Polyconic quilt gone wrong

[LeaseOnLife]

This might be related to RogerD's problems with some Canadian Charts which I don't have.
http://www.cruisersforum.com/forums/...an-148667.html

But I have seen similar behavior, and made it repeatable:

Here is a simple test setup.
Get two US RNC/BSB charts for Lake St Clair and Clinton River in Michigan from Chart Downloader for NOAA RNC&#174

14850_1.KAP
14853_23.KAP

Create a directory and chart group with just these to charts.

In single chart mode (F9), plan a route on 14853_23.KAP thru the Clinton River. Switch on quilting (F9) and the waypoints are moved outside the river.


Dirk

[bdbcat]

Dirk....

Thanks for the tight demo screens and analysis.

OCPN works(today) in Mercator projection only.

We quilt polyconic charts as a convenience factor only.
We do the same for Transverse Mercator charts.

But we do not "re-project" non-Mercator charts in any way. This would be much too slow for raster charts in real time. We use the supplied georef parameters to make the best judgement of how to lay the chart into a Mercator quilt. As you have demonstrated, for some charts and quilts, the OCPN solution is just plain wrong, and unusable. In this particular case, the quilt looks OK, but the georef on the chart is poor. Hmmm...

The safe, conservative solution is to preclude non-Mercator charts from the quilt, globally. We have tried this, and users were not happy. "What happened to my charts which I painfully downloaded? Why can't I see them?" So, we quilt everything we can find, now.

There is a project underway to enable alternative projections in OCPN. This may or may not make it into the mainstream, depending upon interest and developer resources. This will help.

Meanwhile, here we are.

There may be a way to solve the problem case you demonstrate, specifically with routes and points intersecting polyconic charts.

Ill have a look at it.
If it was entered into Flyspray it would get resources for sure....


Thanks again
Dave

[LeaseOnLife]

FS1810, severity=high (plotting positions wrongly on a chart)

FS#1810 : Georeferencing on quilted polyconic charts gone wrong

Thanks

Dirk

[wdkester]

At the very least, a warning should be given that "Only mercator charts are accurately rendered, all other chart projections are shown because of user whining."

Another idea would be to have a check-box in tools>charts to disable this feature.

[wdkester]

I just enabled grid lines and zoomed in /out on Lake Superior which has a lot of polyconic. You will see big variations; however, if you zoom in far enough the charts seem to be spot on.

I'll be using ENC's when I leave the Great Lakes (again) next year.

Earlier this year we had a small brouhaha about warnings for over-zooming. I think this is a more significant problem for G.L. sailors. There really should be a brief, on screen warning, anytime a non-Mercator chart is loaded.

[HappySeagull]

Originally Posted by wdkester

I just enabled grid lines and zoomed in /out on Lake Superior which has a lot of polyconic. You will see big variations; however, if you zoom in far enough the charts seem to be spot on.

I'll be using ENC's when I leave the Great Lakes (again) next year.

Earlier this year we had a small brouhaha about warnings for over-zooming. I think this is a more significant problem for G.L. sailors. There really should be a brief, on screen warning, anytime a non-Mercator chart is loaded.

N-o-o-o-o-o! There's a cone on the chartbar.
I suppose my opinion is because my polyconics are all Historical Documents and Unsupported. I'd be curious as to what the survey date of these is on yours.... Also, for the most part, mine are large scale harbour plans and never quilt with each other.Practically speaking, I don't see how they could, given the nature of the projection.
There is a way to manually recalibrate and make them into quiltable mercs-done that- but it's more than most will bother with.

[wdkester]

Originally Posted by HappySeagull

N-o-o-o-o-o! There's a cone on the chartbar.
I suppose my opinion is because my polyconics are all Historical Documents and Unsupported. I'd be curious as to what the survey date of these is on yours.... Also, for the most part, mine are large scale harbour plans and never quilt with each other.Practically speaking, I don't see how they could, given the nature of the projection.
There is a way to manually recalibrate and make them into quiltable mercs-done that- but it's more than most will bother with.

For the western half of Lake Superior the Polyconic dates were 2014 and 2015, except for a small scale chart with a date of 2002. The Polyconic charts had more recent dates than the ENC's for the same area.

The majority of the Lake Superior charts are Polyconic (and I think this may be true for other Great Lakes). If you use OpenCPN for Lake Superior you will use Polyconic or ENC.

My advice for Great Lakes users of OpenCPN is to use ENCs. If you can't do this, then turn gridlines on and you will see the areas that are rendered poorly.

To be honest I have used OpenCPN and RNCs on Lake Superior and I didn't run into anything; but I have a good set of verified waypoints.

[HappySeagull]

You're kidding! SURVEYED in 2014 -2015 as polyconic? and not just "corrected?

Mine are all originally surveyed many years ago. An example is 3781 surveyed 1911-1983, which suggests the originals were mostly done long ago with the blank spots filled in by 1983. ( I remember blank spots on charts here and I'm not that old.)

another example is 3682,surveyed by Mr. Parizeau, Cmdr. Knight and Assistants 1935-1937
last published in 1987 though there were corrections after.

I would have expected yours to be the same or even older, given the history of the Great Lakes.

[seandepagnier]

Dirk,

Did you try my projections branch? See this thread:
http://www.cruisersforum.com/forums/...st-148103.html

I really need feedback and testing before I can make a pull request.

Originally Posted by bdbcat

Dirk....

Thanks for the tight demo screens and analysis.

OCPN works(today) in Mercator projection only.

We quilt polyconic charts as a convenience factor only.
We do the same for Transverse Mercator charts.

projections branch can quilt all types of charts together and reproject them into many projections as well.

But we do not "re-project" non-Mercator charts in any way. This would be much too slow for raster charts in real time.

The projections branch re-projects in realtime. (requires opengl)

From this work I discovered that the polyconic charts around the great lakes use a conic projection that is different than the polyconic projection used by opencpn. This is not desirable as the coordinates will be slightly incorrect. It might not matter at all (I need feedback) because the error is nothing like what you are seeing now which is really a different problem (corrected by the projections branch) We could detect it. With more work, we could analyse as many conic charts as possible and determine which projections are used and how we can deal with this. My guess is we are simply using the wrong formula.

We use the supplied georef parameters to make the best judgement of how to lay the chart into a Mercator quilt. As you have demonstrated, for some charts and quilts, the OCPN solution is just plain wrong, and unusable. In this particular case, the quilt looks OK, but the georef on the chart is poor. Hmmm...

Please try the projections branch, it should work much better in all cases and in many cases I have observed the master branch is completely broken (course up us ENC quilts)

[seandepagnier]

Dirk,

if you do try the projections branch, you should force update the chart cache to re-generate polyconic borders (which aren't straight lines)

[LeaseOnLife]

I think we are on the wrong track. The quilting / display of polyconic charts lines up pretty well for the Southern Great Lakes, gets worse further North. Look at the two screenshots in post #1, the charts line up just fine, it is the route plotted on top of it that goes wrong.

The issue seems to be the geo-referencing of own-ship, marks, tracks on top of a quilted Polyconic chart, see post #2 from Dave where he says "Hmmm". I don't remember it being that bad. I would have seen this problem on our return to the Great Lakes (08/2012) on whatever old version (3.?.?) I was using back then. Planing a route thru the Detroit River on RNC charts is currently almost impossible: When moving the chart the waypoints move all over, land, water, both sides of the river. I never used S57 charts, so OpenCPN 3.?.? dealt better with this issue. Currently it is horrible, unusable. Telling us to use S57 is a workaround but I like raster charts so much better. Back to SeaClear for this long weekend

Dirk

[LeaseOnLife]

Originally Posted by boat_alexandra

Dirk,

if you do try the projections branch, you should force update the chart cache to re-generate polyconic borders (which aren't straight lines)

Sean

I will test the projections branch some time in future, I really like what you have done, but currently I need a usable setup, no time to tinker with SW.

Dirk

[wdkester]

Originally Posted by HappySeagull

You're kidding! SURVEYED in 2014 -2015 as polyconic? and not just "corrected?

Mine are all originally surveyed many years ago. An example is 3781 surveyed 1911-1983, which suggests the originals were mostly done long ago with the blank spots filled in by 1983. ( I remember blank spots on charts here and I'm not that old.)

another example is 3682,surveyed by Mr. Parizeau, Cmdr. Knight and Assistants 1935-1937
last published in 1987 though there were corrections after.

I would have expected yours to be the same or even older, given the history of the Great Lakes.

You're correct these are the update dates. I couldn't find the survey dates; it doesn't seem to be in the bsb header. The point is moot however; because the majority of the charts in Lake Superior (and I think the rest of the GL) are Polyconic.

In the help file I found two significant references to "Polyconic".

"OpenCPN, supports Mercator Charts, Transverse Mercator Charts, Universal Transverse Mercator (UTM) Charts and Polyconical Charts (used in parts of US and Canada). Charts using other projections will not be displayed. A note about the reason for the display refusal, will be found in the "opencpn.log".

No mention of problems. And -

"The screen is a quilt of two Polyconical charts. Hovering the mouse pointer over the left chart-button highlights the lager scale chart for Manitowoc. Two smaller scale charts that cover the area displayed on the screen are included in the quilt, but are invisible as they are covered by the larger scale charts. To view these charts, right-click anywhere on a chart in the quilt and select "Remove this chart from quilt", one of these charts will now be shown in the quilt.
The Brown right most button shows that CM93 version 2, charts are available for the area, but not currently displayed.
The information box shows relevant information about the chart. We can see that the chart is Polyconic, and because of this a warning is displayed for poor accuracy. The reason is, that although Polyconic charts are allowed to participate in quilting, the result is not totally free from errors, however small."

I couldn't get the error message, for this example, at any level of Chart Zoom/Scale Weighting.

I think there should be:
1. A strong warning in the the first help reference above.
2. One on screen warning that is displayed the first time a Polyconic chart is used in a session and fades after 10 sec. This is for the benefit of people who don't read the manual or the forum.

[mstrebe]

There's a product called "Geocart" that is capable of performing any-to-any reprojection of arbitrary raster sources:

https://www.mapthematics.com/

It's available for Windows and Macintosh. The quick solution would be to reproject the original data to mercator (or is it actually platte-carre?) before importing it into OpenCPN.

[bdbcat]

@LeaseOnLife...

Dirk...

I realize this thread is a little stale, but I come back to it from FS#1810.

I am working to integrate the extended projections methods into the next Beta, and so I try to reproduce your trouble with routes in Lake St. Claire on OCPN 4.1.602.

And I cannot. For me, the plotted route points move by a few meters in quilt mode, but not hundreds of meters as in your example. For me, it is surely useful for navigation, while I agree your example is not.

Any idea what might be going on here?

Thanks
Dave