variance covariance matrix

giulio

Member
Waiting to receive the new tool I would like to remove my doubts: doing an RTK survey with Triumph ls in addition to the coordinates, the pdop, the number of satellites etc. the variance covariance matrix matrices are also memorized for each detected and memorized point, in fact these data are requested by the Cadastre as an index of the quality of the survey.
 

Attachments

  • matrix.JPG
    matrix.JPG
    94.2 KB · Views: 212

Jim Frame

Well-Known Member
Giulio: you can export this information in gfile or Star*Net format. You may have to reformat the data to conform with the format you're required to submit, but that's just a matter of manipulating the ASCII records.
 

giulio

Member
It would be enough for me to go out on a file like that of the promark 500 receiver (from survce), which I had before the navcom (which instead did not include in survce), with the following lines for each detected point.
 

Attachments

  • varia.JPG
    varia.JPG
    61.7 KB · Views: 183

Vladimir Prasolov

Well-Known Member
JAVAD GNSS
Dear Giulio,
We want to add covariance matrix into PDF report for release J-Field version. Please let me know what requirements for covariance matrix is requested by the Cadastre. I could provide that matrix in geocentric XYZ and/or topocentric NEU frame. Also I can use selected length units or SI meters. Please let me know if you need just position covariance matrix excluding time. With time it would be 4x4 symmetric matrix (10 unique values)/ Without time it would be 3x3 symmetric matrix (6 unique values).
 
Last edited:
Vladimir: you may have noticed that the 2021 ALTA/NSPS specifcations here in the United States now include a reference to the covariance matrix for the relative positional accuracy of any two adjacent points in the survey. I'm not sure if what is currently provided in the J-Field ALTA report satisfies that requirement.
 

Vladimir Prasolov

Well-Known Member
JAVAD GNSS
Vladimir: you may have noticed that the 2021 ALTA/NSPS specifcations here in the United States now include a reference to the covariance matrix for the relative positional accuracy of any two adjacent points in the survey. I'm not sure if what is currently provided in the J-Field ALTA report satisfies that requirement.
Dear Glen,
We are providing relative accuracy according Section 3.E.i. It says about how to compute relative accuracy:
'it can also be computed using “the full covariance matrix of the coordinate inverse between any given pair of points.”' It is not about how to represent it. It is represented as usual horizontal, vertical and 3D error. In HTML reports the 95% error ellipse parameters are shown.
 
Last edited:

Jim Frame

Well-Known Member
Does the algorithm take into account centering error somehow? Without that you don't have a result that meets the ALTA specification for the points in the ground, only for the antenna positions. Beware of what you're certifying; GIGO.

P.S. As the data above illustrates, RTK positions close together are generally going to fail the test. I only use RTK for ALTAs on large properties, and never for lines under 1,000 feet long.
 

Vladimir Prasolov

Well-Known Member
JAVAD GNSS
Thanks Vladimir, the matrices required by the "Pregeo" Cadastre program are those in the annex.
Thank you Giulio. We will add matrix too in topocentric coordinates. Unfortunately sometimes easting and northing are denoted as X and Y. So it is not clear from image it is topocentric or geocentric. But Z error is bigger, so I suppose it is topocentric.
 

Shawn Billings

Shawn Billings
5PLS
Does the algorithm take into account centering error somehow? Without that you don't have a result that meets the ALTA specification for the points in the ground, only for the antenna positions. Beware of what you're certifying; GIGO.

P.S. As the data above illustrates, RTK positions close together are generally going to fail the test. I only use RTK for ALTAs on large properties, and never for lines under 1,000 feet long.

Cluster average reduces the magnitude of the error estimates such that the resulting average position can usually pass the standard.
 

giulio

Member
I could snbagliarmi, but in my opinion they are geocentric, the program asks for the Dx Dy and Dz geocentric components so the matrices should be geocentric too. Then subsequently it elaborates the relief and supplies the coordinates etrf2000 (2008.0)
 

Attachments

  • covar.pdf
    135.3 KB · Views: 220

Vladimir Prasolov

Well-Known Member
JAVAD GNSS
Thank you Giulio! I added topocentric covariance matrix as most informative. What you think is better to do: replacing it with geocentric or output both topocentric and geocentric? I think all PLSs will say their opinions too.
 
Last edited:

giulio

Member
The Cadastre's request to insert the matrices relates to an initial indication of the quality of the baseline. Honestly, I couldn't figure out if they want the topocentric or geocentric one. However, they shouldn't be nearly equivalent or am I wrong? Another question is it possible to export a file, to load on the topography program, similar to the * .rw5 file that comes out of Survce? Thank you
 

Vladimir Prasolov

Well-Known Member
JAVAD GNSS
The Cadastre's request to insert the matrices relates to an initial indication of the quality of the baseline. Honestly, I couldn't figure out if they want the topocentric or geocentric one. However, they shouldn't be nearly equivalent or am I wrong? Another question is it possible to export a file, to load on the topography program, similar to the * .rw5 file that comes out of Survce? Thank you
Dear Giulio,
We will output both matrices. They are linked through rotation. Units are different. For topocentric matrix we are using user-defined squared length units. For geocentric covariance matrix we are using SI squared meters due to it is related to global system. In my opinion the 2D or 3D error ellipses/ellipsoids are more informative for people. It is easy to see in what direction the error is biggest. Matrixes are more machine readable. It is difficult to recognize such information from covariance matrix especially if it is geocentric :)

1621416018796.png
1621416110689.png
 
I second Jim's suggestion above about centering error (this has been brought up before). While we're tweaking things, I think that i would be a great option for the operator to be able to enter an estimated centering error, at least for the rover. Granted, if the LS were mounted on a well-calibrated optical plummet tribrach (I suspect that this is hardly ever done), the uncertainty could be as low as 0.005'. But, on a 5 or 6 foot prism pole with a 40' bulls-eye level, it could be several hundredths of a foot. An 8' vial, like on the Javad pole, reduces the uncertainty, especially with a bipod, but it still is a factor that contributes to the overall positional uncertainty.
 

giulio

Member
Thanks Vladimir and Andrey, I agree with you the error ellipse is much more practical and immediately understandable.
 
Top