3D model of tall tower

[Lyle]

We will be flying at a plant to construct a 3D model of its smoke stacks. These stacks are close to 300' tall.

Any recommendations on techniques? Several circles at different altitudes is what I'm thinking.

How about GCPs? I'm using a Phantom 3 so it has the geotagging. Should I go ahead and rent a GPS rover to collect GCPs anyway? I understand (and found out) they are VERY important for larger surveys such as stock piles. But since this will be a small area with a tall object, I'm wondering how important they are. I feel like I should, but my manager doesn't want to charge the client the $300 for the GPS rental if it's not really necessary.

Ideas? Have any of you done this before?

[frank.stremke]

well a gps is not the tool of choice here
i would go with a reflector less capabe total station and shoot in some tatural targets on different levels up the stack
the gps ground points are not good if you need to work 300ft up and probably can not go there or do a gps survey on a wall
but i would definitly go with gcp one way or an other, otherwise you end up with a banana shaped stack
frank

[stihl]

To comment on Frank; if he did his overlap well and made images at different elevations and distances from the tower, photoscan should be able to align the images by image alignment alone.

I'd suggest using flight positions as this will speed up the alignment if you let align by using the reference data.
I'd also suggest upping the default number of tie points from 40k/4k to 80k/60k.. More tie points usually means a more complete dense cloud with more fine detail.

The only upside to using additional GCPS is that the tower will be better georeferenced and will actually be on scale.
If you don't have the possibility of that (because of costs) it's important to bring a measuring tape or two that you can place against or near the tower. If you roll the measurement tapes out to their Max distance and preferably place them perpendicular to each other so you have a scale distance bar for the X and Y axis separately..

If you only use the geotag information from your DJI then the tower will be off of its actual location (not terrible important for relatively measurements on the tower itself) but as there would be no verifyable distance to check its very possible that even relative distances are false.. Without some sort of scale of geo information, Photoscan doesn't know how large a certain object is.

[Lyle]

Cool. Good info and good idea with the total station. Fortunately there is a platform at the 200' level I will physically go to and GPS in a few of Control points. I'm not sure I understand the "... tie points from 40k/4k to 80k/60k...." advice. (I'm relatively new at this and I'm still on a 30 day trial). I guess that's a setting under the Advanced pull down, right?
Thanks!

[stihl]

When you align your photo's, in the alignment window you can set the amount of 'tie-points' and 'key-points'.
Tie-points are points in the photo's that Photoscan deems as useful points. The key-points are a subdivision of the tie-points of points that are actually used for the sparse-cloud.

By default these values are 40.000 and 4.000 for tie and key points respectively. Meaning that it will search for 40k points in every photo and from those 40k Photoscan will select the best 4000 and uses those to align the images.
I prefer upping those values so more tie-points are found. I find this useful for nearly all projects, from aerial surveys to accurate point clouds of objects.