Hershey, Pennsylvania
New York, New YorkBeijing, China
Deep Learning Solutions for Analysis of Synthetic Aperture Radar Imageries

Deep Learning Solutions for Analysis of Synthetic Aperture Radar Imageries

Nimrabanu Memon (Pandit Deendayal Energy University, India), Samir B. Patel (Pandit Deendayal Energy University, India), and Dhruvesh P. Patel (Pandit Deendayal Energy University, India)
Copyright: © 2022 |Pages: 23
DOI: 10.4018/978-1-6684-3981-4.ch008
OnDemand:
(Individual Chapters)
Available
$37.50
Current Special Offers
No Current Special Offers
TOTAL SAVINGS: $37.50

Abstract

The potential of Synthetic Aperture Radar (SAR) to detect surface and subsurface characteristics of land, sea, and ice using polarimetric information has long piqued the interest of scientists and researchers. Traditional strategies include employing polarimetric information to simplify and classify SAR images for various earth observation applications. Deep learning (DL) uses advanced machine learning algorithms to increase information extraction from SAR datasets about the land surface, as well as segment and classify the dataset for applications. The chapter highlights several problems, as well as what and how DL can be utilized to solve them. Currently, improvements in SAR data analysis have focused on the use of DL in a range of current research areas, such as data fusion, transfer learning, picture classification, automatic target recognition, data augmentation, speckle reduction, change detection, and feature extraction. The study presents a small case study on CNN for land use land cover classification using SAR data.
Chapter Preview
Top

Introduction

One of the best examples of an active microwave remote sensing system is Synthetic Aperture Radar (SAR), which includes radiation with 1mm longer wavelengths. Another benefit is the longer wavelength. Objects ten times smaller in size than the microwave wavelength are almost transparent in this region, allowing it to penetrate clouds and operate in all weather conditions.

It has revolutionized many earth observation areas such as agriculture, hydrology, and the study of surface deformation, urban planning, marine applications, space technology, and others. Table 1 shows the applications of SAR at various wavelengths. The ability of the microwaves to penetrate through clouds, atmospheric constituents, and even subsurface penetration has made this technology irreplaceable. SAR sensors in space have improved their ability to discern individual radar characteristics on the earth's surface. In contrast, aerial sensors can create very high-resolution images.

Table 1.
SAR wavelengths and their applications (Podest, 2018)
FrequencyWavelengthApplication
VHF300 kHz – 300 MHzBiomass, Foliage, and Ground penetration
P0.3 GHz – 1 GHzSoil moisture and Biomass
L1 GHz – 2GHzAgriculture sector, Forestry, and Soil moisture
C4 GHz-8 GHzOcean, and agriculture
X8 GHz – 12 GHzAgriculture, ocean, high-resolution radar
Ku14GHzZ – 18 GHzGlaciological applications, snow cover mapping
Ka27GHzZ – 47 GHzHigh-resolution radars

However, it is not easy to deal with SAR images. SAR images are more than two-dimensional images, like in the case of optical data. They are associated with many distortions, and grainy textured speckles affect their quality.

Top

Challenges With Sar Imageries

Radar backscattering response strongly depends on the orientation of targets. Due to slant range geometry, SAR imageries foreshortening, layover, and shadow effects dominate undulating or hilly terrain. These effects limit the interpretation of SAR data.

Complete Chapter List

Search this Book:
Reset