Who is using OpenCPN on the Great Lakes?

[cagney]

There is an important discussion, regarding just how OpenCPN should handle NOAA:s polyconic raster charts for the Great Lakes, going on.
See this link OpenCPN::Tracker All Projects: Tasklist.
What is your experience?
Thomas

[wdkester]

I vote for a project, like the DMA conversion project, to reproject the small scale charts. O would then have to quilt large scale charts (below a certain scale) with the reprojected ones. For Lake Superior coverage I think there are only 3 charts for shure (and maybe 6 others) that need reprojected, the eastern half, the western half and the entire Great Lakes chart. If you have a large scale chart quilted with one of these the results are bizare.

I'm using ENC (but I hate the holes/grey areas) or CM93 until this is resolved.

Given an easy process I would be happy to reproject some charts this winter.

[wdkester]

It looks like 50 or less charts to be reprojected. Of course they should redone when updated. Could somebody point me to the NOAA utility that reprojects polyconic to mercator?

[cagney]

The reprojection tool is here:
Nautical Chart Reprojector | Tools | Digital Coast

Reprojected test chart is available here.. Please right click and use "save as".

Note that OpenCPN, for good reasons, thinks that this chart has poor georeferencing. The chart picture is warped in a way that is not easy to correct. The question is - is it good enough?

Thomas

[cagney]

Originally Posted by wdkester

It looks like 50 or less charts to be reprojected. Of course they should redone when updated. Could somebody point me to the NOAA utility that reprojects polyconic to mercator?

wdkester
Please read the exchange on the quoted tracker thread. You probably have to log in to see it.
Thomas

[wdkester]

I have to download some browser other than explorer to use save as for 14500.

However, when I remove 14500 (scale=1500000) the next smallest scale (for lake superior) is 120000. When I pan this chart around I see an error of maybe 15m. I'm hoping that reprojecting this chart will reduce the error to something acceptable for everyone. I might want to remove 14500 entirely. I'm unlikely to plan any great circle routes between Duluth MN and Buffalo NY.

Will look at reprojected chart tonight.

[wdkester]

I've got to admit that 14500 looks bad on the margins but if you look at the lat/long lines they line up pretty well with the O grid. 14500 is only useful for tracks/routes that go between the Great Lakes. I think each of the lakes has it's own small scale chart (at least some, if not all, of which are merc).

I reprojected 16966_1 (scale 1:120000) and got the georef warning. It doesn't have the ugly margins, lat/long lines match up as well a the standard mercs, and I can't see any problems with WP plotting. I really don't see much difference between the orig and the reprojected version. The O gid and the lat/long are sometimes off in both version by about a line width. In the poly version, with quilting on, I can pan around a WP and get the WP to move relative to some chart feature by a few meters (this is with 14500 disabled). I do not get the same effect quilting off! The WP does not move with respect to the feature.

I forget how to do this but Paul Higgins GE2KAP utility had a way of matching GE shorelines with KAP files. That would be another check.

I'm not sure that there's enough interest in a project to reproject the Great Lakes polys to mercs. If there is I would do Lake Superior and Huron. Otherwise I'll just use ENC's or CM93.

[wdkester]

I just tried plan B: delete 14500, convert any mercators to poly, and use only the largest scale charts to create routes/WP etc or navigate in realtime (which is good practice anyway).

Unfortunately, the conversion routine outputs garbage for mercator to poly, I just get a black screen for the chart 14966_1. So back to ...

Plan C - use ENC's - I'm beginning to like ENC's. Having "groups" makes it easy to look back at the raster charts to see if there is any missing detail.

[bdbcat]

David...

What was wrong with plan A: Convert 14500 to Mercator, and use the others as-is (i.e. in polyconic )?

Dave

[cagney]

Originally Posted by wdkester

I have to download some browser other than explorer to use save as for 14500.

However, when I remove 14500 (scale=1500000) the next smallest scale (for lake superior) is 120000. When I pan this chart around I see an error of maybe 15m. I'm hoping that reprojecting this chart will reduce the error to something acceptable for everyone. I might want to remove 14500 entirely. I'm unlikely to plan any great circle routes between Duluth MN and Buffalo NY.

Will look at reprojected chart tonight.

In a scale 1:120,000 ,1 mm corresponds to 120 meters. If you find an error of 15 m, that corresponds to 0.125 mm. It is of course impossible to produce a paper chart to this accuracy. Re-projecting the chart will not help at all. I think you just proved that OpenCPN does a good job!
I can see 2 charts where re-projecting will help. Apart from 14500 it is 14860 , the only of 5 charts in 1:500,000 that is polyconic.

Thomas

[bdbcat]

Thomas...

I was just going to make that point.

We should remember that the basic feature accuracy of a raster scanned paper chart is no better than 1 mm at the compilation scale. So plotting a 15m accurate GPS position on these small scale charts leaves you a giant blob of uncertainty. On 14500, there is certainly enough accuracy to tell Cleveland from Toledo. But not Cleveland Marina A-dock from B-dock.

Thanks
Dave

[wdkester]

OK, I don't get it. I just moved 14500_m2 into my great lakes folder (but not in a separate folder like the rest of the charts). I removed the 14500 poly from the folder. I checked "rebuild database". I then panned around Lake Superior and got huge errors. Could this be caused by being quilted with the background chart. Singly the charts work just fine.

The charts that you see in the chart bar are: 14966_1, 14661_1, and 14500_m2.

RE PlanA - My goal was to have some small scale charts that quilt with the large scale charts. Making 14500 a merc doesn't help because the next chart in scale is also a merc. A large majority of the charts are poly.

[bdbcat]

David...

Hmmm. I don't get it either.

I have no trouble with 14966_1. If I drop a mark on the light at the south end of the breakwater, it plots fine as I zoom in/out.

Your display does not look like mine. 14966 lays in just fine.

Thomas?

Also, please note that large(er) scale polyconics will be automatically quilted with whatever mercators happen to be available. It does the best it can in placement of the admittedly warped polyconic bitmap.

But I say that the quilting of 1/120,000 14966_1.kap(polyconic) is good enough for navigation to 120 metres.

So, Plan A should work automatically if 14500 is reprojected and 14500(poly) is removed.

Dave

[bdbcat]

Regarding the georef error log message for 14500(reprojected).

I looked into this. It seems that some (all?) of the reference points in the repro chart are in error, badly. An artiact of the NOAA reprojection utility, I think.

We detect this case, assume mercator, and make a best effort fit to the available ref points, using a least-squares curve fitting algorithm. The result is then judged to be OK for navigation.

In the case of 14500, there are over 300 ref points. Plenty for the least squares technique to converge on a good solution.

So, treat this as a warning, for info only.

Dave

[wdkester]

I restarted O And noweverything is OK.

I'm wondering how we communicate Plan A to Great Lakes sailors? An automatically loaded warning layer? A message in the status line. The rollover window? (too brief)