Intensity images and normalization

[Steven F]

I'm attempting to collect ground control points from lidar intensity images for triangulating aerial photos. So far I've tried a few different settings in lasgrid including the highest or average intensity of either first returns or all returns, but the images are still fairly difficult to visually interpret. 

Are there any parameters that tend to work best for producing a nice looking intensity image?

Also, my lidar data is over a very mountainous area so correcting the intensities for range might improve the appearance of the intensity image, and normalized intensities could yield more informative intensity metrics from lascanopy.

Is lastools capable of normalizing the intensity values if I have trajectory files? 

[Martin Isenburg]

Hello,

One answer would be to use RIEGL scanner where "normalized reflectivity" can be exported to the LAS file in form of "extra bytes".

Or i could add this as a feature - likely to lastrack because we would require trajectories - but i do have many more feature requests on my plate. How should such an intensity normalization step be implemented (assuming all is flown by the same scanner)? Suggestions?

Steven told me someone that he was contacted in a private email off-list by a service provider who gave me a little advice. They basically said that intensity normalization usually isn't that important and they aren't aware of any free tools for it. However, they and probably other contractors have custom tools and will provide it as a service if requested by a client.

Steven ended up not needing to perform the normalization for my original purpose, which was to create a good looking image that he could use for georeferencing aerial photos. However, he could see normalized intensity statistics being potentially useful as canopy metrics for forestry. To him it seems like very few people attempt to use intensity metrics when estimating forest structure or classifying trees, but he thinks it's possible that these metrics often don't end up being significant in classifiers partly because they aren't normalized.

He would only suggest adding it if I  think it wouldn't be that difficult and if other people in the community would be interested in it.

Regards,

Martin @rapidlasso

--
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

[Jonah Sullivan]

I have seen a lot of research where normalised intensity was used along with other attributes for point cloud classification (usually with supervised classification through machine learning: support vector machines or random forests).

I have never seen a point cloud with normalised intensity.

I have trouble believing that it can be implemented well considering the complexity of terrain and above-ground objects. 

As an example: a pulse that passes through vegetation will be partially reflected several times before terminating. Each partial reflection will consume a portion of the initial energy. The last return represent either be the remaining portion of the initial energy, or just another small portion with the remaining energy lost in the forest under story. How would you apportion the energy to each of these returns?

Another example: a laser pulse that strikes the side of a building will have a footprint that is a very long ellipse, while a strike on the top of a building will have a much shorter ellipse, and a strike on the edge of a building will be a portion of an ellipse. How would you handle each of these cases?

The following publication may be of interest:

Höfle, B., Detection and Utilization of the Information Potential of Airborne Laser Scanning Point Cloud and Intensity Data by Developing a Management and Analysis System, in Dissertation, Faculty of Geo- and Atmospheric Sciences 2007, University of Innsbruck: Austria.

Jonah Sullivan 

Geospatial Analyst

National Topography Services  |  National Location Information Group

Environmental Geoscience Division  |  GEOSCIENCE AUSTRALIA

Phone:  +61 2 6249 9516    Fax:  +61 2 6249 9999

Email:  Jonah.S...@ga.gov.au    Web:  www.ga.gov.au

Cnr Jerrabomberra Avenue and Hindmarsh Drive Symonston ACT

GPO Box 378 Canberra ACT 2601 Australia

Applying geoscience to Australia’s most important challenges

[Steven F]

It would probably make the most sense to implement intensity normalization as a simple range correction within lastrack. 

Intensity corrected = Intensity*(Range^2 / Standard Range ^2)

This has been used to good effect in a few papers I've come across.

The dissertation that Jonah mentioned uses this method as a basis for comparison for a somewhat more complex method that takes into account angle of incidence and atmospheric thickness. I think the difference in results between these approaches probably wouldn't warrant the additional complexity in most circumstances. 

[Floris Groesz]

I agree that it would be useful to have a basic calibration option in lastrack that corrects for range.
I am looking at some RIEGL data for forestry and my data set only has amplitude (We have a Q-series instrument, not a V-series). It would be very good to be able to do the range calibration in LAStools.
All other corrections like incidence angle and surface type are so complex that I think that they will have to wait.

Floris

[Reginald Argamosa]

Hi Martin,

Is it possible in LAStools to identify which pulse the returns belong to? Also, is it possible to identify the values of intensity per return? I know that lascanopy can compute for the intensity metrics but we want to extract intensity values per return and not in a pixel basis.

see attached image.

Thanks

-Regi

 

[alecsolpenda]

hello regi,

i got the feeling that i might be using the answers to your questions in the near future so please allow me to write something in your post (as martin might also be still paddling somewhere :-) )

have you tried using last2txt for this? there is already a discussion similar to your query and it makes use of the filter "keep_return" (that's for your second question). there is as well a check box option called "point source ID" (to answer your first question?). please let me know if this works,thanks!

alex

[Reginald Argamosa]

Hi Alex,

I tried using las2txt and you are right that it gives the intensity values, gps time, and pt source id. Thank you for that. =) However, i think that the point source id is not the id of the pulse but rather the id of the flight line. What I need is to determine the which pulse the return comes from. 

-Regi

[Martin Isenburg]

Hello Regie,

for single beam LiDAR systems it is the GPS time stamp that is supposed to unqiuely identify all of the discrete LiDAR returns that belong to the same laser pulse. In LAStools you can group these returns together by sorting via the GPS time:

lassort -i lidar.laz -gpstime -odix _sorted -olaz

You can additionally sort them such that the returns from the same pulse (and the same GPS time stamp) are ordered 1 of 4, 2 of 4, 3 of 4, 4 of 4 with: 

lassort -i lidar.laz -gpstime -return_number -odix _sorted -olaz

then you will see that they are grouped together by pulse

las2txt -i lidar_sorted.laz -parse trnxyzi -stdout | more

5881.054885 1 1 277997.16 6122447.51 57.80 30

5881.054895 1 1 277997.17 6122448.19 57.71 23

5881.054901 1 2 277997.56 6122448.66 54.14 24

5881.054901 2 2 277997.94 6122448.80 50.65 17

5881.054906 1 2 277997.54 6122448.99 54.29 31

5881.054906 2 2 277997.94 6122449.14 50.67 10

5881.054911 1 2 277997.54 6122449.33 54.31 10

5881.054911 2 2 277997.94 6122449.47 50.66 64

5881.054916 1 1 277997.54 6122449.66 54.33 28

5881.054920 1 1 277997.54 6122450.00 54.31 18

5881.054925 1 3 277997.57 6122450.35 54.05 12

5881.054925 2 3 277997.75 6122450.42 52.33 27

5881.054925 3 3 277997.94 6122450.50 50.64 18

5881.054930 1 1 277997.93 6122450.83 50.65 102

5881.054936 1 1 277997.93 6122451.17 50.65 125

Regards,

Martin @rapidlasso

--