User Tools

Site Tools



Remote Sensing

Regular course within the MSc Physical Geography at Marburg University.

Course Description

Basically all field methods bear a tradeoff between grain and extend regardless if the tradeoff is primarily caused by man power, finances, logistics or any combination: one either can study selected processes in detail on a very limited number of observations sites or focus on a more landscape oriented survey using more generalized observation parameters. For a good proportion of such studies, this tradeoff can be overcome with the utilization of reliable but cost-effective, spatially explicit but high-resolution remote sensing. As the only spatially explicit measurement technique, remote sensing offers an amazingly toolbox which - in a modern sense - has been explored for more than 150 years but which is far from fully worn. The advent of high performance computing along with machine learning algorithms has just opened up a new chapter in data mining and information retrieval using a wide variety of ground-based, airborne and satellite sensors.

In this course, we will focus on optical and - to a very limited degree - LiDAR remote sensing. Specialized topics including very high resolution and 3D landscape assessments using UAVs, time series analysis and all sorts of up- and downscaling will be covered in other modules.

The individual sessions can be grouped into four sections:

  • LiDAR: in sessions 1 we will have a very quick look on LiDAR data. While 3D information retrieval will become a topic in our UAV-based modules, this session will primarily act as a service session to the Geographical Information System module which will use this data for terrain analysis.
  • Land cover exploration and classification: after a general introduction to optical remote sensing in session 2, we will focus on handling aerial images for visual and semi-automated analysis. Although such images do by far not represent the state of the art in remote sensing systems, it still represents the state of the practice in many areas.
  • Land cover classification from a satellite perspective: in sessions 9 to 12 we jump to satellite based sensors, namely the Sentinel-2 system in order to compare the performance of the latest generation of freely and worldwide available remote sensing systems.

In this module, we will primarily use QGis with the Orfeo Toolbox extension, SNAP and R.

Have fun!


The course has 1 session per week, 3 hours per session.

Session Topic Content
1 Information retrieval using LiDAR observations Introduction to LiDAR remote sensing, conversion of point to raster layers
2 Characteristics of optical sensor data Introduction to optical remote sensing, sensor signal composition, data I/O
Working with very high resolution aerial orthophotos
3 Visualization and equalization of sensor data Digital numbers, grey values, contrast stretch, composites, relative matching
4 Computation of artificial images Data space conversion, dimension reduction, filtering
5 Image classification Training site selection, classification methods, land-cover classification
6 Image classification Machine learning, accuracy assessment
7 Image classification with R Some libraries for R
- T-5 and holding Build-in hold to finish up the classification sessions
8 Image classification revisited Problems, pitfalls, strategies
9 Image classification revisited Feature selection and model training
10 Image classification revisited Segments and objects
11 Image classification revisited Finish
12 Wrap up Feedback and goodbye
courses/msc/msc-phygeo-remote-sensing/description.txt · Last modified: 2017/01/12 20:43 by tnauss