/Drone_image_classification

Classification of drone imagery (RGB and Multispectral) using Random forest with superpixel postprocessing

Primary LanguageJupyter Notebook

Drone_image_classification

Landcover classification of drone UAV imagery (RGB and Multispectral) using Random forest with unsupervised segmentation (superpixel) postprocessing.

Description

In this repository I only store the code (Python). The data is taken from Hilden collection, which is accessible by contacting the Hilden Network.

The code is organized into 4 Jupyter notebooks and one Python script with functions:

  • 0._ - preprocessing of the images. Filling the gaps and resizing the images if needed.
  • 1._ - showing an example of landcover classification on one image
  • 2._ - classifying all the prepared in script 0. multispectral imagery
  • 3._ - classifying all the prepared in script 0. RGB imagery
  • myfunctions - all the custom functions I used

Visualization

If you want to quickly glance at how the analysis is done, just view the 1._script. It contains images and descriptions of each step.

Please feel free to contact me if you have any questions!