本文主要是介绍GEE——使用cart机器学习方法对Landsat影像条带修复以NDVI和NDWI为例（全代码），希望对大家解决编程问题提供一定的参考价值，需要的开发者们随着小编来一起学习吧！
简介
之前发表了两篇关于影像修复的文章，并且制作了APP，大家可以去看以下的两篇博客来了解具体的研究内容和整个方法的有效性：
Google Earth Engine APP——影像条带色差、色调不均匀等现象解决方案Landsat5 NDWI Image Restoration APP_ndwi不能识别泛红水体怎么办-CSDN博客
基于GEE云平台一种快速修复Landsat影像条带色差的方法_gee平台-CSDN博客
影像条带色差产生的主要原因有以下几点：
1. 光学系统问题：光学系统中的透镜、滤光片等元件可能存在偏差或缺陷，导致不同波长的光在传输过程中被聚焦的位置不一致，从而产生色差。
2. 图像传感器问题：图像传感器中的像素单元可能对不同波长的光的响应度不同，导致不同波长的光在图像传感器上形成的图像亮度不一致，从而产生色差。
3. 色彩处理问题：在图像的处理过程中，可能会对不同波长的光进行不同的处理，如增强某个颜色通道的亮度或饱和度，从而导致色差。
4. 环境光影响：在拍摄现场，环境光的波长和强度可能有所不同，对拍摄的影像产生影响，从而产生色差。
总的来说，影像条带色差的产生主要是由于光学系统、图像传感器、色彩处理和环境光等多个因素综合作用的结果。
函数：
本文里面的主要使用的函数众多，包含了归一化函数，直方图统计，机器学习方法以及图形展示等
normalizedDifference(bandNames)
Computes the normalized difference between two bands. If the bands to use are not specified, uses the first two bands. The normalized difference is computed as (first − second) / (first + second). Note that the returned image band name is 'nd', the input image properties are not retained in the output image, and a negative pixel value in either input band will cause the output pixel to be masked. To avoid masking negative input values, use ee.Image.expression() to compute normalized difference.
Arguments:
this:input (Image):
The input image.
bandNames (List, default: null):
A list of names specifying the bands to use. If not specified, the first and second bands are used.
Returns: Image
CLOSE
ee.ImageCollection.fromImages(images)
Returns the image collection containing the given images.
Arguments:
images (List):
The images to include in the collection.
Returns: ImageCollection
ui.Chart.image.histogram(image, region, scale, maxBuckets, minBucketWidth, maxRaw, maxPixels)
Generates a Chart from an image. Computes and plots histograms of the values of the bands in the specified region of the image.
-
X-axis: Histogram buckets (of band value).
-
Y-axis: Frequency (number of pixels with a band value in the bucket).
Returns a chart.
Arguments:
image (Image):
The image to generate a histogram from.
region (Feature|FeatureCollection|Geometry, optional):
The region to reduce. If omitted, uses the entire image.
scale (Number, optional):
The pixel scale used when applying the histogram reducer, in meters.
maxBuckets (Number, optional):
The maximum number of buckets to use when building a histogram; will be rounded up to a power of 2.
minBucketWidth (Number, optional):
The minimum histogram bucket width, or null to allow any power of 2.
maxRaw (Number, optional):
The number of values to accumulate before building the initial histogram.
maxPixels (Number, optional):
If specified, overrides the maximum number of pixels allowed in the histogram reduction. Defaults to 1e6.
Returns: ui.Chart
setSeriesNames(seriesNames, seriesIndex)
Returns a copy of this chart with updated series names.
Arguments:
this:ui.chart (ui.Chart):
The ui.Chart instance.
seriesNames (Dictionary|Dictionary<String>|List|List<String>|String):
New series names. If it's a string, the name of the series at seriesIndex is set to seriesNames. If it's a list, the value at index i in the list is used as a label for series number i. If it's a dictionary or an object, it's treated as a map from existing series names to new series names. In the last two cases, seriesIndex is ignored.
seriesIndex (Number, optional):
The index of the series to rename. Ignored if seriesNames is a list or dictionary. Series are 0-indexed.
Returns: ui.Chart
ee.Reducer.histogram(maxBuckets, minBucketWidth, maxRaw)
Create a reducer that will compute a histogram of the inputs.
Arguments:
maxBuckets (Integer, default: null):
The maximum number of buckets to use when building a histogram; will be rounded up to a power of 2.
这篇关于GEE——使用cart机器学习方法对Landsat影像条带修复以NDVI和NDWI为例（全代码）的文章就介绍到这儿，希望我们推荐的文章对编程师们有所帮助！
