Structure From Motion to create high resolution point clouds

General discussions of measurement techniques and the results of testing of techniques and equipment.

Moderators: edfrank, dbhguru

Post Reply
User avatar
M.W.Taylor
Site Admin
Posts: 530
Joined: Sun May 29, 2011 11:45 am

Structure From Motion to create high resolution point clouds

Post by M.W.Taylor » Fri Jul 05, 2013 5:34 pm

The attached is an example of a point cloud generated using "Structure From Motion" theory and software from University of Washington. The C++ code is open source. I've posted this example on ENTS before.

In this project a plane orbited around the Fetzer Giant, world's tallest and largest known valley oak. Inside plane was pilot Ben Fetzer and photographer Mike Hanuschik. Mike put his camera in photo burst mode so he created a 100+ panoramic series of images of the Fetzer Oak from 360 degrees in an "orbit" patern around the big oak as the central focal point.

But this could be a UAV doing this. My auto pilot UAVs can be programmed to orbit around a 3D waypoint at point a camera at the waypoint in a locked position. Then do photo-bursting. I'll post results of this test soon.

The attached pictures show the orbit "photo burst" use to create the digital elevation model (also known as a point cloud). You can load the point cloud into AutoCad or a free program such as MeshLab and use the ruler tool to measure every tree, object and structure for size/height.

This is the future of tree measurement. This should be in the measurent group workshop in my opinion.

Michael Taylor

WNTS VP
California Big Trees Coordinator
http://www.landmarktrees.net
Attachments
resultant point cloud of orbit pattern photo-burst for the giant Fetzer Oak. You can see the grand oak being measured at 150'+ using the Meshlab ruler tool
resultant point cloud of orbit pattern photo-burst for the giant Fetzer Oak. You can see the grand oak being measured at 150'+ using the Meshlab ruler tool
orbit photo-burst of Fetzer Oak to create lawyered 3 dimensional point cloud
orbit photo-burst of Fetzer Oak to create lawyered 3 dimensional point cloud

User avatar
mdavie
Posts: 150
Joined: Sun Mar 14, 2010 6:51 pm

Re: Structure From Motion to create high resolution point cl

Post by mdavie » Sat Jul 06, 2013 6:35 pm

This is the future of tree measurement. This should be in the measurent group workshop in my opinion.
Absolutely. This is great stuff!

User avatar
dbhguru
Posts: 4464
Joined: Mon Mar 08, 2010 9:34 pm

Re: Structure From Motion to create high resolution point cl

Post by dbhguru » Sat Jul 06, 2013 9:40 pm

Michael,

It will be exciting if American Forests makes room for advanced methods of tree measurement. I'll do my best to create some space for methods such as you are developing, but I think you know the inertia that I will encounter. Adopting such advanced methods would represent more than simply pushing the envelope. It would represent sending the envelope on a journey around the world. We have a good team though. One of the members of the group is an associate professor of forest biometrics, and he does know his stuff. There is stuff brewing with the group that I think you would approve of.

However, I'm not going to win on every issue. I will have to compromise at points along in some of the processes.

Bob
Robert T. Leverett
Co-founder, Native Native Tree Society
Co-founder and President
Friends of Mohawk Trail State Forest
Co-founder, National Cadre

User avatar
Jess Riddle
Posts: 440
Joined: Sun Mar 14, 2010 10:59 am

Re: Structure From Motion to create high resolution point cl

Post by Jess Riddle » Sun Jul 07, 2013 4:33 pm

Wow!

I never thought techniques like this would be available so soon. I had assumed technology like this was at least a decade off.

Jess

User avatar
Don
Posts: 1560
Joined: Tue Mar 09, 2010 12:42 am

Re: Structure From Motion to create high resolution point cl

Post by Don » Sun Jul 07, 2013 9:05 pm

Jess-
One of my fellow Alaskans, a friend, and a lurker (Ken Winterberger) on our forum here has been investigating use of SfM for future natural resource inventories in Alaska. He believes is has a lot of potential and could be a natural pairing with LiDAR. Much like LiDar and satellite imagery can be paired.
And certainly Michael Taylor has gotten with the program!.
-Don
Don Bertolette - President/Moderator, WNTS BBS
Restoration Forester (Retired)
Science Center
Grand Canyon National Park

BJCP Apprentice Beer Judge

View my Alaska Big Tree List Webpage at:
http://www.akbigtreelist.org

User avatar
M.W.Taylor
Site Admin
Posts: 530
Joined: Sun May 29, 2011 11:45 am

Re: Structure From Motion to create high resolution point cl

Post by M.W.Taylor » Fri Jul 19, 2013 11:30 am

Here is another example of a point cloud taken of a forest generated by one of my UAVs using photo-bursting. This particular redwood forest is too remote for me to reach on foot (it would take 3 full days to get there and back). But with the UAV it only took 25 minutes to explore the remote basin. The photo-burst is programmed to activate at specific waypoints in the flight plath. The duration, number of pictures and coverage distance of the photo-burst is completely programmable. Once sccaled, every object in the point cloud can be measured. The attached is the raw point cloud, unprocessed. You can manipulate in Meshlab as a 3D graphic.
Attachments
resultant point cloud of the photo-burst
resultant point cloud of the photo-burst
overhead photo burst of remote redwood forest to measure canopy height above ground
overhead photo burst of remote redwood forest to measure canopy height above ground

User avatar
yofoghorn
Posts: 103
Joined: Tue Sep 04, 2012 12:09 pm

Re: Structure From Motion to create high resolution point cl

Post by yofoghorn » Tue Oct 01, 2013 3:48 pm

Here is a video of a mesh that I did on the Auto Tree in Big Basin that used a point cloud approach. Calibrating this point cloud will allow the most accurate volumetric measurement of this tree and others. Measuring volume on large trees, especially redwoods, will show that their trunks are no where near round at their base. Usually volume over estimation occurs. This method, as developed by Michael Taylor, seeks to eliminate additional volume that is measured but does not exist via the tape diameter "conventional" method (i.e. discrepancies from nooks and crannies that aren't actually part of the trunk).

Bob VanPelt has implemented this technology on the Redwood Climate Change Initiative that was recently completed.

This method is also being used by Tom Stapleton and me for chimeric redwood cuttings he is growing in his greenhouse. We can measure the surface area and ratio of green to white tissue on the chimeras to better understand photosynthesis and the stresses that the plants endure. This is also an important area of research when it comes to redwood climate change, as albinos redwoods can arise due to a stress-related mutation. Studying how these intimate relationships interact with each other might tell us more about redwood climate change and stress than we ever thought possible. This technology I'm sure has other applications, so perhaps brainstorming about "what we can do next" might be the next best thing to think about!

But alas, here is the link to the video:

http://www.youtube.com/watch?v=eh1hjdl7BHs
Zane J. Moore
Plant Biology PhD Student
University of California, Davis

User avatar
M.W.Taylor
Site Admin
Posts: 530
Joined: Sun May 29, 2011 11:45 am

Re: Structure From Motion to create high resolution point cl

Post by M.W.Taylor » Sun Oct 27, 2013 11:18 pm

Zane,

Do you have the min and max basal diameter for that point cloud ? I need some scale to do the volume calculation.

That new mesh overlay looks fantastic !!!

Michael Taylor

yofoghorn wrote:Here is a video of a mesh that I did on the Auto Tree in Big Basin that used a point cloud approach. Calibrating this point cloud will allow the most accurate volumetric measurement of this tree and others. Measuring volume on large trees, especially redwoods, will show that their trunks are no where near round at their base. Usually volume over estimation occurs. This method, as developed by Michael Taylor, seeks to eliminate additional volume that is measured but does not exist via the tape diameter "conventional" method (i.e. discrepancies from nooks and crannies that aren't actually part of the trunk).

Bob VanPelt has implemented this technology on the Redwood Climate Change Initiative that was recently completed.

This method is also being used by Tom Stapleton and me for chimeric redwood cuttings he is growing in his greenhouse. We can measure the surface area and ratio of green to white tissue on the chimeras to better understand photosynthesis and the stresses that the plants endure. This is also an important area of research when it comes to redwood climate change, as albinos redwoods can arise due to a stress-related mutation. Studying how these intimate relationships interact with each other might tell us more about redwood climate change and stress than we ever thought possible. This technology I'm sure has other applications, so perhaps brainstorming about "what we can do next" might be the next best thing to think about!

But alas, here is the link to the video:

http://www.youtube.com/watch?v=eh1hjdl7BHs

User avatar
Don
Posts: 1560
Joined: Tue Mar 09, 2010 12:42 am

Re: Structure From Motion to create high resolution point cl

Post by Don » Mon Oct 28, 2013 1:42 am

Zane/Michael-
Remote sensing-wise, I'm pretty old-school...so I am assuming that you guys are familiar with the use of different wavelengths (such as long used in Landsat) to measure moisture stress. There are landbased, handheld versions that use these and other wavelengths to 'groundtruth' satellite imagery. Could these be used in conjunction with your cloud mappers, either from the ground or from quadcopters, etc?
-Don
Don Bertolette - President/Moderator, WNTS BBS
Restoration Forester (Retired)
Science Center
Grand Canyon National Park

BJCP Apprentice Beer Judge

View my Alaska Big Tree List Webpage at:
http://www.akbigtreelist.org

User avatar
Don
Posts: 1560
Joined: Tue Mar 09, 2010 12:42 am

Re: Structure From Motion to create high resolution point cl

Post by Don » Sun Feb 16, 2014 8:13 pm

Peter-
First off, welcome to the forum, you're sure to find friends and interest in your work here. I meant to reply sooner (got busy...) but was reminded of your post in a discussion of 'how do we find parity in measuring single-stem tree versus multi-stemmed tree (a discussion fraught with peril in some circles, including this one...: > ) and am only now getting back to you.

I'm aware of the techniques such as Structure from Motion as a means of producing 3D reproductions, and am absolutely 'gobsmacked' by the graphics packages available to view them (for example Zane Moore's URL [http://www.youtube.com/watch?v=eh1hjdl7BHs]). I can't even imagine a better way to obtain biomass volume of a 'hollow' tree!

And while the algorithms that it takes to slice the tree, perpendicular to it's longitudinal axis (I'll define that as an estimate as to the pith line) at 4.5' above it's mid-slope base, and determine that cross-sectional area and circumference are do-able, the average forest technician or big tree enthusiast that runs a D-tape and clinometer is likely to be rather intimidated by the complexity of these SfM 3D reproduction techniques.

Where this approach really could shine is with automated solutions ('cloudbursts' from quadcoptered drones obtaining SfM 3D imagery) and user-friendly 'canned' algorithms that provide accurate tree heights/volumes, diameters/circumferences/cross-sectional areas at breast height, crown spreads/volumes, with little more than 'button-push' tasking.

Perhaps you can tell us how you gathered the data that created the image you've included in your post? SfM? Aerial platform (radio-controlled model planes/quad-copter type photogrammetry)? Ground-based stills?

-Don
Don Bertolette - President/Moderator, WNTS BBS
Restoration Forester (Retired)
Science Center
Grand Canyon National Park

BJCP Apprentice Beer Judge

View my Alaska Big Tree List Webpage at:
http://www.akbigtreelist.org

Post Reply

Return to “Measurement and Dendromorphometry”