I have 5 different years (all in the same month) of Landsat TM imagery to perform classification on and I would like to know whether there is a technique to normalise the pixels in each image so my training sites can remain the same for all 5 images.

I understand this is a common technique to perform multiple-date image normalisation using regression but I am unaware of how to apply the technique.

I am using both ERDAS Imagine and Arcmap 10.1 for my analysis.

What tool do I run for this?


I am interpreting your question to ask if you can generate training data from 1 image and apply those statistics to the other 4 images by normalizing the other 4 to the first. This seems dubious to me. While it is possible to normalize atmospheric differences, there are other differences (e.g. phenology, recent weather events) that make normalizing problematic. I assume you are classifying each image separately and because you use the phrase, "training sites" that you are using supervised classification. Consider using the polygons that define your training sites to generate statistics for each image separately. That will fulfill your desire for your training sites to remain the same. I hope I have interpreted your question correctly.

| improve this answer | |

A quick Google search threw up these papers, I am sure there are?

| improve this answer | |

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.