DTM and DSM for forestry

[fpbv]

Hi guys

I have this situation.
My client has a large forestry areas and they want the DTM of it.
I show them the DSM that PhotoScan did in some photos we had in other forestry areas and they liked.
But they want to have the DTM beyond the DSM, because they need to extract the countour lines on the ground, this is more important to them.
So anyone had some idea to do it?
I know aerial laser scan do this but I don?t know how to extract this using PS.
I figured it could to apply mask on the forestry areas (pinus and eucalyptus) and generate the DTM on the ground, but it is quite time consuming.

[gEEvEE]

Well, you just mentioned one of the main reasons of the existence of airborne laser scanning: the ability to filter away the first laser pulses and obtain – if the density of the vegetation and amount of laser pulses/m squared allow – a DTM. Obviously, you can not do this with passive imaging. Almost 100% of a pixel’s contribution comes from the canopy, so there is no way of extracting information on the surface below the canopy. The only thing you can do is – as you mentioned – delete forest zones and interpolate between them. However, do not expect an accurate result in case of extended tree areas. It could approximately work when dealing with single trees. As said: these are the limits of passive imaging and indicates the reason why a discussion such as “What is best: photogrammetry, computer vision or laser scanning?” is ridiculous (although forums and the scientific world are full of them). They all have pro’s and con’s and each of them has its specific application field.

Geert

[Diego]

Of course, time consuming, this is photogrammetry, you need a software like terrain_editor, lidar filters do not work very well, since there is no penetration into the soil of the land.

The problem is that the correlation, data are obtained very fast, but they are useful to generate contour mapping field, you need to edit stereoscopically, as is always the photogrammetric workflow, or better yet, capture the contour lines manually. It is as accurate, time consuming, but ultimately is what the customer needs.

Regards,

Diego

[Wishgranter]

here is a great example for procesing of forest data with lidar. gEEvEE mentioned it precisely, for every use is a specific technology because of specific limitation and trees are my own biggest problem too - Ruins of castles and etc.. Because is problematic to cut the trees, we can shoot when vegetation is not present.......

http://www.youtube.com/watch?feature=player_embedded&v=tiSyA6_KmYs

[Diego]

Okay, lidar technology helps. But before there lidar technology, as it could generate curves in densely populated or high vegetation, for it is the photogrammetry, what happens is not always automatic, but tracing the curves no problem.

[fpbv]

Guys

First of all, thanks for reply.
I already worked with lidar data for forestry and I know how good this is.
I would expect some kind of filter to do it like lidar does, some software to do this DTM filtering.
Since PS export data for las format I thought some other software could do that.
If not, I will have to do in a stereoscopic way, which is, in this case, not a good ideal situation.

[gEEvEE]

But... how would you filter and more importantly: what would you like to filter?

I think that a lot of people here do not understand the basics of remote sensing. Airborne laser scanning (ALS) is an active method. You fire short-wavelength infrared pulses (sometimes near-infrared and even more seldom visible light) that have a certain footprint. When this beam with its finite footprint hits a branch, most of the photons are reflected by the branch and its leaves and are reflected back to the sensor. The sensor records this reflection and you have one measurement (from the forest canopy). The remaining photons of that pulse might - if you are lucky - hit some branches below and if you are really lucky and fire a lot of pulses per m squared, some photons might even hit the ground surfaces. When those are reflected back to the sensor, you have a measurement of the topographic surface (for the sake of simplicity, we are not going into the details of a conventional and full-waveform scanner). Using filters, you might be able to filter away the upper reflections and keep those from the ground surface.

In airborne imaging, you deal with a passive technique. You capture the radiation of the sun that gets reflected by the objects on the Earth’s surface. Again, for the sake of simplicity, we assume that we only capture visible radiation. This reflection of visible light creates a pixel. If you have a dense forest, none of the radiation contributing to one pixel will be primarily of the surface. This means that the pixel values are mainly created by reflection form the tree canopy. In short: the image will display the upper part of the canopy. Period. Nothing else. Using photogrammetry or computer vision, methods can be used to reconstruct the position of the individual pixels in 3D. If the pixels thus visualize the canopy, this is what you will reconstruct. There is nothing else you can reconstruct from them, since you simply do not have any data to do so (again, assuming a dense forest and no pixels visualize the ground surface). So, there is nothing to filter. Even if you export a *.las file, it does not mean it holds the same data as a *.las file created by an airborne laser scanner.

The only thing one can do is cut away the part of the 3D model that contains trees and supplement this with terrestrial measurements or – as Wishgranter correctly mentions: come back in autumn or winter and shoot images at the “leaf-off” condition, so that pixels might be created by the reflection from the ground (and hence we can reconstruct the ground surface). Also stereoscopy will not work with canopy images. What will you see in your stereoscopic view? A 3D view of the tree. That’s it.

Hope it is clear now.

Geert

[fpbv]

Dear Geert

First of all, thanks again for your reply.
Just let me say, in my 14 years as cartographer engineer working and dealing with mapping - aerial and terrestrial photogrammetric, topographic and GPS survey, GIS projects, and now working as consultant for many customers as diverse like mining, forestry, urban cities, agricultural and environment, it is very interesting how come a simple question turn out to weird path.
I assume most of the users of PS know the big difference between laser vs pixel data.
I?ve been looking for a just a alternative solution or, let me say, a better idea to how to overcome this.
When I talk about the las format is because it is a standard format sponsored by ASPRS and ISPRS, as simple as that.
I like when people wants to help others, and I would like to think you wanted to do this.
But I don?t think a lecture on us about the pixel, laser, etc is necessary, or better, the way you put.
Again, I would like to think you are helping us but not in those "kind" terms you put.
Everybody here is dealing with new technologies every day and some of us are dealing with old and new mapping projects techniques for the last 14 years.

I hope is clear now.

[gEEvEE]

Well, no offense. Not wanting to be rude or something. But when I see questions like this or many others on this forum, it just is clear to me that people do want impossible things.
Believe me, I have worked with many people in the field of remote sensing (both students and pros), but when one really starts to delve deeper, it is often apparent that they do not completely understand the very basics of this discipline. Again, no offense and I am certainly not telling that you do not know your stuff. I just wanted to give a more detailed outline of both methods, so that future questions can bear this in mind.
Have a nice day.

Geert

[Wishgranter]

If this field were so easy, client will need not so much money for our work. Pscan create mesh on the points, and if something like tree is present it create just blobs. it could be "cleaned" if you get a realy HIGH resolution but that create another problem with RAM. Therefore using it on foliage is questionable - much easier and better is with LIDAR. the video explain it very clearly what all can be archieved......

It could work on very low atlitude with a high image sensor, but a LOT of post will be nescesary.


 So Pscan is not ideal for foliage if want use as LIDAR. And guys, we love reply like this becasue is a realwork experience that is easy to understand. Im trying explain thing deeper too, because everyone can use the part what is for his field useable. So longer and deeper answers are what we need.

Im reading a LOT of whitepapers (500-600 ) in last few years, but answer like this im get afther reading around 50 papers so a lot of time is saved.
 
I get green light to create a video of capabilites of Pscan, adn hoping that everyone can offer some of it specialization -  GIS, Mapping, Cultural heritage, archeology, urban planing etc. will prepare few outlines what we need for that. Becasue this could save time with experimenting if Pscan is capable for that area......

[Mzee]

Hi,
Download LasTools by Martin Isenburg, specifically LasGround. Needs a dense cloud and gives a promising result which with a few tweaks may help you.
Regards,
Alan

[cstallings]

There are several factors that can make this possible or not. The biggest factor is going to be if it was leaf on or leaf off. Anyone who has done stereo compilation knows that you cannot map a forested area during leaf on conditions. This is why we do most of our acquisition in the early spring when we can have a good sun angle and still have leaf off conditions. With traditional photogrammetry its important that you can see each object from two images. This is the entire basis of the science.  So if you have leaf of conditions you may have many correlated pixels on the ground. I have no idea what density or even of what quality but its a possibility. Next you would need to run a ground macro on the LAS points. Martins tools are very nice but you cannot use them commercially. I imagine if you emailed him and told him what you were up to, he would be very excited to work with it. As for the quality of the product, I could not even begin to speculate however I would be very interested to see it. I seriously doubt it would meet any type of mapping accuracy standards.  For 1 ft contours we usually plan missions to be at least 0.7 meter posting and for 2 ft we shoot for at least 1.4 meter posting. one footers need an RMSE at least 9.25 cm and two footers 18 cm. Good luck and let us know if you have any success.

[Mzee]

Direct from Martins Licence.txt

The following LAStools require a license for commercial or government
  use. They can *only* be used freely for strictly non-military purposes
  in case of non-profit personal, non-profit educational, or non-profit
  humanitarian use, where you can personally (!) convince me that you'll
  better the world (according to my definition). (-:

    * blast2dem

    * lasground
    * lasheight
    * lasclassify
    * lastool
    * lasgrid
    * lasboundary
    * lascontrol
    * lasoverlap
    * lasduplicate
    * lassplit
    * las2tin
    * las2iso
    * las2dem
    * lasthin
    * lassort
    * lastile
    * lasclip
    * las2shp & shp2las.

--------------------------------------------------------------------------

  If you want to use LAStools please contact me for licensing. You can

  (a) license the existing executables for production
  (b) license the existing executables for inclusion in your software
  (c) license the source code to adapt it for your production needs
  (d) license the source code to include it in your software package

[fpbv]

Hi guys

First of all, I must apologize to everybody because I don?t want to offend anyone here.
If I did, I am truly sorry.
I agree with Geert regarding passive and active data. Lidar is very nice for forestry but is a LOT expensive, at least here in Brazil. Believe me.
My client has some low budget so I tried to get some different solution.
Unfortunaly my time to research is very limited so I asked for this.
I already download the LAS Tools but I couldn?t find a way to work properly.
If anyone wants to process some forestry data generated by PS I can send to test.
Just pm and I will put on DropBox.

Thanks in advance.

[Wishgranter]

sounds interesant......