Forum

Author Topic: Classify Ground Points Settings  (Read 4952 times)

RowleyCW

  • Newbie
  • *
  • Posts: 5
    • View Profile
Classify Ground Points Settings
« on: November 19, 2019, 01:45:04 PM »
Hi,

I'm having fun with the Classify Ground Points function and I'm hoping someone can give me a steer in the right direction.

As you'll see from the attached screenshot, it's classifying the first meter or so of the cars, trees and buildings from the ground up, as ground.

The settings i'm using are standard, ie:

Max Angle - 15ยบ
Max Distance - 1m
Cell size - 50m

I'm guessing from reading the manual, that the best place to start is to reduce the Cell size, but any further guidance would be welcome.

Thanks,

Alexey Pasumansky

  • Agisoft Technical Support
  • Hero Member
  • *****
  • Posts: 13729
    • View Profile
Re: Classify Ground Points Settings
« Reply #1 on: November 19, 2019, 02:41:07 PM »
Hello RowleyCW,

I can suggest to reduce max angle and max distance values, for example, to 5 degrees and 0.5 meters and check if it helps to avoid misclassification of the points (you need to reset the previous activation results though).
Cell size 50 meters looks fine here.

Also you can try the ClassifyPoints operation and enable all the available classes and compare the classification results to the Ground Points classification.
Best regards,
Alexey Pasumansky,
Agisoft LLC

toxicmag

  • Jr. Member
  • **
  • Posts: 87
    • View Profile
Re: Classify Ground Points Settings
« Reply #2 on: January 05, 2020, 05:57:45 PM »
Also you can try the ClassifyPoints operation and enable all the available classes and compare the classification results to the Ground Points classification.

Dear Alexey,

what about this feature request (maybe it's already possible but i do not see it)   => 

For the Ground Classification Tool you always define a FROM-class and a TO-class. What about adding "selection" into the FROM-class pulldown menue and let the classification only take part within the selected points? It'll help a lot in manual classification work. As of today the process always runs through the whole densecloud. Even if you do a "manual fake definition" (like i.e. rails) for a small area and work the ground classification process then FROM "rails" to "ground points" it goes through the whole dataset and takes too long.
#multirotor #VTOL #Trinity #Quantum-Systems #a6000 #a7R #RedEdge #ALTUM #RX1II