Charts II: NGA - 2700 Charts

[cagney]

Originally Posted by wdkester

Chart 26128 Kingston Harbor. Opencpn's Long. grid lines don't match the chart. I believe that the corner marks for long. are about 0.5" off This is an error of abt 30mtrs. Not much, but this is a harbor chart. So? For my own use, if I thot that I were going to Kingston, I would "correct" the chart in my files.

I added a comment noting the error in the database.

I downloaded the original jpeg and found, through linear interpolation, that the printed corner longitudes were slightly out. Correcting the SW long to 76 55 57.9 and the NE long to 76 43 05.07 brought the longitude error down to the region of 2.5 m.

Thomas

[dacust]

Originally Posted by Tia Bu

Hey guys:

I was just working on the charts. I changed a few charts that somebody else started. And I noticed some folks have changed a few charts that I previously started (and some I previously had "finished").

And in all those cases, the changes seem to have improved the product.

It occurs to me that what we may have inadvertently begun building here is a chart Wiki. Is that where this project ought to wind up going? Studies show that Wikipedia is more accurate than Encyclopedia Brittanica, even though anybody in the world can change Wikipedia anytime they want to. Could we end up with a site that allows people to append these charts with up-to-date soundings, updates from Notices to Mariners, and other data gathered in the field?

Then the warning on the charts might be-- hey, this is a Wiki chart from a group Internet effort. Be careful! Use it wisely.

Would that be more valuable in the long run? I kind of get the feeling that we're giving some of these charts more careful attention than they've gotten in a long time.

Just a (very) half-baked thought. For what it's worth.

In really like the idea of a chart wiki.

IF the original charts are available for download, as well.

The way I would feel comfortable using wiki maintained charts is if I could toggle back and forth between the original and the updated charts so I could easily see what people had changed/added/deleted. This is another use of the "user defined chart types" in the feature requests.

-dan

[Tia Bu]

Originally Posted by dacust

In really like the idea of a chart wiki.

IF the original charts are available for download, as well.

The way I would feel comfortable using wiki maintained charts is if I could toggle back and forth between the original and the updated charts so I could easily see what people had changed/added/deleted. This is another use of the "user defined chart types" in the feature requests.

-dan

But the original charts aren't available for download. That's why we're doing this. We're worried about quality assurance. But the quality assurance of Wikipedia is ultimately (and counter-intuitively for us old farts) the fact that anybody can change it.

Many volunteers working over the Internet in that case have proven to be more accurate than one of the most conscientious paid editorial staffs on the planet.

What might this mean 1) in terms of a q.a. approach and 2) later, in terms of actually improving on the NGA product?

Wikipedia is not just a wild-west scene. There is an extensive, highly structured volunteer effort behind it that helps maintain quality. Could we emulate it? And would that even be a worthwhile direction?

[dacust]

Originally Posted by Tia Bu

But the original charts aren't available for download. ...

Originally Posted by Tia Bu

...
Could we end up with a site that allows people to append these charts with up-to-date soundings, updates from Notices to Mariners, and other data gathered in the field?

Then the warning on the charts might be-- hey, this is a Wiki chart from a group Internet effort. Be careful! Use it wisely.

...

I was referring to the appending of notice to mariners, etc.

-dan

[recri]

The latest status of the projected aspect ratio test is:

processed: 1203 charts
accepted: 1162 charts
mean ratio 1.00483 stddev 0.00189

problem: chart 83495 ratio 1.02724

Chart 83495 is Iles Wallis and it needs attention from someone who wants to measure all the chart corner coordinates, again.

I've fixed problems with the following charts rather than posting:

54406 62560 83218 83397 91060 63053 63054

There were a couple of misprinted corner coordinates in this set, as well as some hand measured chart corners that didn't work out. One had meticulously measured three edge coordinates to fractions of a second, which were all right, then mistyped the last digit of the one edge coordinate that didn't need measurement. Life is cruel.

My apologies to whoever is working on Karachi for making it move around while you weren't looking.

[wdkester]

Originally Posted by cagney

I downloaded the original jpeg and found, through linear interpolation, that the printed corner longitudes were slightly out. Correcting the SW long to 76 55 57.9 and the NE long to 76 43 05.07 brought the longitude error down to the region of 2.5 m.

Thomas

Thomas,

Should we (the unwashed masses) attempt such corrections? I'm not shure I'm capable. I spent at least 30 min trying to determine whether to add or subtract .5" and I didn't even consider different values for each long.

When we were discussing Puerto Morales, I thought that the consensus was to leave the warts. I suppose, if our download process is responsible for the distortion, we should correct the chart. And if the NGA is responsible for the warts, we leave them.

I notice that you subtracted different amounts from each long, so my intention to add .5" to each would have made things worse

[HappySeagull]

Originally Posted by wdkester

Thomas,

Should we (the unwashed masses) attempt such corrections? I'm not shure I'm capable. I spent at least 30 min trying to determine whether to add or subtract .5" and I didn't even consider different values for each long.

When we were discussing Puerto Morales, I thought that the consensus was to leave the warts. I suppose, if our download process is responsible for the distortion, we should correct the chart. And if the NGA is responsible for the warts, we leave them.

I notice that you subtracted different amounts from each long, so my intention to add .5" to each would have made things worse

the answer must be no.Put on a shirt and wash your hands,at the least!
but there's no consensus that I know on what to do with these cripples beyond running away and letting someone else nurse them-that'd be my first choice with so many healthy ones available to brood.

But otherwise,why not download em and calculate offline and then fill in the fixes online with a note,(put one here too,to keep recri's dogs from savaging it).It may even then take a few tries in the KapCloud but if you stand by,and grab it first to test...well,
If the fledged kap now flies straight,the thing is solved...

the thing is,note it.The fledgling will get more intense scrutiny than any other,I bet.

[cagney]

Originally Posted by wdkester

Thomas,

Should we (the unwashed masses) attempt such corrections? I'm not shure I'm capable. I spent at least 30 min trying to determine whether to add or subtract .5" and I didn't even consider different values for each long.

When we were discussing Puerto Morales, I thought that the consensus was to leave the warts. I suppose, if our download process is responsible for the distortion, we should correct the chart. And if the NGA is responsible for the warts, we leave them.

I notice that you subtracted different amounts from each long, so my intention to add .5" to each would have made things worse

The original error of about 30 m corresponds to an error of about 2mm on a paper version of the chart. If you consider a printing accuracy of ~ 1 mm + some damp affecting the scanned chart, and you have one possible explanation. Normally I would not bother with an error like that, as you cannot count on an accuracy better than 1.5 mm equal to 24 m in this case. So a sensible navigator has margins way more than that. In this case it was a challenge. I wanted to pinpoint what was going on.

The Puerto Morales chart is different. The chart is made using a local datum, that is extremely bad for a continental area (as opposed to an isolated island in the middle of an ocean). It's a chart were the cartographers cannot determine a correction that will make the chart into WGS84. The coastline is "warped" and I think it's impossible to make it agree with GE ( I think that phiggins tried). 26128 is a WGS chart though.

The method I used, is one of several methods that we are forced to use for many chart anyway. There are regular chart without lat/lon printed in the corners, and even more plans or inserts comes into this category.

While it's not necessary to use the corner coordinates for geo-referencing, the are needed for the PLY polygon. It's a big advantage for this project to do all chart in as uniform way as possible, and then we do need the lat/long of the sides of the charts.

It's not all that complicated to do this, but let's wait and see how many chart(let)s there will be that needs this treatment, and how many are left when all other charts are done. I do a few now and then and at least one other user is also in there.

You can use linear interpolation, but be aware of the mercator growing scale away from the equator. OpenCPN can do the job if you pick other reference points. MapCal would be another possibility.

Thomas

[wdkester]

27061 is most peculiar. The corners are entered properly. The NW, NE, and SE corners are OK. The SW corner lat and long are off by more than .5'

[wdkester]

OK, I opened 27061.KAP in MApcal. It puts marks in the corners; but the SW corner mark appears in the inset no the larger chart corner. Mapcal doesn't seem to allow you to edit.

[HappySeagull]

Mapcal will only edit and write to it's CHARTCAL.DIR by starting with an image.A bmp (256 colour only!)or a png.It won't edit a kap.


You could
A)download the jpeg and convert it...if too big,you'll need to resize to 50%.Best to do as a batch conversion,even if it is just one-(saves ram when the huge picture isn't opened)
B)extract the image from the kap with cmd line utilities imgkap or bsb2png(from the good ol' "libbsb-0.0.7" stuff)

[cagney]

Originally Posted by wdkester

27061 is most peculiar. The corners are entered properly. The NW, NE, and SE corners are OK. The SW corner lat and long are off by more than .5'

The ref cross was on the insert, not the chart proper in the SW corner.
Corrected now.

Thomas

[recri]

Speaking of ref points on insets, here are the statistics for pixel squareness, the ratio between the lengths of the two diagonals. The average chart is very square by this measure and the distribution looks normal and narrow.

This histogram identifies the charts that occupy the tails of the distribution. The first four I checked (43376 52042 61472 44400) were on the upper tail at values < 0.9959 and all four had inset corners picked as REF points (though I had to look at 44400 twice to see it).

The rest of the charts on the tails should be checked for inset corners.

mean pxsquare 1.00006 stddev 0.00112
0.9959 - 1 - + 37228
0.9986 - 1 - + 83218
0.9988 - 1 - + 52140
0.9989 - 2 - + 83555 92025
0.9990 - 1 - + 91096
0.9991 - 4 - + 21161 55040 61003 81565
0.9992 - 1 - + 83020
0.9993 - 5 - + 121 29322 63015 81511 82304
0.9994 - 8 - +
0.9995 - 15 - ++
0.9996 - 39 - ++++
0.9997 - 46 - ++++
0.9998 - 115 - ++++++++++
0.9999 - 168 - +++++++++++++++
1.0000 - 367 - +++++++++++++++++++++++++++++++
1.0001 - 187 - ++++++++++++++++
1.0002 - 116 - ++++++++++
1.0003 - 58 - +++++
1.0004 - 18 - ++
1.0005 - 10 - +
1.0006 - 5 - + 145 800 22542 24381 29282
1.0007 - 6 - +
1.0008 - 3 - + 81576 82025 83392
1.0011 - 2 - + 707 82384
1.0020 - 1 - + 44057
1.0023 - 1 - + 401
1.0144 - 1 - + 44058
1.0154 - 1 - + 44070
1.0174 - 1 - + 25482
1.0175 - 1 - + 44073
1.0184 - 1 - + 27084

[paulr]

Originally Posted by recri

Speaking of ref points on insets, here are the statistics for pixel squareness, the ratio between the lengths of the two diagonals.

The rest of the charts on the tails should be checked for inset corners.

Good stuff!

I have corrected:
52140
55040
44073
44070
44058
44057

In each case, the calibrator had mistaken an inset for the corner.

81565, 81511, 81576, 83218 all look OK, but they have in common that they are all old charts with no projection stated.

Paul

[recri]

Originally Posted by paulr

Good stuff!

...

81565, 81511, 81576, 83218 all look OK, but they have in common that they are all old charts with no projection stated.

Paul

Thanks!

Please keep reporting charts that look OK. I'm making a list so that they'll be tagged as "having been looked at" in future reports.