Do we need/want an AIS filter?

[rooiedirk]

OK think I have the target scaling working for testing.
An extra entrance in the option menu.
In the config file the weighing of each used value can be changed, although I believe the default values are not to bad.

AISScaledNumber=5
AISScaledNumberWeightSOG=150
AISScaledNumberWeightCPA=100
AISScaledNumberWeightTCPA=100
AISScaledNumberWeightRange=100
AISScaledNumberWeightClassB=-20
AISScaledSizeMinimal=50

As default the class B targets are reduced a litle earlier then class A.

Coming to git soon https://github.com/RooieDirk/OpenCPNhttp:// Branche Ais-filter

[rgleason]

I like your 'smart alarm' for dueling ships turning in seas.
Perhaps we could add a check to see when the last alert was sent and have that user set?

I like the target scaling, looks good. Could we use the same approach for alarms and Class B?

[rgleason]

Can we discuss FS#1587 briefly - What should be done?
Close it? Guard Zone needed? It's pretty technical.

FS#1587 : AIS Alarm Target proximity alert - (should be closed)

Rick

An interesting question is: how did the boat get into the CPA range without the CPA alert being triggered?
Perhaps the first AIS signal from the boat was not received until the boat was already less than 1NM away?

-- Perhaps they turned their AIS on at the last minute.

I think this should be closed.

Marco response to the same

An interesting question is: how did the boat get into the CPA range without the CPA alert being triggered?
Technically this can be done when transmitted AIS data is incorrect. Theoretical situation: the target is south of us and moving north, but it's COG is 180... It IS getting closer, but CPA won't trigger as according to it's own data it's moving away. A guard range would notice this, CPA calculation won't. A valid question then is... How reliable is the other GPS data when COG is invalid? Cq. What are you actually determining when you raise an alarm?

Invalid headings are quite common. Heading sensors (gyro, GPS compass) that are incorrectly configured. But position, SOG, COG all come from the same device: the GPS. Heading comes from a different device. Heading isn't used in (T)CPA calculation.

Your thoughts would be appreciated.

[rgleason]

Can we discuss FS#807 Briefly? - Marco has some information we should be considering.

FS#807 : AIS Alarm Reduce “Nuisance� AIS Alerts

Steve B:

When your own boat swings and yaws, alerts may be generated, but as the boat corrects, they disappear. In some cases it takes a few attempts to acknowledge them as they disappear too quickly. Then they reappear as new alerts and the process continues.

I suggest another way of addressing this ..... by having a damping factor so that it takes a few seconds of continued alert conditions (based on own boat heading and speed) before the alert sounds. An example can be drawn from the way radar alarms work to reduce false alarms. The first echo is ignored, and only when subsequent scans find something in the same location do they alert - at least that's the way mine works.

Marco:

Damping doesn't sound like it will sove anything (sorry Rick ). GPSes should already send 'stable' data. That means position, COG & SOG should be 'stable'. Data derived from that ((T)CPA) should then be stable as well.

If a ship is 'jumpy', this is because it's GPS is either incorrectly configured (Kalman filter set incorrectly) of not suitable for navigational purposes (like a lot of mouse-GPSes; no filtering at all). See also FS#222 .

RooieDirk are the $2000-$3500 Class A AIS units significantly different than Class B with GPS in some specific ways? IE Kalman filters. What do you think? I liked the ideas you had above.

[rgleason]

One more FS#1881 - AIS Alarm - Proximity alert

FS#1881 : AIS Alarm - Proximity alert

Rick

Under Options > Ships > AIS
"No CPA if Target Range is greater than (nm)"
This is the same as a Radar Guard Zone.
This item should be closed now. Hotrod would you please close it now? Bdcat want OP to close.

Frode Lynum

Well this feature has not been fulfilled, it only set a given distance not a area.
Look at the picture, the sones are the sectors at given distance for outer and inner part together with the width of the sector.

What has been requested is an alert when a vessel passes trough a set guardzone/area both no matter what direction the other vessel/vessels move trough that area.
So this is not the same as a radar guard zone.

What should be done about this?

[Paul Elliott]

Marco said:

Damping doesn't sound like it will sove anything (sorry Rick ). GPSes should already send 'stable' data. That means position, COG & SOG should be 'stable'. Data derived from that ((T)CPA) should then be stable as well.

Perhaps they should, but define "stable". I've got GPS units (including one on my Class-B transponder) that certainly show my COG swinging around as my boat yaws in the big swells. I typically put a simple smoothing filter on the COG before using it for AIS alarm processing (using NavMonPc's configurable filters).

As I mentioned, NavMonPc doesn't use ROT in it's AIS alarm processing, but that sounds like a good way to help filter out transient CPA/TCPA alarms. Class-B transponders don't send ROT, but Class-A units do.

[rgleason]

Thank you Paul, that is my experience too. We need a damper and I vote for your approach or RooieDirk's (provided it does not bog down the AIS with too many calcs). Class A and Class B may be two different instruments, but Opencpn AIS should be easily configurable for both uses. The "hockey puck GPS" have improved accuracy starting with sirf chips that surpassed anything before.

[rooiedirk]

Running into a small coding problem.
When tying to implement a cycle for the AIS button on the toolbar, I need to add extra icons. I can make these icons as ...svg files, but does anybody how to implement this in O?
Couldn't find it in developers manual.

[rgleason]

Jon ocpn_draw_pi has been dealing with this.
He had rotating buttons for awhile, then created a popup that seems to work fine.
He has two main buttons for ocpn_draw.
Perhaps you can look at his code.
I know he uses small jpgs that get installed.
I will leave a note on his thread about this.
http://www.cruisersforum.com/forums/...ml#post1978262

I believe Pavel is working on a new system when he has time, to convert all this to some other way of rendering button icons that is more compatible with all the platforms, which would also apply to all the nav aid symbols, but that is something in the distant future.

I see https://github.com/RooieDirk/OpenCPN now works.

[rooiedirk]

Draw, that is in a plugin. For a plugin it is different then in O itself.

[jongough]

The Draw plugin was originally coded into the mainline, but Dave wanted it extracted and put into a plugin, so that is what happened.

The icons for the toolbar are all stored in the same png file, toolicons_*.png, in the src/bitmaps directory. There are three of these files that have different icons to give a different look to OCPN. You will need to add your icons into these files then you have to tell OCPN how to find the icons. The styles.xml file defines the icon name, its location within the png file and the size of the icon. The icon is added to the toolbar by adding code into chart1.cpp/Frame::CreateAToolBar(). Look for "Hide AIS Targets" and use this as a prototype.

My original mainline code for boundaries is still around. If you want to have a look you will find it here:
https://github.com/jongough/OpenCPN/tree/boundary
It 'should' compile OK and you 'should' be able to run it. It is nearly a year old now so is rather out of date.

Hope this helps
Jon

[rgleason]

Thankyou Jon

[bdbcat]

Dirk...

We don't (yet) draw SVGs in OCPN. This is something that is being worked now, probably make it into 4.2.

For now, you will find some scripts that convert from SVG to PNG in the src/bitmaps directory. The scripts call inkscape to do the heavy lifting.

Example:
opencpn/src/bitmaps/32x32_svg_src/cursor/create_all_32x32.sh

These scripts are run manually against the SVGs, and the PNGs are stored in the git build tree.

Good Luck
Dave

[Hakan]

Dirk..
If you don't mind I'm curious and have tried to build your AIS-filter branch, merged with recent OCPN Git code, on Windows/VS2013.
No hurry at all but if you're interested I received the following error:

OCPNPlatform.obj : error LNK2001: unresolved external symbol "long g_ShowScaled_Num" (?g_ShowScaled_Num@@3JA)

Håkan

[rooiedirk]

Originally Posted by Hakan

Dirk..
If you don't mind I'm curious and have tried to build your AIS-filter branch, merged with recent OCPN Git code, on Windows/VS2013.

Ah a small slip. I can't check under windows, and under linux it is working, But I think if you change line 161 in OCPNPlatform.cpp into:

extern int                          g_ShowScaled_Num;

(is now wrongly extern long g_ShowScaled_Num
I will also update in git.
Dirk