Extract types of segments from Raster-Polyline layer

1094
5
07-26-2019 09:17 AM
DanielBrenner1
New Contributor III

See attachment.

I have a fairly large Polyline layer of Highway Ramps that has a high amount of linework errors (missing segments) that are showing up in the ESRI Base Map Imagery layer.  Instead of manually comparing the Polyline (ramps) layer against the Raster layer), I was thinking of converting the Raster layer into Polylines, extracting ramps from this polyline layer, and then comparing the new Polyline layer against my ramps layer to indicate where the missing ramps are.  

When I was doing Raster-Polyline, I was guessing working with Environment Settings might help but I didn't see anything that would help classify the types of routes in the imagery into polylines (i.e., roads versus ramps).

Is there a way to extract types of segments from a Raster-Polyline layer?  

(If you wanted to play with this, think of how Image Classification gives us a Number of Bands (R,G, and B); why not have a number of types of features of certain feature classes?. i.e., Polylines:  Highways, Ramps, Streets) 

Thanks.

0 Kudos
5 Replies
DanPatterson_Retired
MVP Emeritus

An 'image' (aka, picture) knows nothing about its contents.  You have to perform 'image classification' to derive a classification scheme if possible.

Overview of image classification—ArcGIS Pro | ArcGIS Desktop 

0 Kudos
DanielBrenner1
New Contributor III

Yes I will try imagery classification but how would this separate out

pixels that equal the shape of a ramp? Wouldn't there need to be some cell

attribute values based on certain conditions associated with imagery

patterns equal to a ramp?

0 Kudos
DanPatterson_Retired
MVP Emeritus

For sure... pattern recognition and spatial association are all parts of certain types of image classification

Understanding segmentation and classification—Help | ArcGIS Desktop 

You have to weigh the number of times that you need to do this versus the time you need to invest in learning the methodologies and interpreting them.  Also, human-ware is a pretty fast and accurate classifier, even if you had a couple hundred off-ramps to digitize

0 Kudos
DanielBrenner1
New Contributor III

This isn't a matter of digitizing; its a matter of creating a layer from

the raster to compare a pre existing layer with and determining where the

gaps are in the pre existing layer.

Part of the issue I can see is that asphalt can create x type of pixel

value whether its a ramp or a parking lot. How do I indicate before the

classification that I need asphalt cells = ramps?

0 Kudos
DanPatterson_Retired
MVP Emeritus

You can't by just the spectral response associated with asphalt.  That is where the other classification methods come into play.. ie. the pattern.  All you can do with simple image classification is separate asphalt from grass and other surface types.

0 Kudos