The main purpose of this tutorial is to use random forest classification and other classifications for analysis. The main purpose here is that we use functions such as errorMatrix(actual, predicted, order ) for analysis in addition to conventional land classification, but RMSE and R2 need to be done by ourselves Write code analysis, which also includes the calculation of covariance.
Here is the image band:
planet（GREEN,BLUE,RED,NIR）
Image (4 bands)
type:Image
bands:
List (4 elements)
0:"b1", float ∈ [0, 6.553500175476074], EPSG:32634, 7618x4878 px
1:"b2", float ∈ [0, 6.553500175476074], EPSG:32634, 7618x4878 px
2:"b3", float ∈ [0, 6.553500175476074], EPSG:32634, 7618x4878 px
3:"b4", float ∈ [0, 6.553500175476074], EPSG:32634, 7618x4878 px
function:
ee.Reducer.covariance()
Creates a reducer that reduces a certain number of one-dimensional arrays of the same length N into a covariance matrix of shape NxN. This reducer uses the one-way covariance formula from Sandia National Laboratories technical report SAND2008-6212, which may lose accuracy if values span a large range.
return: reduce
ee.Kernel.square(radius, <