bahboubacar99/Coastal_Plastic_waste_detection_Sentinel2

In this project, we will detect plastic waste on the coastline of Jamestown in Accra, Ghana, using Sentinel-2 images. The S-2 images used in this project have been corrected for surface reflectance using the ACOLITE atmospheric correction model. In addition to the S-2 bands, we have spectral indices as well as bands obtained by spectral unmixing. # Coastal Plastic waste detection by Sentinel-2 In this project, we will detect plastic waste on the coastline of Jamestown in Accra, Ghana, using Sentinel-2 images. The S-2 images used in this project have been corrected for surface reflectance using the ACOLITE atmospheric correction model. In addition to the S-2 bands, we have spectral indices as well as bands obtained by spectral unmixing with the reference reflectances of our classes of interest. Roughly twenty variables were used in this project to train a Random Forest model, namely: coastal, blue, green, red, redE1, redE2, redE3, NIR, NIR_A, MIR_1, MIR_2, NDVI (Normalised Difference Vegetation Index), KNDVI (Kernel Normalised Difference Vegetation Index), PI (Plastic Index), NIR_prime, FDI (Floating Debris Index), plastics_bottles_abundance, fish_net_abundance, blue_bags_abundance, water_abundance, HDPE_abundance, wooden_abundance. # Database construction In this project, we collected training and validation data for our classes of interest across five different sites where we know the location of plastic waste. Polygons were digitised at the location of our classes of interest at these five sites to extract the values of our variables of interest in order to build our database. # Prerequisities The codes in this notebook are written in R, so you need to install the R kernel and the libraries required to run the notebook. ``` bash conda install -c conda-forge r-base ``` ``` bash conda install -c conda-forge \ r-ggplot2 \ r-tidyr \ r-lattice \ r-caret \ r-raster \ r-rastervis \ r-randomforest \ r-iml \ r-xgboost \ r-kernelshap \ r-shapviz \ r-ranger \ r-mlmetrics \ r-proc \ r-dplyr \ r-data.table \ r-shapr \ r-jpeg ``` # Clone projet You can clone this project with the following line of code ``` git lfs install ``` ``` git clone github.com ``` **Before executing, change the absolute path of dir_base replace it by the obsolute path to the i …