UAVs and airplanes currently provide imagery with very high/super high spatial resolution. These data sources are increasingly being utilized for detailed vegetation studies in nature conservation, agriculture, forestry and other research areas. Pre-processing and analyzing the data is challenging at this spatial level, especially when dealing with change detection analysis. It involves addressing multiple sources of errors, noise, and inaccuracies. In relation to recent studies conducted by the TILPEC research team, which focus on the detection, change detection, and health status evaluation of primarily natural but also cultivated vegetation, the proposed PhD project should aim to improve vegetation classification/change detection accuracy by testing various machine learning/deep learning methods. The methods will be tested based on case studies for different habitats/types of vegetation – peat-bogs of relict arctic-alpine tundra, grasslands, meadows with invasive species or others.
Various machine learning/deep learning approaches should be tested and compared to achieve very high detection/change detection accuracy (over 90 %) for individual species or even individuals of selected species. The PhD project leads to the proposal of a final, highly accurate processing chain, taking into account various factors such as the quantity/spatial distribution of training/validation data, variable illumination during data acquisition, influence of terrain, number of species within the habitat, species composition and density/abundance etc.
Applicant should have advanced knowledge in remote sensing and should be ready to work in the field. Experience with machine learning and deep learning methods and publications demonstrating this experience are an advantage and will be considered in the selection process.
Deadline is closed
