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 - Tapejara

Pages: [1]
1
Feature Requests / Re: Alpha in PNG files should be treated as a mask
« on: January 14, 2014, 03:23:17 AM »
That is good. Thanks.

2
Feature Requests / Alpha in PNG files should be treated as a mask
« on: January 04, 2014, 10:19:50 PM »
In creating a model from JPeg photos, I went through the trouble of editing out all of the parts of the photos that were not part of the model (leaving Alpha=0 "transparent") and stored the edited photos to PNG files. In my mind this should have told PhotoScan to treat the transparent parts of the photos as a sort of mask to make the solver ignore those areas and consider them to not contribute to the solve of the model. But the solve created a lot of snow-like looking white stuff where there should have been no geometry. My photos had no white yet a lot of "white stuff" ended up in my model. I can only presume that the solver was ignoring the Alpha in the PNG files and was placing a default white background behind the photos and then interpreted it as part of the substance of the model. This should not have happened.

3
General / Having trouble realigning photos
« on: September 29, 2013, 08:04:48 PM »
I have a project with 84 photos. On one side of the object, PS got lost and the cameras are in wrong places. The manual says "Reset alignment for incorrectly positioned photos using Reset Photo Alignment command from the photo context menu." If I select one of the cameras there is no "Reset Photo Alignment" command in the photo menu in which all of the commands are grayed out. There is also no "Align Selected Photos" in the Photo menu as described in the next step in the manual.

4
General / Re: Document describing camera data export (XML) format?
« on: September 19, 2013, 07:10:28 AM »
I just figured it out. The transformation matrices that are produced by using the Tools/"Export Cameras..." option will transform from camera local space to model global space. Doing a matrix multiply of the transformation matrix times the origin vector [0,0,0,1] results in the camera's location in model global space. Since the PhotoScan Z Axis is opposite that of Lightwave I have to flip the Z Coordinate and the cameras end up in the right places relative to the model. The model export and import features in PhotoScan and Lightwave know how to convert between axes but my program that does the transform wasn't doing the axis conversion. Now I know what to do to fix it.

Problem solved. Thanks all for your help.

5
General / Re: Document describing camera data export (XML) format?
« on: September 19, 2013, 05:31:29 AM »
Basically, in a nutshell, what I would really like to know is what kind of transformation is done by using the transform matrices generated in the XML file when you do a Tools/"Export Cameras...". In other words, when you multiply the matrix by a coordinate vector:

1) What space is the original vector describing a point in before the matrix multiply?
2) What space is the new vector in as a result of the matrix multiply?

Is it global to camera (local) or camera (local) to global or something else?

6
General / Re: Document describing camera data export (XML) format?
« on: September 19, 2013, 12:00:44 AM »
I want to extract the camera's coordinates and its angular orientation from this matrix. This information doesn't seem to be explicitly provided anywere else in the file.

Update: I have done a matrix multiply of the cameras' transform matrices timed the origin vector [0,0,0,1] and obtained [X,Y,Z,1] coordinate vectors for the cameras. The locations are in circles just like I took the pictures but they are not oriented around the origin as I expect them to. Take a look at the screen shot I have of the scene. I would expect the cameras to surround the object I was photographing but that does not appear to be the case. Either my understanding of what information is in the camera transform matrix in the XML file is wrong or something is wrong with the export transform matrix feature of PhotoScan. In the Layout picture you can see that the origin of the scene is outside of the ring of cameras. In the PhotoScan picture you can see that the origin and model is in the center of the ring of cameras. Am I doing something wrong or is there a problem with the export feature? (There is also a possibility that the program I wrote has a problem.)

I have also done this same thing but using the inverse of the camera transform matrix and my results were similar but worse. The two concentric rings were tilted relative to eachother and they are not that way in PhotoScan.

7
General / Re: Document describing camera data export (XML) format?
« on: September 18, 2013, 07:45:30 PM »
I am assuming that the camera's transform matrix transforms any point's absolute [X,Y,Z,1] model/point cloud coordinates into a new space relative to the camera's local coordinate system. If I take the inverse of this matrix and then multiply it by the origin for the camera's local space augmented with a 1 to make a 4-dimension vector [0,0,0,1] then I should get the augmented camera's coordinates [X,Y,Z,1]. Can someone verify if this is correct?

Or is the camera's transform matrix already the inverse of the matrix that transforms camera relative to absolute coordinates in such case that I don't need to take the inverse of the matrix?

8
General / Re: Document describing camera data export (XML) format?
« on: September 17, 2013, 10:06:10 PM »
Can I derive the camera's location in X,Y,Z model/point cloud space from the camera's transformation matrix?

9
General / Re: Document describing camera data export (XML) format?
« on: September 17, 2013, 03:45:15 AM »
Are the matrix numbers listed such that rows are high-order and columns are low-order as they would be in C/C++ code? (FORTRAN order is the reverse)

10
General / Re: Document describing camera data export (XML) format?
« on: September 16, 2013, 03:27:44 PM »
From my understanding, there are many kinds of transformations. Is this the "perspective projection" transformation matrix for the camera?

11
General / Document describing camera data export (XML) format?
« on: September 16, 2013, 01:11:37 AM »
Hello,

The XML file produced when selecting Tools/"Export Cameras..." contains sixteen floating point numbers under the tag of <transform>. Is there a document somewhere that describes the meanings of each of these numbers. They are just a list and are not given any attribute names. The other information in the file seems to be fairly self-explanatory.

12
General / Re: Proper orientation of Model (and other related issues)
« on: September 15, 2013, 03:36:33 AM »
I think I can derive the information I need from the camera XML file. Thanks.

13
General / Proper orientation of Model (and other related issues)
« on: September 13, 2013, 03:15:25 AM »
After successfully digitizing an object I am now thinking about scanning environments. When I processed the photos the software produced a model that had some arbitrary size and orientation. This is okay for freeform model building but I see a lot of good information being wasted. In particular, the PhotoScan knows the relative orientations of the camera perspectives when the photos were taken. It would be nice if I could tell the software how far appart the camera positions were and what their inclinations were. This would help in setting up a coordinate system that could be used with my other software. With this information the software could make the object the right size and oriented correctly. Then it would be nice if PhotoScan could optionally export a marker for each camera position in the model. This would help in aligning the scanned environment with the CG virtual environment.

In anticipation for assisting in this process, I have purchased a geared camera head that can be leveled. I can control the three rotational orientations and measure them to a fraction of a degree. I also have tape measures for measuring short distances and a LASER range finder to measure distant objects up to about 1000m away. It would be nice to be able to feed this information to PhotoScan and have it solve my coordinate system for me (I will also need a compass). I understand that many cell phones have GPS in them and even include that information into the EXIF meta data associated with the photos. I do not have a cell phone and will be using a higher quality camera for taking all of my photos and will not have GPS data (which is only acurate to ~3m anyway).

Is it possible to do all of these things in PhotoScan or will I have to orient and scale everything manually?

Pages: [1]