Machine Learning untuk Analisis Land Use dan Land Cover

Materi:

• Klasifikasi untuk analisis Land Use dan Land Cover di QGIS
• Klasifikasi untuk analisis Land Use dan Land Cover dengan Google Earth Engine 

Software yang digunakan:

• QGIS
• Orfeo Toolbox
• Google Earth Engine

Langkah-langkah:

• Install QGIS

Sign in to Google Earth Engine

• Sign in to Google Earth Engine, klik sign up untuk pertama kali

• Setelah suskes sign up, selanjutnya lakukan sign in, dan coba masuk ke halaman Platform -> Code Editor

• sd

Tambahkan Orfeo Toolbox di QGIS

• Untuk membaca langkah-langkah download dan menambahkan Orfeo Toolbox, silahkan download ebook disini 
• Download Orfeo Toolbox, klik https://www.orfeo-toolbox.org/download/
• Unzip file hasil download di C, misal C:\OTB-7.0.0-Win64 dimana folder ini memuat Library Orfeo.
• Selanjutnya download folder lain yang memuat file configurasi Orfeo Plugin di QGIS. Download zip folder di https://gitlab.orfeo-toolbox.org/orfeotoolbox/qgis-otb-plugin dan unpack it.
• Buat folderdi drive C:\qgis-plugins\qgis-otb-plugIn and copy semua file hasil download diatas, sehingga seperti ini.

• Selanjutnya buka QGIS, lalu pilih menu Settings -> Options. Selanjutnya pilih menu Processing -> Providers lalu pilih OTB. Set OTB application folder dengan C:\OTB-7.4.0-Win64\lib\otb\applications dan OTB folder dengan C:\OTB-7.4.0-Win64 lalu klik OK.

• Maka Orfeo Toolbox siap digunakan di QGIS.

Analisis supervised dengan Orfeo Toolbox  di QGIS, tahapan yang akan dilakukan adalah:

• Melakukan analisis Unsupervise (K-means) dengan OTB di QGIS.
• Menggunakan Random Forest, SVM, dan Decission Tree di QGIS.
• Melakukan performasi assesment akurasi dan compute main accuracy measures.

Analysis di GEE:

• Import dan visualisasi Sentinel image di GEE

//import the satellite data from european space agency

var S2 = ee.ImageCollection("COPERNICUS/S2");

//filter for Dubai

S2 = S2.filterBounds(Dubai);

print(S2);

//filter date

S2 = S2.filterDate("2020-01-01","2020-05-11");

print(S2);

//visualize image

var image = ee.Image(S2.first());

Map.addLayer(image, {min:0, max:3000,bands:"B4,B3,B2"}, "DubaiTrue Color");

Map.addLayer(image, {min:0, max:3000,bands:"B8,B4,B3"}, "Dubai False Color");

• Import dan Visualisasi Citra Landsat 8 di GEE

//import the satellite data from european space agency

var L8 = ee.ImageCollection("LANDSAT/LC08/C01/T1_TOA");

var spatialFiltered =L8.filterBounds(Cirebon);

print('spatialFiltered', spatialFiltered);

var temporalFiltered = spatialFiltered.filterDate('2015-01-01', '2015-12-31');

print('temporalFiltered', temporalFiltered);

// This will sort from least to most cloudy.

var sorted = temporalFiltered.sort('CLOUD_COVER');

// Get the first (least cloudy) image.

var scene = sorted.first();

Map.centerObject(scene, 9);

Map.addLayer(scene, {}, 'default RGB');

var visParams = {bands: ['B4', 'B3', 'B2'], max: 0.3};

Map.addLayer(scene, visParams, 'true-color composite');

//Create training dataset

var training = scene.sample(

  { region : Cirebon,

    scale : 20,

    numPixels : 5000

  }

  );

  

//Start Unsupervised clustering algorithm and training it

var kmeans = ee.Clusterer.wekaKMeans(3).train(training);

//Cluster the input using the trained clusterer.

var result = scene.cluster(kmeans);

//Display the cluster with random colors

Map.addLayer(result.randomVisualizer(),{}, "Unsupervised kmeans classification");

//Export the Image to Drive

Export.image.toDrive({

  image :result,

  description : 'kmeans Cirebon',

  scale : 20,

  region : Cirebon

  });

 

• Unsupervised analysis di GEE

//import the satellite data from european space agency

var S2 = ee.ImageCollection("COPERNICUS/S2");

//filter for Dubai

S2 = S2.filterBounds(Dubai);

print(S2);

//filter date

S2 = S2.filterDate("2020-01-01","2020-05-11");

print(S2);

//visualize image

var image = ee.Image(S2.first());

//Map.addLayer(image, {min:0, max:3000,bands:"B4,B3,B2"}, "DubaiTrue Color");

Map.addLayer(image, {min:0, max:3000,bands:"B8,B4,B3"}, "Dubai False Color");

//Create training dataset

var training =image.sample(

  { region : Dubai,

    scale : 20,

    numPixels : 5000

  }

  );

  

//Start Unsupervised clustering algorithm and training it

var kmeans = ee.Clusterer.wekaKMeans(5).train(training);

//Cluster the input using the trained clusterer.

var result = image.cluster(kmeans);

//Display the cluster with random colors

Map.addLayer(result.randomVisualizer(),{}, "Unsupervised kmeans classification");

//Export the Image to Drive

Export.image.toDrive({

  image :result,

  description : 'kmenas Dubai',

  scale : 20,

  region : Dubai

  });

 

• Supervised analysis di GEE