Ungraded Assignment
Design a project aiming in the computation of a comprehensive image (raster data stack) collection for species prediction, and heterogeneity mapping.
Skills and Competences
At the end of your ungraded assignment project you should be able to:
- process optical raster data to compile a comprehensive dataset for your area of interest
- compute artificial images which highlight spectral, structural or spatial properties using the cpabilities of R and more (e.g. OTB)
- design and implement medium complex workflows related to raster based image computations
- work with collaborative project and software development tools
- sufficiently document your code and analysis workflow for re-usability and scientific review
- compute land-cover classifications using R packages like caret to build a model for the prediction of the selected tree species classes
- discuss the principal logic and required steps of training a machine learning model,
- understand the importance of appropriate training site selection strategies,
- discuss strategies on which predictor variables to include in a classification model,
- communicate your work in a poster presentation.
Comprehensive dataset for Classification purposes
Please compile a comprehensive dataset based on spectral and structural information derived from both multi-channel aerial and LiDAR datasets which is suitable for the parallel and upcoming project work mentioned above. The dataset should provide information on:
- spectral properties
- structural properties
- spatial patterns
Tree species Classification
Please design and perform a classification workflow to identify the tree species in MOF on pixel resolution using a ML approach. You should provide:
- tree species classification for Marburg Open Forest
- training and validation strategy