Loading...
Remote sensing data analysis in R-Studio: A machine learning perspective
Brief Description

Given this era of the fourth industrial revolution (hereafter, referred to as 4IR) the technological developments we see around us in most industries (such as mining, agriculture, engineering and geospatial science) rely on big data acquisition, processing and analysis in-order to produce end-user products and automated systems. In particular the processing and analysis of big data often require the use or adoption of certain advanced algorithms such as artificial intelligence and machine learning. This remote sensing data analysis course cover the physical principles of satellite remote sensing and remote sensor data processing using machine learning methods. In particular, this includes: (i) examining the characteristics of spectral signatures across regions of the electromagnetic spectrum (EM) spectrum for various surface targets, (ii) criteria for interpretation of remote sensing images, (iii) image pro-processing in R-studio, (vi) supervised classification of images using machine learning methods in R-studio such as the Artificial neural network (ANN), Random forest (RF), Support vector machine (SVM) and Classification and regression trees (CART); followed by verification of classification results by constructing an error assessment matrix. Furthermore course will also cover the comparison between non-parametric linear (Stepwise multiple linear regression) and non-linear (K-NN) regression methods for the estimation or modelling of terrestrial biophysical and/ or biochemical properties.

Learning Outcomes
  • Know how to setup and use R-studio for satellite image data processing and analysis i.e. do basic programming
  • Carry out satellite image pre-processing (including atmospheric correction) in R-studio and assess image quality using image statistics in R-studio
  • Carry out supervised classification using different machine learning algorithms on a series of raster layers and to do validation of the classification results
  • Carry out linear and linear non-parametric methods to make estimations and/or predictions terrestrial biophysical variables like the Leaf area index (LAI), Chlorophyll and Fractional Vegetation Cover.
  • Carry out statistical analysis of error for various modeling scenarios
Course Content

Day

Topic (lecture)

Time-duration

Lab exercise

Time-duration

1.

Physical principles of remote sensing

3 hrs

Lab 1: Image quality assessment and pre-processing in R-studio

5 hrs

2.

Advanced supervised image classification methods (Machine learning)

2hrs

Lab 2: Supervised classification using machine learning methods (SVM, RF, ANN, CART) in R-studio

6 hrs

3.

Parametric vs non-parametric modelling for estimating terrestrial biophysical and/or biochemical properties

2 hrs

Lab 3: Empirical statistical models: Linear vs non-linear non-parametric models

6 hrs

Entry Requirements

The candidate must have:

a) basic statistics background and

) worked with at least one GIS or Remote sensing software.

Course Number:
P007854
Catalogue and Category:
Environmental Management and Geophysics
Who Should attend:
Natural science practitioners, GIS technologist, GIS professionals
Delivery Mode:
Contact Sessions
Contact Days:
3