Dear Stanley,

Google Earth Engine does indeed have the ability to perform supervised classification, using a variety of algorithms (including Naive Bayes, CART, Rifle Serial Classifier, SVM). For an example, see this video from IGARSS 2014 that describes how Earth Engine was used to estimate urban extent.

The number of training points needed for a "reasonable result" depend on many factors, including:

• the classification algorithm that you are using;
• the number of classes that you interested in estimating;
• how spectrally distinct the classes are (if classifying using image spectra);
• the SNR of the image;
• the homogeneity of the image.

These factors need to be considered regardless of what software you use to perform the classification.

To clarify, Google Earth and Google Earth Engine are two distinct technologies. Google Earth is a desktop 3-D visualization tool, while Google Earth Engine (https://earthengine.google.org) is a cloud-based geoprocessing platform. 

http://www.igarss2014.org/ShowRecording.asp?C=D924D9B2

Dear Stanley,

Google Earth engine itself does not have any classification algorithms unless if you want to do it manually. In addition, it is not clear what do you want to classify because in this case you need to decide the resolution of the image and this limit the area seen on the screen e.g. land use requires high resolution image to differentiate between different classes such as major roads, primary roads, secondary roads, industrial buildings, residential buildings etc...

unless if a general classification of land cover (forests in general, agriculture lands, urban settlements ) in this case I would do it manually on Google Earth.

In general, you can capture an image representing the 80 Sq, Km and use a commercial software to classify it  (ENVI, Erdas) or open source software that is the best solution.

Dear Paul and Awad, Thank you for your response, actually the intension is to train and classify into different land cover classes(total=5), e.g forest, agricultural area, built up etc. And the training point I intend using google earth placemark to produce a fusion table via kml. The total area is approximately 80000 sq km not 80 sq km.

Dear Stanley,

Google Earth Engine does indeed have the ability to perform supervised classification, using a variety of algorithms (including Naive Bayes, CART, Rifle Serial Classifier, SVM). For an example, see this video from IGARSS 2014 that describes how Earth Engine was used to estimate urban extent.

The number of training points needed for a "reasonable result" depend on many factors, including:

• the classification algorithm that you are using;
• the number of classes that you interested in estimating;
• how spectrally distinct the classes are (if classifying using image spectra);
• the SNR of the image;
• the homogeneity of the image.

These factors need to be considered regardless of what software you use to perform the classification.

To clarify, Google Earth and Google Earth Engine are two distinct technologies. Google Earth is a desktop 3-D visualization tool, while Google Earth Engine (https://earthengine.google.org) is a cloud-based geoprocessing platform. 

http://www.igarss2014.org/ShowRecording.asp?C=D924D9B2

What about unsupervised classification? Possible in Google Earth Engine?

What about unsupervised classification? Possible in Google Earth Engine?

Yes. Unsupervised classification is now possible on Google Earth Engine.

you may refer this code link...

https://code.earthengine.google.com/c098d203000ac089414d4a636eaaade5

could you please share the changing the date and downloading codes of the bellow classification.

https://code.earthengine.google.com/c098d203000ac089414d4a636eaaade5