在谷歌地球引擎中对图像集合对图像执行 PCA

Question

我需要对图像集合的每个图像执行PCA。然后，我只想保留主组件轴 1，并将其作为波段添加到图像集合中的每个图像中。最终，我想导出一个.csv文件，其中 GPS 采样位置位于行标题处，图像 ID 作为列标题，平均主成分轴 1 作为值。这样做背后的想法是，我希望在R中进一步的统计分析中使用一个代理(光谱异质性(。

这是我到目前为止的代码：

//Create an test image to extract information to be used during PCA
var testImage =ee.Image('LANDSAT/LC08/C01/T1_SR/LC08_168080_20130407')
.select(['B2', 'B3', 'B4', 'B5', 'B6', 'B7'],
['Blue', 'Green', 'Red', 'NIR', 'SWIR1', 'SWIR2']);
// Define variables for PCA
var region = Extent;
var scale = testImage.projection().nominalScale();
var bandNames = testImage.bandNames();
Map.centerObject(region);
// Function for performing PCA
function doPCA(image){
// This code is from https://code.earthengine.google.com/7249153a8a0f5c79eaf562ed45a7adad
var meanDict = image.reduceRegion({
reducer: ee.Reducer.mean(),
geometry: region,
scale: scale,
maxPixels: 1e9
});
var means = ee.Image.constant(meanDict.values(bandNames));
var centered = image.subtract(means);
// This helper function returns a list of new band names.
var getNewBandNames = function(prefix) {
var seq = ee.List.sequence(1, bandNames.length());
return seq.map(function(b) {
return ee.String(prefix).cat(ee.Number(b).int());
});
};
// [START principal_components]
var getPrincipalComponents = function(centered, scale, region) {
var arrays = centered.toArray();
var covar = arrays.reduceRegion({
reducer: ee.Reducer.centeredCovariance(),
geometry: region,
scale: scale,
maxPixels: 1e9
});
var covarArray = ee.Array(covar.get('array'));
var eigens = covarArray.eigen();
var eigenValues = eigens.slice(1, 0, 1);
var eigenVectors = eigens.slice(1, 1);
var arrayImage = arrays.toArray(1);
var principalComponents = ee.Image(eigenVectors).matrixMultiply(arrayImage);
var sdImage = ee.Image(eigenValues.sqrt())
.arrayProject([0]).arrayFlatten([getNewBandNames('sd')]);
return principalComponents
.arrayProject([0])
.arrayFlatten([getNewBandNames('pc')])
.divide(sdImage);
};
var pcImage = getPrincipalComponents(centered, scale, region);
return (pcImage);
}
// map PCA function over collection
var PCA = LandsatCol.map(function(image){return doPCA(image)});
print('pca', PCA);

Extent是我的投资回报率，而LandsatCol是预先处理的图像集合。此处的代码在尝试将 PCA 映射到图像集合(倒数第二行代码(时生成错误。错误显示："数组：参数'值'是必需的"。

关于如何处理这个问题的任何建议？以及如何将主成分轴 1 添加为图像集合上每个图像的波段？

Answer 1

我想通了。错误"数组：需要参数'值'"与稀疏矩阵有关，稀疏矩阵是过滤、裁剪和特殊化区域内以执行 PCA 的产物。地球引擎不能与稀疏矩阵一起工作。

这是工作代码。LandsatCol是我预先处理的图像集。

// Display AOI
var point = ee.Geometry.Point([30.2261, -29.458])
Map.centerObject(point,10);
// Prepairing imagery for PCA
var Preped = LandsatCol.map(function(image){
var orig = image;
var region = image.geometry();
var scale = 30;
var bandNames = ['Blue', 'Green', 'Red', 'NIR', 'SWIR1', 'SWIR2'];
var meanDict = image.reduceRegion({
reducer: ee.Reducer.mean(),
geometry: region,
scale: scale,
maxPixels: 1e9
});
var means = ee.Image.constant(meanDict.values(bandNames));
var centered = image.subtract(means);
var getNewBandNames = function(prefix) {
var seq = ee.List.sequence(1, 6);
return seq.map(function(b) {
return ee.String(prefix).cat(ee.Number(b).int());
});
};
// PCA function
var getPrincipalComponents = function(centered, scale, region) {
var arrays = centered.toArray();
var covar = arrays.reduceRegion({
reducer: ee.Reducer.centeredCovariance(),
geometry: region,
scale: scale,
maxPixels: 1e9
});
var covarArray = ee.Array(covar.get('array'));
var eigens = covarArray.eigen();
var eigenValues = eigens.slice(1, 0, 1);
var eigenVectors = eigens.slice(1, 1);
var arrayImage = arrays.toArray(1);
var principalComponents = ee.Image(eigenVectors).matrixMultiply(arrayImage);
var sdImage = ee.Image(eigenValues.sqrt())
.arrayProject([0]).arrayFlatten([getNewBandNames('sd')]);
return principalComponents.arrayProject([0])
.arrayFlatten([getNewBandNames('pc')])
.divide(sdImage);
};
var pcImage = getPrincipalComponents(centered, scale, region);
return ee.Image(image.addBands(pcImage));
});
print("PCA imagery: ",Preped);