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.


Topics - toxicmag

Pages: 1 [2]
16
Dear PS-Community  :)

i started processing a RedEdge dataset with Agisoft Photoscan following the pdf-guide.
http://www.agisoft.com/pdf/PS_1.4_(IL)_Refelctance_Calibration.pdf

Imported as "create multispectral camera from files as bands" and after reflectance calibration using the Micasense panel (one single shot with all 5 bands) and entering the calibration values manually, checking "use reflectance panels" and starting the process all of a sudden the TIFs turn black in workspace panel. :/

Any ideas welcome.

Thanks
Alex

17
Sometimes I experience display issues with geoTIFF in QGIS (also had this once in TatukGIS). Now i have a situation in latest QGIS running on Linux KDE. The image is kind of disturbed (s. example). If you move it it randomly gets different defects.

On another QGIS station it's shown perfectly. Also in other photo applications like GIMP, or even simple windows viewers.

What might be the reason? Looks like a rendering, or openGL issue or something like that.

Exported from 1.4.0 as a 85% compressed JPG (This is my standard value and viewed absolutely correct in other situations).

Thanks a lot.
Alex

18
Dear community,

i am working on a high detail (GSD < 1mm/px) project in our specific local coordinate system.
It's a german specialty and called Gauß-Krueger 4 (EPSG: 31468).

A typical GCP here has values of 7 digits starting with 4 or 5. Here is an example of the csv file from GNSS:

509   4575003.7167   5347169.3807   348.3360

In Agisoft Photoscan i set this coordinate system in the reference settings from the beginning, imported the GCP and later built the orthomosaic in the same EPSG code setting.

Nevertheless the tiled ortho has strange values in the tfw-file (and also appears in wrong places in QGIS).
Instead of values typical for the EPSG-code it says for example:

0.00069588
0.0
0.0
-0.00069588
-381.603842103895
21124.811800215


In my former projects this never happend. I always export to tiled orthos and the tfw file is correct.

What could be the reason from your point of view?


Greetings

Alex

19
Dear community,

i created various orthomosaics from one location - having now a big one for the whole area and various spots out of this big area but a lot more detailled.

The big one is georeferenced - the smaller detailed orthos are not (in the small areas not enough GCP were available).

I would like to reference the small ones now afterwards to have them at their correct position in the big ortho.
It would be great to have that done without having to re-process all orthomosaics. Just telling them where they are placed inside the big area and export them again.

QGIS would also be available if that helps....

Thanks in advance.

Alex


20
General / Tie points // areas of poor overlapping
« on: October 18, 2017, 06:08:21 PM »
Dear members,

i would like to create an orthomosaic of a wall.
The 3D-structure of the wall is not of any interest, just a large stitched photo with the possibility to measure distances is needed (GCPs exist).

So overlapping was not a focus (~20%) but for sure i need a clean mesh to project on.
Unfortunately there are too few tie points (and because of that holes in the mesh that cannot be closed) and the final orthophoto has a lot of gaps. I tried to improve this fact be placing markers manually but the tie points seem to exist already from one photo to the other, because i got recommended positions for the marker on the other photo.

Does anyone have a solution to get the holes closed?
Or can i somehow put a detailled scale on a stitched photo from different software?

Please take a look on the two screenshots (1 -Photoscan with holes; 2- AutoPano when just stitched).

Thanks a lot

Alex

21
Dear Photoscan-Team,

this feature already was discussed here in the past and in Berlin i talked to you about the possibility to:

# have an orthomosaic projected not just on plain structure, but also on bounded ones (i.e. a chimney)
# cylindrical would be nice, even godlike it would be to define a free plane in space

Thanks for your answers and keep up the great work.


Alex

22
General / Orthomosaic // discrepancy between markers in PointCloud
« on: March 16, 2017, 12:13:57 PM »
Dear users,

i'm building an orthophoto of a buidling facade and i'm using a local coordinate system for that.
Everything is fine, but when it comes to buidling the orthomosaic (using the specific projection pane for each side of the building) there is a shift of all the marker positions.

It's fine in pointcloud view and also mesh, so it must be some mistake in projection.

Type is "planar" and surface is mesh. Blending mode Mosaic.


Any ideas?   :/

Thanks

Alex

23
General / Cleaning point clouds // define field of view; line of view
« on: March 15, 2017, 12:20:13 PM »
Dear PS-user,

how do you deal with complex point clouds of i.e. buildings with facades and roof?
From top view the buidling is shaped like a "U" and on top the roof as well.

To get a nice mesh it is necessary to clean the point cloud, especially at all the windows and other reflections.

But besides Agisoft PS being a little weak in handling big point cloud data smoothly there is also the problem that you always have points of other parts of the building in view.

To delete and clean the points of the facade for example it would be great to define a "line of view". So you could "blank out" the rooftop for example and edit the facades points. By setting this "line of view" you could easily go through the whole project and clean up your point cloud.

Later again you simply go back to your standard view and all points are visible again and ready for meshing.

What do you think about it? Any solutions that are available right now?

Greetings

Alex

24
General / Meaning of errors in report // GeoMapping
« on: December 12, 2016, 05:58:36 PM »
Dear community,

do you know what is the exact description and definition of the error values which show the accuracy of the GCP in the final report?

To what is X, Y, Z related to as an offset?
The pointcloud?
The average above all photos and their GCP positions?

And what about the value in the last column?
0.405 pix should be definitely smaller than 2 cm....

I know that the topic is roughly described in the manual, but for GCP it says for example: "Error (m) - root mean square error for X, Y, Z coordinates for a GCP location"....

But related to what?

Thanks a lot.

Alex

PS: Example below.

25
Bug Reports / Coordinate System Conversion => theory and practice
« on: December 08, 2016, 03:27:42 PM »
Dear Support-Team and Users,

i experienced an error in the conversion of the coordinate systems.

Many of my clients in Germany still use the oldfashioned Gauss-Krueger-Format for their GCPs.
My aerial footage stores UTM positions in the photos EXIFs.

I was working on a site which is ~100m above sea level. My flying altitude was 100m and so the correct positioning of my EXIF data is at 200m for each photo.

To base this data on the same system i converted to Gauss-Krueger according to the EPSG-code my client is using (31467) and by definition the projection for Gauss-Krueger is based on a Geoid and the height-value is ~49m lower. So my flying altitude now is at 151m - at least the position of the photos.

This is not correct and does not match with the GCPs of my client which were recorded in Gauss-Krueger and the height value on ground-level is also 100m in their system. So when i imported their GCPs into my pre-processed model based on my EXIF data, their GCPs all were "floating" in 50m height.

I experimented with the "ground altitude (m)" in the Reference Settings Dialog but without any solution.


26
General / geomapping and water reflections
« on: November 28, 2016, 03:07:22 PM »
Hello Photoscan Community,

how do you deal with reflecting water-surfaces while geomapping?

I'd like to map bigger piles of sand which are surrounded by water (river, lake etc.) and are moved from one location to another to have a kind of monitoring of the sand movements.

Reflecting surfaces do not result in proper points and so i masked the water area in every photo. It is clear that i have no points in this parts of the modell to compare one mapping with the following one.

What do you suggest in that case?

Thanks a lot.

Alex

27
General / Accuracy of pointcloud // geoMapping 30acres without RTK/PPK
« on: November 08, 2016, 10:44:59 PM »
Dear colleagues,

i got very good results in precisly and manually georeferencing my 153 photos with Geosetter.

Though not using any RTK/PPK solution at all the accuracy of the point cloud seems to be good.
I measured distances within the dense point cloud and they match reality quite well.

Since global accuracy is not needed for the project i can go with the inner accuracy only.

Can you help me out with ansering these two questions:

1) What inner accuracy can one expect from this process?
2) If RTK would be used the global accuracy will be higher of course. But will also be inner accuracy within the points be better?

Here's some report data related to that flight:

Number of images: 153
Flying altitude: 95.6 m
Ground resolution: 1.74 cm/pix
Coverage area: 0.121 km²

Camera stations: 153
Tie points: 253,623
Projections: 1,430,764
Reprojection error: 1.27 pix

Camera Locations:
X error (m)  Y error (m)    Z error (m)     XY error (m)       Total error (m)
1.64238        0.841022       0.656978         1.84519               1.95866

DEM:
Resolution: 6.95 cm/pix
Point density: 207 points/m²

Thanks for your support.

Alex

28
General / 3D model (mesh) export // reduce to points only?
« on: November 03, 2016, 02:56:11 PM »
Dear members,

i'd like to export the mesh of a survey to the wide field of autodesk (mainly AutoCAD) tools for my client.
I decided to use DXF 3DFaces but the scale of the model was lost and there was no reference left.

I have two questions now:
1) Is there an export format which keeps these information?

2) Is it possible to reduce the mesh information on it's referenced points only (and having AutoCAD doing the meshing)?


Thanks a lot.

Alex

Pages: 1 [2]