Google Earth Engine端元分解导出影像全黑值为0问题求助

问题描述 #

尝试使用Google Earth Engine代码进行光谱端元亚像元占比估算，但导出的影像在ArcMap中显示全黑且像素值为0，代码如下：

// The purpose of this script is to estimate sub-pixel fractions
// of identifiable spectral "endmembers".  This involves finding
// "pure" areas to estimate the endmembers, some matrix algebra
// followed by the mapping of the fractional cover.

// Use the reflective bands. 
var bands = ['B2', 'B3', 'B4', 'B5', 'B6', 'B7'];

// First, let's find a cloud free scene in our area of interest.
// Make a point using the geometry tools and name the import 'point'.
// Import Landsat 8 TOA data and name the collection 'l8'.
var image = ee.Image(l8
    .filterBounds(point)
    .filterMetadata('CLOUD_COVER','less_than', 2)
    .filter(ee.Filter.calendarRange(2020,2020,'year'))
    .filter(ee.Filter.calendarRange(10,12,'month'))
    .first())
    .select(bands);
    
  print(image)
Map.addLayer(image, {bands: ['B4', 'B3', 'B2'], max: 0.3}, 'image');

// Now, delineate polygons of 'pure' regions.  Click +New Layer for 
// each polygon.  Name the imports 'bare', 'vegetation' and 'water'.
// Get the mean spectrum in each of the endmember polygons.
var bareMean = image.reduceRegion(ee.Reducer.mean(), bare, 30).values();
var waterMean = image.reduceRegion(ee.Reducer.mean(), water, 30).values();
var vegMean = image.reduceRegion(ee.Reducer.mean(), vegetation, 30).values();
var snowMean = image.reduceRegion(ee.Reducer.mean(), snow, 30).values();

// Optional: plot the endmembers
print(ui.Chart.image.regions(image, ee.FeatureCollection([
    ee.Feature(bare, {label: 'bare'}), 
    ee.Feature(water, {label: 'water'}),
    ee.Feature(vegetation, {label: 'vegetation'}),
    ee.Feature(snow, {label: 'snow'})]), 
  ee.Reducer.mean(), 30, 'label', [0.48, 0.56, 0.65, 0.86, 1.61, 3.2]));

// Turn the endmember lists into an array that can be used in unmixing.
// Concatenate the lists along the 1-axis to make an array.
var endmembers = ee.Array.cat([bareMean, vegMean, waterMean, snowMean], 1);
//print(endmembers)

// Turn the image into an array image, in which each pixel has a 2-D matrix.
var arrayImage = image.toArray().toArray(1);

// Perform the unmixing in array space using the matrixSolve image method.  
// Note the need to cast the endmembers into an array image.
var unmixed = ee.Image(endmembers).matrixSolve(arrayImage);

// Convert the result from an array image back to a multi-band image.
var unmixedImage = unmixed.arrayProject([0])
    .arrayFlatten([['bare', 'veg', 'water', 'snow']]);

// Display the result.
Map.addLayer(unmixedImage, {}, 'fractions');

// Constrained:constraining the result to be non-negative and sum to one.
var constrained = image.unmix([bareMean, vegMean, waterMean, snowMean], true, true);
Map.addLayer(constrained, {}, 'constrained fractions');

//Export output to Google Drive
Export.image.toDrive({
  image: constrained,
  description: 'unmix',
  scale: 30,
  region: point,
  maxPixels: 1e9,
  fileFormat: 'GeoTIFF'
});

解决方法 #

• 修正导出区域参数：当前导出用的region: point是单个点，仅会导出1个30m分辨率的像素，导致影像无有效数据。需将region替换为实际研究区范围，比如：

  • 手动绘制研究区多边形并命名为studyArea，修改Export参数为region: studyArea
  • 或者使用image.geometry()导出整个Landsat影像覆盖范围，修改为region: image.geometry()

• 验证端元质量：检查bare、vegetation、water、snow等端元多边形是否为真实纯像元区域，通过print(bareMean, vegMean, waterMean, snowMean)查看端元均值是否合理，若某类端元均值异常（如全0）会导致解混结果失效。

• 调整ArcMap显示设置：解混输出的像素值为0-1之间的小数，ArcMap默认拉伸规则无法识别。右键图层→属性→符号系统→拉伸，选择"最小值-最大值"拉伸，或手动设置显示范围为0到1即可正常显示。

• 确认源影像有效性：通过Map预览和print(image)输出，检查加载的Landsat影像波段值是否正常，排除源影像全0的异常情况。

内容的提问来源于stack exchange，提问作者Mbariz