mgmanalili/WFP_UNIDO

A WFP-UNIDO collaborative project on Acacia mapping for Namibia --- ## World Food Programme - UNIDO project --- ### Application of Machine Learning and geospatial techniques for individual Acacia tree species mapping #### The project aims to develop a novel approach of automating the detection of Acacia species from very high resolution satellite imagery. The system would then do image segmentation to identify individual Acacia tree crowns. A correlation betweek crown area and volume was then applied to infer volume of Acacia over a large area. The script is written in pure R using geospatial libraries to automate the task. ### Detailed documentation will be provided soon. This is a work in progress --- #### Main Developer: Michael Manalili (WFP) #### Remote Sensing Specialist: Sirio Modugno, PhD (WFP) #### Project Manager: Ashraf Abushady (UNIDO) --- #### Changelog Error: Kappa metric values are missing Issue: Data error (raster input) Solution: Please Check the Satellite Image or the Drone image and make sure it is in Multi-Color 3 Bands only and not in 4 Bands! If you are processing your image under ESRI ArcGIS, the software would automatically add a 4th Band called “Alpha” which improves the image for visualization. However, the script is designed to calculate the Kappa metric values for 3 Bands only as that is the common standard in GIS. --- Implemented parallel processing Issue: Since the task task a while to process, this version adds parallel processing (raster and R processing) to deliver results raster Solution: A library called parallel was implemented for module 02(Data Processing) and 03(Tree Segmentation) in order to speed up the processing. In this way, we may not need to clip the rasters by chunks. If you have a dedicated machine to process only this task, you can use ncores-1 to run moodules 2 and 3. This speeds up the processes by more than 75%. ---