Messages

46

General / Re: Importing DEM to improve geomtry accuracy

« on: February 14, 2018, 03:15:14 PM »

The DEM overwrites the DEM that's already generated in the report or is simply being used as the DEM to project the images onto. That step happens at  Build Orthomosaic, where you can choose which Surface to use (Mesh or DEM).

The order of steps is;
1. Align images in proper coordinate system
2. Geo-reference the project either by geotagged images or GCPs
3. Optimize the tie-points with the info from step 2
4. Import external DEM in the same coordinate system
5. Build Orthomosaic with Surface set to DEM

47

General / Re: Large Error in GCPs

« on: February 14, 2018, 03:10:22 PM »

Isnt the disabled option of alignment the best choice as far as accuracy goes vs generic or reference? Although it does take much longer.

Generic down-samples the images before trying to find matches in the tie-points. This speeds up the alignment because Photoscan will know which images are taken consecutively to each other which speeds up the estimating camera locations step.
I'd only advice to use disabled mode either with small data sets (<100 photos) or data sets that will not properly align on Generic mode.
If you have geo-tagged images it's best to use reference mode. In this mode, Photoscan will use the coordinate information to predict which images are consecutive.

48

General / Re: Altitude or height.

« on: February 03, 2018, 08:25:01 PM »

You can't just convert AGL to AMSL because AGL is a height based on an arbitrary reference, i.e. the elevation of the project terrain. If there's no correlation between the references used there's no real way to convert the values without knowing the elevation of the ground in the coordinate system you want to work in.

What's best would be to record the start-off elevation with a high or low accuracy GPS receiver. Then change the flight log elevations in Windows Excel by adding the recorded value to the AGL elevations to get ASML or WGS84 elevations (based on your recorded value).

They also mention geoids but Lord knows how you'd figure that out without documentation.

There are courses and studies you can take to deal with these subjects.

49

General / Re: Improve Orthomosaic, HELP!

« on: February 03, 2018, 08:14:29 PM »

Create a polygon around the area of interest and right click and select set as outer boundary. Then you can export the Orthomosaic as a TIFF file and it will be cropped to this polygon outline.

50

General / Re: 18mm on full frame

« on: January 16, 2018, 04:18:19 PM »

Hi Julien999,

Built-in lens or sensor image stabilization can change interior camera geometry and alter interior orientation values between consecutive images when Photoscan might expect the same parameters. This in theory decreases the potentially achievable accuracy.

51

General / Re: GCP Error as function of distance from GCP

« on: January 12, 2018, 05:34:34 PM »

From empirical testing I found out that for projects with a GSD of 2.5 cm the maximum distances I want GCPs to be apart is around 175 meters. Beyond that, the model will start to concave.

52

General / Re: Work Flow for GCP

« on: December 21, 2017, 03:16:55 PM »

Whilst I agree that technology is getting better and more sophisticated each year, it's easy to forget the most error prone part of a project, the human factor. Because of that there's always a need for an independent check. There's no guarantee that the base station used for your RTK/PPK UAV solution is correctly set up without proper checks. If this part of the chain goes wrong the results are practically useless without checkpoints to verify your results.

We always verify our GCP coordinates compared to other reference data since a simple mistake like a wrong antenna height input before the measurement *can* happen. It doesn't matter if all goes well in 999 out of a 1000 projects. If stuff goes wrong and you have no way to check for errors or verify your findings with an independent method then you really need to consider whether you're a serious surveying company or not.

They are simply accepted to be correct.

Perhaps if your client is satisfied with sub meter accuracy. Most clients will want adequate proof of your blackbox principle height model.

53

General / Re: Work Flow for GCP

« on: December 20, 2017, 04:44:07 PM »

You're correct. It's still necessary to verify your deliverables with an independent check. Whether that's by checking the XYZ displacement on on-site measured check points or by comparing the model to another trusted accurate model usually means having to go out in the field with a RTK system.

54

General / Re: Calculate DEM resolution

« on: November 29, 2017, 01:34:09 PM »

Hi Emanuele,

That's quite simple. If you have your aligned project, simply go to File > Generate Report and save this pdf file. Open it and go to page 2. Look at the value behind the Ground resolution.

If it for instance gives you a value of 2.0 cm/pix you can calculate the native DEM resolution by multiplying the dense cloud quality levels with the ground resolution.

Ultra high  = 1 * 2.0 = DEM resolution of 2 cm
High          = 2 * 2.0 = DEM resolution of 4 cm
Medium    = 4 * 2.0 = DEM resolution of 8 cm
Low          = 8 * 2.0 = DEM resolution of 16 cm
Lowest     = 16 * 2.0 = DEM resolution of 32 cm

55

General / Re: Agisoft PhotoScan 1.4.0 pre-release

« on: September 29, 2017, 03:32:49 PM »

Appreciate it and I will try it immediately! Is there a more specific reason that the previous 50 pair limit got lifted? I mean,  can selecting more pairs lead to visibly greater results but only in certain cases, like say small object modeling?

For some projects, like building scanning from the copter (also with a lot of overlap) the limit caused the dense cloud reconstruction artifacts along the building corners.

Can anyone please confirm whether or not this will reduce the True Ortho artefacts near buildings/objects? It seems like this should. Longering the processing time but increasing the quality is definitely interesting.

56

General / Re: Don't apply masks to dense point cloud building

« on: September 27, 2017, 05:45:56 PM »

Masked areas will also be excluded from the Dense Point Cloud.

I understand what you want to do but your approach could be improved. Letting Photoscan use the entire image for the alignment is fine because after the alignment you can use the gradual selection to filter out bad tie-points. Alternatively if your terrain is very flat you can also manually select and filter vegetation tie-points. From experience however I can tell you that this barely makes a difference and only using the gradual selection tool helps much more.

If you're uncertain about the quality of your tie-points I would advice to use the 'Image Count' filter option in the gradual selection menu (found under Tools > Gradual selection). There you can select tie-points that only show up in 2 images, 3 images, 4 images etc.. That way you can be certain that your tie-points are of high quality. I would also advice to check the other filter options to reduce the max and RMS tie-points reprojection error to at least 1 pixel (error is found by right clicking the chunk and choosing 'Show Info').

After building your dense point cloud you can run the automated ground classification or manually classify points in the point cloud. These can then be turned on or off during the DEM or Mesh building process which allows you to filter out objects from your final model.

57

General / Re: Rolling shutter option detail

« on: September 26, 2017, 11:21:42 PM »

Thanks Dave for clearing that up about the mirror. I was looking at super slow motion videos of the shutter blinds of DSLRs mirror cameras and cameras like the a7r which are mirrorless and together with your post it cleared up a lot.

If there is a small and neglectable rolling shutter distortion with cameras that use a shutter blind, can these be considered global shutter cameras?

58

General / Re: Rolling shutter option detail

« on: September 26, 2017, 11:08:36 AM »

Mechanical shutters are always global shutters. The mechanical shutter is found in DSLR cameras and as far as I know refers to the glass mirror inside the camera that flips up or down to let light fall onto the sensor all at once, completely eliminating the rolling shutter distortion because the sensor is read all at once.

In general you'll either have a camera with a CCD or a CMOS sensor. CCD cameras are nearly always equiped with a global shutter mechanism but since CMOS sensors have a couple of advantages over CCD, manufacturers have started building CMOS sensors with a global shutter. Those prices have dropped down a lot and now you can find CMOS global shutter mechanisms in cameras like the sony alpha 7r.

The DJI Zenmuse X4S camera uses a small mechanical leaf shutter which, although the camera is very compact and uses a CMOS sensor, eliminates the rolling shutter distortion.

59

General / Re: Migrating Markers between Chunks

« on: September 25, 2017, 05:19:03 PM »

Tools > Export > Export Markers will generate a Agisoft XML file that can be loaded into any new Photoscan project which automatically marks all the images with the appropriate markers.

Then create a new project and go to Tools > Import > Import markers.

The only condition for this to work is that the new project has to have exactly the same files (image file names) as the project from which this file was exported from.

60

General / Re: Cordinates GCP

« on: September 21, 2017, 05:25:01 PM »

White areas means missing data areas. It seems like the bounding box is tilted.

Can you go to Tools > Reset Region and then rebuild the Orthomosaic with a geographic projection?

The wobbly roof edges are 'True Ortho' artefacts. This is due to improper geometry of the buildings due to an incomplete dataset.