Author Topic: Classifying the Ortho Photo  (Read 1863 times)


  • Jr. Member
  • **
  • Posts: 59
    • View Profile
    • AccuPixel Ltd - Dealer and Training Centre
Classifying the Ortho Photo
« on: June 27, 2019, 12:11:28 PM »
I'm aware we can classify objects and features in the dense cloud. Automated tools exist in Metashape to automatically identify roads, vegetation etc.

But is anyone aware of a means to identify or mark the same object types/surfaces in the ortho photo?

Basically, I have a massive ortho photo and have lots of objects distributed within and would like to automatically identify their location - perhaps by adding a layer to the image and adding coloured shapes to indicate their location visually - in the ortho photo.

This is, I think, outside of Metashape (for now...) so open to all suggestions.
Agisoft endorsed online Metashape training - see:


  • Jr. Member
  • **
  • Posts: 55
    • View Profile
Re: Classifying the Ortho Photo
« Reply #1 on: July 03, 2019, 08:45:57 AM »
You could use ArcGis Desktop with the spatial analyst extension to achieve this. Create training samples of the various surface types you want to classify then use the maximum likelihood algorithm on the ortho. There is also the SCP plugin for QGIS which can do what you’re after, however I have found the results to be more consistent with ArcGis.

Vaidutis Zutautas

  • Newbie
  • *
  • Posts: 10
    • View Profile
Re: Classifying the Ortho Photo
« Reply #2 on: July 03, 2019, 02:59:02 PM »
Hi Simon,

The possibilities of your wishes very much depend on quite a few parameters, e.g.  the kind of objects/areas you would like to extract and classify from your orthophoto as well as the actual spatial resolution of the imagery and available tools and skills.

Based on these main aspects, you should be looking at two methodological realms, first - land-cover classification (LCCS), which is applicable to various types of areas (i.e. landcover). In most cases such task is achievable by utilizing supervised, non-supervised or object-based classification methods, based on available data and tools. The first two are mainly based on colour information and are available in most of the GIS packages (personal preference is Geomatica PCI), whereas object-based analysis hangs on more complexities and could be achieved in e.g. Trimble's eCognition.

The second method -object extraction- concerns more complex objects in the imagery where not only colour but shape or position play a role in distinguishing particular features. Tasks of such nature require image segmentation and later classification based on available/wished traits.

Although Metashape is not an image-processing software per se in its nature, you could achieve -ish- results of basic features if you establish given classes in the dense point cloud, export it and create class-based footprints in some external software.

« Last Edit: July 03, 2019, 03:01:55 PM by Vaidutis Zutautas »