How to Average Multiple Occupations of the Same Base

Kelly Bellis

ME PLS 2099
5PLS
Matt's announcement (#6) and my guilt for unintentionally hi-jacking Darren's thread precipitates this new one as I seek further understanding from behind the veil.
 

Matt Johnson

Well-Known Member
5PLS
Use CoGo Average to average the desired base station coordinates together. Then shift all the desired base station sessions to this averaged point with M-Local.
 

Shawn Billings

Shawn Billings
5PLS
What Matt has suggested is exactly what I would suggest as well Kelly. I just attempted it with a project I did a few weeks ago and hit a couple of bumps though. Theory meets Practice. It seems that the solution type for the base point restricts the base point's contribution to the average. I have three points, two are flagged as ST while one is flagged as FX. The average seemed to only use the FX point. We'll investigate further.
 

Kelly Bellis

ME PLS 2099
5PLS
Matt's metthod (even if it worked) doesn't address the idea. Simply averaging three points by adding them up and dividing by three treats them all as having equal weight, and maybe we just settle for that, but the original idea broached in February allowed for each sessions statistical uncertainty to be involved - I hear Alexey groaning all the way from Moscow to Ellsworth.

For any type of averaging to be successful, regardless of methodologies, the list of base files, I think, would need to be similarly presented as they are in the DPOS screen:
BACKGROUND-TASKS-MANAGER_20160519-09.31.42.png


Note in this example screen shot that 17U was occupied in no less than four separate recent projects just on this page alone. The others prefaced by my go-to default autonomous naming, GNSS1, are all uninteresting, boring and irrelevant real survey jobs. The recent projects all tied to 17U all relate to the past three months of my vetting of Phodar and Pix4D involving a small railroad bridge crossing a nearby narrow street. 17U might also be helpful in testing JField's potential for producing Uber-Control.
 

Shawn Billings

Shawn Billings
5PLS
The average tool has performed a weighted average since its inception. It is not an unweighted mean unless it is an average of design points (which have no accuracy estimates).
 

Kelly Bellis

ME PLS 2099
5PLS
The average tool has performed a weighted average since its inception. It is not an unweighted mean unless it is an average of design points (which have no accuracy estimates).

Is this true even when the base station is in various projects? I've always thought that Averaging was done on a project by project basis.
 

Matt Johnson

Well-Known Member
5PLS
Is this true even when the base station is in various projects? I've always thought that Averaging was done on a project by project basis.

Currently I don't believe there is a way to average points from multiple projects. Is this a real situation that you need to average points from multiple projects?
 

Kelly Bellis

ME PLS 2099
5PLS
Real? I don't understand. Yes, these are real honest to God paying client type projects in addition to my mad scientist experiments, which is where I prefer to do testing first before getting REAL jobs all messed up, like yesterday's project. I'm still hoping to salvage that one, which doesn't yet involve averaging beyond what JField is currently capable of doing, which will be perfectly fine for multiple shots on monuments or just wire fence remains.

After I get GeoCalc 2016 installed on my new machine, I'll be able to assess if in fact that there was any significant damage done by the current stable release 2.0.0.219 DPOS bug. I'm hoping that there wasn't. Oops, sorry, guess I'm blending threads.
 

Shawn Billings

Shawn Billings
5PLS
Sorry to be dense, but what are you attempting to "salvage" and what "damage" are you concerned with?

Matt wasn't asking if you had real projects or real clients, he's trying to assess whether the request merits the resources to implement. Do you commonly find yourself needing to average points in multiple projects? If it's a sincere need I'm sure it will be met with a sincere solution.
 

Kelly Bellis

ME PLS 2099
5PLS
I sincerely wish that the spread in ellipsoidal differences seen thus far in the range of nearly 5cm could be handled by JField. Mr. Frame uses StarNet and the G-File exports afforded by JField for example, so I'm not the only user that is concerned with this issue.
 

Kelly Bellis

ME PLS 2099
5PLS
Sorry to be dense, but what are you attempting to "salvage" and what "damage" are you concerned with?

Shawn, you're one of the least dense people I know. The damage is probably, hopefully limited to just my lack of confidence in knowing what I'm doing fueled by seeing the numbers relating to 95% Confidence Ellipse:

BASEROVER-STATISTICS_20160521-14.08.35.png
 

Shawn Billings

Shawn Billings
5PLS
I sincerely wish that the spread in ellipsoidal differences seen thus far in the range of nearly 5cm could be handled by JField. Mr. Frame uses StarNet and the G-File exports afforded by JField for example, so I'm not the only user that is concerned with this issue.

I'm confused. I thought we were talking about the ability to perform a weighted average of base coordinates from various projects.

You're now interjecting an OPUS vs DPOS height difference and a bug that shows high RMS values in a statistics screen in J-Field. I'm just struggling to keep up. I don't know what to tell you about OPUS. We don't have any control over OPUS. If it's a concern for you, I'd recommend contacting the OPUS team. I promise we will provide any relevant information we can from our end, but OPUS is an independent service.

I can say that the RMS values being incorrectly reported in the statistics screen has been fixed in a Pre-Release version of J-Field. I can understand your hesitation to upload a Pre-Release version, but this issue has been addressed and should be coming to you soon. I believe you can look at the Process Point Information screen (which shows the various solutions for a particular point) and see the proper RMS values. It's good that you pointed this issue out. It's been needing correction for a while in the test versions, it apparently fell between the cracks, but I am fairly certain that it is cosmetic for the most part, the exception being weighted averaging and possibly the export of vectors, although I'm not altogether sure about the effect, if any, on weighted vector export.

Just to be clear, is there a need to be able to do a weighted average of points that are stored in various projects?
 

Kelly Bellis

ME PLS 2099
5PLS
I'm confused. I thought we were talking about the ability to perform a weighted average of base coordinates from various projects.

That's correct.

You're now interjecting an OPUS vs DPOS height difference and a bug that shows high RMS values in a statistics screen in J-Field. I'm just struggling to keep up. I don't know what to tell you about OPUS. We don't have any control over OPUS. If it's a concern for you, I'd recommend contacting the OPUS team. I promise we will provide any relevant information we can from our end, but OPUS is an independent service.

I never mention OPUS, but you have. I prefer not to bring that into this discussion, and I didn't mean for that to have been suggested and I'm sorry if that's what confused you.

What I was referring to is that the differences in height which are produced by DPOS for a single base station when it has been occupied multiple times and on each of those occasions, for a lengthy session; i.e., several hours worth.

Just to be clear, is there a need to be able to do a weighted average of points that are stored in various projects?

Exactly. Base points in particular.
 

Shawn Billings

Shawn Billings
5PLS
Ok. I see. Thanks, Kelly.

It may be best to allow users to use the clipboard to copy survey points from one project to another. Currently we can copy a survey point to the clipboard, but it becomes a design point, losing its statistics in the process. If you could copy a base point, BaseA from Project A into Project B, then you could average BaseB and BaseA to come up with a better base coordinate in Project B. For Project A, you would need to copy BaseB into Project A and average BaseA and BaseB to come up with a better base coordinate in Project B.

I think there are times that importing survey points could be helpful for various tasks, it just hasn't been a high priority to include them in the clipboard.
 

Shawn Billings

Shawn Billings
5PLS
I sincerely wish that the spread in ellipsoidal differences seen thus far in the range of nearly 5cm could be handled by JField. Mr. Frame uses StarNet and the G-File exports afforded by JField for example, so I'm not the only user that is concerned with this issue.

Tonight I ran a test. I set up a base with autonomous position. I then collected 9 points from the base. The base was named Base1, the points were L001 through L009. I then moved the base to L003, which I called Base2 and then collected the point my base was previously occupying (Base1) and named it L010. I then collected points on L001-L002 and L004 through L009 (I collected points on each point except the point my base was now occupying.

I exported vectors as a G-File and exported a pdf file.

I then processed the points using DPOS. Once I obtained solutions from DPOS, I forced all of the points to the post processed solution.

I exported the (now post processed) vectors as a G-File and also exported a pdf file.

Comparing the pdf files, everything looked exactly right. The solution type changed in the pdf report and the coordinates were all slightly different as expected

I then processed the RTK vectors in Carlson Software's SurvNET. I then separately processed the Post Processed vectors in SurvNet. The RTK yielded slightly better results, but I saw no glaring issues. I did have to massage the point naming a bit to get the files to behave properly, but I expect that with least squares software.

If anyone is interested, I've attached the pdf files and the SurvNET reports are compressed.
 

Attachments

  • DPOS Vector Test SurvNET reports.zip
    8.5 KB · Views: 361
  • 2016-05-21 A-160521 Ppk.pdf
    11.1 MB · Views: 358
  • 2016-05-21 A-160521 Rtk.pdf
    11.1 MB · Views: 340

Shawn Billings

Shawn Billings
5PLS
Shawn - did you mean to post this under this thread?

It seemed as appropriate as any other.

I sincerely wish that the spread in ellipsoidal differences seen thus far in the range of nearly 5cm could be handled by JField. Mr. Frame uses StarNet and the G-File exports afforded by JField for example, so I'm not the only user that is concerned with this issue.

It appears that the G-file export works without issue. I'll investigate a bit more after submission to DPOS with base shift and report if I find any issues.
 

Kelly Bellis

ME PLS 2099
5PLS
It seemed as appropriate as any other.

Are you saying the answer as to how to average multiple occupations of the same base, be they in one or multiple projects, is using third party software, be it StarNet or Carlson? That's fine, and is what I understood from past discussions, but still wondered if I could find a way in JField 2.
 

Shawn Billings

Shawn Billings
5PLS
Are you saying the answer as to how to average multiple occupations of the same base, be they in one or multiple projects, is using third party software, be it StarNet or Carlson? That's fine, and is what I understood from past discussions, but still wondered if I could find a way in JField 2.

No. Perhaps we aren't communicating well. For this I apologize. You seemed to have concerns about pdf export not showing the buckets properly and coordinates not properly updating. You also seemed concerned about vector export based on the mention of Jim Frame's use of Star*Net. I see that this is not the issue.

Currently there is not a simple solution for averaging base sessions from multiple projects in J-Field. There is a solution for base averaging within a project, once one last issue is resolved. I have an idea for a solution across projects, but currently it doesn't exist. The test above demonstrates that vector export is working properly with the new multi-solution approach in J-Field 2.0. It also demonstrates that the pdf report does indicate which solution type was used and updates the coordinates correctly based on the selection in the Point Solution Information screen in the Points menu.
 
Top