Take the 2-minute tour ×
Geographic Information Systems Stack Exchange is a question and answer site for cartographers, geographers and GIS professionals. It's 100% free, no registration required.

Prior to image classification, collection of training sites and obtaining a signature file is the most critical step. Field visits with a reliable GPS Unit is usually the first step followed by overlaying the points on the to-be-classified image. Other than this when identifying training sites on satellite images, especially the coarse resolution ones what are the rules normally followed by professionals - like minimum number of pixels in a training site, distance between two training sites for same/different classes, how improvements are made after preliminary classification etc. Standard references are also welcome.

share|improve this question

3 Answers 3

up vote 3 down vote accepted

I wont be able to give a great deal of information on this, but there are certain rules I think we need to follow for training sites:

  • Training sites should be spread all over the image i.e. should be covering a large area and not just focussed in a region
  • Just as Georeferencing, training sites should be identified at the corners and at points which give maximum sampling options
  • The objective is to identify a set of pixels that accurately represents spectral variation present within each information region

Hope it helps...

share|improve this answer

The article below highlights some concepts regarding number of pixels per training site, and number of training sites per class for sampling (pg 2). Hawth's tools can be used to generate random stratified training points if you are using ArcGIS.

Improving image classification accuracy: a method to incorporate uncertainty in the selection of training sample sets

share|improve this answer

Firstly a little more detail about your application would help return a better answer. Second, generally speaking in natural systems it is rare to find linear relationships and so you cannot rely on having two points make a line. You need more. But most importantly you need points "in the middle". If you are lucky the relationship between the remote sensing output and the feature of interest will be a modest curve and the midpoint of that curve will be easier to define if you have some or many points near the middle.

With regard professionals... I say that all they know is how to cut the corners, so don't do what they do! Use best practice and find a suitable statistical test which tells you when you have gathered enough data. Plotting stuff up on the fly is a good start.

My contribution at least

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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