I have a multispectral camera with 5 bands (r, g, b, red edge, NIR) and I would like to classify plants object based with Orfeo ToolBox. I am flying with the drone close to the plants and have an image resolution of about 0.01 m per pixel. I guess I have to do a segmentation first, then an object based classification. What steps and models in OTB do I have to follow?
There are four steps to perform an object-based classification in OTB:
1) Image segmentation
note that there is also an external module with the Baatz and Schape criterion : http://tully.ups-tlse.fr/lassallep/grm/tree/master
2) Feature extraction in order to compute object-based variables for the classification
3) training the classifier to build a model based on training samples
4) Applying a model to the feature
Here is a detailed description of single steps: Object-based classification (Tutorial)