Obia classification steps
Following #133 (closed)
Branch available here
After obia learning steps, the next steps are the classification step, and then I also add a additionnal step to reassemble subtiles (from "split segmentation by tiles and regions" step) and get the same pattern file as in the original workflow. Then the processing chain continue with the mosaic step, etc, until the end.
So there is two new steps:
- Obia classification
- Reassemble parts
1. Obia classification
Do the classification using otbVectorClassifier. Outputs are saved in classif folder. The data table contains the predicted class and the confidence score. I choose to store the output files (one per subtiles) in tile subfolder (ex: classif/T38KPD, classif/T38KPE etc) to keep the main folder tidy.
2. Reassemble parts
Per tile, every parts are rasterized two times, one with the class value, one with the confidence value, according to the segmentation raster spacing. Then it merge classif rasters to get one raster per tile,model and run (ex: Classif_{tile_name}_model_{nb_model}seed{nb_run}. tif), and merge confidence rasters in the same way (ex: {tile_name}_model_{nb_model}_confidence_seed{nb_run}.tif). These files have the same patterns that in the originial workflow, so they are found for next steps.