GPS Site Calibration/Localization Explanation

Suggest as a starting point using the search function in the upper right hand corner to search past threads on 'localization', aka 'calibration'. You'll find a number of threads on the subject. I'll just add that the mapping projection being used can play a significant part in how well this will work out for you depending on what you're out to accomplish. Good luck!

The simple version would be that a calibration connects the Lat, Long, Height of the GPS to the North, East, Elevation of the worksite. That's the VERY simple version.

I was just reading What is "localization", exactly? at
https://surveyorconnect.com/community/gnss-geodesy/what-is-localization-exactly/ . Good stuff, except I would love to see some diagrams with numbers from an actual project. Better yet, I would love to see a project where someone screwed it up because they let the software do the thinking for them.

For construction layout, depends on how big an area you're going to be working with.

Single point calibration will bring you to an elevation that you can base the project on if it's not huge.

Then you could do a static session and get the accuracy across the site for you to work with.

Again simple version like JARO said.

Or you can make life really interesting and set up an adjustment that tilts the plane. Unintentionally it's not fun.

Either way, not attacking you, you're wise to seek additional info.

But your boss is insane telling you about going out to do any site calibration unless you told them you had years of experience with that process and it seems like you're just getting started with it.

We won't let you fall on this technological sword without a good explanation at the very least.

?ÿ

?ÿ

So in my short experience with site calibrations/localizations for GPS grading models, I have always been provided the control by the original design firmƒ??s surveyor. Iƒ??ve also typically localized/calibrated using Topcon hardware & Topcon Pocket 3D where the program has preset tolerances that are acceptable for control points (software is for contractors). These tolerances can be adjusted as well. So at a minimum Iƒ??ll ask for 4 control points from the design firmƒ??s surveyor along the outer boundary where it is free from any obstructions/canopy & outside of the areas to be graded whenever possible. Proposed corners or existing corners are normally good for control. If any points are more than say 0.03-0.04ƒ?? off either horizontally or vertically then I re-observe/measure and I let the localization observations run for 180 epochs/seconds (seconds according to Pocket 3D). I have had some points be a 0.1ƒ?? or so off before and when thatƒ??s the case I try to either have them re-verify the point or I try to work without it if thereƒ??s more than 4 points & it is not an outer extremity control point. This might not be the best way to do it but it has yielded independent verified as-builts on the grading in various areas like ponds or site features like swales that were within 0.1ƒ?? or less. Itƒ??s always a good idea though to try to check behind yourself with a level occasionally and check back in on several control points after starting up the receiver in between breaks

What kind of equipment are you using? Calibration and Localization mean different things to Trimble and the other Brands.

Calibration is a way to merge Geodetic coordinates to existing XYZ control. The better way to obtain XYZ control is to project Geodetic coordinates to a mathematical "plane". Some of these would be State Plane projections, Universal Transverse Projection, Low Distortion Projections.?ÿ

If there is existing ground control that doesn't "fit" a defined projection, software developers created programs for GPS units that will "adjust" the GPS values to merge that data to the existing ground control.?ÿ

However, this will distort the GPS file and it's always important to have tight ground control with any calibration.?ÿ

I recommend that it should only be done for small sites. The larger the site the more error the calibration will create.?ÿ

I never do it unless there is no other way, it's something from the early GPS days and was done mostly to control the vertical. There were only rudimentary Geoid models early on. Today the Geoid models are far more accurate than a calibration. I'm thinking it may have made a comeback cause of VRS and machine control which seems to be set-up to only work with a calibration.?ÿ

I created this video a few years ago to explain how localizations are performed in Javad's J-Field software. I explain how to identify problems with data used for a localization.

JaRo gave a great short answer. I'll add to it a bit but try to keep it brief. A localization determines translation, rotation and scale between two Cartesian coordinate system by using least squares to minimize the residuals between two sets of grid values (source and destination). It begins by determining the centroid of each set of coordinates (centroid of source and centroid of destination). The centroid is the arithmetic mean of the two coordinate sets (average the northings of the source coordinates, average of the eastings of the source coordinates and the same for destination). Rotation is weighted based on the distance from source centroid to each source coordinate, the longer the distance, the higher the weight. Scale is also based on the difference in distances from centroid target to each target coordinate divided by the centroid source to each source coordinate. The longer the lines the higher the weight applied to the scale factor.

?ÿ

Localizations are a Helmert 7 parameter transformation. Translate NEU, Rotate around N axis, E axis, U axis, scale (7 parameters). I prefer how our software treats localizations because you have the control to disable some parameters, particularly I prefer to disable rotation around N axis and E axis and scale.?ÿ

?ÿ

It's a powerful tool with a proper place, but like most things, when used with abandon can create serious problems.

@shawn-billings

Would disabling specific parameters not be beneficial only depending on the specific job?

I have minimal experience with localization but it's always worked pretty well for me. It's always been fairly small job areas, make sure I'm inside the localization and spread out the points used around the job in each direction.

One job I did did not need accurate horizontal, it was only for grades, shot in specific areas but I wanted to use the GPS and the site was not integrated. Using only 2 points for horizontal and one known good benchmark for elevation, I did a check to a known point just out of curiosity. It was out 5 or 6cm. It didn't matter for what I was doing that day but I would never do it that way on a typical job. I was just curious to see what the check would be like. It was better than I expected.?ÿ

@williwaw

Thanks for the response, I checked it out a few hours ago which helped tremendously

@jitterboogie

Thanks for the response. Iƒ??m still somewhat new to the company and the subject of a localization got brought up because he wanted me to tag along with him sometime next week to do one. ?ÿEven though we discussed what it was, I left the conversation a little confused which is why I reached out on here to get a different explanation. Appreciate the response

The golden rule for me when using a localization is to use a minimally constrained localization and then add points, checking their residuals before including them in the localization. I'm not sure if all software allows this. A minimally constrained localization would include two points for horizontal, disabling the scale factor, and one point for vertical. A single point will get the translation, but a second point is needed for rotation. The vertical point can be one of the two horizontal points or a third point. The single vertical point will work for this minimally constrained localization provided that you use a Geoid model. The Geoid model gives the difference between the ellipsoid height (which is native to GNSS positions) and sea level. Because the geoid and sea level are not parallel (or concentric) with one another, the separation between them changes from one location to the next, and the Geoid model estimates this separation.

?ÿ

The minimally constrained localization is just to make sure that subsequent points added to the localization are within tolerance. If a point is not within tolerance, then the problem could be in the added point or in one of the points in your initial minimally constrained localization.

?ÿ

I have seen the four point localization fail. It was a strange set of circumstances that took me a while to sort out with a customer in another State. He localized to four points and the residuals were golden. Three points create a plane and a fourth can provide a check on the plane most of the time, so we struggled to figure out an issue he had. He localized, holding the four points immediately. He then tied into another point with an elevation and missed it by several tenths of a foot (maybe 0.4 if memory serves). After a lot of head scratching, I had him start the localization over holding a single point and checking the elevation on the other three. So he held point A (let's call it) and looked at the elevation for B, C, and D. B looked pretty good, but C and D were off by 1 foot. As it turns out, the elevations for A and B were NAVD88 and the points for C and D were NGVD29 and in his State the difference between NAVD88 and NGVD29 are about 1 foot. It was easily spotted using the minimally constrained approach, but when putting all four in and looking for residuals it was impossible to identify. This is because, usually with the four point approach, we expect if we have a problem, the problem will be with one point, like a table leg that is too short, any three points will make a plane and the one bad point will be easy to identify. But when two legs are short, the four legs all touch the floor, but the table is sloped (which is what happens to your localization).?ÿ

?ÿ

The geoid in my area is very good, and I suspect in most parts of the country it is good. If I were to go to a third world country, I'd use localization over small areas to produce my own geoid by tying into leveled points. But in the US, I trust the geoid, so for most localizations, I turn off the tilt for elevations. I typically use a single point, which only controls translation in the vertical. Sometimes I may use more than one, but with the tilts turned off, I am still only getting a vertical translation, but now it is using an average of the points held for vertical. One of the benefits of this approach is that I am not constrained to working "inside the box". If you are using the vertical to tilt the plane of your localization, you must be careful to work within the polygon formed by your vertical constraints. This can especially be true in corridors. Imagine that you are working along a highway and you have three vertical points, two are on the West side of the highway and one is on the East. The point on the East is 100 feet perpendicularly offset from the two on the West. Lets say there is a 0.1 foot error in your observation on the East point and your observation was low. The localization tilts the plane down by the 0.1 foot error. You find out later that there is a benchmark a mile to the East and you think it would be a good idea to tie it in. How much error are you likely to see on that benchmark? Had you just used the Geoid and disable the tilt, then you wouldn't see any error outside of your ability to measure and the accuracy of the benchmark to your project elevations.

?ÿ

Localizations are awesome. I highly recommend them. I think they are particularly outstanding tools for boundary reconstruction of precise terrestrial surveys (such as with a total station). They can save an incredible amount of time in evaluating points. But localization is a professional's tool and needs to be used carefully by people who understand the basic mechanics of them.

@skeeter1996

Our field crews are equipped with Topcon Hiper VRs with MAGNET Field

@shawn-billings

Sorry, I posted and then started watching your video. I got 2/3 through and then had to go get dinner. I will finish watching and read this post and take it all in. It's pretty interesting stuff. I wish more people made good videos like that about surveying.?ÿ

I will say that topcon magnet field has some pretty similar work flows to what you were showing. I would find 2 to get started and help me find the rest. See how they all affected the results. I need to look into it more and check for the "parameters".?ÿ

Thanks again.?ÿ

@mightymoe

Thanks for the response and I believe a big reason why our surveying department ?ÿlocalizes all of our projects is due to the fact we utilize machine control.?ÿ

@shawn-billings

Wow, that was one of the best explanations ever. I'll have to look into our local benchmark datum. As far as vertical goes I've only ever used one benchmark and checked to another (in construction sites using a robot usually) but with GPS I'd either be setting up on a known coordinate with a known elevation, or in the case of the few localizations I either set my base right on the benchmark I wanted to use then did localization for horizontal only or I transferred an elevation to where I wanted to set up and once again user horizontal only holding the elevation I was set on and doing a check to another benchmark.?ÿ

?ÿ

@surveyordave96

Trimble calls Localization "Calibration". It basically works the way Shawn is describing Localization. I always hold the scale factor to 1. You can get some crazy solutions if you let the software calculate a scale factor and I don't know of anyway to detect it because everything will fit perfectly in the results.

I don't do calibrations, mainly because I don't do stakeout or projects where I need coordinates in real time. I prefer to do all adjustments in the office. In the days before good geoid models, and when we were using NAD27 for control, I would solve for translation, rotation (both in azimuth and N-S and E-W) and scale, but always in the office in the LSQ adjustment. Now that we have CORS and good geoids, I can't remember the last time I solved for these "auxiliary parameters". In Geolab it is easy to include or solve for this, in Star*Net not so much. I also note that in Trimble office software these parameters are turned on by default (at least they used to be, not sure with newer versions).?ÿ

I have seen calibrations horribly misused. For example, using some benchmarks that are all on the same line (like along a highway). The axis along the highway is well determined, but the axis at 90?ø to that can have a pathological tilt. The surveyor then went and started surveying photo control located off of the highway with errors in the 10 to 20 cm range in height.?ÿ

I have met surveyors who ALWAYS do a calibration (that is what they were told to do) even though the control is NAD83 (2011) and they have a good geoid.?ÿ?ÿ

Calibrations do have their uses, like tying in a local grid system, etc. But they should be used with caution.?ÿ

@surveyordave96

I've been told that very thing by machine control people, they have to do a calibration to make the equipment work.?ÿ

Some I've been told do a calibration every morning.?ÿ

?ÿ