Synthetic Aperture Radar (SAR) imaging systems have the ability to acquire images of the terrain surface in all weather conditions and all day times. Digital Elevation Model (DEM) can be generated from two or more SAR images, and is considered essential in various recent applications. Acquired SAR images are often exposed to speckle noise, which has a negatively bad effect on the processing and interpretation of the SAR images, and hence on the DEM generation process. In this paper, a Convolution Neural Network (CNN) based preprocessing layer is suggested in the DEM generation process from SAR images. The main purpose of the suggested CNN based preprocessing layer is removing speckle noise from input SAR images, from which an enhanced DEM can be generated. Extensive experiments are carried out on SAR images, and different DEMs are generated from original SAR images and from despeckled ones. Comparative analysis is figured out, and results show significant enhancements in despeckled SAR images and in the subsequent generated DEMs.
Digital Elevation Model Enhancement using CNN-Based Despeckled SAR Images
01.03.2020
1472164 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Digital Elevation Model Mosaic of Mercury
NTRS | 2001
|Research on Digital Elevation Model Using GNSS-IR Technology
British Library Conference Proceedings | 2021
|