Charts II: NGA - 2700 Charts

[r.fairman]

Quality control

How do we quality control the calibrated images and final .kap files.
Data entry is notoriously error prone.
Could we generate a four waypoint gpx file from the data entry on the server.
Opening the chart and the gpx file should result in four waypoints sitting exactly at the chart corners.
If the gpx file had the same number at the chart, quality control would be relatively easy The QA guy could open a batch of charts in OpenCPN and open the associated gpx files and zoom to the max possible and QA each chart one by one.
We have ,moved into real cartography with all the pain and responsibiliies but without the pay
I still feel we have to apply the NTM's otherwise we are issueing charts that we know have errors and we know some of the errors

I feel like the devils advocate. Don't shoot the messenger

regards

Richard Fairman

[Netsurfer]

Hi folks ...

Originally Posted by r.fairman

Quality control

How do we quality control the calibrated images and final .kap files.
Data entry is notoriously error prone.
Could we generate a four waypoint gpx file from the data entry on the server.
Opening the chart and the gpx file should result in four waypoints sitting exactly at the chart corners.

Sure we could. But what will be the result? As you build the GPX file from the data entered by users for calibration they always will match but do not say anything about precision.
IMHO the "quality" could only be checked manually (or in very rough way we can do a logical check of the entered calibration data). Therefor we might set up an easy to use webform to report errors.
So the most important thing is that chart calibration is done as accurate as possible!

I still feel we have to apply the NTM's otherwise we are issueing charts that we know have errors and we know some of the errors

This would be a very huge work ...!
IMHO we should bundle our first KAP files with the original NTMs which we could grab from the NOAA site.
And if users have corrected charts we sure can replace it.

Maybe some kind of "quality meter" with a voting option and the possibility for comments for each chart may be helpful?

Summary:
First let's come to a basic set of kap files which means all charts need to be calibrated and all other infos for the header need to be gathered.
That should be our next step. And it will take already a lot of time. In the meanwhile we could think about the rest.

Gunther

[r.fairman]

I think we are running too fast Too fast to market springs to mind The danger is we produce a 2700 kap charts with errors we know and can describe in detail but are reluctant to (or too busy) to implement. The NTM's are easy to apply to paper charts but nigh on impossible to apply to the final kap charts
I would be more happy to wait and navigate with a chart with the NTM's applied than an out of date chart with a list of NTM's that I cannot apply to the chart.

The QA of the data entry and resultant precision must be tested on each and every chart we produce.
I understand the entered data will be the same for the gpx file and the sopposed chart corners.

We need confirmation that the chart will open in OpenCPN and that the calibration is within tolerances

Surely a quick assessment of whether 4 waypoints visually line up with their entered postion on chart would confirm the precision and validity of the data entry This is or course manual process
If anyone else can suggest a better method and what level of precision is acceptable

richard

[Netsurfer]

Richard

Originally Posted by r.fairman

I think we are running too fast The danger is we produce a 2700 kap charts with errors we know and can describe in detail but are reluctant to (or too busy implement. The NTM's are easy to apply to paper charts but nigh on impossible to apply to the final kap charts

Well, before we are able to "produce" any kap charts we need the calibration data and some other data that could not be gathered automatically. And as you already pointed out the entered calibration data need to be checked back for accuracy. But applying the NTM's corrections would also need to be cross-checked.
Doing all this in one step before producing a kap chart for checking I fear we will never come to an end ...

The QA of the data entry and resultant precision must be tested on each and every chart we produce. Surely a quick assessment of whether 4 waypoints line up with their entered postion on chart would confirm the precision and validity of the data entry

I guess you didn't get my point ... Which data will you take for the waypoints? There are only the ones entered by users. So even if they are totally wrong they will always match ....

If anyone else can suggest a better method and what level of precision is acceptable

IMHO the precision of a raster chart could be easily tested by moving the cursor to a known point on the chart and compare with the coordinates given by OCPN.
I don't know which level of precision is acceptable. And I also think that this might be difficult in general as it e.g. depends on the underlying image. Therefor we should or could only try to get the maximum level of precision for each chart. Each user has to answer the question for himself whether the level is sufficient or not.

Gunther

[r.fairman]

I guess you didn't get my point ... Which data will you take for the waypoints? There are only the ones entered by users. So even if they are totally wrong they will always match ....

They are meant to match the chart corners If they do not line up with the chart corners there must have been an error

[allanpeda]

I dunno, but no amount of curating, aside from re mapping the area, can fix old charts. They are only useful as a starting point to generate approximations which must be confirmed from other sources.

They are useful, but you can't polish a ... well you know the expression i hope.

[Netsurfer]

Richard ...

Originally Posted by r.fairman

They are meant to match the chart corners If they do not line up with the chart corners there must have been an error

what I do not understand is where you want to get the coordinates for your waypoints from?
The only possibility I can think of is to get them from the entered calibration data. But by doing so this will not show any errors as the values will always match and therefor the waypoints will always match the corners of the chart.

What am I missing?

Gunther

[wdkester]

Originally Posted by r.fairman

I think we are running too fast Too fast to market springs to mind The danger is we produce a 2700 kap charts with errors we know and can describe in detail but are reluctant to (or too busy) to implement. The NTM's are easy to apply to paper charts but nigh on impossible to apply to the final kap charts
I would be more happy to wait and navigate with a chart with the NTM's applied than an out of date chart with a list of NTM's that I cannot apply to the chart.

The QA of the data entry and resultant precision must be tested on each and every chart we produce.
I understand the entered data will be the same for the gpx file and the sopposed chart corners.

We need confirmation that the chart will open in OpenCPN and that the calibration is within tolerances

Surely a quick assessment of whether 4 waypoints visually line up with their entered postion on chart would confirm the precision and validity of the data entry This is or course manual process
If anyone else can suggest a better method and what level of precision is acceptable

richard

My 2 cents -

I vote for saving the NTM updates to a final phase. Most NTM's seem to be light/buoy changes. Major feature changes, such as cruise ship facilities, can be seen on Google Earth. Yes, GE can be out of date but the more populus areas (like harbors) are kept more up to date. I think it's a good practice to look at GE for any unfamiliar areas.

On the chart registration issue I can see 2 different QA checks-
1. View the chart on Ocpn; zoom way out; place the cursor on the ends of the lat/lon lines; this will show that the chart is (or isn't) registered properly.
2. Place way points on several prominent physical features; save in gpx; import to Google Earth and check their coincidence with the map features.

[HappySeagull]

A quick updateable way?maybe make a gpx file for the NTM's...Useful in OpenCPN for sure,as a layer...volunteer-creators could post these on the site with the charts.

Calibration to Kap is Primary.
"Notices to Mariners" are always being updated.That's why there can be several years between new charts being issued.

[LeaseOnLife]

Here are another $0.02

NTMs
I was looking at oceangrafix, an outfit providing papercharts, including NGA charts. There doesn't seem to be anything done regarding NTMs on NGA-charts they deliver.

Another place, landfallnavigation.com, using the oceangrafix print service, is a bit more clear in the fineprint: "NGA charts are uncorrected, current edition".

We are downloading the current edition. Like oceangrafix or landfallnavigation, we should link to the NTMs on NOAA/NGA websites.

QA:
We got 2700+ charts to calibrate, then check. We have to realize, these are not always the best charts for any given location. How about a web-tool, where any user, at any time, can indicate confidence in

• a basic review (chart is in the right location, quilting looks OK)
• calibration to NGA grid
• calibration to WGS84 by NGA
• detail of chart, compared to other sources, like local charts
• Is the downloaded chart the most current one available from NGA?
• ...

Wikipedia is doing something similar for their articles, it could be called peer-review.

I think the NTM and QA issues should not halt the next step, the calibration of the simple, square charts.

[Surveyor]

Originally Posted by HappySeagull

A quick updateable way?maybe make a gpx file for the NTM's...Useful in OpenCPN for sure,as a layer...volunteer-creators could post these on the site with the charts.

Calibration to Kap is Primary.
"Notices to Mariners" are always being updated.That's why there can be several years between new charts being issued.

This is the way to go IMHO. (if NTMs are provided at all)

* Will not hold up the process
* Easier than photoshop or gimp
* Universally compatible
* Easily edited and/or amended.

Happy, you are a genius!

[dacust]

NOTMs as layers work fine as long as they are OpenCPN users. But do we want to limit the use to only OpenCPN?

Having said that, the layer idea, is, of course, a very good idea.

But, let's say we don't go the layer route.

NOTMs come out all the time. We need a way to update them every time. Applying NOTMs will not change the calibration. So, as long as the calibration is saved, NOTMs can be applied to the .png (or however the final stitched, re-sized, reduced color, etc. is stored) and then a simple script can create the new .kap

Doing it this way, the calibration is done first. A .kap is created from the non-updated chart.

Then when/if the NOTMs are applied for a particular chart, a new .kap is created.

For those that would rather not rely on community updated charts, or want the original to compare it to, I think the original non-updated .kaps should remain available.

That's how I would want them. I'd want a full set as original, and the latest updated ones so I could compare them to see what changed. The changed items I'd look at extra careful to make sure they really changed as charted.

-dan

-dan

[r.fairman]

I vote for Dans strategy It is a realistic way forward for all including non OpenCPN users. In any case to will be easier to apply NTM's once the chart has been calibrated as most NTM are based on lat and long positions with update to depths, lights etc. I am not sure what the QA strategy will be does dan have any suggestions
regards

richard

[HappySeagull]

...part of my thinking per gpx is that the NTMS would show up on ALL your charts ,if wanted.

This,as a layer,withgpx marks using custom icons, is a great feature of O....Other apps will follow OpenCPN and read her GPX or wither and die in their proprietary formats(curse them!)

That said,anyone that wants to can extract any kap into an image and rebuild it,per freely available Notices to Mariners can do so right now.

So!-Those of you that've rebuilt a chart every time a new NTM comes out,raise your hands!

[r.fairman]

QA

The QA using GE works well as suggested .

Drop 4 waypoints in landmarks on your chart
export as GPX
Open using GE
Do they line up Y/N
If Y calibration is ok
If N recalibrate and QA

As an aside it would be nice to use the GE plugin for this. Can waypoints created in OpenCPN display in the plugin without having to close the plugin, open GE and import

An alternative QA would be to move the boat to known landmarks and compare its position in the GE plugin if that is possible

regards

Richard