I would like to perform supervised classifications in R on a large dataset of photographs to assess land cover change. Has anyone attempted this for landscape photographs in R? And is there value in using the random forests model process for such image classifications? Thanks!
From my experience of handling classification problems, I would recommend you to go with K-nearest neighbors classification. Trying different distance functions in this regard can come up with interesting insights.
Besides the aforementioned K-NN classifier, I would also suggest a MLC (Maximum Likelihood Classifier). A meta algorithm, as ADABoost [1] can also an option.
As the classification is intendend to be done in R, RStoolbox [2] is probably a good starting point, as it is devoted to remote sensing image processing and analysis, and provides several tools for supervised classification.
[1] Article A Decision-Theoretic Generalization of On-Line Learning and ...
[2] https://cran.r-project.org/web/packages/RStoolbox/index.html
Convolutional Neural Networks are the best for image classification (as Saied also mentioned). You can use Keras with TensorFlow - eay to use and powerful - implemented in R (see https://keras.rstudio.com/).
- Badges
- Science topic
- Similar topics
- Computer Science
- Images