Date: Tuesday, March, 10th, 2015; 12:30 – 13:30
Place: Techno-Z, Z_GIS-meeting room: SC30EG0.E07 (Schillerstraße 30, building 15, ground floor)
Presenter: Emilia Cojoc (PhD candidate, Department of Systems Ecology and Sustainable Development, University of Bucharest)
Title: Spatial connectivity effects on mountain grassland biodiversity and carbon storage
The effects of spatial connectivity on biogeochemical processes and biodiversity dynamics is relatively poorly understood. We address the influence of bacterial community and plant species diversity on carbon storage and primary productivity for different types of boundary conditions. We quantify total carbon in soil and vegetation in semi-natural mountain grassland connected to forested landscape in the Carpathian Mountains. We also evaluate plant species diversity, plant biomass and quantify cellulolytic bacteria in soil. This allows us to test the hypothesis that higher spatial connectivity enhances plant diversity and consequently carbon storage and primary productivity.
Presenter: Ovidiu CSILLIK (PhD candidate – 1st year, West University of Timisoara, Romania, supervisor: Dr. Lucian Dragut), CEEPUS stay (1st March – 30th April 2015)
Title: Improving image segmentation results with automated refinement of objects
Increasing spatial resolution of remote sensing imagery has led to challenges of extracting meaningful information thereof. While methods for objective parameterization of image segmentation have recently been proposed, over- and under-segmentation still remains an issue for parts of the target image objects, therefore identification and further refinements of these objects are necessary. We implemented an automated tool for refinement of image objects within the eCognition software, that takes into account and reduce the effects of the main errors of image segmentation: over- and under-segmentation. Promising results were obtained by applying the tool on various datasets (WorldView-2, QuickBird and DSM). The tool proposed has the potential to improve the quality of segmentation as well as the classification accuracy and it can be successfully applied as a continuation of ESP2 tool, for refining the automated extracted objects.