Charts II: NGA - 2700 Charts

[r.fairman]

I dont know how the web based calibration coding is going but thanks to nohal and net surfer in anticipation Has anyone considered using OCR to fully automate the process. I have captured a corner of a chart and run though a free ocr program and converts the lats and longs ok if i rotate the image I know this is a simplistic idea but maybe worth consideration. I now need a ocr that will convert horizontal and vertical text on one go.

regards

Richard

[phiggins]

Can anyone tell me how MapCal or OpenCPN figures out the lat/lon for the upper left corner of a chart

For instance if you take chart 25001 and the calibration data is:

[25001.PNG]
NA=Caribbean Sea-Eastern Part 25001
FN=25001.PNG
GR=0
PY=1
QU=0
SC=1300000
BC=4
B1=9.99810000000001,-72.3333333333333
B2=10.0158500028536,-59.0154999992096
B3=18.7684833333333,-59.0096166666667
B4=18.7555166673491,-72.3254166664701
CC=4
C1=7094,641,18.766666666667,-59.000000000000
C2=170,5327,10.000000000000,-72.333333333333
C3=3459,3212,14.000000000000,-66.000000000000
C4=5535,1058,18.000000000000,-62.000000000000
GD=WGS84
NU=
PR=1
DS=0,0
WI=7360
HE=5699
DX=201.65
DY=201.65

if use the first two calibration points:
X1=7094 Y1=641 LAT1=18.766666666667 LON1=-59.000000000000
X2=170 Y2=5327 LAT2=10.000000000000 LON2=-72.333333333333

Then use this formula for degrees/pixel:
yres = (lat2 - lat1) / (y1 - y2)
xres = (lon2 - lon1) / (x2 - x1)

I get: XRES=0.00192566917003654 YRES=0.001870820884905463

then if I use this formula to figure out the upper left lat (lat3) and the upper left longitude (lon3) with x3 and y3 = 0

lat3 = lat1 + (y1 - y3) * yres
lon3 = lon1 + (x3 - x1) * xres

the answer it gets is:
lat3=19.96586285389140 = 19 57.951 N
lon3=-72.66069709223921 = 72 39.641 W

but Mapcal and OpenCPN says its lat=19.55.385 N lon=72 38.544 W

Both are off quite a bit.

Does anyone know the formula?

[sinbad7]

MY method of converting NGA bitmaps,whatever it's worth, is as follows:

1. Reduce the enormous file sizes to 60-70% of originals using 'BatchImageResizer'
2. Convert the files to 127 col in .png format using GIMP.
3. Calibrating the files using MAPCALII
3. Converting the files to .kap format using GE2KAP

For me,this produces a precise and sharply defined chart of acceptable size of 1-3 MB. Checked against the grid in OCPN it produces an accuracy more than adequate for me.

Tore

[sinbad7]

Here is a typical .png file spec. after preparation in GIMP2.
Anything specifically wrong with it?

Tore

[Meark56]

I wanted to download the map 52046 for calibration, but I got the image of the map 52045. I think there is a problem of incorrect link or incorrect Jpeg.
To correct and to help you, I have the right dezoomify 52045 and placed on the server project NGA charts.

Regards

[paulr]

Originally Posted by Meark56

I wanted to download the map 52046 for calibration, but I got the image of the map 52045. I think there is a problem of incorrect link or incorrect Jpeg.
To correct and to help you, I have the right dezoomify 52045 and placed on the server project NGA charts.

Good find.

There appear to be at least 4 other incorrect files on the ftp server:

26247, 57082, 61200, 92220

I say "at least" because I found these by checking duplicate file sizes and it is possible for there to be duplicates with different file sizes, due to the various download and stitch scripts in use.

I am getting the 4 files above from NOOA, but others with faster connections are very welcome to step in and upload them to the ftp server first!

Paul

[Tia Bu]

Paul:

I'm not at all sure this is correct, but it seems to me the problem you are encountering is because these charts are Mercator projections. So you can't make a linear equivalence between lat/lon and the x/y dimensions. In other words, at the top of the chart, a minute of latitude or longitude occupies a different number of pixels than at the bottom of the chart.

I don't know how Mercator projection works mathematically, but I'm getting ready to Google it :-)

M.

[Tia Bu]

Google and Wikipedia lead quickly to the formula for converting lat/lon to x/y coordinates in a Mercator projection, but my calculus is too far back in the past to make much sense of it: Mercator projection - Wikipedia, the free encyclopedia

There are a good many charts in the Miscellaneous region that aren't Mercator projections. The Great Circle charts and polar charts are a conical or gnomonic projection, I think. Can OpenCpn or other chart plotting programs even display those?

[phiggins]

Tia Bu,

Yes I realize that the Meractor projection is the problem. I googled it to death and tried a dozen solutions that I found for what is called the Gudermannian formula, but none gave the answer.

Anyway I think I found the code in the OpenCPN source so I'm going to steal/borrow it and try it.

To me it looks like the OpenCPN code handles many different projections.

Thanks for your help.

[Tia Bu]

Originally Posted by sinbad7

MY method of converting NGA bitmaps,whatever it's worth, is as follows:

1. Reduce the enormous file sizes to 60-70% of originals using 'BatchImageResizer'
2. Convert the files to 127 col in .png format using GIMP.
3. Calibrating the files using MAPCALII
3. Converting the files to .kap format using GE2KAP

For me,this produces a precise and sharply defined chart of acceptable size of 1-3 MB. Checked against the grid in OCPN it produces an accuracy more than adequate for me.

Tore

That looks great. Have you done any charts with insets yet? If so, how did you handle those?

M.

[phiggins]

I don't know if sinbad7 has done any insets but I have: chart 24402 has two insets that I just removed by putting in more border points to go around them with MapCal. Its a bit tricky but first put the 4 around the outside starting in Lower left corner then save the calibration then move the 2nd point back and then start inserting new border points.

And you can do both of his steps 1,2 with Nconvert, much easier than firing up Gimp.

Also, I just uploaded GE2KAP.V6.0.10 which handles images up tp 5Gig. Of course you have to have the virtual memory to support that size. It also fixes a few things Tore found.

[paulr]

Originally Posted by paulr

There appear to be at least 4 other incorrect files on the ftp server:

26247, 57082, 61200, 92220

I have now uploaded the correct versions of these files to the FTP server.
Along with Meark56's correct version of 52046, they are all in the incoming folder, along with July and August files.

I think I am correct in saying that they only become available for download once they have been relocated on the server.

Paul

[sinbad7]

If you are not going to use the insets for navigation,why not just ignore them?
They can be handy for a visual inspection of the harbour in more detail.
The charts paper corner points remain the same.

For some reason,none of my progs. including NConvert would open many of these huge NGA charts,which is why I used BatchImageResizer which can batch reduce large files without opening/displaying them first.

Tore

[wdkester]

Originally Posted by sinbad7

If you are not going to use the insets for navigation,why not just ignore them?
They can be handy for a visual inspection of the harbour in more detail.
The charts paper corner points remain the same.

For some reason,none of my progs. including NConvert would open many of these huge NGA charts,which is why I used BatchImageResizer which can batch reduce large files without opening/displaying them first.

Tore

Tore,
Which BatchImageResizer are you using. The one that I downloaded from CNET doesn't seem to have a % reduction.

[Tia Bu]

Originally Posted by sinbad7

If you are not going to use the insets for navigation,why not just ignore them?
They can be handy for a visual inspection of the harbour in more detail.
The charts paper corner points remain the same.

For some reason,none of my progs. including NConvert would open many of these huge NGA charts,which is why I used BatchImageResizer which can batch reduce large files without opening/displaying them first.

Tore

Yes, I wound up just ignoring them as I made some charts. But some of the charts are really nothing but a collection of insets, many of which would be very useful for navigation. On a cruise up the U.S. East Coast and back this summer, I wound up using the NOAA insets a great deal in OpenCPN. At least with the NOAA charts, where there's an inset, you generally need it.

The best way to deal with insets I've found so far is to cut them out of the image using GIMP and make them a separate chart. But surely there's a better way?