RTK test, Aerial pictures accuracy, and OSM Database Accuracy

Posted by StephaneP on 11 September 2017 in English (English)

RTK accuracy

Since 1 or 2 years, I'm testing some low-cost GNSS receivers with RAW output. The goal is to get a cm accuracy. One way is to store the raw data, then post-process it with the open-source software RTKLIB. I had various fails and success and I finally find a point to place my own reference station, my "base" : base

One test was to put the "rover" on my car go back to my home. RTKLIB gave me a solution with "FIX" for a big part of the record : global solutionorange is "float" and green is "fix" (best accuracy)

An interesting part is a new roundabout, too new to see it on any aerial picture : New roundabout

Ok, but what about the accuracy ? So, zoom in, zoom in, ..... zoom more !! individual location Each square is 1 cm. Yes, the accuracy is about 1.5 centimeters !!

RTKLIB gave me a very good accuracy, but is this real ? I can't answer for this individual point, but with the results I got on surveys points from the French national geographic institute (IGN), I think I can say that the accuracy should be at about 5 or 10 cm, as my base coordinates are not perfect.

RTK vs Aerial Imagery accuracy

We are in September 2017, and the IGN just published new aerial imagery, and they usually do a very very good job. Finally I can see the roundabout I draw on OpenStreetMap 1 year earlier.

Let's compare my recording and the pictures: aerial vs RTK

Hey ! Not bad !! Now I must train to drive with a centimeter accuracy :-)

OpenStreetMap database accuracy

I speak about gnss trace accuracy, aerial imagery accuracy, but what about OpenStreetMap accuracy ? I heard that with 7 decimals, the coordinates stored in the OpenStreetMap database get only a 10 cm accuracy. Don't you think that It'll be a problem sooner or later ?

Perhaps we should consider adding a 8th decimal ?

Comment from tyr_asd on 12 September 2017 at 07:27

Very impressive!

Btw: 7 decimal digits already correspond to about 1.1cm.

Hide this comment

Comment from SimonPoole on 12 September 2017 at 07:47

Just a further note on increasing the possible accuracy: the 7 decimal digits are a consequence of using 32bit signed integers for the coordinates, changing that would be fairly involved.

In any case effects due to movements of the continental plates are larger (see ) than the limits on accuracy due to our coordinate representation and if anything should be addressed first.

Hide this comment

Comment from StephaneP on 12 September 2017 at 08:29

@SimonPoole Yes, the plates are moving, but it's a good idea to have a good relative accuracy. If a plate was moving at 10 meters each years, I still wan't to know if an object is 1 meter distant from another.

But I've calculate some distances on and tyr_asd is right, with the about 1 accuracy, so there is no problem. I think we have some time before we need a millimeter accuracy :-)

Hide this comment

Comment from Rinaldum on 12 September 2017 at 15:00

Dis, j'aimerais pas te rencontrer a la sortie d'un rond point :-D

Hide this comment

Comment from StephaneP on 12 September 2017 at 15:21

Ah bon ? Pourquoi ?

Hide this comment

Comment from philippec on 12 September 2017 at 20:11

He would miss you just by 1 cm.

Hide this comment

Comment from philippec on 12 September 2017 at 21:08

Where is the button I must push ?

Hide this comment

Comment from zstadler on 23 September 2017 at 09:52

Which low-cost RTK GNSS receivers are you using?

Hide this comment

Comment from Geonick on 24 September 2017 at 14:31

I’ve also heard from low-cost RTK GNSS which use Precise Point Positioning (PPP). Using these type of GPS mice (e.g. from u-blox) you don’t need a rover/base station and they reach sub-meter accurracy – especially when post-processed with RTKLIB: See e.g. .

Hide this comment

Comment from jcr83 on 25 September 2017 at 15:05

Broadcom has announced a new GPS receiver chip (BCM4775X) that uses two frequency bands. They claim a 30 cm precision. See:

If true, this precision would be good enough for OSM.

Hide this comment

Comment from StephaneP on 25 September 2017 at 16:28


PPP is usually used with dual frequency receiver (L1 + L2), and the RTKLIB dev doesn't recommend to use it with one one frequency. I've tested it but the results were not very good. And more important, PPP should be used in static mode, so you can't use it when you move. I've never read user reviews on this special "Ublox PPP", if you find one, i'm interested to read it.

@jcr83 Yes I'm aware of this chip ( I hope we will see it soon in a final product. And other medium-cost L1/L5 receiver are already available (Siwft Navigation, Tersus).

There is a very interesting blog on precise satellite localization, but I don't understand everything :

Hide this comment

Comment from Adrian 2 on 5 October 2017 at 22:04

The u-blox NEO-7P does precise point positioning. I wrote about it briefly on talk-fr I am impressed with it, provided that you have a good view of the sky; and provided that, after startup, you wait for ten minutes for it to reach the best accuracy, before recording tracks.

Hide this comment

Comment from philippec on 5 October 2017 at 22:27

OSM has gone no further as to make first of april jokes on this delicate subject. I cannot laugh with such stupidity. The other kind of jokes are about agreements with Google. I am waiting for Galileo and a Bluetooth device to obtain a 2 meter accuracy. It may cost something.

Hide this comment

Comment from StephaneP on 6 October 2017 at 05:26

@philippec Could you explain our stupidity and make us smarter ?

Hide this comment

Comment from philippec on 6 October 2017 at 07:28

Maps are no laughing matter. Sending fallacies about them into the world can have serious consequences, even lead to loss of life. That is what these harsh words try to convey.

Hide this comment

Leave a comment

Parsed with Markdown

  • Headings

    # Heading
    ## Subheading

  • Unordered list

    * First item
    * Second item

  • Ordered list

    1. First item
    2. Second item

  • Link

  • Image

    ![Alt text](URL)

Login to leave a comment