Charts II: NGA - 2700 Charts

[bcn]

After a first lot of generated/calibrated charts the biggest stumble stone are corners with pixels out of range for calibrating. I might do this manually in Gimp, but that means to download the jpg first.

JPGs rotated 90 ot 270 degrees: has somebody tried to rotate those with Irfanview's lossless rotation?

BTW. in OpenCPN the chatr look quite good. The 1:300.000 seems to be not so perfect, though.


Hubert

[HappySeagull]

I've been commenting if there's a problem like that.but it'd be easier if the list indicated a partial completion...?so I needn't open the dialogue to find out why?
Oh,and be sure to wait a few seconds before giving up on the corners...sometimes it just takes awhile to refresh and show 'em.

[cagney]

Originally Posted by HappySeagull

TiaBu,those with bits sticking outside the rectangle?better not do them yet,in Phase One.They'll need a more complicated border...

I think that another approach is possible - ignore it! The reason that these charts have a rectangle like that, is after all that they originally were paper charts. Think "quilting", and the reason for a rectangle like that just evaporates.

Thomas

[HappySeagull]

Originally Posted by cagney

I think that another approach is possible - ignore it! The reason that these charts have a rectangle like that, is after all that they originally were paper charts. Think "quilting", and the reason for a rectangle like that just evaporates.

Thomas

but quilting would chop off a bit of useful chart ,wouldn't it?,..if this chart was the only "good" one you had.But sure,you'd have the piece back with quilting off.
I wouldn't mind drawing a better (8 point)border for 'em,unless it turns out to be a problem online but even then I'd still download em and do it...My respect for Hydrographers(and programmers!) is Boundless

[ChuckSK]

On some charts, 81002 for example, the white paper margins are too wide for the 1:3 scale to reach the chart corners.

Chuck

I see that bcn has already noted the problem.

[bcn]

Two issues:

- 51154: the download link for the .kap does not work (link to .htm)

- 51146: will not open correctly in Opera - no calibration possible. Will try later with FF.

Hubert

[nohal]

Originally Posted by bcn

Two issues:

- 51154: the download link for the .kap does not work (link to .htm)

Regenerated. Should be OK now.

- 51146: will not open correctly in Opera - no calibration possible. Will try later with FF.

Same in FF, I've already seen a chart like that in region 1. My suspicion is we handle special characters badly somewhere - Will consult it with Netsurfer.

@all:For the charts, where the corners are not possible to calibrate, please set the status accordingly so that we know that we have to return to those charts and don't try to produce the KAPs from them until we do so.

Pavel

[bcn]

The next case:

51261: incomplete chart (seems to be just a quarter or so)

Hubert

[bcn]

51380- incorrect generated kap file.
The image file was quite skewed.

[HappySeagull]

14018: won't allow any saving and won't calibrate whatsoever
21161: corrupt image

Polyconic harbours in Western Canada-not sure we should have done these.I did one,someone did another They are NAD27 datum so be careful to use these kaps as "eyeball aids" only 'til otherwise corrected.

[HappySeagull]

51013:Scale is wrong.Given as 1/89!
52045: (Gibraltar harbour) Kap doesn't work in OpenCPN. Corner calibration as Minutes + Decimal fraction a possible cause?


52045 edit:Nope.Decimal fraction works perfectly.A wrong input on corner...probably me!

[bcn]

Hi Happy Seagull,

perhaps we should organize who is taking care for which part of the charts to optimize our efforts?

If somebody could start with the 5xxxx series with 54000-54300 that would be fine.
I would look at the beginning after the charts 50000-53999.
Is that ok?

Hubert

Hubert

[paulr]

Not wishing to add to the project leaders' work, but some formal instructions on accuracy and quality might be useful. For instance, what is the effect of the calibration being a few pixels out? Is the attached calibration, for example, sufficiently accurate?

[bcn]

Dear Paul,

the example would not be acceptable for me.
The mark should be in the inner corner of the lines - trying to get it where degradation/anti-aliasing starts.
If you have a look on the pixel values while moving the cross hair you will see that you go quite a distance for a change in pixel values.
So perhaps you will get the same values as placing the mark in the inner corner, but for repeatability and a common procedure you should aim to push it precisely to the point.

There are a lot of charts where the corners are burred or doubled by marks or having artifacts. Then you might use other lines as references as well.

Finally compare the resulting pixel values: ideally you will have four pairs of identical values. If they are 5 pixels off I wouldn't worry. But we have charts with a difference of 80px and more. These charts are skewed and you might add a comment to alert about this fact.

Hubert

[paulr]

Originally Posted by bcn

Dear Paul,

the example would not be acceptable for me.

Finally compare the resulting pixel values: ideally you will have four pairs of identical values.

The calibration in message #538 is from chart 51103 (not one I did). I noticed several others in the same section with slightly 'off' calibrations.
The question, which I guess Pavel can answer easily, is what effect this actually has on the accuracy of the chart, in real life.

I'm not sure that calibrators should be paying too much attention to the resulting pixel values though. I have already seen one chart where the calibration appears to have been done to make the values identical, at the expense of positioning the lines accurately. This just guarantees that the chart remains skewed, whereas, if I understand correctly, Pavel has put huge amounts of skill and work into correcting skew based on precise corner positioning.

I think quality is going to be a big issue and we need to either formalise QA somehow or else use a random 'many-eyes' approach. If everyone doing calibrations checked two others for every one they did, we would surely pick up on lots of errors.

As a case in point, chart 51100 is recorded as "no soundings datum", but the datum is there on the chart - LLW. This needs correction.

Paul