0

I've tried using the classifiers in ArcMap which only supports one ancillary dataset and I tried using Orfeo in QGIS 2.18, though the Orfeo Image Classification tool does not accept any ancillary raster datasets. I can create a training model using multiple bands and images in Orfeo but Image Classification only accepts one raster file so I can't feed it the extra ancillary data I used during the training.

Besides the original satellite image, I want to add in NDVI, NDWI, and DEM's to help the classifier just so you guys have an idea of what I'm trying to do. I haven't come across something like this besides creating a completely custom classifier.

How do I Classify With Multiple Ancillary Datasets using QGIS?

EDIT: I've figured it out. I just make a new composite raster image which includes the ancillary data as additional bands or create a virtual raster catalog. This way I can use the new composite raster with classifiers in any Remote Sensing software which often accepts only 1 raster file at a time. I've already tested this with my low spectral resolution data and it helped my classification output by adding NDVI and a NDWI as extra bands.

Esparko
  • 31
  • 5
  • Maximum Liklihood in ArcGIS allows for the use of multiple bands: http://desktop.arcgis.com/en/arcmap/10.3/tools/spatial-analyst-toolbox/maximum-likelihood-classification.htm. Also, I would recommend performing the classification using R, which allows you tremendous freedom in setting up your parameters. Here is something to get you started: https://gis.stackexchange.com/q/39021/8104. – Aaron Jan 06 '19 at 15:47
  • I reduced the scope of this question by retrofitting to its first answer. If you also wish to ask how to do it using ArcGIS Desktop then please do that in a separate question. – PolyGeo Jan 07 '19 at 01:00
  • I just realized, is it a sensible thing to include supplementary data as additional bands in Remote Sensing Data then do that in MLC? @Aaron

    For example I have a 4 band image then I just create a new composite image with NDVI and Radar as band 5 and 6 respectively?

     – Esparko Jan 07 '19 at 10:32
  • @pbroto You should be able to include the individual bands including NDVI and radar without creating composite bands. – Aaron Jan 07 '19 at 13:40
  • Ah I see now, thanks for the tip. Though from what I can tell this is only available for the MLC classifier in ArcMap? The other methods such as Random Forest/SVM or classifiers in other software might have to use a composite image of bands and ancillary data by the looks of it. – Esparko Jan 07 '19 at 15:17

1 Answers1

1

Unless I am misunderstanding your needs, I can recommend the use of Semi Automatic Classification plugin in QGIS, which is actively developed and very well documented, and in which a complete dataset composed of any amount of bands of any number of files works as input for the classification.

PolyGeo
  • 65,136
  • 29
  • 109
  • 338
Gabriel De Luca
  • 14,289
  • 3
  • 20
  • 51
  • Thanks for suggesting this plugin, it's my first time using it and the classification process is interesting. What I wanted to do is combine RGB+NIR bands for bands 1-4. I'm trying to experiment adding ancillary data to the classification methodology by adding it as a raster band. For example, NDVI values for band 5 and Radar for band 6. – Esparko Jan 07 '19 at 15:21
  • @pbroto, the result of a classification is a shapefile with the polygons resulting from the classification. The origin is a dataset of bands of one or more files, it is not necessary that all data be in a single file. Maybe what you're trying to do is create a virtual raster catalog before? – Gabriel De Luca Jan 07 '19 at 15:55