Hi
My goal is to distinguish forest tree species (in dense forests) with drone images. I did not get acceptable results using spectral features. Because the pictures were taken in the fall and the trees have taken on a lot of colors. I have decided to do this using the texture features in thee Cognition software. There are about five tree species in the area. Can you guide and explain the steps?
If you have another method, please?
Thanks a lot.
Hi Mohammadhassan Naseri, since you appear to be detecting multiple trees on one image, it would make sense for me to look into image segmentation models.
A college of mine successfully implemented UNets to detect houses on satellite images (similar project: https://www.ibm.com/cloud/blog/the-xview2-ai-challenge). Once the image is segmented, the ROIs could be further classified to tree type.
Cheers,
Raoul
Maybe you could also try to pretrain a very shallow autoencoder (like a residual network approach) with your dataset. This might be enought for a simple feature extraction. Anyway, if the dataquality is very bad, it becomes hard to get anything meaningful out of it.
Hi,
Seems like a very good task for a Deep Learning model. Already tried with a CNN?
https://towardsdatascience.com/wtf-is-image-classification-8e78a8235acb
- Badges
- Science topic
- Similar topics
- Data Mining
- Classification