Hi
I am working with eCognition Developer software. resently, I used it for UAV imagery segmentation. This image is for the fall season and it has 4 bands( Green, Red, Reg and NIR).
My purpos is classify trees speces of forest dense. But, classification does not done well. becuse, each tree is classified into several classes. My openion is segmentation does not done well. I tried lots of for improved segmentation, But I did not achieve acceptable results.
Do you have an opinion on this?
Thank you all dear ones
You could try using Unets for image segmentation: Article U-Net: Convolutional Networks for Biomedical Image Segmentation
That should give you better performance and pixel-by-pixel image segmentation. I think that would be very useful for classifying trees in your project.
Cheers, Raoul
hi.. you did not mention which platform you have used. However, for segmentation, training samples size and purity as well as parameteres also take a vital role. Try to fine tune the parameters and training samples.
Can you post a sample image to see resolution, contrast, etc. and look at image quality?
- Badges
- Science topic
- Similar topics
- Geoscience
- Remote Sensing