open topic
Here are listed some possible themes for a master or bachelor thesis. Own proposals are welcome.
Detecting dynamic changes in vegetation patterns in areas of semi-natural Vegetation in southern Chile
The objective of the Master-Thesis is to classify a series of historical aereal photographes due to vegetation cover and to analyse temporal changes in vegetation patterns. These information should be used to identify areas of progressive degradation of forest to shrubland, areas of forest succession due to abandonment, and aereas with a dynamic conversion between shrubland and human-induced types of land cover, in particular pastures.
Preprocessing of optical satellite data for tropical forest monitoring
Description: Multispectral satellite images of tropical forests are often difficult in use due to frequent and heavy cloud coverage. You will produce a sequence of preprocessing steps for common and most recent satellites which provide best possible data for further interpretation. The processing includes techniques like cloud masking or radiometric and atmospheric correction. The study sites are located in the Democratic Republic of Congo and Fiji.
Classification of tropical forests based on multispectral satellite data
Description: The classification of tropical forests is an important basis to derive further information like biomass or carbon sequestration. You will develop automated or semi-automated methods using image processing software to classify forests on our study sites in the Democratic Republic of Congo and Fiji. Knowledge of tropical forests is therefore of advantage.
Development of a forest biomass model from combined hyperspectral and lidar remote sensing data
The mapping of forest biomass is of great interest for several reasons. Forest biomass plays an important role in the development of global Carbon-cycling models as well as for regional bioenergy production. In the proposed thesis the student will develop species-specific biomass models based on previously-existing biomass formulas. The remote sensing data will be used to derive height and density information of the forest stands as well as tree species classifications. This information serves as input to the abovementioned existing biomass formulas on single tree level and will be then extrapolated to stand level.
Using R for K-nearest neighbor classification of categorical forest variables
In pattern recognition, the k-nearest neighbors algorithm (k-NN) is a method for classifying objects based on closest training examples in the feature space. In k-NN, the function is only locally approximated and all computation is deferred until classification. An object (e.g. an environmental entity) is classified by a majority vote of its neighbors, with the object being assigned to the class most common amongst its k nearest neighbors.
Visiting professor Dr. John Weishampel in Freiburg
EARSeL - Workshop on Temporal Analysis of Satellite Images