*NOW FIXED* Heading Errors in the MIO Camera Database

M8TJT · Posted: Sat Jun 30, 2007 4:27 pm Post subject:

I have now checked just over 2000 cams in the GPSW database that have direction info. The cams that I have checked are all those South of 52N and all cams East of Zero E.

I have submitted these changes to PGPSW, who inform me the data will be in the next release of their database. :D

The majority of the data was within 10deg of the road azimuth, but a significant number were more than 40deg off az. Sad

The highest proportion of large errors were where the published data was one of the Cardinal points (N, S, E and W) where I guess that the person submitting the data was travelling in (say) a southerly direction, so entered 180 in the database, but the road was actually NW/SE (135 deg) causing a 45deg error. This seemed to occur, albeit to a lesser extent, on the Sub-cardinal points as well.

There were a few that were obviously set to the wrong direction, but I could not, with any certainty, establish the real, or even most likely direction, so rather than not alerting the GPS at all, I have set these to non-directional.

I would like to see a notice on the cam input page to the effect that 'No data is better than wrong data' as the idea of the direction data is so that drivers will not be alerted for cameras 'pointing the other way'. If incorrect data is entered for a camera then the GPS may not alert at all, Sad

and I for one would rather be 'false alarmed' by a camera 'pointing the other way' than to drive past an unanounced camera.
Hope this helps all you guys (and gals). Watch this space
Trevor

Sallyann · Lifetime Member Joined: Jun 23, 2006 Posts: 768

Trevor,
You seem to have done all the right things and it should make a big improvement to the database. I know that relatively few GPS units are equipped to use the directional data at present, but the big manufacturers must surely follow before long.
I look forward to the next release of the database with some assurance that the directions will now be correct - at least those south of 52N.
This must have taken you many hours' work, and IMO you deserve a lifetime subscription for your efforts :D .

Sal

Andy_P · Posted: Sat Jun 30, 2007 7:53 pm Post subject: Re: Well Done

M8TJT · Posted: Sat Jun 30, 2007 8:10 pm Post subject:

Sal, Andy.
Thanks for your support. :D I was going to do some more directions, but got tied up in a couple of forums.
Hey Ho. Get on with it Trev.

Trevor

repsaj · Occasional Visitor Joined: Dec 16, 2006 Posts: 11

Darren · Posted: Sun Jul 01, 2007 2:24 pm Post subject:

M8TJT Of course your continuing hard work deserves recognition and so I have pleasure in welcoming you to the select group of lifetime subscribers Smile

It's worth mentioning that we had already awarded him an extra year a while back for his earlier work Smile

We're actually in the process of finalising details of an expansion to the lifetime membership qualification scheme.

All being well an announcement will be made next week but the important part is that there will be vastly more opportunities to qualify. Being the first to supply or correct heading data and speed data are just some of the new criteria.

Thanks for all your hard work Smile

_________________
Darren Griffin

M8TJT · Posted: Sun Jul 01, 2007 3:20 pm Post subject:

Darren.
Thanks for your kind comments and lifetime. :D :D :D :D
I deliberately did not mention the previous year extention on forum but did thank MaFt for it privately.

I'm now doing some more as I 'watch' the French GP and the superbikes.
I will send PGPS the results as soon as I have a have done a significant number of cams more.
Thanks again :D
Trevor
PS what was wrong with the PGPSW site earlier today?

Darren · Posted: Sun Jul 01, 2007 3:26 pm Post subject:

The web server hung and MikeB who is the website guru is in the US at present and so was fast asleep at the time.

I'm now in possession of the relevant info to restart the web services if it occurs again Smile

_________________
Darren Griffin

Andy_P · Posted: Sun Jul 01, 2007 8:37 pm Post subject:

Did you learn how to "switch it off and on again" then? Laughing

M8TJT · Posted: Sun Jul 01, 2007 9:09 pm Post subject:

If it's Microsoft, that's the way to do it Embarassed

Darren · Posted: Sun Jul 01, 2007 9:26 pm Post subject:

M8TJT · Posted: Sun Jul 08, 2007 7:58 pm Post subject:

I have at last ('so soon' I hear you say) finished checking all 4427 directionalised (yes, I did make that up) cameras in the PGPSW database. I have sent the update file to MaFt, so with any luck and the wind behind, the corrected directions will be soon included in the database.

I have also done some simple analysis on the results, and my suspicions mentioned in a previous post were confirmed. The largest direction errors, both in magnitude and quantity, occur wher the reported direction is one of the cardinal points, and to a lesser extent, the sub cardinals. The exception to this seems to be around north, where nearly all of the directions were within a few degrees of the true road direction. The opposite is true of the cams recorded as having a 180 degree direction (with the exception of the more than 10 deg error, which points to the fact that the errors were larger in magnitude than this).

The results are as shown below
Db Az 90 180 270
10 Deg Err 67% 60% 62%
20 Deg Err 38% 41% 36%
30 Deg Err 19% 29% 19%
40 Deg Err 9% 15% 9%
50 Deg Err 4% 9% 3%

The percentage figures show the number of cams that had an error of larger than the LH col (Deg Err) shown as a percentage of the cams that had a published direction in the top row.
i.e. Of the cams that had a published direction of 180 deg, 41% of them had a direction error of 20deg or larger.

With the MIO acceptance angle of about plus/minus 40 deg, it would not detect 15 % of the cams that had a direction of 180, 9% of the cams that had a direction of 90 etc.

On cameras for which the direction was obviously wrong, but where I could not, with any certainty, correct the direction, I have made them omni directional, on the basis that it must be better to have a false alarm on a cam on the other side of the road, rather than not be alerted to it when it is on your side of the road. There were not a lot of this 'type' of cam.
From the above, it should be plain to see that people submitting direction data should be more careful with the direction, and if they cannot accurately (within say 20Deg) dermine the direction, then don't put it in the database. An alternative to this could be achievd by someone doing a 'sanity' check on the directions before inclusion in the database. I have offered my services to do this.
Trevor

Sallyann · Lifetime Member Joined: Jun 23, 2006 Posts: 768

That's a tremendous achievement Trevor. I admire the job you have done on this and we should all be grateful to you for it. Now it's down to MaFt to do his bit!

I'm sure you're right to be cautious and stick to the omni if there is any doubt at all.

Sal
PS Thanks for the emailed file. I will try it out tomorrow.

M8TJT · Posted: Tue Jul 10, 2007 7:49 pm Post subject:

Seems that MaDt HAS done his bit in the just released PGPSW cam database (Thanks MaFt).
I have checked out a random sample of four of the worst cams, and they are now A OK, so hopefully non-alerts will not happen, and false alarms will be reduced.
:D

MaFt · Posted: Wed Jul 11, 2007 9:24 am Post subject:

	PocketGPS
	Web