Session 2 Clustering_GEE_SOLVED

The link to the solved exercise can be found here: https://code.earthengine.google.com/e99d8135f7976608ed3cc73aabedcddb

///////
//+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
//S. Aparício (Solenix), I. Alonso (Solenix)  
//
//MIT LICENSE CC BY EUMETSAT
//+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
//❄️☀️
// Session 2 “Hands-on: Clustring two image of different seasons 
// for change detection"
// PARTICIPANT'S VERSION
//
//
//+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
//      EXERCICE 2: CLUSTER A S2 CLOUD FREE IMAGE OF ENDALEN (SVALBARD) 
//  FROM TWO DIFFEREN SEASONS AND APPLY A CHANGE DETECTION BETWEEN THEM        
//+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+


/////////////////////////////////////////////////////////////////////////////
//////////////////////   1. DEFINED AND VISUALIZE AOI           /////////////
/////////////////////////////////////////////////////////////////////////////


//Create a geometry over Endalen, in Svalbard - and name it AOI and visualize it


//Center the visualization there
Map.centerObject(AOI,10);

//////////////////////////////////////////////////////////////////////////////
//////////////////////    2. PREPARING SATELLITE IMAGERY     /////////////////
//////////////////////////////////////////////////////////////////////////////

//Define a "summer range date" with ee.Date could be from 2022-06-01 to 2022-08-28 > e.g. ee.Date('2022-01-01')
//e.g. ee.Date('2022-01-01')

var start = ee.Date('2022-06-01');
var end = ee.Date('2022-08-28');

//Define a "winter range date" could be from 2022-05-30 to 2022-07-02 
var start2 = ee.Date('2022-05-30');
var end2 = ee.Date('2022-07-02');

//Create a summer collection
var summerCol = ee.ImageCollection('COPERNICUS/S2')
                    .filterDate(start,end)
                    .filterBounds(AOI)
                    .filter(ee.Filter.lt('CLOUDY_PIXEL_PERCENTAGE',10))

//Create a winter collection
var winterCol = ee.ImageCollection('COPERNICUS/S2')
                    .filterDate(start2,end2)
                    .filterBounds(AOI)
                    .filter(ee.Filter.lt('CLOUDY_PIXEL_PERCENTAGE',5))

var summerImage = summerCol.first()
var winterImage = winterCol.first()

var visRGB = {
  min: 1400,
  max: 9500,
  gamma: 2.24,
  bands: ['B4', 'B3', 'B2']
};

Map.addLayer(summerImage.clip(AOI), newVis, 'Sentinel-2 Summer Image')
Map.addLayer(winterImage.clip(AOI), visRGB, 'Sentinel-2 Winter Image')

Export.image.toDrive(summerImage)
//////////////////////////////////////////////////////////////////////////////
////////////    3. Create image of absolute differences        ///////////////
////////////////////////////////////////////////////////////////////////////// 


var changeCover = summerImage.subtract(winterImage).clip(AOI)
Map.addLayer(changeCover.clip(AOI), {}, 'Reflectance Abs. differences Image')


//////////////////////////////////////////////////////////////////////////////
//////////////    4. Collecting training data        /////////////////////////
//////////////////////////////////////////////////////////////////////////////


//During the supervised classification we started at this point to
//collect data about the differnt clases.
//Here we won't be doing it manually but we will create a function
//to give as 100 random points in the area.
//Ideally should be more than 100 points, but this is just an example to speed things up


//This sample function will return a FeatureCollection with 100 points
//containing the spectral value of the image they are located on


var training_data = changeCover.sample({
                  region: AOI,
                  scale:20,
                  numPixels: 100
})

//////////////////////////////////////////////////////////////////////////////
//////////////    4. Create a cluster and train it   /////////////////////////
//////////////////////////////////////////////////////////////////////////////


// The next stop is to create a clusterer object from one of the different
//algorithms available:
//var clusterer = ee.Clusterer.wekaXMeans()
var clusterKmeans = ee.Clusterer.wekaKMeans(3);

// This is still need to be trained. This is done by
//applying the train() function on the clusterer and passing in the
//training data

var trainedKmeans = clusterKmeans.train(training_data);


//This cluster has now looked at all this points, and group them by similiarity
// while these groups are as distinc as possible from other groups of points


//When we apply this cluster to our image, it will take all pixels of the image
//and determine, to which of these groups the pixel is most similar



var clusteredImage = changeCover.cluster(trainedKmeans)


//////////////////////////////////////////////////////////////////////////////
//////////////    5. Visualize results              /////////////////////////
//////////////////////////////////////////////////////////////////////////////


var visCluster = {
  min: 0,
  max: 1,
  palette: ['white', 'red']
};

Map.addLayer(clusteredImage.clip(AOI), visCluster, 'Clustered Change Image')

///////////////////////////////////////////////////////////////////////
/////////////// 6. EXTRA: EXPORTING WITH NICE VISUALIZATIONS///////////
//////////////////////////////////////////////////////////////////////////////

var finalImage = clusteredImage.clip(AOI)

var visToExport = clusteredImage.visualize({
  min:0, 
  max:1,
  palette: ['white', 'red']
})

Export.image.toDrive({
  image: visToExport, 
  description: 'Exporting_10m_resolution'
});