences, we propose two methods for reducing this gap and performing the results for all input representations, models, and dataset preprocessing.

show abstract

“…In this study, I showed how amplitude data-derived scattering mechanisms can aid in sinkhole-related anomaly detection. Auto-encoder based methods, such as Adversarial Auto Encoder (AAE), have been used for anomaly detection in amplitude image [128]. I recommend that phase-related anomaly detection methods used in Chapter 5 should be combined with self-supervised anomaly detection meth-…”

Section: • Further Use Of Amplitude Data In Anomaly Detectionmentioning

confidence: 99%

InSAR Time Series Analysis for Sinkhole Detection using Deep Learning

Kulshrestha

View full text Add to dashboard Cite

Footprints on the sands of time.' This Ph.D. thesis is a mark that I lay in this incredible and diverse field of SAR remote sensing, about which I am extremely passionate. I saw and followed the footsteps of many people, got inspired, gained knowledge, found wisdom, and finally became capable of leaving my mark on 'the sands of time'. Somehow, I feel incredibly connected to the grayscale, speckled SAR amplitude images and continuous rainbow-like interferogram images. I would like to thank all scientists, starting from Dr. Carl Wiley to all the ones active today who have contributed towards the development of this field.I want to thank the promoter of my thesis, Prof. dr. ir Alfred Stein. His guidance has helped me to stay on track and finish my tasks on time. He helped me in the preparation of my papers, in their submission and rebuttal process. It was an honor for me to be his student and to be mentored by him. He showed me incredible insight and new perspectives on research and on life in general. Next, I want to thank Dr. Ling Chang, my co-supervisor. She was the one who showed faith in me by selecting me for this Ph.D. project. Starting from our first meeting to the last, she kept opening new gateways of knowledge in front of me during each encounter. She supported and acknowledged me when things worked, but also supported me when they did not. I admire her humility and patience throughout my Ph.D. track. I would also like to thank Prof. dr. ir. Ramon F. Hanssen, Dr. Freek van Leijen, Dr. Valentyn Tolpekin and Dr. Wietske Bijker for mentoring me during my Ph.D. journey.I want to thank my parents who kept nurturing me and supporting me during my PhD journey. Their love and affection never let me feel lonely. They carried me through a few difficult times but also rejoiced with me during my successes. My mother kept instilling positivity in my outlook. My father ensured that I took care of everything around me. Together, they helped me ensure that I stayed healthy and happy. My brother Abhishek made sure that my sense of humor stayed alive and well-polished. My fiancé, Reshma Jeswani, who kept on supporting me with her love, kindness and affection. She always helped me realize my potentials and never gave up on me. In addition, my extended family members including, Ratna Shil, Divyadarix shan Shil, Asha Shil, Anjali Kulshrestha, Siddhi Shil, and Shreya Shil, showed pride in curiosity in my Ph.D. project. This motivated me to be proud of my work and also in helping me to improve scientific communication to the general public.Next, I want to thank my friends, who became my family, and stuck with me through all the good and bad times. I got an incredible amount of satisfaction from cooking for them and seeing them enjoying it with me. I want to thank Abhisek Maiti, who always responded to my requests by 'chal', meaning -let's go do it. Michael Marshall, Mariana Belgiu and Ville Lehtola were incredible in supporting me in every way possible, both professionally and personally. Yolla Asmar's cheerful smile and helpfu...

show abstract

Self-Supervised Learning Based Anomaly Detection in Synthetic Aperture Radar Imaging

Cited by 4 publications

References 74 publications

Self-Supervised SAR Anomaly Detection Guided with RX Detector

Self-Supervised SAR Anomaly Detection Guided with RX Detector

Impact of PolSAR Pre-Processing and Balancing Methods on Complex-Valued Neural Networks Segmentation Tasks

InSAR Time Series Analysis for Sinkhole Detection using Deep Learning

Contact Info

Product

Resources

About