Spaceborne sensors operating in the microwave spectrum including Synthetic Aperture Radars (SAR), Scatterometers, and Radiometers have been used for derivation of various geophysical parameters in a wide range of applications such as soil moisture change-detection, water bodies mapping, floods and droughts monitoring, snow cover mapping, freeze/thaw detection, vegetation phenology, crop classification, and urban areas identification. Microwave remote sensing, in range of 1 cm to 1 m wavelength, in both active and passive forms, have special properties because of low frequency of microwaves compared to visible and infrared. They can penetrate through clouds, rain, and dust with no or negligible attenuation as the longer wavelengths are not involved in atmospheric scattering. This unique property allows measuring the backscattering energy under almost all weather and environmental conditions. The microwave sensors at different frequencies are used for various land, ocean, and atmospheric applications. In this tutorial, the focus will be land applications of the active microwave sensors (radars): an overview from basics to applications, satellite data processing techniques and retrieval algorithms.

Overview

Teacher CV

Dr. Vahid Naeimi

Overview

This tutorial introduces SITS data and its categorization into multivariate and univariate SITS, which is a novel categorization of this data. In addition, we discuss the characteristics, limitation and the methods used to classify both categories of SITS data. Different representations of SITS data are introduced in this tutorial. In details the following topics will be discussed:

1- Theoretical background
    a. Definition of SITS
    b. Different types of SITS data and their difference
    c. SITS data application

2- Univariate SITS classification
    a. Advantage and disadvantage of univariate SITS data for classification
    b. Challenges of univariate SITS data classification
    c. Different features and representations of Univariate SITS data for classification
    d. Extracting information from trend of univariate SITS for classification

3- Multivariate SITS data classification
    a. Advantage and disadvantage of Multivariate SITS data for classification
    b. Challenges of Multivariate SITS data classification
    c. Different representations of multivariate SITS data for classification

The use of satellite images is becoming more and more common in the field of Earth Sciences: wide areas (thousands of sq miles) can nowadays be safely observed at once, revealing geophysical parameters in a totally new perspective. Microwave signals in particular can penetrate the atmosphere and acquire images from satellites day and night, providing unique archives of observations of the Earth surface. In this context, Synthetic Aperture Radar Interferometry (InSAR) has proved being an advanced technology capable of recording ground movements at unprecedented scales (inches over hundreds of miles). However, it is still little known that, by exploiting long series of SAR images, the precision of the InSAR technology can be pushed to detecting millimetric displacement of single objects. Thanks to advanced algorithms it is nowadays possible to monitor also infrastructures as bridges, roads, dams, railways. In this workshop InSAR concepts will be presented together with the most multi-temporal processing techniques. Several case studies will be shown, from wide areas subsidence monitoring up to the analysis of settlement of urban structures. Examples of ground movements caused by water extraction, seismic activities, landslides, water pressure (on dams), thermal expansion (on building/bridges) detected by SAR satellites will be discussed. The speaker will also briefly introduce the software he developed for analyzing InSAR data, SARPROZ (www.sarproz.com).

Overview

Teacher CV

https://engineering.purdue.edu/~perissin/

Overview

This tutorial covers a wide range of applications of kernel-based learning methods for different analyses of remote sensing data. In this tutorial, both the basic concepts of the algorithms and their applications will be discussed. Although, using the kernelbased classification algorithms such as support vector machines are very common, but the focus of this tutorial will be on the other applications of the kernel-based algorithms. In details the following topics will be discussed:

1- Theoretical Background
    a. Kernel function and data representation
    b. Kernel characteristics and different types of kernel functions

2- Kernel-based learning methods for remote sensing data classification and regression
    a. Kernel-based supervised classification (Support Vector machines and Support vector domain
       description)
    b. Kernel-based unsupervised classification
    c. Kernel-based regression (support vector regression)

3- Kernel-based learning methods for dimensionally reduction
    a. Kernel-based feature extraction (Kernel- principle component analysis)
    b. Kernel-based feature selection (Hilbert-Schmidt independence criteria)

4- Kernel-based image fusion
    a. Multiple kernel learning for data fusion

5- Practical issues using the kernel-based learning methods
    a. Normalization issues
    b. Role of parameters in the performances of the algorithms
    c. Optimization algorithms for kernel-based learning methods
    d. Introducing available toolboxes for kernel-bases analysis

It is noteworthy that some of these topics such as “kernel-based image fusion” and “kernel-based feature selection” are among the most recent fields in the kernel-based analysis in machine learning community which also have been used by remote sensing researchers.