Forum

Author Topic: Point cloud density over trees  (Read 8933 times)

Radbert

  • Newbie
  • *
  • Posts: 9
    • View Profile
Point cloud density over trees
« on: May 21, 2015, 07:19:09 PM »
Hi -

I am wondering if anyone here might be able to help explain a counter-intuitive result I obtained when using different qualities of dense point clouds for 3D mapping over forested vegetation. 

My sparse point cloud from 80% forward and 70% side overlapping UAV photos is as expected least dense over areas with trees (likely due to both less overlap with taller objects plus a 3D branch structure where potential tie-point targets shift rapidly with small changes in perspective).

When I create low, medium, and high quality dense clouds from this, I am surprised to see that the high quality dense cloud has the lowest density of points over treed areas despite having a much larger number of total points.  Does this have something to do with the precision requirements for mapping points at different qualities? (e.g. more points are mapped for trees in the low-quality cloud because their location has a lower requirement for matching points among photos)?

Pdf file is attached showing these outputs.

Thanks,
Rob

aggieair

  • Jr. Member
  • **
  • Posts: 93
    • View Profile
Re: Point cloud density over trees
« Reply #1 on: May 21, 2015, 08:11:51 PM »
I get the same results Rob - Lowest actually populates the trees but sparsely.  Posting to follow along for more info!

When I make orthophotos, I actually classify all the ground points and make my mosaic on that so not including the trees is OK for those types of orthophotos (which will eliminated the wavy or drop off appearances with tress/buildings).

StevenF

  • Newbie
  • *
  • Posts: 39
    • View Profile
Re: Point cloud density over trees
« Reply #2 on: May 22, 2015, 01:03:28 AM »
I get the same effect. I doubt it's because of any precision requirement in the software. The quality setting specifies the image resolution used during dense matching: Ultra = original pixel, High = 1/4 original, Medium = 1/16 original, etc. At a fine scale features may appear very different in two images, but when you coarsen the image those features are likely to appear more similar in the separate images because you've reduced the local variability.

The result of using a lower quality setting on rough surfaces like tree canopies is a lower point density (because there are fewer pixels to match), but you often a greater coverage of points (because it's easier to find matches). Higher quality setting are likely to yield more accurate point locations which can be important if you want to get accurate height measurements when working with coarse imagery. For UAV applications where the approximate ground sampling distance is less than 10cm this is probably less important, but when working with imagery that has a gsd >30cm I've found my height measurements for trees are significantly more accurate with Ultra quality.

What still baffles me is that I often get better coverage with Aggressive filtering. I wish someone would explain the filtering settings to me more clearly.

Radbert

  • Newbie
  • *
  • Posts: 9
    • View Profile
Re: Point cloud density over trees
« Reply #3 on: May 22, 2015, 03:24:42 AM »
Thanks Steven - that is a very helpful response and makes a lot of sense.  Perhaps one could merge two point clouds based on a height criterion  - using a low quality cloud for trees and high quality cloud for everything else.

Rob

bigben

  • Sr. Member
  • ****
  • Posts: 406
    • View Profile
Re: Point cloud density over trees
« Reply #4 on: May 22, 2015, 07:38:10 AM »

What still baffles me is that I often get better coverage with Aggressive filtering. I wish someone would explain the filtering settings to me more clearly.

I haven't done any searching on this yet, but yes, I think I need a better understanding of what it does and why I'd use certain settings.  Not sure if it's new (don't remember seeing it) but I just tried with filtering Disabled and got a pretty complete dense cloud with a few odd patches of noise. Previous attempts had practically no forest on the hills in the middle.

[edit]Forgot that I'd also used a lower quality setting for that dense cloud, which accounted for fewer holes.  Running some comparisons at different settings to suck it and see.
« Last Edit: May 22, 2015, 10:42:44 AM by bigben »

Radbert

  • Newbie
  • *
  • Posts: 9
    • View Profile
Re: Point cloud density over trees
« Reply #5 on: May 22, 2015, 05:51:10 PM »
What still baffles me is that I often get better coverage with Aggressive filtering. I wish someone would explain the filtering settings to me more clearly.

This is an interesting and valuable observation.  I just tried aggressive filtering on my low quality cloud and got much better point cloud coverage over trees compared to mild filtering (see pdf).  If I disable filtering, the cloud coverage looks almost as good at first glance, but not after removing all the noise lying below the ground surface.

jmos

  • Newbie
  • *
  • Posts: 16
    • View Profile
Re: Point cloud density over trees
« Reply #6 on: December 17, 2015, 03:24:24 AM »
Hi,
just wondering if there has been any clarification or improvements on creating detailed Point cloud with no holes?
Using Version 1.1.6 build 2038

Been experimenting and I also have had best results with Aggressive depth filtering and medium or low Quality
Tie Points set to 10000, GCP's only accurate to 2m.
I did try No depth filtering, you get everything, not yet tried cleaning it up yet.

Regards

loda5678

  • Newbie
  • *
  • Posts: 1
    • View Profile
Re: Point cloud density over trees
« Reply #7 on: December 17, 2015, 06:44:45 AM »
Thank you

Artisan S

  • Newbie
  • *
  • Posts: 31
    • View Profile
Re: Point cloud density over trees
« Reply #8 on: December 17, 2015, 11:14:36 PM »
Interresting......now I know what those quality indicators mean, I guess that figures, I've seen weird results as well in some of my scans. This explains it.....thanks.

Greets, Ed.

jmos

  • Newbie
  • *
  • Posts: 16
    • View Profile
Re: Point cloud density over trees
« Reply #9 on: January 15, 2016, 04:31:51 AM »
Had another go with some new imagery at 10cm GSD, quite happy with the results and no holes in Dense Cloud.
Previous Images processed were 17cm GSD, less GCP's and less image overlap.

Processed at High/Aggressive, ran out of memory when trying Ultra High.

« Last Edit: January 15, 2016, 04:38:52 AM by jmos »

jmos

  • Newbie
  • *
  • Posts: 16
    • View Profile
Re: Point cloud density over trees
« Reply #10 on: January 15, 2016, 04:33:29 AM »
See report Attached