0

I have a problem with random forest classification, receiving an error message after inputting the training dataset. My code includes a long preprocessing for Sentinel1 and Sentinel2, but here I'll simply summarize the code here (starting from line 248) GEE code: https://code.earthengine.google.com/3e844f7bd2cfa702709853fc78497dc4

//input the training data
var trainingData = ee.FeatureCollection('projects/ee-2023swedenagb/assets/Swe2017sample_ROI1');

//using sampleRegion to extract the values from S1 and S2 image
var s2_sample = s2L2017c1_6_Veg.sampleRegions({
  collection: trainingData,
  scale: 10,
  properties: ['AGB_DW'],
  projection: 'EPSG:3006'
});
print(s2_sample)

var s1_sample = s1L20171_6_predb.sampleRegions({
  collection: trainingData,
  scale: 10,
  properties: ['AGB_DW'],
  projection: 'EPSG:3006'
});
print(s1_sample)

//there are 211 pixels being extracted as the training data separately from S1 and S2
//merge the output from sampleRegion S1, S2 
var S1S2_2017_sample = s2_sample.merge(s1_sample)
print(S1S2_2017_sample)
//now there are 422 pixels being printed, which means the merge() function merges the S1 S2 but in different rows of data (but the 211 pixels are the same)


//using addBands function to combine the whole S1, S2 images
var S1_2017_VV = s1L20171_6_predb.select('VV');
var S1_2017_VH = s1L20171_6_predb.select('VH');

var S1S2_2017 = s2L2017c1_6_Veg.addBands(S1_2017_VV).addBands(S1_2017_VH);
print(S1S2_2017);

//selecting all the bands I need to input into the random forest model
var variables2017 = S1S2_2017.select(['VV','VH','B1','B2','B3','B4','B5','B6','B7','B8','B8A','B9','B11','B12','NDVI','RVI','EVI','NDVIn','NDVIre','DVI','Vigreen','GNDVI','SAVI','MSAVI','ARVI','NDI45']);

//using random forest as the first trial to calculate the image, also where the error occured
var classifier = ee.Classifier.randomForest(10).train({
  feature: S1S2_2017_sample,
  classProperty: 'AGB_DW',
  inputProperties: variables2017
})

var classified = variables2017.classify(classifier)
Map.addLayer(classified)

and the error says

Line 315: Required argument (features) missing to function: Classifier.train(classifier, features, classProperty, inputProperties, subsampling, subsamplingSeed)

Trains the classifier on a collection of features, using the specified numeric properties of each feature as training data. The geometry of the features is ignored.

I learn this method from a video from Google Earth (https://www.youtube.com/watch?v=6KIJB4A6VbI), but I'm not really sure why the error occurred.

1 Answer 1

0

After searching for helps from people, I learn that in this case, I should use the sampleRegion function after addbands, then it is required to use the new version of RF model, which is Classifier.randomForest -> Classifier.smileRandomForest. The link below is the announcement of the code update from GEE: https://groups.google.com/g/google-earthengine-announce/c/rCu4FP_Cn08/m/DqC192X9BAAJ?fbclid=IwAR3zvvG3CHzsHrhrTQspifwsNnk8wIBAgK_C3OjQMn_pvB4tS7ULa7p0qdc

var trainingData = ee.FeatureCollection('projects/ee-2023swedenagb/assets/Swe2017sample_ROI1');

//merge S1 and S2 data(image) in terms of their bands


var S1_2017_VV = s1L20171_6_predb.select('VV');
var S1_2017_VH = s1L20171_6_predb.select('VH');

var S1S2_2017 = s2L2017c1_6_Veg.addBands(S1_2017_VV).addBands(S1_2017_VH);
print(S1S2_2017);


var s1s2_sample = S1S2_2017.sampleRegions({
  collection: trainingData,
  scale: 10,
  properties: ['AGB_DW'],
  projection: 'EPSG:3006'
});
print(s1s2_sample)

var variables2017 = S1S2_2017.select(['VV','VH','B1','B2','B3','B4','B5','B6','B7','B8','B8A','B9','B11','B12','NDVI','RVI','EVI','NDVIn','NDVIre','DVI','Vigreen','GNDVI','SAVI','MSAVI','ARVI','NDI45']);

var classifier = ee.Classifier.smileRandomForest(10).train({
  features: s1s2_sample,
  classProperty: 'AGB_DW',
  inputProperties: variables2017
})

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.