Geometry going wrong in 'samey' areas

[hanparker]

I'm not sure if that's what's happening but here's the situation..

I'm modelling big long stretches of road. I'm using python script to set the settings and batch up the photos, set the reference and camera calibration etc. The models seem pretty good through villages. As I move into big long boring stretches with hedges on both sides and nothing much to differentiate (I'm guess this might be the problem) - the models geometry stops matching the real world..

I was having problems with rotation of my models but then had it pointed out that yaw pitch roll isn't used. So I started providing photos from both sides of the road. If I check the model window before I start Alignment - I can confirm I've got all the photos for both sides of the road - nothing odd or different going on from the models that look pretty good in the built up area. I'm only modelling relatively small chunks at a time with around 600 photos in each. Using 10000 tie points, high accuracy and "reference" pair selection

So my hypothesis is that it's struggling because of lack of interesting features.. Does this sound likely and is there anything I can do to help it out?

[hanparker]

Update:

I've now allowed it to use the masks while aligning photos - that's helped quite a bit but I am still hoping to perfect further. My survey data is a ground based survey. I've so far avoided using Ground Control Markers - I've been reading http://adv-geo-research.blogspot.co.uk/2014/03/agisoft-photoscan-crash-course-v-102.html and wonder if GCM might help my sections that are going wrong?

[James]

I guess the trouble with a hedge is that it might look completely different from different angles if you are too close to it. I mean a brick wall is nice and flat and whatever angle you look at it from you will see the same faces of the same bricks, but if you stand 3m from a hedge to take a photo, then move 4m along the hedge and look back, you will see the leaves etc (whatever hedges have) probably on their other sides, and not much will match. The matches you do get may be false ones.

The people doing aerial forestry scans probably know more about this sort of thing, but i know another problem with vegetation is that it moves between photos!

Also like you say it's all a bit samey.

Perhaps you should increase the frequency of images in the non-urban areas.

What sort of camera setup are you using, is it fisheyes, wide angles etc? And what interval of shooting and speed of travel are you using?

Ground control markers might help align images that otherwise wont align or align badly, but that's going to introduce a lot of manual work. Otherwise markers will help you scale and orient your data, and if you have enough of them you can optimise to remove non-linear distortions caused by slight errors in the alignment, but i dont think it would have much impact on the shape of your hedges.

[hanparker]

Hi James,

Thanks for that info. I stumbled across a couple of the forest canopy threads too and had a read of those. We used an external company to do the survey. The used a Leica Pegasus. It's not fish eye or wide angle - I'm setting the "Camera Type" to Frame within the camera calibration settings. This is what the survey company said about frequency "Max frame rate is 8fps - we usually cap it at 5m spacing".

It's not the shape of the hedges that's a problem. I don't think we're that bothered about them.. A hedge is a hedge for this purpose.  It's more that it loses alignment with the real life geometry - so the road isn't where the road should be. It's too high, too low or rotated. I would think I had something wrong with the way I am feeding the data to it - but most of the time it works - I just get some long bland stretches that go haywire.

[James]

At 8fps they would have to be doing about 90mph to only get images every 5m!

I assume they mean that at slower speeds they don't use the full 8fps.

If they are observing the village speed limits, then they are also probably going a bit faster in the country lanes (but still getting images every 5m) so perhaps the increased speed is introducing motion blur, as well as the other problems mentioned to do with shooting mostly vegetation.

I would suggest instead of trying to process 600 images at once, just take 10 or 20 images in a dodgy area to process and see if you can identify where and why it fails.

I have a feeling that on narrow country lanes 5m might be too great an interval due to the difference in viewing angles between one shot and the next.

[hanparker]

Thanks - I'll give it a go!

[Paulo]

Hi,

i was wondering, what format image are you getting from the pegasus survey? Are they just 4 Mpix images (2000 by 2000 pix format) or some stiched images from the multiple camera pegasus setup?

Because if you are just getting images from one of the side looking cameras then you may not be getting enough image overlap berween images to get a good photo alignment. And then the problem remains that basically you are getting a single line of images that will have rotational problems in photo alignment.

Pity that Photoscan does not take the orientation (yaw, roll, pitch) into account for alignment as the Pegasus probably outputs  precise orientation and position info for each image given its survey grade GPS/IMU system...

[hanparker]

They 2046 x 2046. There a 6 photos in a hexagonal configuration (and a sky facing fish eye lens which I ignore... wonder if using that would help... ). There is a some lovely yaw pitch roll info but yes  I have to ignore that. It often fails to align the camera that is right beside the side of the road, facing straight into the hedge - which is kind of understandable - but there are usually 11 other pictures from more or less the same spot.

I've had to shelve sorting it out with Photoscan for now as I'm pushed for time.

Anyone know if there are any future plans to include yaw pitch and roll is the processing?