Select to view content in your preferred language

Extracting parts of raster based on an equation

2652
5
10-10-2014 02:43 AM
NimaSen
New Contributor

Hello everyone,

I have been struggling with a raster dataset that I have and I wanted to ask your opinions.

Basically, I have a raster which is acoustic data and I want to isolate parts of it because certain signals relate to different things. So, for example, for species1, I want the values that are high (approx 200 decibels and higher, light color), but only those that are surrounded by low values (less than 80 decibels, dark color). Also, I want only the ones that form streaky lines, as you can see in the picture. Similarly, for another species, I would like to extract the areas which are dark in color, but surrounded by light colors (animals in the sediment, the sediment also has a similar light color as the first species mentioned). I am attaching an overview so you can see the two types of animals I am trying to differentiate between in this dataset (there are more than two species, but lets start with that).

 

I have tried a number of things, but with no success. I would appreciate it very much if someone has an idea about how to tackle this issue.

Many thanks

0 Kudos
5 Replies
JohannesBierer
Regular Contributor

Create a selection raster and use Extract by mask?

0 Kudos
NimaSen
New Contributor

Thank you, Johannes for your response. I've tried this, but it doesnt actually help. It just helps to extract but I need the extraction to be very specific, not just a range of values or within a certain area, but a range of values based on the neighboring cells.

0 Kudos
JohannesBierer
Regular Contributor

I would select for example 200 decibels and use expand tool to select the under 80 decibel areas in the surrounding.

For streaky lines I have no solution with raster data. So I would propose to convert this areas to Polygons, calculate the shape index an reconvert selectet features with a small shape index to raster.

Sound complicated but isn't that grave ...

0 Kudos
JohannesBierer
Regular Contributor

Shape index is a indicator for the compactness and can be calculated like this:

([SHAPE_Area] * 4 * Atn ( 1 ) ) / ([SHAPE_Length] * [SHAPE_Length])

0 Kudos
DanPatterson_Retired
MVP Emeritus

You have big issues with too much garbage in your image...I would get rid of it first using extractbyAttributes amongst several tools.  Ultimately you want to get down to areas that form the desired shapes you can work with....not necessarily the final list.  For the shapes of the remaining, I would suggest using zonalgeometry with the ellipse option.  From their you can refine your criteria and extract what you want

0 Kudos