Forum

Author Topic: What Vertical Accuracy Should I get?  (Read 7746 times)

sarko

  • Newbie
  • *
  • Posts: 23
    • View Profile
What Vertical Accuracy Should I get?
« on: May 26, 2012, 12:08:26 AM »
Hello,

I have been a Photoscan user for some time, but typically use older, course resolution imagery.  I have recently received a set of about 300 high-resolution images collected from a helicopter.  They were collected with a Nikon D3 camera and generally are very high quality.  I have run through my typical workflow, which is:

1.  Align photos on high
2.  Run a low geometry model for quick verification
3.  Insert ground control.  Ground control for this project were collected by a professional survey crew and should be accurate to a few centimeters. 
4.  Adjust ground control marker in images to ensure markers are properly placed
5.  Optimize model
6.  Generate Medium heightfield model
7.  Output Orthoimage and DEM (output resolution is 1.1cm on the ortho and 7cm on the DEM)

I have used 6 gcps spread throughout the image area and have about 10 reserved as check points.  When I perform my accuracy analysis on the 10 check points, I get a horizontal CE95 of about 4.4cm.  That is very good and certainly good enough for my application.  However, when I run the analysis on the DEM, I get a LE95 of about 19cm.  This is not as good as I would like and worse than I was expecting.  Given other people's experience with high resolution imagery, is this a reasonable value for elevation error? 

Also, would it help to run the model as high or ultrahigh?  I typically don't see much advantage to doing this with the historic imagery, but perhaps it is more useful with high-resolution imagery. 

Thanks,
Scott

Diego

  • Full Member
  • ***
  • Posts: 167
    • View Profile
Re: What Vertical Accuracy Should I get?
« Reply #1 on: May 26, 2012, 03:33:29 AM »
Hi,

For me, this is not a high-resolution camera. Not to mention the lens that you find used. If you want high precision, then you must use professional equipment, such as a medium format metric camera. For the camera you used the accuracy is reasonable. This is photogrammetry and PhotoScan is good but not do magic. The input is the most important, after all, is only correlation of images, and its effectiveness depends on many factors, shadows, etc.. This is not a lidar.

Regards,

Diego

fpbv

  • Full Member
  • ***
  • Posts: 102
    • View Profile
Re: What Vertical Accuracy Should I get?
« Reply #2 on: May 26, 2012, 06:36:20 PM »
Sarko

It depends of a lot factors, such:

- lens: fixed or zoom;
- flying height above ground;
- final user: civil engineering, forestry, mapping, etc;
- which accuracy are you expecting;
- contour lines: 1 m, 5 m, 10 m,
- no a troubled flying with a lot inclination...

The camera you got the imagery is very good, it has some good capabilities, but like you saw the altimetry it is the main concern.
For planimetry is fine, but you should expect no more than 12cm  to 16 cm.
You can have it good altimetry but you should look at all those things.
As soon you tell us about all this so we can find properly what is wrong.

Fabricio

macsurveyr

  • Jr. Member
  • **
  • Posts: 69
    • View Profile
Re: What Vertical Accuracy Should I get?
« Reply #3 on: May 26, 2012, 11:39:11 PM »
There is no reason you cannot meet or exceed lidar accuracies with photogrammetry - yes even with a lesser camera than a Nikon D3. When I say photogrammetry, I mean any photogrammetry based software and that includes PhotoScan, but there are rules that must be followed with regards to overlap of the images and more importantly, the base to height, or the geometry of the camera to subject. The focal length of the lens is the key to forming good geometry. You did not mention the lens used, but the best geometry is created using a lens in the range of 24 mm to 35 mm - expressed in full size 35 mm sensor equivalent terms. The only other source of error is the combination of lens and camera sensor distortions and their stability. A very high quality camera calibration can be derived if you have enough good photos (not blurry) with consistent settings, proper geometry, adequate overlap, and enough redundancy - overlapping strips or cross strips - so that lens parameters can be modeled correctly. I don't know if your set of photos meet all the requirements for PhotoScan to solve for a high quality calibration, but it is certainly possible with proper photos.

The differences you are seeing in your analysis of the DEM may be misleading. What are the errors shown in the Ground Control panel in PhotoScan?

Tom

Oli63

  • Jr. Member
  • **
  • Posts: 70
  • www.geospy.at
    • View Profile
    • GeoSpy Aerial Imaging & Mapping GmbH
Re: What Vertical Accuracy Should I get?
« Reply #4 on: May 27, 2012, 06:05:48 PM »
My last project with 160 photos produced the following results with 10 GCPs and 5 control points.
The GCPs are the 0s:

Error(m)         X error         Y error      Z error
0.000000         0.000000         -0.000000   0.000000         
0.019324         -0.003320      0.003339      -0.018741         
0.046620         -0.030979      0.010814      0.033117         
0.052064         -0.007972      0.021518      -0.046734         
0.000000         0.000000         -0.000000   0.000000         
0.000000         -0.000000      -0.000000   -0.000000         
0.081491         0.034416         -0.034463   0.065336         
0.000000         -0.000000      -0.000000   -0.000000         
0.000000         0.000000         -0.000000   0.000000         
0.000000         0.000000         -0.000000   -0.000000         
0.000000         0.000000         -0.000000   -0.000000         
0.000000         0.000000         -0.000000   -0.000000         
0.000000         0.000000         -0.000000   -0.000000         
0.000000         0.000000         -0.000000   0.000000         
0.084857         0.008853         0.015547      -0.082950

I  used a 16 mm lense and a 24 MP camera. Certainly important is the overlapping of photos. I used 65% in both dimensions.

What's interesting to me is: how do you find out the CE95 and LE95 values? It would be great to have a tool in PS for them. Customers are interested in an objective criteria for the accuracy.
Greetings
Oliver
GeoSpy Aerial Imaging & Mapping GmbH
www.geospy.at

sarko

  • Newbie
  • *
  • Posts: 23
    • View Profile
Re: What Vertical Accuracy Should I get?
« Reply #5 on: May 29, 2012, 07:04:19 PM »
Thank you to everyone for the inputs.  I will use them and see if I can improve my results.  As far as the CE95 and LE95, I calculated these from the check points I had using the FGDC formulae.  That is what many users in the U.S. use to quantify the quality of a geospatial data product.  It would be nice if Photoscan included these, but perhaps they are not standard around the world and so might not be of great use. 


Thanks,
Scott