Skip to Main content Skip to Navigation
New interface
Conference papers

SAR Anomalies Detection based on Deep Learning

Abstract : This paper proposes an anomaly detection method for SAR imagery based on deep learning. It does not require ground truth of anomalies, which addresses a recurrent problem in remote sensing: the lack of labeled data to train neural networks. The proposed model combines an adversarial autoencoder followed by a statistical change detector based on the covariance matrix. A despeckling step is first performed, which allows to filter the speckle noise and to significantly improve the detection performances.
Document type :
Conference papers
Complete list of metadata

https://hal-centralesupelec.archives-ouvertes.fr/hal-03790615
Contributor : Jean-Philippe Ovarlez Connect in order to contact the contributor
Submitted on : Wednesday, September 28, 2022 - 3:04:51 PM
Last modification on : Friday, October 28, 2022 - 3:18:56 PM

File

Max - Gretsi 2022.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-03790615, version 1

Citation

Max Muzeau, Chengfang Ren, Sebastien Angelliaume, Mihai Datcu, Jean-Philippe Ovarlez. SAR Anomalies Detection based on Deep Learning. XXVIIIème Colloque GRETSI 2022, Sep 2022, Nancy, France. ⟨hal-03790615⟩

Share

Metrics

Record views

10

Files downloads

5