GEE实现图像随机森林分类

迷南。 2023-09-23 20:07 123阅读 0赞

对图像进行土地利用分类，因此下面是监督分类的流程以及代码案例。  
1.首先分类最开始应该建立样本数据集，在这里我分了四类，然后就开始自己的采样，设立好分类后，对目标进行分类。  
![在这里插入图片描述][7034a66b186247d785e284ea873a89d1.png]  
然后对每个样本进行颜色选择和属性定义

![image.png][]

//选择需要裁剪的矢量数据 
    var aoi = ee.FeatureCollection("users/yangyao19960805/NewFolder");
    //加载矢量边框，以便于在边界内选取样本点
    var empty = ee.Image().toByte();
    var outline = empty.paint({
        
     featureCollection:aoi, // 行政边界命名为fc
     color:0, //颜色透明
     width:3 //边界宽度
    });
    Map.addLayer(outline, {
        palette: "ff0000"}, "outline");
    //Function to mask the clouds in Sentinel-2
    function maskS2clouds(image) {
        
      var qa = image.select('QA60');
    
      // Bits 10 and 11 are clouds and cirrus, respectively.
      var cloudBitMask = 1 << 10;
      var cirrusBitMask = 1 << 11;
    
      // Both flags should be set to zero, indicating clear conditions.
      var mask = qa.bitwiseAnd(cloudBitMask).eq(0)
          .and(qa.bitwiseAnd(cirrusBitMask).eq(0));
    
      return image.updateMask(mask).divide(10000);
    }
    
    //Build the Sentinel 2 collection, filtered by date, bounds and percentage of cloud cover 
    var dataset = ee.ImageCollection('COPERNICUS/S2_SR')
                      .filterDate('2019-01-01','2020-12-31')
                      .filterBounds(aoi)
                      .filter(ee.Filter.lt('CLOUDY_PIXEL_PERCENTAGE',10))
                      .map(maskS2clouds);
    print("Sentinel 2 Image Collection",dataset);
    var dem = ee.Image("NASA/NASADEM_HGT/001")
    // Construct Classfication Dataset
    // RS Index Cacluate(NDVI\NDWI\EVI\BSI)
    var add_RS_index = function(img){
        
      var ndvi = img.normalizedDifference(['B8', 'B4']).rename('NDVI').copyProperties(img,['system:time_start']);
      var ndwi = img.normalizedDifference(['B3', 'B8']).rename('NDWI').copyProperties(img,['system:time_start']);
      var evi = img.expression('2.5 * ((NIR - RED) / (NIR + 6 * RED - 7.5 * BLUE + 1))', 
      {
        
        'NIR': img.select('B8'),
        'RED': img.select('B4'),
        'BLUE': img.select('B2')
      }).rename('EVI').copyProperties(img,['system:time_start']);
      var bsi = img.expression('((RED + SWIR1) - (NIR + BLUE)) / ((RED + SWIR1) + (NIR + BLUE)) ', 
      {
        
        'RED': img.select('B4'), 
        'BLUE': img.select('B2'),
        'NIR': img.select('B8'),
        'SWIR1': img.select('B11'),
    
      }).rename('BSI').copyProperties(img,['system:time_start']);
    
    
      var ibi = img.expression('(2 * SWIR1 / (SWIR1 + NIR) - (NIR / (NIR + RED) + GREEN / (GREEN + SWIR1))) / (2 * SWIR1 / (SWIR1 + NIR) + (NIR / (NIR + RED) + GREEN / (GREEN + SWIR1)))', {
        
        'SWIR1': img.select('B11'),
        'NIR': img.select('B8'),
        'RED': img.select('B4'),
        'GREEN': img.select('B3')
      }).rename('IBI').copyProperties(img,['system:time_start']);
      return img.addBands([ndvi, ndwi, evi, bsi, ibi]);
    };
    var dataset = dataset.map(add_RS_index); 
    var bands = ['B2','B3','B4','B5','B6','B7','B8','B8A','B11','NDVI','NDWI','BSI'];
    var imgcol_median = dataset.select(bands).median();
    var aoi_dem = dem.select('elevation').clip(aoi).rename('DEM');
    var construct_img = imgcol_median.addBands(aoi_dem).clip(aoi);
    //分类样本
    var train_points = cropland.merge(grassland).merge(city).merge(forest).merge(water);
    var train_data= construct_img.sampleRegions({
        
      collection: train_points,
      properties: ['landcover'],
      scale: 10
    });
    //精度评价
    var withRandom = train_data.randomColumn('random');//样本点随机的排列
    var split = 0.7; 
    var trainingPartition = withRandom.filter(ee.Filter.lt('random', split));//筛选70%的样本作为训练样本
    var testingPartition = withRandom.filter(ee.Filter.gte('random', split));//筛选30%的样本作为测试样本
    //分类方法选择随机森林
    var rf = ee.Classifier.smileRandomForest({
        
      numberOfTrees: 20,  
      bagFraction: 0.8
    }).train({
        
      features: train_data,
      classProperty: 'landcover',
      // inputProperties: inputbands
    });
    //对哨兵数据进行随机森林分类
    var img_classfication = construct_img.classify(rf); 
    //运用测试样本分类，确定要进行函数运算的数据集以及函数
    var test = testingPartition.classify(rf);
    //计算混淆矩阵
    var confusionMatrix = test.errorMatrix('landcover', 'classification');
    print('confusionMatrix',confusionMatrix);//面板上显示混淆矩阵
    print('overall accuracy', confusionMatrix.accuracy());//面板上显示总体精度
    print('kappa accuracy', confusionMatrix.kappa());//面板上显示kappa值
    Map.centerObject(aoi)
    Map.addLayer(aoi);
    Map.addLayer(img_classfication.clip(aoi), {
        min: 1, max: 4, palette: ['orange', 'blue', 'green','yellow']});
    var class1=img_classfication.clip(aoi)
    //导出分类图
    Export.image.toDrive({
          
           image: class1,  
           description: 'rfclass',  
           fileNamePrefix: 'rf',  //文件命名
           folder: "class",  //保存的文件夹
           scale: 10,  //分辨率
           region: aoi,  //研究区
           maxPixels: 1e13,  //最大像元素，默认就好
           crs: "EPSG:4326"  //设置投影
       });

![在这里插入图片描述][66713f1d3c9e443799129d5e559d3ff8.png]

[7034a66b186247d785e284ea873a89d1.png]: https://img-blog.csdnimg.cn/7034a66b186247d785e284ea873a89d1.png
[image.png]: https://img-blog.csdnimg.cn/ff85f5d3cda846edaedb32540148545c.png
[66713f1d3c9e443799129d5e559d3ff8.png]: https://img-blog.csdnimg.cn/66713f1d3c9e443799129d5e559d3ff8.png