13006
Python and Java API / Re: Estimate image quality
« on: June 18, 2014, 06:44:21 PM »
Sure, James, my fault.
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.
13007
General / Re: Texturing problem/question
« on: June 18, 2014, 03:02:41 AM »
Hello William,
With the interpolation disabled you get a lot of isolated mesh components, thus texture atlas is very scattered and consists of many tiny pieces.
I can suggest to enable intepolation for mesh reconstruction.
With the interpolation disabled you get a lot of isolated mesh components, thus texture atlas is very scattered and consists of many tiny pieces.
I can suggest to enable intepolation for mesh reconstruction.
13008
Bug Reports / Re: Crash and corruption when saving large project
« on: June 17, 2014, 06:21:43 PM »
Hello frnx,
The crash may be related with the hardware issues, so we recommend to check RAM modules using Memtest86 utility.
The crash may be related with the hardware issues, so we recommend to check RAM modules using Memtest86 utility.
13009
Python and Java API / Re: Estimate image quality
« on: June 17, 2014, 12:05:50 PM »
Hello SAV,
For the quality estimation you should use the following method applied to the chunk instance:
Then to access the quality value you need to do the following:
For the quality estimation you should use the following method applied to the chunk instance:
Code: [Select]
chunk.estimateImageQuality()Then to access the quality value you need to do the following:
Code: [Select]
camera = chunk.cameras[0]
quality = camera.frames[0].meta["Image/Quality"]
if quality < threshold:
camera.enabled = False
p
13010
Python and Java API / Re: Acessing calibration parameters
« on: June 17, 2014, 09:55:22 AM »
Hello Patrice,
If you need to access camera calibration parameters of the active chunk you need to use the following:
In your sample you just haven't applied the imported calibration to the chunk from the project and use newly created (and therefore empty and not connected to the project) sensor instance.
If you need to access camera calibration parameters of the active chunk you need to use the following:
Code: [Select]
doc = PhotoScan.app.document
chunk = doc.activeChunk
calib_in = PhotoScan.Calibration()
calib_in.load(path)
sensor = chunk.sensors[0]
sensor.user_calib = calib_inIn your sample you just haven't applied the imported calibration to the chunk from the project and use newly created (and therefore empty and not connected to the project) sensor instance.
13011
Other Languages / Re: Фильтрация модели
« on: June 10, 2014, 03:30:32 PM »
Добрый день,
К сожалению, в настоящее время PhotoScan не позволяет избирательно оптимизировать модель.
Выделение крыш в отдельный класс плотного облака, с его последующим игнорированием при построении полигональной модели может немного помочь, однако, потом потребуется закрыть отверстия перед текстурированием.
В качестве другой альтернативы, можно использовать сглаживание модели с помощью Python команд из панели Консоль:
К сожалению, в настоящее время PhotoScan не позволяет избирательно оптимизировать модель.
Выделение крыш в отдельный класс плотного облака, с его последующим игнорированием при построении полигональной модели может немного помочь, однако, потом потребуется закрыть отверстия перед текстурированием.
В качестве другой альтернативы, можно использовать сглаживание модели с помощью Python команд из панели Консоль:
Code: [Select]
PhotoScan.app.document.activeChunk.smoothModel(3)данная команда применит Лапласиановское сглаживание с шагом 3.13012
Python and Java API / Re: Remove Unnecessary Photos
« on: June 09, 2014, 02:31:11 PM »
Hello Mark,
Unfortunately, there's no such a script available.
Unfortunately, there's no such a script available.
13013
Python and Java API / Re: Question About Python Export Dense PointCloud
« on: June 09, 2014, 02:29:40 PM »
Hello ppkong,
Can you please post a screenshot of the dense cloud view in PhotoScan window for the corresponding project?
Can you please post a screenshot of the dense cloud view in PhotoScan window for the corresponding project?
13014
General / Re: poor results - any hints?
« on: June 09, 2014, 12:48:57 PM »
Hello onera,
Here's the problem is mostly in the object, I think that the same shooting scenario with more textured object would give much better results.
The workflow followed:
Align Photos - High accuracy, preselection Disabled, 40000 points per photo,
then resized Bounding Box to cover the mug and the trunk only,
Build Dense Cloud - Medium quality, Mild depth filtering,
Build Mesh - Arbitrary surface type, soure - Dense Cloud, enabled interpolation and 200000 polygons,
Build Texture - Generic mapping mode and Mosaic blending mode, texture atals size/count 8192/1.
Here's the problem is mostly in the object, I think that the same shooting scenario with more textured object would give much better results.
The workflow followed:
Align Photos - High accuracy, preselection Disabled, 40000 points per photo,
then resized Bounding Box to cover the mug and the trunk only,
Build Dense Cloud - Medium quality, Mild depth filtering,
Build Mesh - Arbitrary surface type, soure - Dense Cloud, enabled interpolation and 200000 polygons,
Build Texture - Generic mapping mode and Mosaic blending mode, texture atals size/count 8192/1.
13015
General / Re: poor results - any hints?
« on: June 09, 2014, 12:03:03 PM »
In general, apart from the texture of the object itself, I can suggest to use the image frame more effectively and to capture less background (such as far away trees, sky and etc.), use even less ISO value (but even on 200 the quality of images is very good) and increase f-stop values to provide focal depth (if you are using a tripod the latter two points should not be a problem).
13016
General / Re: poor results - any hints?
« on: June 09, 2014, 12:00:34 PM »
Hello onera,
The results are quite good for this type of object.
The cup itself is untextured and PhotoScan could not find sufficient number of feature points of such surfaces, however, there's some dirt on the objects and it actually helps to estimate depths more or less adequately.
You can see even in this dataset, that tree trunk is very textured object and PhotoScan finds a lot of feature points on it, even most of cameras are not focused on the trunks' surface.
The results are quite good for this type of object.
The cup itself is untextured and PhotoScan could not find sufficient number of feature points of such surfaces, however, there's some dirt on the objects and it actually helps to estimate depths more or less adequately.
You can see even in this dataset, that tree trunk is very textured object and PhotoScan finds a lot of feature points on it, even most of cameras are not focused on the trunks' surface.
13017
General / Re: KMZ export quality?
« on: June 07, 2014, 12:45:14 AM »
Hello Fernandoker,
Are you speaking of orthophoto or model export? In the first case you can specify the export resolution and in the latter one, the texture should be generated in the desired quality to fit expectations. However, note that extremely large texture atlas may be unrecognized by Google Earth.
Are you speaking of orthophoto or model export? In the first case you can specify the export resolution and in the latter one, the texture should be generated in the desired quality to fit expectations. However, note that extremely large texture atlas may be unrecognized by Google Earth.
13018
General / Re: Classify Ground Points not selectable
« on: June 06, 2014, 04:14:55 PM »
Hello schurli,
As it is mentioned in the corresponding tutorial (http://www.agisoft.ru/tutorials/photoscan/08/), it is necessary to reference the model in real-world dimensions to use automatic classification tool.
As it is mentioned in the corresponding tutorial (http://www.agisoft.ru/tutorials/photoscan/08/), it is necessary to reference the model in real-world dimensions to use automatic classification tool.
13019
Bug Reports / Re: matrix invert: singular matrix
« on: June 06, 2014, 12:58:29 PM »
Hello Andy,
Such behavior might be caused by the presence of the photos with too few (or even no) valid tie points on them. They can appear after manual tie point removal from sparse cloud, for example, using gradual selection tools.
So I can recommend to sort the cameras in the Ground Control pane using Projections column and to reset alignment for those with too few projections compared to others (usually t could be a few tens of points).
Such behavior might be caused by the presence of the photos with too few (or even no) valid tie points on them. They can appear after manual tie point removal from sparse cloud, for example, using gradual selection tools.
So I can recommend to sort the cameras in the Ground Control pane using Projections column and to reset alignment for those with too few projections compared to others (usually t could be a few tens of points).
13020
Python and Java API / Re: Why is exported DEM so much smaller than OrthoMosaic?
« on: June 06, 2014, 11:18:21 AM »
Hello nickponline,
Default export resolution is different for orthophoto and DEM export (you can check it in the corresponding export dialog). Effective orthophoto resolution is close to the effective resolution of the original images, whereas default DEM export resolution depend on the model reconstruction quality.
Default export resolution is different for orthophoto and DEM export (you can check it in the corresponding export dialog). Effective orthophoto resolution is close to the effective resolution of the original images, whereas default DEM export resolution depend on the model reconstruction quality.
