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 - Alexey Pasumansky

Pages: 1 ... 867 868 [869] 870 871 ... 880
General / Re: ortophoto export
« on: December 14, 2011, 03:36:16 PM »
Hello Wishgranter,

In PhotoScan it is possible to set up the known reference distance between two points (with markers placed on their locations) of the model or point cloud. In this case if there are no GCP information the local coordinate system would be scaled using the reference distance info.

 Is that an answer for your question or you mean something more complicated dedicated for orthophoto tif files processing?

General / Re: performace question
« on: December 14, 2011, 03:22:30 PM »
Hello Wishgranter,

PhotoScan uses single precision arithmetics in the GPU processing part (actually the most time consuming calculations are done even using integer arithmetics).

So the double precision performance limitations of the consumer video cards doesn't affect the processing speed at all. Consumer GeForce video cards should be as fast as professional (Quadro or Tesla) ones for PhotoScan.

General / Re: 8 vs 16 bit Images
« on: December 07, 2011, 07:50:42 PM »
Hello Wishgranter,

Currently 16 bit tiffs could be uploaded into PhotoScan but there would be treated as 8 bit, because by now PhotoScan support only 8 bit images.

Feature Requests / Re: Automatic Cropping
« on: December 06, 2011, 01:17:24 PM »

PhotoScan can import masks from alpha channel. If you can apply alpha to the unwanted background for the whole set of images in the external software it could be treated by PhotoScan as mask (Photo menu -> Import Mask).
Also you can propagate mask from the single photo to the whole chunk or workspace, for example cropping the center area containing the object.
Maybe these options could be used in your workflow.

But what do your mean by automatic detection of the region of interest? Just automatic Magic Wand tool or anything more "intellectual" such as detecting of the object's borders on the complicated background.

General / Re: Pre-sales query
« on: December 05, 2011, 03:59:13 PM »
Hello Mr. Curious,

If the stickers are too small it would not help, because there would still be large areas.

I think there could be a few alternative ways. For example you can project some kind of pattern on the surface or cover the car with something that could be easily washed off (whitening or flour, for example). Maybe placing some kind of textured materials inside the car to the windows would also be helpful.
Or you can take a photos of heavily mudded vehicles.

General / Re: Pre-sales query
« on: December 05, 2011, 11:29:14 AM »
Hello Mr. Curious,

Wishgranter is right - there could be some problems with geometry reconstruction of large shiny and untextured areas.
But we could also try to process your sample photos of the parked cars.

General / Re: whats the best set-up
« on: November 30, 2011, 01:18:19 PM »
Hello tbone,

Concerning the RAW data please refer to the following topic:
As it is a closed format PhotoScan supports .CR2 files only based on our experience and opened information. So we recommend to convert .CR2 files to JPEG or TIFF format using camera-native software before processing in PhotoScan.

The processing workflow of the images taken by synchronized cameras does not differ of the standard workflow. And in general several synchronized cameras are useful when shooting moving objects.

Background could be useful if PhotoScan doesn't find enough points on the object's surface (for example untextured or shiny objects) and it could be taken into account at Align Photos stage. Plain background is good if you want to mask everything but the object or even it is not required in case of plain and untextured background.

Добрый день, Дмитрий,

Спасибо за предложение, мы подумаем над тем, как лучше всего реализовать подобную возможность. Насчёт сроков реализации пока сложно сказать что-то конкретное.

Bug Reports / Re: Model failure
« on: November 28, 2011, 03:53:01 PM »
Hello John,

How many points are there in the point cloud and do you rebuild the cloud each time before reconstructing the geometry?

If it is possible please provide us this project with two photos (please send it to and we'll try to investigate the problem.

General / Re: forest alignment problem. HELP
« on: November 28, 2011, 03:08:01 PM »
Hello bagus,

The problem with this data set is that the areas seen by three cameras at once are too small.

The possible solution for these six images is to split the project by chunks of two photos (first with second, second with third and etc), add markers using the same labels for different chunks so that there are at least three of them in the overlapping area of three photos (from two consecutive chunks). After that is done for all the six chunks proceed to Align Chunks command using marker based option and merge chunks at least.

We've managed to build a geometry for your data set using the described way (see screenshot attached), but still there could be accuracy problems because the same cameras from different chunk do not match.

If you wish we can send you the project (saved in PhotoScan 0.8.4).

General / Re: import mask
« on: November 28, 2011, 01:41:49 PM »
Hello kiteman,

PhotoScan allows to import mask from alpha-channel. This feature is available from Photo menu by Import Mask command.

Bug Reports / Re: Model failure
« on: November 25, 2011, 06:07:45 PM »
Hello John,

Are you using PhotoScan on 32-bit  system? And does "Not enough memory" message appears when building geometry fails?

Feature Requests / Re: Visual DTM Alignment for DEM Export
« on: November 25, 2011, 12:32:52 PM »
Hello Tyler,

I think that you can try to align the chunks containing these two models. If the alignment is successful the reference between chunks could be better then you have with GCPs.
But if the difference between models is too big it wouldn't work if chunks are being aligned based on points.

Добрый день,

В принципе, это возможно реализовать, однако, насколько мне известно, wrl-формат позволяет привязывать снимки только целиком, без учёта рельефа и коррекций дисторсий, что скорее всего скажется на точности привязки и правильности перекрытий.
Правильно ли я понимаю, что предложенный вариант аналогичен накидному монтажу?

General / Re: Moving object
« on: November 25, 2011, 12:16:41 PM »
Hello rurik,

Could you please provide a screenshot of the camera positions after alignment for better understanding the case?
I think there could be some problems if you have too much shots made from the same positions or have several photos which sensor axes lay on the same line.

Pages: 1 ... 867 868 [869] 870 871 ... 880