This file contains a script to perform Remote Sensing image classification through Support Vector Machines (SVM) algorithm.
Open the script in the Google Earth Engine (GEE) code editor: https://code.earthengine.google.com/ or https://signup.earthengine.google.com/#!/
Clicking on RUN button the script will be executed by GEE.
What this script makes in some basic steps:
1- Defines the samples for each class. These classes are defined by the user, so modify the code in order to achieve realistic results.
2- Creates a point in a defined geographic place and so creates a buffer. The radius of the buffer must be defined by the user.
3- Selects Landsat 8 dataset of images and apply a median-based reduction. The user must select the start and end date for the images.
4- The selected images are clipped based on the buffered area created earlier.
5- A composite of bands is displayed on the screen.
6- The samples selected are transformed to a Feature Collection in order to feed SVM model.
7- The pixels values are collected, for each pixel/point, the information of each band used in the composition are retrieved.
8- Defines SVM parameters.
9- Trains the algorithm and classify all the buffered area. The result is displayed on the screen.
10- Exports the classified image to Google Drive.
In the repository i put the exported result. The file name is classification.tif.