Using existing feature class for training samples--not working

1455
2
Jump to solution
12-03-2022 09:31 AM
Labels (1)
AnninaRupe1
New Contributor III

Hello all!

I am trying to perform an image classification using a supervised, pixel-based configuration. I'm using ArcPro 3.0.2

I have a classification schema and training samples dataset created. Here is the schema: classname: Marsh, classvalue: 0; classname: Water, classvalue: 1. I have created these fields in my training samples dataset; classname as text and classvalue as long integer. The dataset was created by creating random points and then two people interpreted the imagery to classify the points as marsh or open water. This was done for classification accuracy purposes. It was my understanding that these data could be input into the image classification workflow as the training samples rather than manually drawing them in the moment.

In the Image Classification Wizard, I can get to the Training Samples Manager section, but can't get any farther. Here is the issue: the schema loads great but my training samples dataset doesn't show up in the bottom half of the window. There is no place even to manually load the data. I've also tried accessing the Training Samples Manager tool by itself (outside of the Image Classification Wizard) but am running into the same issue. I can navigate to my training samples data set and press open, but nothing happens. I have tried point and polygon layers here.

AnninaRupe1_0-1670088364429.png

 

0 Kudos
1 Solution

Accepted Solutions
AnninaRupe1
New Contributor III

In case anyone else has this issue...

The polygon must surround an entire pixel for this to work. So, what I did was create a new feature class, zoom to each point, and create a square around the pixel that the point was in. 

In future iterations, I'll probably buffer the point more so I won't have to manually create the training set.

View solution in original post

2 Replies
AnninaRupe1
New Contributor III

In case anyone else has this issue...

The polygon must surround an entire pixel for this to work. So, what I did was create a new feature class, zoom to each point, and create a square around the pixel that the point was in. 

In future iterations, I'll probably buffer the point more so I won't have to manually create the training set.

annierosen
New Contributor

In case anyone runs into this problem again (I did), you should also be able to use the buffer tool to create polygons without drawing them manually.

0 Kudos