DIC04-P01: Digital Image Classification Practice

Practical: Exercise Information

The main objective of the practical is go through the Digital Image Classification procedure. That consists on image uploading and pre-processing, training and clustering, image classification and assessment.

The technique of classification in QGIS is based on the material taught in the lectures but differs in the sampling technique. It is based on the application of a clustering distance from a center in the feature spaces and an "on-the-fly" classification allowing the user to interact quickly with the system to get a better result.

Moreover it works with the concept of "macroclasses" and sub-classes of these macro's. I.e. A macroclass would be "Vegetation" and "Subclasses" could be grass, wheat, maize, etc... This allows better and easier grouping of classes when the subclasses can't be identified by spectrum.

Learning objectives

After this exercise you are able to:

  • Use the spectral signatures of pure objects to recognize surface elements
  • Develop a training set for classification.
  • Make a simple supervised classification of a Sentinel-2 image using the Semi-Automatic classification plugin SCP of QGIS

Resources

Background Information

Prior to the execution of this exercise, the student is requested to attend the 4 short lectures in Digital Image Classification and the corresponding quizzes.

Video of the practical

Support material

  • See the printed practical guide in Digital image classification.

Equipment

  • PC with QGIS 3.4 and excel software (optional)

Data for the exercise

DOWNLOADS:

Products

Digital Classified Image in QGIS.

Assessment of the classification and feedback

Due to limited time, the assessment will be done visually and using the data collected in the field. This part will be guided in class