I want to know the complete step to perform the validation of segments obtained from the OBIA in e-Cognition or in ArcGIS if possible, of LISS-IV image, i.e., to find out the "goodness of fit", i.e., D-value of maximum and minimum over lap of segments.
I have exported the shapefile of segmented layer and also made the reference layer.
ROC curve
Go through this link.
https://www.int-arch-photogramm-remote-sens-spatial-inf-sci.net/XLI-B7/241/2016/isprs-archives-XLI-B7-241-2016.pdf
confusion matrix
This is easier with ENVI ZOOM or ENVI 5.3.
1- Image Segmentation
2. Draw the correct parts from the image
3- Use the confusion matrix to extract the segmentation accuracy
You can also use the software help.
Object based validation should at least provide 2 information. Classification's accuracy as well as segmentation accuracy in terms of how well the generated boundary fitted with the reference data.
I've used area based similarity measurement described by Tim Whiteside,
Article Area-based and location-based validation of classified image objects
it is easier to understand and very well explained.
You can also check my paper applying such method, through this link.
Article A support vector machine object based image analysis approac...
As mentioned by Zylshal , you need to perform an area-based comparison of the segmented results with your reference data. If you have multiple classes, you can possibly create thematic maps for each class and compare it with corresponding reference data.
Chills
- Badges
- Science topic
- Similar topics
- Computer Science
- Images