Forum

Author Topic: Problem with Dense Cloud Classification  (Read 5371 times)

Mohabon

  • Newbie
  • *
  • Posts: 48
    • View Profile
Problem with Dense Cloud Classification
« on: June 23, 2014, 10:14:58 AM »
Hello,

i would like to receive some feedback by expert users about a dense cloud classification tool on crop fields (as far as i know the worst possible scenario to apply the algorythm).

The crop field area i want to classify is 250m by 150m. The crop is 1.5m tall from ground level since it is june and it is almost ready to be cut! See "Crop field.jpg"

 I used standard parameters like 8.5°, cell=100m and height = 1m in the classify ground point.

As you can see, there also some "trenches" which are supposed to help the algorythm.

After using the algorythm, i find a very strange result: trenches are classified as "vegetation", while crop is classified as ground, as you can see in the other jpg.

Question 1:How can this happen? Since ground is the lowest level, shouldnt this problem never happen?

Question 2: I have manually classified the crop field as "Low vegetation" and i have re-run the Dense Clould Classify tool. I was very careful not to select the trenches.
I thought that this "boost" would have helped the algorythm in refining the classication and maybe sort the problem. I have reapplied the classification tool and i got same results: looks like that manual intervention is ignored by algorythm. Is this confirmed? Can it be changed by any chance? It would improve a lot the results.

Thanks



« Last Edit: June 23, 2014, 10:19:02 AM by Mohabon »

3DFranz

  • Newbie
  • *
  • Posts: 44
    • View Profile
Re: Problem with Dense Cloud Classification
« Reply #1 on: June 23, 2014, 11:27:03 AM »
Hi Mohabon,

you have used wrong parameters.
Take a look at : http://www.agisoft.ru/tutorials/photoscan/08/

The second Parameter (Max disctance (m) ,..dertmines the assumption for the maximum cariation of the ground-ELEVATION at a time. 1m is for your purpose too much. Try 0.5 or something like this.
The Third parameter ..Cell size with 100m is also too high.

In case the result of automatic dense cloud classification is not acceptable the procedure can be re-run using adjusted parameters (for example, if some on-ground objects like stones and small bushes were classified as ground points it is reasonable to reduce Max angle and Max distance parameter values).

Note: It is important to reset the assigned classes for the dense cloud prior to starting the procedure again with the new set of parameters.

Hope that is a useful information for you.
Best regards

Mohabon

  • Newbie
  • *
  • Posts: 48
    • View Profile
Re: Problem with Dense Cloud Classification
« Reply #2 on: June 23, 2014, 03:56:44 PM »
Hello 3DFranz and thanks a lot for your interest.

I would also appreciate some extra information about the parameters you suggest in the scenario i described = think crop field as tall as 1.5m with some trenches inbetween.
You suggest H=0.5, but what cell size? and Angle?

And at last: howto reset the dense cloud classification? I cannot see a reset button or anything related with classes in the project.

Kind regards,

MoH

stihl

  • Sr. Member
  • ****
  • Posts: 410
    • View Profile
Re: Problem with Dense Cloud Classification
« Reply #3 on: June 26, 2014, 10:47:08 AM »
Hi Mohabon,

I've been playing around with this tool as well. From I know the angle isn't that important in your particular case. If I were you I'd try different values for the cell size and height. It takes some times to get a feel what they all do. For me it's usually a hit or miss, but 'fine tuning' it helps.

As for resetting your ground points before trying new parameters; select ?ll the points with a selection tool (I usually use the circle tool) and go to Tools > Dense cloud > Assign class > to 'Created (never classified)' .. this resets the filter and allows you to try new parameters.