2

Just theoretical (and maybe very silly) question - Is it possible to do crop classification (supervised or unsupervised) from multi-spectral camera or SAR without direct nexus with training points or sampling areas?

For example, observing NVDI and trying to find some patterns related to phenological stages?

1

There are several potential approaches here: 1) an unsupervised approach and 2) expertly selecting classes in your image to use as training data, 3) feature extraction, 4) differentiating classes by statistical measures such as texture.

With unsupervised classification, you still need to determine the number of classes, combine similar classes, and assess the results. Expertly selecting pixels is a better approach, second to field training data, if you have the knowledge to differentiate between classes on-screen. If you are extracting shapes such as crop circles from your remotely sensed data, you will want to use feature extraction algorithms, such as OpenCV's Hough Circles.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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