Single Tree Crown Extraction and Species Identification with Lastools

[anto]

Hello,

I am learning how to handle lidar data using Lastools. My primary interest is forest mapping so at the moment I am using lidar data found on the web to generate DEM and canopy height models (CHM).

Let's say that iwhat I would like to do is to extract from raw lidar data number of trees and trees species (i.e. oaks vs pine) in a certain region of interest.

I am wondering if Lastools can be used to accomplish this task.

If these is not possible via only Lastools, what other softwares I would need to do that ?

Something like Lastools + Grass + Qgis?

or Lastools + ArcGis?

Thank you in advance,

Antonio D.

[Anahita]

Hi Antonio,

You could do anything that you want from lastools. 

The first step is create a CHM and then extract tree location. Then, you could extract the tree height and crown structure shape from CHM.

In your case, you need to extract lidar intensity which is very useful for tree species classification. If you look at literature review, you could see researchers used lidar intensity and height percentile for identifying different species with lidar data.

Good luck

Anahita

  

Sent from my iPhone

--
Download LAStools at
http://lastools.org
http://rapidlasso.com
Be social with LAStools at
http://facebook.com/LAStools
http://twitter.com/LAStools
http://linkedin.com/groups/LAStools-4408378
Manage your settings at
http://groups.google.com/group/lastools/subscribe

[Anahita]

Hi Paulo,

according to literature review, structural variables and Intensity characteristics derived from LiDAR data are useful for for classifying species. The intensity features of the FIRST echoes can differ more between species than the other echo categories. i attached two articles as sample which used intensity for species classification. the first step is that you can extract high quality intensity ( -first_only) from Lastools and then use it as ancillary data or base data for species identification.

Best regards

Anahita

From: João Paulo Pereira <joaopaul...@florestal.eng.br>
To: anian...@yahoo.com 
Sent: Monday, 14 October 2013, 14:25
Subject: Re: [LAStools] Single Tree Crown Extraction and Species Identification with Lastools

Hi Anahita,

I am very interested about what you sent to Antonio. I was wandering if you could suggest a tutorial or works regarding this methodology you said about species classification using LiDAR intensity returns.

Best regards.

-- 

Engº. João Paulo Pereira

Forest Engineer

Master's Student in Forest Engineering

State University of Santa Catarina - UDESC

Researcher in the Silviculture and Forest Management Group 

Computer Technician and Software Support

Phone: +55 049 3225-4305

Cell Phone: +55 049 9986-1751

[simaya35]

Hai Dear,

My task find out the Tree species crown and tree identification , Any auto las point classification tool possible for challenge task tree species classification 

any Lastool available this type of classification.

Advance thanks

simaya 

[Terje Mathisen]

simaya35 wrote:
> Hai Dear,
>
> My task find out the Tree species crown and tree identification , Any
> auto las point classification tool possible for challenge task tree
> species classification
>
> any Lastool available this type of classification.

The problem here is mainly the LiDAR point density:

You _must_ have many points per tree crown in order to have any
possibility to determine where each tree is located and then what kind
it can be.

Assuming you do have the usual ~10 points/sq m for forest
classification, I would start with the usual pipeline to determine the
ground surface, then make a copy of all the tiles using replace_z to get
just object height, then I would generate a dense set of ISO contours
based on first hits only, maybe using something like 25 cm contour interval.

Each tree should then turn up as a knoll, in the form of multiple
concentric rings.

All this can be easily done with the standard LAStools toolbox
(lastile/lasground/lasheight/las2iso), so the only thing that remains is
to take the output of las2iso and sort the contours so that you
determine the highest ring for each tree crown, then look at the closest
surrounding rings in height-sorted order:

The maximum height plus the distribution of crown area as a function of
height below the top will probably give you the ability to make a pretty
good estimate of the tree species. You train your recognizer by loading
a set of the above values for several (as varied as possible) tres of
each species.

Is this a research project/Masters/PhD?

Good luck!

Terje

>
> Advance thanks
> simaya
>
>
>
> On Thursday, October 10, 2013 10:50:22 PM UTC+5:30, anto wrote:
>
> Hello,
>
> I am learning how to handle lidar data using Lastools. My primary
> interest is forest mapping so at the moment I am using lidar data
> found on the web to generate DEM and canopy height models (CHM).
> Let's say that iwhat I would like to do is to extract from raw
> lidar data number of trees and trees species (i.e. oaks vs pine)
> in a certain region of interest.
> I am wondering if Lastools can be used to accomplish this task.
>
> If these is not possible via only Lastools, what other softwares I
> would need to do that ?
>
> Something like Lastools + Grass + Qgis?
> or Lastools + ArcGis?
>
>
> Thank you in advance,
> Antonio D.
>

> --
> Download LAStools at
> http://lastools.org
> http://rapidlasso.com
> Be social with LAStools at
> http://facebook.com/LAStools
> http://twitter.com/LAStools
> http://linkedin.com/groups/LAStools-4408378
> Manage your settings at
> http://groups.google.com/group/lastools/subscribe

--
- <Terje.M...@tmsw.no>
"almost all programming can be viewed as an exercise in caching"

[Morteza Shahriari Nia]

Hi,

Did any of you guys have any success singling out tree crowns? I am new to this field and any help is appreciated. Similar to Antonio I need to know height distributions of each tree species.

Best,

Morteza

[Antonio Ruiz]

One possibility is to compute a raster Canopy Height Model as CHM = DSM-DTM. Then you can change the sign of the elevations -CHM and import this model into a GIS (ArcInfo, GRASS...) and apply the hydrographic modeling tools to detect watersheds. 

You can assume that each watershed is a tree crown. The GIS assigns a label to each watershed. This label can be attached to each lidar point. If you create a raster with the watershed labels, this assignment can be fast, but you have to program it. 

After that you can compute some lidar metrics for each tree. I did it some time ago. I computed 15 lidar-derived parameters similar to that that are computed by Fusion in raster cells, but tree by tree and we used them to classify tree species. This was published in Spanish, 

http://www.icc.cat/index.php/content/download/48928/340523/file/lidar_dmc_casi_llenyosos.pdf

but there is also a short (and worse) version in English:

http://folcomuns.firabcn.es/Globalgeo_Proceedings/Proceedings/Fotogrametr%C3%ADa%20y%20Teledetecci%C3%B3n/Active%20radar%20and%20LiDAR%20sensing/Tree%20species%20classification....pdf

Regards,

Toni

[Martin Isenburg]

Hello,

for single tree detection you should *not* compute a CHM as a simple
difference DSM-DTM because the resulting raster tends to be full of
"pits" if you use standard algorithms for DSM and DTM generation (e.g.
rasterization of a TIN generated from the first return (DSM) or ground
return (DTM) of the LiDAR). . These pits hamper subsequent single tree
detection as we have shown in a joint paper with paper with ITC [1]
and there is a try-it-yourself blog article here that demonstrated the
stark differences that may occur visually:

http://rapidlasso.com/2014/11/04/rasterizing-perfect-canopy-height-models-from-lidar/

How bad the difference is depends on the penetratability of your
vegetation, the number of flightlines, the off-nadir angles used, the
grid resolution, the scanner used, and a few more factors. But DSM-DTM
is always less good ... sometimes a little less good, but often really
horrible compared to the pit-free algorithm.

Regards,

Martin @rapidlasso

[1] Khosravipour, A., Skidmore, A.K., Isenburg, M., Wang, T.J.,
Hussin, Y.A., 2014. Generating pit-free Canopy Height Models from
Airborne LiDAR. PE&RS = Photogrammetric Engineering and Remote Sensing
80, 863-872.



Th

[Barend Erasmus]

I like the watershed approach…but has anyone tried it in a savanna environment? My experience there was that the mix of standalone trees and bush clumps really made it difficult. Mixed bush clumps, with different heights and species, with the watershed method, do provide local crown maxima (after changing the sign to +CHM) but that does not correspond very well with individual trees. Perhaps easier in a boreal forest?

Barend

 

Prof Barend Erasmus

Exxaro Chair in Global Change and Sustainability Research and

Director:  Global Change and Sustainability Research Institute (GCSRI)

University of the Witwatersrand

 

+27 11 717 6602 (office)

This communication is intended for the addressee only. It is confidential. If you have received this communication in error, please notify us immediately and destroy the original message. You may not copy or disseminate this communication without the permission of the University. Only authorised signatories are competent to enter into agreements on behalf of the University and recipients are thus advised that the content of this message may not be legally binding on the University and may contain the personal views and opinions of the author, which are not necessarily the views and opinions of The University of the Witwatersrand, Johannesburg. All agreements between the University and outsiders are subject to South African Law unless the University agrees in writing to the contrary.

[Rombouts, Jan (FSA)]

Hi Barend,

 

Chen uses a watershed approach in California savanna:

 

Chen, Q., Baldocchi, D., Giong, P. and Kelly, M. (2006) Isolating individual trees in a savanna woodland using small footrpint Lidar data. Photogrammetric Engineering & Remote Sensing, 72, 923-932.

 

There are many alternative algorithms for extracting individual tree information. Here is a good paper comparing them:

 

Vauhkonen, J., Ene, L., Gupta, S., Heinzel, J., Holmgren, J., Pitkänen, J., Solberg, S., Wang, Y., Weinacker, H., Hauglin, K. M., Lien, V., Packalén, P., Gobakken, T., Koch, B., Næsset, E., Tokola, T. and Maltamo, M. (2012) Comparative testing of single-tree detection algorithms under different types of forest. Forestry, 85, 27-40.

 

 

Jan

This communication is intended for the addressee only. It is confidential. If you have received this communication in error, please notify us immediately and destroy the original message. You may not copy or disseminate this communication without the permission of the University. Only authorised signatories are competent to enter into agreements on behalf of the University and recipients are thus advised that the content of this message may not be legally binding on the University and may contain the personal views and opinions of the author, which are not necessarily the views and opinions of The University of the Witwatersrand, Johannesburg. All agreements between the University and outsiders are subject to South African Law unless the University agrees in writing to the contrary. </table

[Antonio Ruiz]

I have not tried in savanna but in Spain we have dehesas that are similar (with isolated trees). What I do is to limit the extension of the tree crown area to the watershed area deeper than 1/3 or 1/4 the tree height. After that, if the tree height is smaller than some minimum tree height, the tree is rejected. 

I agree with Martin. The DSM smoothing is critical. 

Toni

[Martin Isenburg]

Hi Antonio,

Not smoothing. At least not standard Gaussian smoothing (of a DSM rasterized from a TIN of first returns) as sometimes suggested. That has actually a worsening effect on the canpony for the purposes of single tree detection as we (or rather Anahita's elaborate omission and comission error experiments) have shown in [1]. The selective removal of only the pits without affecting the canopy overall is what did the trick ...

Martin @rapidlasso

[1] Khosravipour, A., Skidmore, A.K., Isenburg, M., Wang, T.J., Hussin, Y.A., 2014. Generating pit-free Canopy Height Models from Airborne LiDAR. PE&RS = Photogrammetric Engineering and Remote Sensing 80, 863-872.