Messages

436

General / Re: inconsistent results

« on: September 19, 2014, 10:13:51 AM »

GPS on board of the cameras are usually crapy, and have refresh rates that sometimes do not match UAV frame interval. Can you export camera centers just after loading photos and overlay them onto your shp before alignment?. That could help to understand what happened.
Also, if that was not the reason for the so called inconsistency, could you show us your settings in the GCP pane?

437

General / Re: How to measure elevation on specific terrain points

« on: September 15, 2014, 11:09:33 AM »

The question is not whether Photoscan DEM is accurate enough or not, but if your photos, the project design and the ground control can deliver the resolution that you require in the model.
Lets say it does. The easyest way to extract spot elevations (coordinates) from the model inside Photoscan is to pin markers wherever you want in the model, and then export estimated values for such markers.

438

General / Re: Advice on making my flat field "flat"

« on: September 13, 2014, 02:25:29 PM »

 
Jose Martinez
www.geobit.es

439

General / Re: Advice on making my flat field "flat"

« on: September 12, 2014, 06:59:06 PM »

Hi all
Thermal imagers are usually low resolution devices. It's quite dificult to obtain good dense point clouds due to the poor quality in the photographic sense (we usually consider that quality stands for crispness, fidelity, etc). It surprises me you managed even to have them oriented.
If you thing your termal images are well oriented (and that is a must), and they share ground refference with that of the rgb ones, why don't you just import the mesh model that came out from the rgb  chunk (export it as fbx, for example) and the create the thermal orthoimage just directy onto it?... it works, I've done it many times
Regards

Jos? Mart?nez
GEOBIT

440

General / Re: Photogrammetry and restoration project

« on: September 04, 2014, 12:42:07 AM »

Hello.

I'm actually working on a project where I need to fit different 3D parts. Which program allows such a task. The idea would be, the the individual 3D objects (pieces) cannot overlap.

Any ideas?


PS: is this possible with Cinema 4D?

This can be done by means of programs developed for laser scanner data management (Polyworks, Rapidform, Geomagic, 3dReshaper... even with Meshlab it could be possible). But usually they're not designed to fit pieces but scans. So you'll have to manage how to force the program to think that the two pieces you want to fit are two scans of a single one. It's not straightforward but absolutely possible.
If the pieces to be fitted together keep sufficient contact surface, the contact surfaces have not suffered significant wear, and your model is accurate enough, the best-fit alignment will deliver by far better results than manual approach.
(PS: I can guide you through that process by means of Polyworks but outside of this Agisoft's forum)

441

General / Re: Referenced chunck. Problems with Z axis

« on: June 26, 2014, 10:12:56 PM »

I agree with you
I have asked for the same and made a feature request long time ago. For some reasons it seems not to be a issue for most people wether bounding box z is exactly aligned with the actual vertical or not.
http://www.agisoft.ru/forum/index.php?topic=2552.msg13560#msg13560

I'd love some clarification by Alexey. why it is not a direct option?

442

General / Re: Is it possible to measure window openings with Ps?

« on: June 25, 2014, 01:12:31 PM »

I would not be very optimistic unless the surface texture showed the best imaginable texture for feature extraction, and that sounds not so likely to happen when it comes to new building materials.
On the other hand, measure ground control points better than 2mm is not just piece of cake. I would say sfm based photogrammetric programs like Photoscan are not the best choice for such a survey unless you had many things under tight control. It would be hard to achieve the that accuracy even using scissor platform instaid of a copter. The key points to be measured to determine the openings, should be marked and pined one by one.
Try first your setup at the ground level.
Good luck

GEOBIT

443

General / model lower than ground control

« on: June 24, 2014, 12:48:09 AM »

Hello:
I've done a project (aerial work) in wich an average deviation of -0.07m in dense cloud with respecto to ground control points (targets) is annoying me.
It seems that Photoscan underestimates elevations of all automatic matchings causing an error lowering results by some amount. Note that none of the ground control points exhibits a deviation with the opposite sign, and that makes it unlikely that the error came from a bad adjustment of the bundle block.
In this project all photos suffer from noticeable defocusing due to a bad focus set but they still  appeared to be useable.
Does someone have experienced something like this?

JMR

444

General / Re: Obtaining pixel coordinates

« on: June 10, 2014, 11:49:01 PM »

1 make sure you are selecting the right coordinate-system that applies to your country. Go to the settings button (the rightmost one) in the tool bar on top of ground-control pane.
2 update the adjustment, check residuals and optimize in case needed (this will mean redoing dense cloud and mesh)
3 export the orthophoto by choosing the coordinate system of your work. If you use geotiff, all geo-reference parameters will be embedded in file header so the tiff will be just what you want, but if you prefer png export, make sure you select "save world-file", this will tell Pscan to create a tiny file containing the same geo-information.

In case the application used for index calculation is not a geo-tool but just a generic image processor such as Photoshop, orthoimage geotag may get wiped out from the tiff header, so make sure you save the world-file as it will be read by most GIS applications if contained in the same folder as the image itself.

Split into tiles (blocks) is also an option possible in the same export ortho dialog.

Regards

445

General / Re: Obtaining pixel coordinates

« on: June 10, 2014, 08:39:33 PM »

well then my assumption was wrong, you dont need gps coordinates for a pixel in image plane but simply you want a multi-band geo-referenced orthophoto. don't you? once its done you can create a new band with the index out from maths between RED and NIR chanels by means of a suitable image processing software. Am I missing something?. Do you need index values backprojected onto original frames for some reason?

446

General / Re: Obtaining pixel coordinates

« on: June 10, 2014, 02:12:35 PM »

GPS coordinates fo each pixel makes little to zero sense for me but you if you think they are usefull you could obtain them out from pixel coordinates and camera external parameters and calibrated focal lenght.
I'd firstly transform pixel coordinates Xp,Yp (2D) in computer standard form (origin in top left pixel and +Y axis downwards) into a 3D camera centered system asumimg for each pixel
X'=Xp-(Width/2)
Y'=-(Yp-(height/2))
Z'=f
then tranfoming those coordinates into GPS is just a matter of applying omega phi kappa that corresponds a a given camera.

with regard to the other "wish" you wrote, yes, it is actually posible to use the geotags in exif if they are present as a way to speed up orientation calculations (pair preselection ground control mode) allowing Photoscan to use those coordinates as a loose constraint or intital guess.
Regards

447

General / Re: general questions

« on: June 07, 2014, 02:29:21 AM »

why not stills? if you want to make use of video frames you'll face problems caused by rolling shutters, relatively low resolution and image compression artifacts. You can also find some decent cameras that can take stills (full) during video in case the latter is a must. why not two cameras one taking photos and the second for motion
what will be your subject? the coast? the sea bed?

448

General / Re: Does this sound practical? High res scanning of over-sized prints.

« on: June 03, 2014, 12:26:26 PM »

I would recommend using a macro lens better than a tele lens (much better for camera alignment). On the other hand you should try to keep camera as perpendicular to the drawings as if it was an aircraft photo-flight.
Long time ago I built a setup for a mummy photo scan using a linear guidance system http://medias.ina.de/medias/en!hp.ec.br/LFS?#LFS for a carriage http://medias.ina.de/medias/en!hp.ec.br/LFL..-SF?#LFL..-SF mounted on an aluminium profile. http://www.boschrexroth.com/en/xc/products/product-groups/assembly-technology/basic-mechanic-elements/index

The rail was fixed on two trestles allowing smooth and regular motion for the camera.

The setup has been useful in lots of projects after the mummy, so it's an investment that is worth.

449

General / Re: Importing meshes: workflow options.

« on: June 03, 2014, 11:18:14 AM »

Very nice. The thorns & cross are CAD additions?

The thorns have actually been scanned too. The results are good as the scanner was set to highest resolution posible (about 85 points per mm along the laser line). The photo model is very bad, in this case, for reasons you surely understand. Anyway the mesh reduction has messed the thorns and makes them look awfull in close-up, but its still nice from the distance.

450

General / Re: Importing meshes: workflow options.

« on: June 02, 2014, 07:05:41 PM »

here you are an Example of a model that results from laser scanning with a Perceptron V5 on a Romer Infinite arm, textured in Photoscan from 70 photographs taken with an Olympus E3.
The ground control is shared by the two techniques and that makes just straightforward the alignment of both datasets.
The cross was removed during scans and photos. It has been CAD modeled out from scans taken with a Faro Focus3D.
The scanning and photos took 2 days, and the mesh cleanning and edition to deliverables took 1 week.
Enjoy!
https://skfb.ly/zWFP