Transmission line profile mapping

[Santiago]

HI,

A company asked me to do the following job for them, but I really don't know if it could be possible with photogrammetry. We have a Hexacopter and a Lehmann Lp960 for aerial mapping equipped with a sony nex 5n 16 mp. The job consists in:

Create a profile on some transmission lines (power lines) that extend 150 km. They need to know were the lines are with respect to the terrain (including vegetation so no need to remove it). What I don't know is if the photogrammetry techniques can accurately model this transmission lines. Here's an example of what they need.

http://www.powerlinesystems.com/pti/ptipaper.html

I think that this job could only be done accurately with terrain techniques or LIDAR, or if anyone here has had any experience with a similar job It would be great to know your experience.

Thanks

Regards,

Santiago

[Lui]

Hi Santiago,

from my experience (6 years in this business) the only option to get 3D wires is by LIDAR. And even here some wires can be tricky to get sometimes due to insulation material properties. To get an accurate results for danger object/vegetation analysis some additional realtime observation should be performed like temperature, wind and powerline load. Only with those parameters and 3D wire powerline normalization is possible and we can later predict wire sag in extreme operational conditions.
So I think that there is no real hope for a photogrammetric approach.
If you need some additional information regarding powerline mapping and normalization, you can always mail me.

Best regards

Lui

[phisan]

Hi,
i am interesting to generate point-cloud from images as well, esp. for transmission line 3-d mapping. It happens XeosImaging could do it.
http://www.xeosimaging.com/products/photogrammetric-point-clouds/?lang=en


Semi-global matching is magic word!   E.g. you could use  3-d correlator of SimActive
http://www.simactive.com/.   To my understanding , 3-d correlator is not 'full SGM'.
Since it claims generation of 4 image pixel to 1 point cloud.
This performance is similar to AgiSoft could do.


For 3-d mapping from airborne imaging uav or manned one , we event believe we could
get denser point cloud than lidar.  One drawback is that point-cloud is 'single return'
Where Lidar is multiple-return data and have powerful penetrating through
loose canopy.


Br,

[Diego]

Semi-Global Matching is very different Agisoft algorithm, the main difference is the redundancy, the great disadvantage of algorithm agisoft in the point cloud generation, is generated based points in pairs of images, while Semi-Global Matching points generated based in all possible images to determine the point, this gives more accuracy to the point clouds, I had an opportunity of testing Intergraph Semi-Global Matching, and while agisoft, generated a very large point cloud (30 GB), a cloud of Semi-Global Matching smaller points (1.2 GB), but much more accurate and consistent, without gaps in the point cloud, the accuracy was less than an inch, clear the test was done with a camera UltraCam Xp.

[Santiago]

Thanks guys,

Lui I agree with you, the best option here would be LIDAR, but I have to say what Phisan posted is truly interesting. Although when looking at the accuracy XEOS can achieve even when using ground control points (20 cm vertical and horizontal), can't compare to LIDAR.

I'll look more into SimActive's Correlator 3D, it seems interesting, but I get a feeling it's not design for every UAV with a common camera out there. Unlike Agisoft or Pix4d it does ask for orientation parameteres like angular values, which I can't obtain from my UAV's.

Regards,

Santiago

[Oli63]

I agree with Lui too. No chance to solve this problem with photogrammetry. Lidar is the better choice in this case.

Semi Global Matching seems to be an interesting concept, especially concerning the problem of taking photos of buildings and anything with vertical elements in it. Here's a good paper about it:
http://www.dlr.de/rm/en/PortalData/3/Resources/papers/modeler/cvpr05hh.pdf