How do you analyze two images?

BrandonPrice1 · ‎01-09-2019

Is there an automated way to analyze two imagery layers in arcmap?

DanPatterson_Retired · ‎01-10-2019

It isn't going to return you anything useful. You basically have pictures, so you would just get the differences in color values.

It would take way less time to create a point file of the burn locations and tie those into a municipal lot file.

If you need the actual areas, then manually digitizing them would probably still be faster than performing a before and after image classification, cleaning the data up, doing some testing of the classification.

Human-ware is a pretty good tool, often under-rated and under-utilized

View solution in original post

DanPatterson_Retired · ‎01-09-2019

Are you referring to classification?

Overview of image classification—ArcGIS Pro | ArcGIS Desktop

or are they already classified and you are looking to do something with them?

BrandonPrice1 · ‎01-09-2019

I think they are classified. I would like to analyze one that has burn images with the other that doesn't. Whatever the best and most automated way is.

DanPatterson_Retired · ‎01-09-2019

What do you want to analyze? You could subtract them to begin with. Explain those differences.

It isn't automated thought, you have to use the Minus tool from arctoolbox

BrandonPrice1 · ‎01-09-2019

I am trying to create a vector polygon for the changes in the burned houses. Like red for the burned and green for not. I tried to use the difference button on the image analysis to show the changes in raster form. It might need some fiddling with. I then want to somehow that to vector.

BrandonPrice1 · ‎01-09-2019

Result of using the difference button. Not much contrast to identify the burned houses.

DanPatterson_Retired · ‎01-09-2019

As suspected, the images aren't classified.

At this stage, I would just recommend creating a point file and mark a point for each house that was burned.

At least then, you could reference the locations to a map that has lot or subdivision plans

XanderBakker · ‎01-10-2019

The possibilities to automate the process depend highly on the type of information you have available. If the spectral response of a burned house is different from a house that wasn't burned, you could classify both images and compare the results. It would help to have a dataset with the building footprints to limit the comparison to only houses.

Some reading here: Understanding segmentation and classification—Help | ArcGIS Desktop

If that doesn't work, there are ways to use Deep Learning to detect burned houses, but that will be pretty complex.

BrandonPrice1 · ‎01-10-2019

Thanks guys. There is no way to use the difference button on unclassified images in the image analysis window?

BrandonPrice1 · ‎01-10-2019

I don't any image classification training would accurately detect what is burned and what isn't