GPS to orthophoto discrepancy

[EricJohnson]

Not sure if this is the correct forum for this question.

We are beginning to build a GIS for a smallish combined utility system (water sewer roads stormwater) with several hundred each ov valves, manholes, etc. We are collecting data with a Trimble GeoXH, but have noticed that even after differential correction, the results are a bit off (up to 3 feet) when laid over a recent georeferenced aerial.  We can't see everything on the aerial, or we would use just that to build from, but we are trying to get the best accuracy and precision for our points. So I am trying to figure out my best next step: if I move the points to the apparent correct locations on the map, I get a warning about losing the GPS data, and it asks about rebuilding, which puts the point back where it started. I guess I could try a hybrid approach and not GPS the points I can see, but that could get confusing. Any ideas for a workflow for this?

Eric Johnson

[DanielErklauer]

Not sure if this is the correct forum for this question.

We are beginning to build a GIS for a smallish combined utility system (water sewer roads stormwater) with several hundred each ov valves, manholes, etc. We are collecting data with a Trimble GeoXH, but have noticed that even after differential correction, the results are a bit off (up to 3 feet) when laid over a recent georeferenced aerial.  We can't see everything on the aerial, or we would use just that to build from, but we are trying to get the best accuracy and precision for our points. So I am trying to figure out my best next step: if I move the points to the apparent correct locations on the map, I get a warning about losing the GPS data, and it asks about rebuilding, which puts the point back where it started. I guess I could try a hybrid approach and not GPS the points I can see, but that could get confusing. Any ideas for a workflow for this?

Eric Johnson

if you are using a trimble geoXH and post processing the position I'm assuming you are using pathfinder office which will tell you the accuracy of each position taken.  You can't overlay points on an aerial image and determine the precision of a GPS measurement there is allot more that goes into determining positioning accuracy than that.  The most likely issue here is that your gps data exceeds the resolution of the imagery .  Can you supply some information on where you got the imagery and its resolution and projection.  Whats your data WGS84 or did you do a transformation?  Also if you move your points manually you have blown your survey and mine as well use a tape and compass instead.

[EricJohnson]

Our GIS data is all in NAD83 Nevada State Plane West 2703 feet, which lets us be compatible with others around here. We are running Arcpad on the GPS unit, and when we check the data back in, we run the differential correction using GPScorrect. I had been told once to let the GPS collect in it's native WGS1984 and it would convert when we did checkout and in so that is what we have been doing.

The photo we are using as our background is an (forgive me if I don't have the right lingo) orthophoto which was set up from the start in the same coordinate system, and is 6" resolution. Which is why I figured that our points should be pretty tight between the two.

Eric

[EricRice]

Hi Eric,

I don't have the right answer for you, but I have some things for you to think about.

1. It's common for horizontal accuracy of pixels to be measured by pixel size/2, which would put your orthophoto at a 3" horizontal accuracy. This would be amazingly accurate. This measure of accuracy is more typical of satellite imagery though.

2. Your orthophoto as the name implies, was orthorectified using a DEM. The resolution of that DEM is critical to know. A 30m v 10m v 1m DEM could easily account for the pixel not being where your GPS point data indicates it should be.

3. What was the accuracy of the source data the image was georeferenced to?

4. Did you really 'georeference' the image in ArcGIS? What was the RMSE? What was the transformation used? How many control points? Or did a vendor supply an image that included georeferencing information?

5. Do you have other datasets to add to the mix to compare? These should have known accuracies.

6. Is your point shift uniform - like all to one direction? If yes, I would suspect the points in this case, but if not then I would suspect the image is slightly wrong.

7. Does the provider of the orthoimage provide metadata regarding its accuracy? (if applicable)


I'm inclined to think your points are more correct. Perhaps you could re-georeference the image using highly accurate gps points once they are all collected. Just a thought.

Best Regards,
Eric

[EricJohnson]

Hi Eric,
     Those are a lot of questions, and I am not certain how to answer, but I will try.

About the aerial: I do not know exactly who did the referencing, however what this was was a contracted flight to create pictometry, each section therefore has 5 views: birdseye (which I am using) and north, south, east, and west obliques. I acquired it from our county, who oversaw the process. They use the coordinate system, so I am assuming that that was how it was delivered, but I could be wrong.

Direction is not consistent, however this is based on the few valve locations which are visible in the photo (steel lids are hard to pick up from asphalt) I do not have any other solid locations to check against yet, but could go grab GPS points for the sewer manhole lids which show up pretty well. As to resolution, I can measure, in arcmap, the diameter of the water valve boxes in the areas where they show with decent contrast, and they are right about 1 foot in diameter.

You lost me with RMSE and the other parts of your question about the imagery.

Eric

[DanielErklauer]

Eric, what type of corrections is your handheld unit receiving (are you using an RTK network, WAAS, DGPS?).  And what type of environment are you operating in (heavy tree canopy, open field, within a city)?

[EricJohnson]

For correction, we are post-processing. Our terrain varies but yes it can have heavy tree cover, however for the places which made me ask the question, the view is open (thus the clear aerial locations) Our GeoXH is used with the external antenna on a pole.

Eric

[DanielErklauer]

Eric, I have to give you the aswer you probably do not want to hear.....

If the 3 feet is a horizontal bust outside of the scope of your work you are going to have to do one of two things if not both....

1.  Do you have an an independent data set in the area of question where the horizontal acuracy of it is known (such as survey control monuments in the area)?  If so overlay the data in and compare it to determine if your issue is with the GPS or the ortho.

2.  Go back to the site and set several test control points and use that to verify accuracy (this is the best method).  Do this on points already collected.