Forum

Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.


Messages - JMR

Pages: 1 ... 24 25 [26] 27 28 ... 34
376
General / Re: Do I need to calibrate cameras?
« on: March 31, 2016, 08:01:48 PM »
You are welcome, avp:
Given that your small cameras have fixed optics, I guess you can rely on self-calibration if object and cameras setup geometry are aproppriate.
But once is asumed you can consider your cameras are true fixed-lens cameras, why do you reject to calibrate them with an optimal object/scene?
Even though two camera modules might look identical, they should be considered as diferent cameras and therefore, each corresponding output image sets should be grouped in the calibration dialog.
To calibrate easily, take series of convergent photos of an object with lots of texture (a granite stone wall might work great) at similar working distance you will use with your piles. Create an independent chunk for each camera module and store your calibration parameters for future use once you are satisfied with the cameras alignment.
Light sources should not influence calibration unless the key points are strongly linked to shadows borders and the light sources are in movement during photo scanning. This may lead to bad structure estimation thus may make self-calibration biased.
You can trust a regular dslr with a fixed lens (fine legacy glasses are the best choice in my opinion) if focus ring is taped or locked by any means. I myself have drilled my voightlander focus ring in three points and locking headless bolts have been added. The entire lens mount has been glued to the body side with epoxi and I don't plan to detach this lens any more unless it becomes absolutely necesary. Some cheap lens adaptors are not recommended, as they are usually machined with lower specs and the lens mount can shift sideways quite a lot.
It might sound exagerated, but, please, apply some force from side to side to your lens near the mount and will probably notice it shifts at least some tenths of mm. Lets say 0.2mm. For a usual pixel size of five microns, this means principal point is shifting about 40 pixels which is not a negligible amount. If you detach your lens, you won't be capable to mount it again in the very exact same place, and that's why I glued mine.

Best

377
General / Re: Do I need to calibrate cameras?
« on: March 31, 2016, 12:41:14 AM »
Precalibration is a need under certain circumstances. For example an object that cannot fill a frame due to its linear shape (a knife)
Also can be recommended if object's poor texture gives little opportunity for a reliable selfcalibration
If actual ground control in object side is imposible but object metrics can be defined by camera positions or scale bars between cameras, I would say it is highly required to work with calibrated (fixed) camera too.
Keep in mind that regular cameras have many moving parts. Lens barrel and groups are loose enough to allow tiny motors to move them by applying little torque. Many cameras reset/collapse their optics during the turn off/on operation and it is hard to believe they will return to any specific position from time to time.
The lens mount, the sensor suspension and the entire body are also weak points when it comes to micron level stability if you want to consider that a camera can be seriously calibrated.
Also precalibration makes sense only if the camera/lens settings (zoom and focus) you calibrate, are usable with the object you want to photoscan. So please don't try to calibrate with a chessboard on your laptop screen the camera you will use for aerial photogrammetry.
Automatic calibration works fine in most cases. It is accurate enough if you use the camera consistently, not touching zoom and focus, all image stabilization features are disabled, your shots are well distributed, with adequate overlap and the object texture collaborates. But still remember that your model's accuracy and camera calibration accuracy as well strongly depend on the ground control.
Entering camera positions a priori will not tell Photoscan not to calibrate. The only way to avoid self-calibration during the align cameras step is to set camera parameters as "fixed"

Best

378
General / Re: Getting a DEM with aerotriangulated images
« on: March 28, 2016, 10:55:47 AM »
I think there must be some inconsistencies between the camera model used by your aerotriangulation, and the one used by Photoscan. take care with units and check if you are using mm or pixels. If your photos are tagged as NC, then you must set the pixel size in mm if you want to enter lens parameters in mm.   Make sure also that you are using the same reference system.
if all is correct, then your seps should be
1 load cameras
2 import camera calibration and set as fixed
3 import externals
4 set camera accuracy in m and deg as zero or almost zero
5 build tie points

379
General / Re: Post Processing Software?
« on: March 28, 2016, 10:33:38 AM »
give a try to Virtual Surveyor https://www.virtual-surveyor.com/
Regards!

José Martínez

380
General / Re: Turntable issues.
« on: March 21, 2016, 02:15:11 PM »
Turntables are acceptable as they can actually emulate a moving camera changing its position with respect to object. This means that in each frame all objects should change accordingly to the camera movement emulation. If your rotary stage and background are starring the scene and are actually steady, then you will not get the program fooled into thinking that it is the camera that has moved.
So you need to mask every thing but the object to allow this "confusion".
Get closer to the object to fill the frame as much as posible... and better if you try with a texture rich subject. why the hell most people want to plushes end plastic toys? pick up a rough stone

381
General / Re: Aligning top and bottom of an object
« on: March 09, 2016, 08:19:31 PM »

382
can you tell us details like camera/lens, altitude over ground and distance between photos? are your cameras followed by NA,NC or just NA?

383
General / Some tutorials in Spanish
« on: March 01, 2016, 03:48:14 PM »
Hola amigos: Os invito a ver algunos tutoriales en español creados para un curso de virtualización de Patrimonio en el que participamos hace dos años. Están un poco obsoletos y fueron creados para la versión 1, a pesar de ello, espero que sigan teniendo algún valor para ustedes.

Hello friends: I invite you to a set of tutorials in Spanish that were created for a course about Cultural Heritage Virtualization two years ago. As long as they were created for Photoscan 1.0 they are a little outdated,  even though I hope you find them usefull.

Best regards
www.geobit.es (Agisoft Reseller)

Enjoy our Point cloud publication service at www.pointbox.xyz

https://www.youtube.com/playlist?list=PL_cDdjQPbX1GXV4AC3Y3dYVwTPh4kxnk-

384
General / Re: Reference to original photos
« on: February 20, 2016, 03:36:34 AM »
Two alternative ways:
a.) Right clic in the point cloud and choose filter photos by point
b.) make a small selection around the desired point and right clic, then filter photos by points
both cases will force the photos pane to show just a subset of photos accordingly.

GEOBIT

385
General / Re: Vertical and horizontal accuracy of point cloud?
« on: February 19, 2016, 11:32:44 PM »
Hello:
From my experience you cannot rely on regular any UAV on board gps unless it can deliver precise positioning by means of diferential GPS corrections. Not at all.
Not to mention many UAVs tagg photos with altitude over home (not over ellipsoid or sea level) in round meters.

386
General / Re: Export markers - ticked coordinates
« on: February 15, 2016, 10:44:31 PM »
Sounds plainly like export estimated values in ref panel by the flopy drive button

387
General / Re: Build Orthoimage option & Export orthoimage
« on: January 23, 2016, 06:07:01 AM »
30000x30000 with 24bits color depth gives about 2,6 gigabytes. Don't you think its insane for any kind of file?
It is commonly accepted that human vision power goes up to 0.2mm, so your sub-mm ortho image must be delivered (printed or showed) at a scale bigger than 1:3 in order your eyes can outresolve this image (see as pixels). It is worth then this giant effort?

Saving storage makes sense generaly and is good for the entire Earth sustainability. But that said, if you still think you have a good reason to enter the sub-mm world without purchasing memory, then you can use a script to split your project into smaller chunks. This will give you the chance to create one orthomosaic per chunk and export them as if tiles were created in the old version fashion.

Find this script file in the python section

Best
GEOBIT

388
General / PointBox launch
« on: January 12, 2016, 11:33:23 PM »
Hello friends:
GEOBIT is pleased to announce PointBox, an online pointclouds publishing tool allowing you to interact in very agile and easy way with large point cloud models in the network.

With PointBox you can show your work by means of an attractive and user friendly interface and send your customers and friends, links for them to visualize your point clouds from anywhere through Internet connection.

Please, enter www.pointbox.xyz sign up and create your portfolio.

The viewing platform is at an advanced stage of testing and we believe that it already works rather satisfactorily. However you may still have some minor flaws we'll address as soon as we can. Be patient if you suffer from any kind of problem and, please, let us to know about it. We need your feed back

The display engine is based on WebGL technology, an already well-established standard for displaying 3D data on the Internet. Therefore, for your enjoyment it is necessary that both your browser and your graphic's hardware supports this technology. If you receive error messages that refer to the lack of these features, we regret to say that you can not enjoy the cloud display in your computer.

Show us your masterpieces!, enjoy sharing!

Best regards!

389
Feature Requests / Re: Size of Marker ID text
« on: December 09, 2015, 11:56:08 AM »
You are absolutely right my request had nothing to do with Simon's  :-[ ...I should have read more carefully but original subject should have been related to targets not to markers ID. Sorry, Alexey.
Thanks

390
Feature Requests / Re: Ortho at given resolution
« on: December 09, 2015, 12:43:50 AM »
Should I turn this feature request into a bug report?

Pages: 1 ... 24 25 [26] 27 28 ... 34