Share:
Notifications
Clear all

Help to Improve Localization Practices

tkc1
 tkc1
(@tkc1)
10+ posts Member

Morning all, I have a couple questions about a localization I am working on.

It is a road reconstruction project (~1km in length), running similar to if you imagined a right angled triangle - the long side, cutting the corner of the two more major cross streets. Currently, I have 4 property bars and two cut crosses with coordinates (along the road on either side, not surrounding), that work well together and show low residuals. The scale of the project is coming out to 0.99999. 

Walking on site, I seem to align with existing structures closely. My first question is if I should now create some additional control points of my own surrounding the road (on side streets). I do not have any information about property bars etc. on other streets, or coordinates to apply to them, so my feeling is that it won't help much. I may be missing something though.

My next question, which is bothering me more, is that the consultant supplied me with 4 vertical control points that aren't working together as nicely as I would like. My typical practice is to select one, set that as my vertical control and check the others. When I do this, two of these at the opposite end are out by 4-5 cm. I don't normally use more than one point vertically in a job, but should I try to shoot 4 and see what kind of residuals show up? Could my layout of horizontal control points affect this at all? I have no experience using geoid models, maybe the consultant shot these points in a way I am unfamiliar with? 

Any help is appreciated! 

Quote
Topic starter Posted : March 5, 2020 5:49 am
MightyMoe
(@mightymoe)
5,000+ posts Supporter

First level the control.

If the 4-5 cm is still there that is your answer for the vertical, then: 

Apply metadata to use a projection and geoid model. 

Occupy the control and use the given coordinates, geographic locations, hold them. 

Then check the control points, see if the levels and geoid model and horizontal control work, if not sit down and look at each point to find one that may be disturbed or can be eliminated somehow. Sometimes it's the one with the base receiver that has issues. 

It's easy to push buttons to localize until it isn't. 

 

 

ReplyQuote
Posted : March 5, 2020 7:16 am
JKinAK liked
Norman Oklahoma
(@norman-oklahoma)
5,000+ posts Member

The effect of earth curvature in 1 km is about 8cm. Refraction effect tend to negate that, in small part. This is likely a large part of what you are seeing.  If you were traversing (or levelling) you would be reproducing those effects. If you are using GPS you are not. Use all of the supplied control elevations in your localization.  

  

ReplyQuote
Posted : March 5, 2020 7:20 am
Rover83 liked

tkc1
 tkc1
(@tkc1)
10+ posts Member

@mightymoe So, if after levelling I am still seeing a 4-5 cm discrepancy then that is when a projection and geoid would need to be applied? I can find out the projection from the Civil3D file; however, how would I know what geoid model to use or how to obtain it? Thanks.

 

ReplyQuote
Topic starter Posted : March 5, 2020 8:09 am
tkc1
 tkc1
(@tkc1)
10+ posts Member

@norman-oklahoma Thanks. I have two of those vertical control points shot at the moment. On the drawing the difference between the two is 0.53m. When I look into my GPS measurements on each, the heights differ by 0.48m. Can you help me further process what’s happening here? If I was to use both of these and the others in my localization will I not see errors? I’m on another project at the moment or I would go try but just trying to wrap my head around it for tomorrow. 

 

ReplyQuote
Topic starter Posted : March 5, 2020 8:17 am
Rover83
(@rover83)
500+ posts Member
Posted by: @tkc1

So, if after levelling I am still seeing a 4-5 cm discrepancy then that is when a projection and geoid would need to be applied?

No, that would mean whoever supplied those elevations did not do their job, and the most likely course of action would be to contact the project manager. They should either (1) contact the firm that ran initial control and get them back on site to fix their screwup, or (2) work out an arrangement for you to do it correctly and update the values. If you were not provided with enough control around and throughout the site to perform a decent calibration, then raise that point as well.

 

Posted by: @tkc1

On the drawing the difference between the two is 0.53m. When I look into my GPS measurements on each, the heights differ by 0.48m. Can you help me further process what’s happening here? If I was to use both of these and the others in my localization will I not see errors?

Well, that all depends. If you have run levels between those points and they check out, you may have a problem. Or not. Depending on your GNSS procedures and site conditions, you may very well have up to 5cm of total error when measuring between two points. Are you using a geoid model? If so, how good is it?

There are a lot of options for calibrations, i.e. with or without a projection, with or without a geoid, etc. I would read the manual/tutorial for whatever field software you are using before going any further. Otherwise you are just asking for trouble.

 

 

ReplyQuote
Posted : March 5, 2020 9:13 am

Norman Oklahoma
(@norman-oklahoma)
5,000+ posts Member
Posted by: @tkc1

....On the drawing the difference between the two is 0.53m. When I look into my GPS measurements on each, the heights differ by 0.48m....

There could be several things at work here, and probably some combination of all of those things is what you are experiencing. Two biggies are:

  1. Curvature and refraction which may not be properly accounted for. Possibly by both you and the design surveyor.
  2. Old fashioned imprecision in measurement. Again, possibly by both you and the design surveyor.

If you really want to nail down elevations you will have to do it with a level - preferably a digital level. That can get expensive, but may be worthwhile.  There are, obviously, tons of details here that I have no knowledge of, but I expect that a properly weighted localization using good quality GNSS measurements will likely be sufficient for your needs.

BTW, 2 crosscuts and 4 boundary monuments for control in a kilometer of intersecting road seems like very little, like by about half to a third. Are you sure that there is no more?     

ReplyQuote
Posted : March 5, 2020 9:56 am
tkc1
 tkc1
(@tkc1)
10+ posts Member
Posted by: @rover83

Depending on your GNSS procedures and site conditions, you may very well have up to 5cm of total error when measuring between two points. Are you using a geoid model? If so, how good is it?

There are a lot of options for calibrations, i.e. with or without a projection, with or without a geoid, etc. I would read the manual/tutorial for whatever field software you are using before going any further. Otherwise you are just asking for trouble.

As of right now, I am not using a geoid model. I have never used a geoid model, but this is the longest stretch of road I have worked on so I think I am starting to see the issues that can arise. I run Pocket3D, and I know where to go to apply that data, but I have never put it to practice so I don't know how the steps change. Along with the Pocket 3D, I am working with a Hiper VR base and rover. Could you enlighten me to what should be done after I load a geoid model into the project? Thanks for the help.

Posted by: @norman-oklahoma

There are, obviously, tons of details here that I have no knowledge of, but I expect that a properly weighted localization using good quality GNSS measurements will likely be sufficient for your needs.

BTW, 2 crosscuts and 4 boundary monuments for control in a kilometer of intersecting road seems like very little, like by about half to a third. Are you sure that there is no more?

There are several property bars along the road, but they are not all shown on the consultant's plans - so I shot the ones that were, and subtracted 2 or so that seemed slightly off with the rest. 

ReplyQuote
Topic starter Posted : March 5, 2020 10:11 am
MightyMoe
(@mightymoe)
5,000+ posts Supporter

@tkc1

If you still see the 4-5cm then the control has issues, the geoid will be of no use.

The geoid model is basically a topographical model, along the contour lines the ellipsoid and geoid models will be parallel, if you run 90 degrees to the geoid model then you will see the ellipsoid and the geoid model be non-parallel and the geoid height will increase or decrease. If your project is 90degrees to a geoid model contour line then it isn't unusual to see 4-5cm/km change.

I see below you mention .5m which would be 50cm. 

Not sure if we are discussing 5cm or 50cm, big difference of course, totally different discussion. 

If it's 5cm the levels will show you, and a level survey is a geoid survey, a GPS survey without a geoid is an ellipsoid survey, which doesn't pick up the geoid slope. 

 

ReplyQuote
Posted : March 5, 2020 10:13 am

StLSurveyor
(@stlsurveyor)
1,000+ posts Supporter

I would run a level through all the control and then report the facts, maybe one is out or has been hit since the design phase and that one is messing up the localization -  can you look at each residual with your GNSS unit if so, do you see an outlier?

Next, if you are localizing the a "local" system then why try to use a Geoid?

If you set new control outside on the localization then the further you go the worse it will get. If the site is set up like a right triangle than the hypotenuse side should be "okay" assuming the control exists at both end points with at least one point at the 90.

ReplyQuote
Posted : March 5, 2020 10:17 am
Rover83 liked
tkc1
 tkc1
(@tkc1)
10+ posts Member

@mightymoe 

Ok, that's what I was thinking but I have no experience with geoids so I had to ask. It's 5cm of error, and I was explaining that the difference in elevation values between two of these points on the drawing is 0.53m compared to the GPS measurements on my data collector saying the difference is 0.48m between them. 

 

ReplyQuote
Topic starter Posted : March 5, 2020 10:21 am
tkc1
 tkc1
(@tkc1)
10+ posts Member

@stlsurveyor

I normally run with one point as my vertical control, so I will need to shoot the remaining 3 to see the residuals. I've just passed by and checked my elevation on them (using the one as the control and not shooting these in as control points), so I will need to take those measurements. 

The hypotenuse side (so the road in the project) is where all of the control is. I don't have anything off of the road or at the 90. 

 

ReplyQuote
Topic starter Posted : March 5, 2020 10:30 am

Rover83
(@rover83)
500+ posts Member

It's worth noting that not all site control is established with localization/calibration in mind - a lot of projects need the accuracy of total station work for layout so calibrations really aren't considered during the survey/design phase.

For a sizable chunk of our design topo projects we never break out the GNSS. We supply redundant control for the construction sheet set, but it is rare for it to encompass the site in such a way to get a really great calibration. If the client wants us to develop a calibration or space out additional control to support one, we can do that, but it is not a standard practice.

Of course, if the local site was developed from a geodetic basis, a good consultant would provide transformation parameters in addition to all the usual coordinate and elevation values.

But you may be looking at a project that requires you to break out the total station. As Norman pointed out, you don't have nearly enough control for a project of that size.

ReplyQuote
Posted : March 5, 2020 12:23 pm
StLSurveyor
(@stlsurveyor)
1,000+ posts Supporter

@rover83

I agree. Time to bust out the total station to traverse around the the site and then run some levels. 

 

ReplyQuote
Posted : March 6, 2020 3:39 am
Share: