Messages

16

General / Re: Aligning top and bottom of an object

« on: March 09, 2016, 04:20:01 PM »

Thank you all for your replies. I tried James's method and it works great, so now I have a full 3D model with both top and bottom. (BTW, I did not change the background, but will try that tactic for the next go round.)  The workflow blog referred to by Kiesel is also admirably clear, so thanks for that too.

My question still remains, though. How does chunk alignment proceed - sparse or dense point cloud, mesh? Like I mentioned, I removed the supporting surface from both the upright and inverted models in my original attempt, but got that weird alignment. I would have thought having just the point clouds and meshes of the pot itself would have allowed the two chunks to align.

Thanks again for your help.

17

General / Re: Bad imagery alignment

« on: March 09, 2016, 01:12:31 AM »

Do mean align rather than import? If the former, probably because of the varying reflectance patterns of the water from different camera angles.

18

General / Aligning top and bottom of an object

« on: March 09, 2016, 12:00:39 AM »

I would like to align the top and bottom of an artifact, in this case a pot with three legs, to create a full 3D model. The surface has a painted design plus cracks, and the legs are modeled, so the object is not at all a regular or featureless surface.

Rather than create masks (too time consuming for production purposes), my idea was to take the object upside down (42 photos) and create a model, than do the same with the pot right side up, 45 photos. In both cases the photos from the two chunks aligned very well, so I had one chunk resting upright on a flat surface, the other inverted on the flat surface. I made sure to take a circle of photos around the side in both cases.

I had hope that by removing both the dense and sparse clouds of the support layer (the flat surface), plus a bit more, the two chunks would align. They aligned, but with the legs from the inverted model pointing up, and those from the normal model pointing down (see attached).

Can anyone explain the error of my ways and why Photoscan will not align the two correctly (or even why it will align it the way it does!). I thought alignment was performed on the point cloud.

Thanks

19

General / Re: Difficulties measuring using scale bars

« on: March 07, 2016, 01:43:47 AM »

Thank-you Alexy for your response, but I am still having trouble, although I made sure I had checked the scaling scale bar (points 1-2). Attached are screenshots. As mentioned, Filter by Photos has nothing for any of the marker, but if I click Show Info for each of the markers, the related photos do appear (4 for the first marker, 10+ for the others).

Regarding your observation concerning coordinate systems, the manual (p. 48) states: "Model coordinate systems must be also initialized before the distance measurements can be performed. Alternatively, the model can be scaled on known distance (scale bar) information to become suitable for measurements." This seems to indicate that a projection is NOT necessary for measurement when a (known) scale bar distance is employed. Is the manual incorrect?

N.B. Answered my own question. Coordinate system need not be defined, but object can be scaled with a single scale bar. My mistake was not hitting the Settings button after defining the scalebars. Maybe the manual should be updated to reflect that necessity and the "settings" button have a more appropriate name.

Is there a way to reset scaling, say if one wishes to use a different scalebar as the primary scale factor?

20

General / Difficulties measuring using scale bars

« on: March 06, 2016, 10:53:08 PM »

I am having difficulty using a scale bar to measure between two points, and the manual is not very clear on the procedure. I am able to set marker points (both on photos and directly on the model) and then can create an initial scale bar using two of the points. I then can also click on View Source and enter the distance between a pair of the points. Any time I toggle back to View Source, the value is still there, the line between them is highlighted, as are the points. So, up to this point all is OK.

My difficult comes in the next step. I go to the Reference tab, click View Estimated. I select another pair of marker points, right-click on them, and select Create Scale Bar. The bar is created and is selected. However, no Distance est value appears next to it. I double click on the area - nothing. The boxes next to both scale bars is unchecked. Can't figure what I am doing incorrectly, and in fact it worked once when I was clicking randomly about. Unfortunately I couldn't replicate it.

I also notice that if I right-click on the marker point and select Filter Photos by Marker, nothing appears (though the markers do in fact appear in multiple photos).

Using v. 1.2.3 Pro, Windows.

Any help would be greatly appreciated,  I really need the information.

Thanks!

21

General / Align selected cameras

« on: March 02, 2016, 01:01:59 AM »

Can anyone explain why "align selected cameras" works on (some) photos following an alignment, but these same photos are not aligned in the original alignment pass? In other words, why does alignment initially fail for a given photo, then succeed during "align selected cameras"? Shouldn't the program be "smart" enough to make a second pass during full alignment to see if unaligned photos can be aligned before finishing?

22

Python and Java API / Re: Extracting profiles

« on: December 02, 2015, 11:47:31 PM »

Thank you Alexey for your quick reply. Your solution was ingenious and I was able to extract a clean edge from the bounding box. However, my fill hole still had problems. I am enclosing some screen shots to show the problem. The edge is indeed straight and planar, but there are still some annoying and incorrect facets. Is there a way to explode the fill hole so that some triangles can be omitted? or is there a problem with the algorithm?

Thanks again for your help.  Bill

23

Python and Java API / Extracting profiles

« on: December 02, 2015, 06:13:49 PM »

I would like to extract the profile of an object, such as a pot, from my Photoscan model. My idea is to slice off half the pot at midline, then use fill holes to provide the profile (I want a solid representation of the cross-section). It works pretty well, but the edge produced by slicing the mesh is rather ragged, which can lead to spurious facets on the profile. This is true even with mesh reconstruction at the highest settings.

Has anyone devised a script to provide a "clean" linear cut through a mesh?

24

General / Difficulties with align chunks using cameras

« on: November 03, 2015, 05:20:15 PM »

I have been using align chunks with cameras with success to preview alignments. However, I am having difficulty aligning chunks created from photos that I downsampled using Photoshop. The photos align well enough in individual chunks, but I can't get the chunks to align using camera positions. I have tried building "bridge" sections to tie chunks together better; the object is a long low building and the "bridge" extends its entire front, but no luck.

I have made no attempt to modify the settings of the downsampled photos, other than to note that they exist (though the dimensions are correct, some fields appear unaltered from the original). Could this be causing the problem? Like I said, I have built bridges using photos from two other chunks, but they will then not align.  Downsampled from 5184x3456  to 3456x2304.

25

Python and Java API / Re: identifying non-aligned photos

« on: October 23, 2015, 03:07:49 PM »

Thank you James, worked perfectly.

26

Python and Java API / identifying non-aligned photos

« on: October 23, 2015, 02:29:52 PM »

I am new to Python scripting. What is the property that identifies a photo/camera as unaligned (N/A)? I searched the documentation but couldn't find it. I would like to run a script to list all photos/cameras that are not aligned.

27

General / Re: Export pdf - I only get a tiny portion of the model

« on: August 20, 2015, 01:31:50 PM »

To create a static PDF, you would export an orthophoto in either TIFF or JPG format (if you have PS pro). Then you can open that in Acrobat directly and it will import it to a page. If you wish more control over the import, you should import the orthophoto into something like Illustrator or Canvas, format the page, and export as a PDF. In a pinch you could even use Word or Powerpoint.

28

Feature Requests / Re: Adding new photos to existing chunk

« on: May 18, 2015, 05:32:18 PM »

Could this be acted upon, for all the reasons given by the other authors? It would be a tremendous time-saver.

29

General / Alignment questions

« on: May 17, 2015, 09:05:01 PM »

I have a question about alignment procedures and strategy. I made a medium-level alignment of 178 photographs and all of them aligned. I then added 30 images and performed another medium alignment. 195 aligned, but 14 of the original images failed to align (so 9 of the new additions did not align as well). I then performed a high-level alignment on the 208 photos. Only 183 aligned, but 13 of the 14 images from the first round returned. In all cases the aligned photographs appear to be where they should be.

Since I would like to use medium alignment as a preview in the field to see if my coverage is sufficient, I wonder if I could get an explanation of why photos are rejected in medium vs. high level, and why aligned photos might get rejected when additional photos are added.

30

General / Re: Why this is occuring?

« on: January 05, 2015, 12:22:52 AM »

As noted in the help files, Photoscan does not do well with shiny or smooth surfaces, the former I would imagine because the reflection will differ depending on the angle of the photograph, the second because alignment depends on contrasts in surface texture. There are several messages concerning how to take cars and other shiny rounded objects you might consult. Have you tried using markers?