Remote Sensing

Guide to AI and Data Courses · Part III: Signals, Images and Audio
Satellites and sensorsSpectral bandsCorrectionsClassificationChange detectionRelated courses
Sercan KÜLCÜ · Giresun University · Computer Engineering

Introduction

What Question Does It Ask?

What can satellite and aerial images tell us about the Earth?

Remote sensing extracts information about the Earth without touching it, by measuring reflected and emitted energy with sensors on satellites, aircraft or drones. An image is not just a color photo but multiband data, where each band carries a different physical measurement.

At a glance

Approach

physical sensing and image analysis

Input

multispectral satellite imagery

Output

land cover map, change

Success measure

map accuracy, kappa

Math load

medium

Data needs

plenty; labels limited

Introduction

A Short History

1858

Nadar photographs Paris from a balloon: the first known aerial photograph.

1960

The first weather satellite, TIROS-1, is launched; clouds are imaged routinely from space.

1972

Landsat 1 is launched: the first civilian program to monitor the Earth in multiple bands.

1978

Seasat becomes the first civilian satellite to collect synthetic aperture radar (SAR) images from space.

2008

The US Geological Survey (USGS) opens the Landsat archive to everyone for free.

2015

Sentinel-2A is launched; Copernicus open data starts an era of frequent, free imagery.

The Course

Core Concepts

Electromagnetic spectrum

Visible, infrared, thermal and microwave bands.

Passive and active

Optical sensors measure sunlight; radar sends its own signal.

Types of resolution

Spatial, spectral, temporal and radiometric.

Spectral signature

Water, plants and soil reflect differently in each band.

Band indices

NDVI for vegetation, NDWI for water.

Corrections

Atmospheric, terrain and geometric correction.

Land classification

Labeling each pixel as forest, field, urban or water.

Change detection

Comparing images from different dates.

The Course

A Typical Workflow

Sensesatellite,aircraft, UAVCorrectatmosphere,geometryComputebandsNDVI,compositeClassifyland cover,objectsMapGIS, changereport
Example: Sentinel-2 images taken before and after a forest fire are corrected for the atmosphere. A burn index is computed from the near-infrared and shortwave infrared bands; the difference between the two dates reveals the burned area. The result is measured in hectares and drawn on a map.

The Course

What Does the Course Cover?

  1. Electromagnetic radiation and surface interaction
  2. Sensors, platforms and types of resolution
  3. Radiometric and atmospheric correction
  4. Geometric correction, coordinate systems, projections
  5. Image enhancement and colour composites
  6. Spectral indices: NDVI, NDWI
  7. Supervised and unsupervised land-cover classification
  8. Accuracy assessment and ground truth
  9. Change detection and time-series analysis
  10. Thermal sensing and LiDAR
  11. Radar (SAR) imagery and interferometry
  12. Deep learning for satellite imagery

Skills you will gain

raster and vector layers in QGIStime series in Google Earth Engineprocessing imagery with GDAL and rasterioland-cover maps with accuracy tablesinterpreting indices such as NDVI

A typical semester; the order and coverage vary between universities.

Try It

Measuring Forest Loss from Space

Two satellite images of the same area, from two dates, have red and near-infrared (NIR) bands. Find vegetation with NDVI, slide the threshold and measure how many hectares of forest were lost; click a pixel to see its value.

Remote sensing turns bands the eye cannot see into a physical indicator. The course teaches band ratios, masking disturbances such as clouds, and measuring change as area.

Context

Prerequisites

Courses worth knowing before this one, and why they are needed.

CourseWhy it is needed
Image Processingmultispectral images, correction
Computer Visionland cover classification, change detection
Physicselectromagnetic waves, reflection and absorption
Geographic information systemscoordinate systems, map projections and layers
Programminglarge array and file operations in Python

Context

Are You Ready for This Course?

Ask yourself; for every "no", the right-hand column shows where to start.

Ask yourselfIf not, start with
Can you explain the visible and infrared parts of the electromagnetic spectrum?Electromagnetic waves from high-school physics
Can you explain latitude/longitude versus projected (metre) coordinates?A short introductory GIS tutorial with QGIS
Can you apply a histogram and thresholding to an image?Image Processing
Can you train a classifier and interpret its confusion matrix?Machine Learning
If you can say "yes" to most of these, you are ready; one or two gaps can be closed during the semester.

Context

Neighbouring Courses

subfieldbuilds onshares methodsapplication

Overlap

Overlapping Courses

RelationCourseShared ground
applicationThis course applies the methods of Image Processingmultispectral images, correction
applicationThis course applies the methods of Computer Visionland cover classification, change detection
shares methodsShares methods with Time Series Analysis and Forecastingmultitemporal image series

Differences

Where It Differs (1/2)

CriterionRemote SensingImage ProcessingComputer Vision
Core questionWhat can satellite and aerial images tell us about the Earth?How is an image enhanced, transformed and divided into regions?How does a computer understand what is in an image or a video?
Approachphysical sensing and image analysispixel-level transformationsextracting meaning from images
Inputmultispectral satellite imageryimageimage or video
Outputland cover map, changeprocessed image or regionslabel, box, depth, 3D structure
Success measuremap accuracy, kappavisual quality, PSNRaccuracy, mAP, IoU
Math loadmediummedium: convolution, Fourierhigh: geometry, linear algebra
Data needsplenty; labels limitedlittle; learning not requiredmany labeled images
  • Image Processing: Image processing provides the general methods; remote sensing applies them to multiband satellite images that carry physical measurements.
  • Computer Vision: Vision recognizes objects in ordinary photos; remote sensing maps the Earth using spectral signatures and location.

Differences

Where It Differs (2/2)

CriterionRemote SensingTime SeriesMedical Imaging
Core questionWhat can satellite and aerial images tell us about the Earth?How do we find structure in measurements over time and forecast the future?How do X-ray, CT and MRI images help diagnosis and treatment?
Approachphysical sensing and image analysistime-dependent statistical modelsimage processing plus clinical knowledge
Inputmultispectral satellite imagerytimestamped measurementsCT, MRI, X-ray, ultrasound
Outputland cover map, changeforecast, trend, seasonalitysegmentation, measurement, diagnostic suggestion
Success measuremap accuracy, kappaforecast error (MAE, RMSE)sensitivity, specificity, Dice
Math loadmediummedium-highmedium-high
Data needsplenty; labels limitedlong historylittle; labels are expensive
  • Time Series: Time series analysis provides the general methods; remote sensing tracks change in images of the same place over years.
  • Medical Imaging: Both treat an image as a physical measurement; medical imaging measures the body, remote sensing the Earth.

Practice

Applications

Agriculture

Crop health, yield prediction, irrigation.

Forestry

Burned areas, deforestation monitoring.

Disaster management

Damage maps after earthquakes and floods.

Urban planning

Construction and urban sprawl.

Climate and environment

Monitoring glaciers, sea surface and drought.

Defense

Reconnaissance and target area analysis.

Practice

Risks and Ethics

Surveillance and privacy

High-resolution satellite and drone imagery can track people's homes, vehicles and movements without consent.

Maps without ground truth

A classification map never checked in the field can mislead farming or disaster decisions despite high reported accuracy.

Clouds and shadows

Unmasked clouds and shadows look like forest loss or flooding in change detection.

Models that do not transfer

A model trained on European data fails silently in other climates and farming patterns; each region needs validation.

In depth: Explainable AI and AI Ethics.

Practice

Tools and Datasets

Tools and libraries

Google Earth EngineQGISGDALrasterioESA SNAPTorchGeo

Well-known datasets

Landsat archiveSentinel-1 and Sentinel-2EuroSATBigEarthNetSpaceNetUC Merced Land Use
Satellite imagery is plentiful and free, but accurate labels are scarce: ground truth means going into the field.

Practice

Common Misconceptions

Common beliefIn fact
A satellite image is an ordinary color photo.It also has infrared, thermal and radar bands that the eye cannot see.
Higher spatial resolution is always better.For wide areas and frequent revisits, medium resolution is often more suitable.
If there are clouds, the satellite sees nothing.Optical sensors cannot; radar (SAR) sees through clouds and at night.
This is just an image processing course.Sensor physics, the atmosphere, map projections and field validation are also part of it.

Next Steps

After This Course

Courses that build on or extend this one

This course is mostly an end point: it builds on the others.

Related careers

Remote sensing specialistGIS specialistAgricultural technology engineerEnvironmental and climate data analystGeomatics engineer
Remote sensing is image processing and computer vision applied to the Earth; it joins with time series analysis and big data systems. From here you can move on to climate and agricultural data science, geographic information systems, or satellite image analysis with deep learning.

Next Steps

Further Reading

  • Lillesand, Kiefer and Chipman, Remote Sensing and Image Interpretation, 7th ed., Wiley, 2015.
  • Jensen, Remote Sensing of the Environment: An Earth Resource Perspective, 2nd ed., Pearson, 2007.
  • Campbell and Wynne, Introduction to Remote Sensing, 5th ed., Guilford Press, 2011.
  • Richards, Remote Sensing Digital Image Analysis, 5th ed., Springer, 2013.

The End

Chapter 21: Remote Sensing
← Previous chapter · Map of the field · Next chapter →
1 / 1 Sercan KÜLCÜ. All rights reserved.