Object based image classification using ArcPy or Python

Question

I am looking to classify several Landsat 5 TM scenes using object based image analysis within ArcMap either using a built in tool, writing script in ArcPy, or writing a separate Python code (this is less ideal since I am not sure how to run this code in ArcMap). I have been running pixel level classifications on these scenes but my images are not as accurate as they should be. I am fairly new to writing Python script but feel more comfortable using this language than eCognition Developer (which has poor documentation regarding utilizing the program as a beginner and running this type of analysis).

Do tools exist within ArcMap to separate the scene into objects?

Can ArcPy code be written to segment the scenes?

If so, are there any effective resources out there regarding this?

Answer 1

ArcMap has the Segment Mean Shift tool which requires the Spatial Analyst License. This tool will group pixels into objects that have similar spectral attributes. If you look at the tool overview in the link attached it will give you code snippets to help you implement with ArcPy.

http://desktop.arcgis.com/en/arcmap/10.3/tools/spatial-analyst-toolbox/segment-mean-shift.htm