GEE ——多种机器学习方法（随机森林、cart、svm等）进行土地分类用光谱指数、植被、土壤、雪和烧毁区

//加载研究区样本点信息
var rois = ee.FeatureCollection(train_p1); 
var rois2 = ee.FeatureCollection(testing_p1); 

//地形参数
var ALOSDSMDEM = ee.Image('JAXA/ALOS/AW3D30_V1_1'); 
var ALOSDM = ALOSDSMDEM.select('AVE').toDouble().clip (puna).rename("ELEVACION"); 
var SLOPE_F = ee.Terrain.slope(ALOSDM).toDouble().clip (puna).rename("PENDIENTE"); 
var ASPECT_F = ee.Terrain.aspect(ALOSDM).toDouble().rename("ASPECTO"); 
var SIN_ASPECT = 
ASPECT_F.divide(180).multiply(Math.PI).sin().toDouble().rename("ASPECTO"); 

//光学影像
var IMGLandsat8= ee.ImageCollection ('LANDSAT/LC08/C01/T1_SR') 
 .filterDate ('2018-01-31', '2018-06-30') //fechas disponibles ('2013-04-11' - actualidad) 
 .filterBounds (puna); 
 
 //影像去云
function maskLXsr(image){ 
 var cloudShadowBitMask = ee.Number(2).pow(3).int(); 
 var cloudsBitMask = ee.Number(2).pow(5).int(); 
 var qa = image.select('pixel_qa'); 
 var mask = qa.bitwiseAnd(cloudShadowBitMask).eq(0) 
 .and(qa.bitwiseAnd(cloudsBitMask).eq(0)); 
 return image.updateMask(mask).divide(10000).clamp(1e-6, 1) 
 .copyProperties(image, ["system:time_start"]); 
} 
//影像去云映射和中位数合成
var lXmask = IMGLandsat8.map(maskLXsr); 
var Landsat8_NRG1 = ee.Image(lXmask.median()); 
//波段替换和转化为双精度
var Landsat8a = Landsat8_NRG1.clip (puna) 
 .select(['B2', 'B3', 'B4', 'B5', 'B6', 'B7']) 
 .rename(['blue','green','red','nir','swir1','swir2']) 
 .toDouble(); 
//
var Landsat8 = (Landsat8a.select('blue','green','red','nir','swir1','swir2') 
 .divide((Landsat8a.select('blue','green','red','nir','swir1','swir2') 
 .reduce(ee.Reducer.sum())).divide(ee.Number(6)))) 
 .multiply(ee.Number(100)); 
 
//添加指数
//添加 NDVI 
var addNDVI=function(imagen){ 
 var NIR='nir'; 
 var RED='red'; 
 var ndvi= imagen.normalizedDifference([NIR, RED]).rename('NDVI'); 
 return imagen.addBands(ndvi).select('NDVI')}; 
var ndvi = addNDVI(Landsat8).select('NDVI').toDouble().rename("NDVI"); 
var NDVI_FIN = ndvi.clamp(-1, 1); 
//print(NDVI_FIN,"NDVI_FIN") 
//Map.addLayer(NDVI_FIN,{min: 0, max: 1},"ndvi"); 

//添加NDBI
// Normalized Difference Builtup Index (NDBI) NDBI = (SWIR1 – NIR) / (SWIR1 + NIR) 
var addNDBI=function(imagen){ 
 var NIR='nir'; 
 var SWIR1='swir1'; 
 var ndbi= imagen.normalizedDifference([SWIR1,NIR]).rename('NDBI'); 
 return imagen.addBands(ndbi).select('NDBI')}; 
var NDBI = addNDBI(Landsat8).select('NDBI').toDouble().rename("NDBI"); 
var NDBI_FIN = NDBI.clamp(-1, 1); 
//print(NDBI_FIN,"NDBI_FIN") 
//Map.addLayer(NDBI_FIN,{min: 0, max: 1},"NDBI"); 

//添加NBR
//Normalized BURN RATIO "(NIR - SWIR1) / (NIR + SWIR1)" 
var addNBR=function(imagen){ 
 var NIR='nir'; 
 var SWIR1='swir1'; 
 var nbr= imagen.normalizedDifference([NIR, SWIR1]).rename('NBR'); 
 return imagen.addBands(nbr).select('NBR')}; 
var NBR = addNBR(Landsat8).select('NBR').toDouble().rename("NBR"); 
var NBR_FIN = NBR.clamp(-1, 1); 
//print(NBR_FIN,"NBR_FIN") 
//Map.addLayer(NBR_FIN,{min: 0, max: 1},"NBR"); 

//添加NDWI
// Normalized Difference Water Index (NDWI) NDWI = (Green – NIR) / (Green + NIR) 
var addNDWI=function(imagen){ 
 var NIR='nir'; 
 var GREEN='green'; 
 var ndwi= imagen.normalizedDifference([GREEN, NIR]).rename('NDWI'); 
 return imagen.addBands(ndwi).select('NDWI')}; 
var NDWI = addNDWI(Landsat8).select('NDWI').toDouble().rename("NDWI"); 
var NDWI_FIN = NDWI.clamp(-1, 1); 
//print(NDWI_FIN,"NDWI_FIN") 
//Map.addLayer(NDWI_FIN,{min: 0, max: 1},"NDWI"); 


//添加NBR NDSI
// Modified Normalized Difference Water Index (mNDWI) or Normalized Difference Snow Index (NDSI) 
// A common index for detecting water (mNDWI) or snow (NDSI), where mNDWI = NDSI = (Green – SWIR1) / (Green + SWIR1). 
var addNDSI=function(imagen){ 
 var GREEN='green'; 
 var SWIR1='swir1'; 
 var ndsi= imagen.normalizedDifference([GREEN, SWIR1]).rename('NDSI'); 
 return imagen.addBands(ndsi).select('NDSI')}; 
var NDSI = addNDSI(Landsat8).select('NDSI').toDouble().rename("NDSI"); 
var NDSI_FIN = NDSI.clamp(-1, 1); 
//print(NDSI_FIN,"NDSI_FIN") 
//Map.addLayer(NDSI_FIN,{min: 0, max: 1},"NDSI"); 

// 添加 MSAVI 
var MSAVI = Landsat8.select('nir').multiply(2).add(1) 
 .subtract(Landsat8.select('nir').multiply(2).add(1).pow(2) 
 .subtract(Landsat8.select('nir').subtract(Landsat8.select('red')).multiply(8)).sqrt() 
 ).divide(2) 
var MSAVI_FIN = MSAVI.toDouble().rename("MSAVI"); 
//print(MSAVI_FIN,"MSAVI_FIN") 
//Map.addLayer(MSAVI_FIN,{min: 0, max: 0.5},"MSAVI_FIN"); 

// 添加 Enhanced Vegetation Index (EVI) 
//An "optimized" vegetation index designed to enhance the vegetation signal with improved 
sensitivity 
//in high biomass regions. Landsat EVI = 2.5 * (NIR – Red) / ( NIR + 6*Red – 7.5*Blue + 1). 
var EVI1 = ((Landsat8.select('nir').subtract(Landsat8.select('red'))) 
 .divide(Landsat8.select('nir').add(((Landsat8.select('red').multiply(6)) 
 .subtract(Landsat8.select('blue').multiply(7.5))).add(1)))).multiply(2.5) 
var EVI = EVI1.toDouble().rename("EVI"); 
//print (EVI, 'EVI') 
//Map.addLayer(EVI, {min: -1, max: 1}, 'EVI'); 

//添加光谱变异植被指数 Spectral Variability Vegetation Index (SVVI) 
var SVV1a = Landsat8.select('blue','green','red','nir','swir1','swir2').reduce(ee.Reducer.stdDev()); 
var SVV1b = Landsat8.select('nir','swir1','swir2').reduce(ee.Reducer.stdDev()); 
var SVVI = SVV1a.subtract(SVV1b).rename("SVVI"); 
//var bands = ['blue','green','red','nir','swir1','swir2'] 
//var bands = ['NDVI','NDBI','NBR', 'NDWI','NDSI','MSAVI','EVI','SVVI']; 
//var bands = ['blue','green','red','nir','swir1','swir2', 'NDVI','NDBI','NBR', 'NDWI','NDSI','MSAVI','EVI','SVVI'] 
//var bands = ['blue','green','red','nir','swir1','swir2','ELEVACION', 'PENDIENTE', 'ASPECTO']; 
//var bands = ['NDVI','NDBI','NBR', 'NDWI','NDSI','MSAVI','EVI','SVVI', 'ELEVACION', 'PENDIENTE', 'ASPECTO']; 
//var bands = ['blue','green','red','nir','swir1','swir2','NDVI','NDBI','NBR', 'NDWI','NDSI','MSAVI','EVI','SVVI','ELEVACION', 'PENDIENTE', 'ASPECTO']; 

//多波段融合
var predictors_base = Landsat8.addBands(NDVI_FIN).addBands(NDBI_FIN).addBands(NBR_FIN) 
 .addBands(NDWI_FIN).addBands(NDSI_FIN).addBands(MSAVI_FIN) 
 .addBands(EVI).addBands(SVVI) 
 .addBands(ALOSDM).addBands(SLOPE_F).addBands(SIN_ASPECT); 
 
//选择不同波段用于训练模型
var predictors = predictors_base.select(bands); 

//这里有两个一个是训练样本一个是测试样本
//提取样本点这里直接变成矢量集合 
var trainingPartition = predictors.sampleRegions({ 
 collection: rois, 
 properties: ['Clave'], 
 scale: 30 
}); 
var testingPartition = predictors.sampleRegions({ 
 collection: rois2, 
 properties: ['Clave'], 
 scale: 30 
}); 

//多分类器集合为一个变量 为下一步应用不同模型进行准备 
var classifiers = [ee.Classifier.libsvm(), 
 ee.Classifier.minimumDistance(), 
 ee.Classifier.smileCart(), 
 ee.Classifier.smileGradientTreeBoost(100), 
 ee.Classifier.smileNaiveBayes(), 
 ee.Classifier.smileRandomForest(100)]; 
 
//进行随机森林训练
var classifier = ee.Classifier.smileRandomForest({numberOfTrees: 100,}).train({ 
 features: trainingPartition, 
 classProperty: 'Clave', 
 inputProperties: bands 
}); 
//描述经过训练的分类器的结果。
var dict = classifier.explain(); 
//变量重要性重要性发分析 
var variable_importance = ee.Feature(null, ee.Dictionary(dict).get('importance')); 

//加载变量重要性图表
 var chart = 
ui.Chart.feature.byProperty(variable_importance) 
.setChartType('ColumnChart') 
.setOptions({ 
title: 'Random Forest Variable Importance', 
legend: {position: 'none'}, 
hAxis: {title: 'Bands'}, 
vAxis: {title: 'Importance'} 
}); 
var votes = ee.Image().clip(puna); 

//这里遍历所有分类方法得出结果
for(var i in classifiers) { 
 votes = votes.addBands(predictors.classify(classifiers[i].train(trainingPartition,"Clave")))} 
print(votes, "votes"); 
var votes1 = votes.select('classification_4').toDouble() 
print(votes1, 'vot1') 
 
 //导出分类结果
*Export.image.toDrive({ 
 image: votes.select('classification','classification_1', 
 'classification_2', 'classification_3', 
 'classification_4', 'classification_5'), 
 description: 'C1', 
 scale: 30, 
 folder : 'COVER_NYC', 
 maxPixels: 8000000000, 
 region: puna});