4

I am working on a project where we have made a RGB orthomosaic using a drone which I want to classify using QGIS. I am not a fan of unsupervised classification, and I started exploring the possibilities of object-based classification. I tried the trial of eCognition; just a wonderful workflow.

Sidenote:
the rasters pixels are 1.5 cm x 1.5 cm. The total area is around 3 hectares. So a lot of data. I exported the raster to a raster pixel width/height of 5.0 x 5.0 cm,

Target:
classify the image using object-based analysis of the (I suspect) the Orfeo toolbox

Workflow: segmentation (meanshift) OTB:

processing.runalg("otb:segmentationmeanshift","G:/Veld1/veld1_5cm.tif",0,5,30,0.1,100,100,0,0,"G:/Veld1/veld1_5cm.tif",True,True,100,0.1,"layer","ID",512,1,"None","G:/Veld1/SEGMENTATION/segm_5cm_tile512,size1000.shp") 

This works fine. I want to classify the image using the polygons from the segmentation mean-shift; just like in eCognition.

Maybe make a sort of ruleset like an ROI in the semi-automatic classify plugin for QGIS? Steps:

  • add fields and rule polygons in certain classes
  • classifier

Does anyone know a good workflow? To give a feeling for the data; I uploaded an image with the image and the segmentation output.

segmentation output

  • Hello Chris. I wondered if you had made any progress with this and found a workflow that works for you? – JPD Sep 15 '16 at 15:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.