SRL Positioning System (OFFICIAL THREAD)

[Flight]

Now that SPS works quite fast, we could try using bigger areas, that'd be more user friendly? instead of 500x500, we could try 1000x1000?

I definitely agree. I wonder if this is part of my problems I've been having with walking paths. Maybe bigger areas would mean more accuracy, less chance to get points mixed up with surrounding areas. I'm for this, Marpis.

[marpis]

I definitely agree. I wonder if this is part of my problems I've been having with walking paths. Maybe bigger areas would mean more accuracy, less chance to get points mixed up with surrounding areas. I'm for this, Marpis.

Well, it probably won't add accuracy, and accuracy is not a problem anyway. There's something else going wrong there then. It just make it easier to pick areas.

[Flight]

Well maybe or maybe not. It seems to have problems when a new map region is loaded.

[Coh3n]

Now that SPS works quite fast, we could try using bigger areas, that'd be more user friendly? instead of 500x500, we could try 1000x1000?

Wouldn't that decrease accuracy?

Well maybe or maybe not. It seems to have problems when a new map region is loaded.

You mean when a new surface is loaded?

I've been having a little trouble with the Dwarven Mine. It works really well sometimes, but other times GetMyPos returns something completely wrong. Marpis, the MatchesFound addition would help with this would it not? Part of it could be that the mine is a pretty consistent colour, but on the other hand, each part of the map/minimap has a unique outline that distinguishes it from other areas.

[Flight]

Wouldn't that decrease accuracy?

You mean when a new surface is loaded?

Yes.

[marpis]

Wouldn't that decrease accuracy?

You mean when a new surface is loaded?

I've been having a little trouble with the Dwarven Mine. It works really well sometimes, but other times GetMyPos returns something completely wrong. Marpis, the MatchesFound addition would help with this would it not? Part of it could be that the mine is a pretty consistent colour, but on the other hand, each part of the map/minimap has a unique outline that distinguishes it from other areas.

Very unlikely that it'd decrease accuracy.

Well, the MatchesFound addition could help, but it would probably just return (-1, -1).
You can try smaller tolerance.

[Coh3n]

Yes.

What kind of problems?

Very unlikely that it'd decrease accuracy.

Well, the MatchesFound addition could help, but it would probably just return (-1, -1).
You can try smaller tolerance.

Well from Harry's issue with the water/maze, and my issue with it repeatedly clicking black on the minimap, couldn't we somehow have it check the colour on the minimap with the colour on the SPS area? If they're no where near each other, something is wrong?

[Flight]

As I explained previously it just walks to random spots about half way through my path (I figured it was very close to where the new area is loaded), and counts those wrong points as part of the path array, so it sill thinks it walked the path successfully.

[Coh3n]

As I explained previously it just walks to random spots about half way through my path (I figured it was very close to where the new area is loaded), and counts those wrong points as part of the path array, so it sill thinks it walked the path successfully.

Sorry, I don't really understand. When I say "Surface" I mean Runescape Surface, Essence Mine, Dwarven Mine, etc. When I say "Area" I mean the pieces of the map, the 13_12 and 10_9 that are called upon in SPS_Setup.

The areas are all loaded at once, not as you're walking the path. Your problem may be resolved if you use the new maps, I'm not sure. I've created a new branch with the new maps that can be downloaded here. Keep in mind that you have to remake your path, and if you're using the PathMaker, use the new RS Surface map in SPS/img/runescape_surface/.

[Flight]

Sorry, I don't really understand. When I say "Surface" I mean Runescape Surface, Essence Mine, Dwarven Mine, etc. When I say "Area" I mean the pieces of the map, the 13_12 and 10_9 that are called upon in SPS_Setup.

The areas are all loaded at once, not as you're walking the path. Your problem may be resolved if you use the new maps, I'm not sure. I've created a new branch with the new maps that can be downloaded here. Keep in mind that you have to remake your path, and if you're using the PathMaker, use the new RS Surface map in SPS/img/runescape_surface/.

Wonderful, I'll give those new maps a whirl. Also, thanks for the reminder about the loading the new img into the path-maker.

[marpis]

Now that SPS works quite fast, we could try using bigger areas, that'd be more user friendly? instead of 500x500, we could try 1000x1000?

I take this back!

I have an idea how to recognize which are the most likely areas you're on! I'm going to write some code when I have time.
On one area there are many many minimap views, depending on where on that area you are. Each minimap view has different RGB ratios depending on your environment. In forests there is lots of green but little blue and red. In swamps there's also lots of green but there might be medium amount of blue and low red.
We have 27 profiles for every kind of minimap view:
(LowR, LowG, LowB)
(MedR, LowG, LowB)
(HigR, LowG, LowB)
(LowR, MedG, LowB)
(LowR, HigG, LowB)
and so on and so on.

Now I will make a script that will analyze every area and tell me which kinds of minimap view profiles it contains. One area may have many profiles, so it will appear as a suggestion in many places. When we have profiled our current minimap view, we only use areas from a premade TStringArray that only has areas which contain views like ours.

The problem that will probably rise here, is that there could be way too many areas in every array. I know that this will happen when in woods, but there's just not much you can do about it.

To prevent this, we could also profile the areas the same way using HSL.
That will give us 729 different profiles, which will most likely mean less areas in each array. To do this, we need to modify the plugin a bit.

The goal is to have 1-3 areas in each array, excluding forests.

EDIT: Ok so I made this script. For area '8_10' I got 8 profiles. Now I'm going to try dividing RGB ratios to 5 categories instead of 3 and see how it goes.

[Nava2]

Coh3n and I were talking about doing HSL as well.. Its not that complicated and we could do it.

Just store the information in a text file alongside the full pictures. We can "profile" every map and then just build records instead of the just the TIntegerArray. We could then cross reference the values of the current area on the minimap (..ugh speed?) and the stored values. Make sense?

Btw, I'm running your profiler from the other thread.. this needs to be plugin-ized. BAD.

[marpis]

Coh3n and I were talking about doing HSL as well.. Its not that complicated and we could do it.

Just store the information in a text file alongside the full pictures. We can "profile" every map and then just build records instead of the just the TIntegerArray. We could then cross reference the values of the current area on the minimap (..ugh speed?) and the stored values. Make sense?

Btw, I'm running your profiler from the other thread.. this needs to be plugin-ized. BAD.

Well, yeah I just don't even have lazarus on my laptop now.
And my idea is to have an Array [0..444] of TStringArray because there are 125 different profiles, each profile is presented as a number, ranging from 0 to 444. The formula can be found in the profiler script. Then it's simple as this:

var
  CurrentViewProfile: Integer;
  MostProbableAreas: TStringArray;
  ProfiledAreas := Array [0..444] of TStringArray;


begin
  CurrentViewProfile := SPS_ProfileMinimap();
  MostProbableAreas := ProfiledAreas[CurrentViewProfile];
end;

Then just run the regural FindMapInMapExx using MostProbableAreas.

[Nava2]

But can we profile the minimap to be similar to the full chunk? It seems that the mark might be off dependant on where you are in the chunk.. Thus the part of the map might be better off being smaller.. We would have to implement some form of tolerance or a range of values since the average when you are on the docks of karamja it will be very different than if you are in the plantation.. Just something to think about.

I hope that made sense..

[Yago]

Can we do something where you dont have to be facing north?

[Nava2]

Can we do something where you dont have to be facing north?

You could do that with rotating the bitmap before giving it to sps -- it will be slower.

I believe someone else was looking into this.. It's bit necessarily hard, but just a bit intensive.

[marpis]

But can we profile the minimap to be similar to the full chunk? It seems that the mark might be off dependant on where you are in the chunk.. Thus the part of the map might be better off being smaller.. We would have to implement some form of tolerance or a range of values since the average when you are on the docks of karamja it will be very different than if you are in the plantation.. Just something to think about.

I hope that made sense..

My technique takes this into consideration, as it finds multiple profiles for each area.
It goes through every possible "minimap view" in that area and makes a profile, adds it to an array. In the end all double profiles are cleaned from the profile array.

[Nava2]

My technique takes this into consideration, as it finds multiple profiles for each area.
It goes through every possible "minimap view" in that area and makes a profile, adds it to an array. In the end all double profiles are cleaned from the profile array.

There are (total_width-minimapwidth)*(total_height-minimapheight) possibilities, per SPS map. Is that what you took into account?

Remember, the minimap could shift by just one pixel.. :/

[marpis]

There are (total_width-minimapwidth)*(total_height-minimapheight) possibilities, per SPS map. Is that what you took into account?

Remember, the minimap could shift by just one pixel.. :/

Try to read the code in the other post, see if you can understand it.
When I make a profile for our current minimap, I count all the pixels together, to get a total sum of the red, green and blue counts. Then I rate how much there is of each color, with a scale from 0 to 4. When I rate all 3 colors in such manner, we get 125 different combinations of these profiles. For example this is one profile:

Red count very low: 0
Green count medium: 2
Blue count: rather high: 3

My script goes through every possible minimap view in every area, calculating what all profiles there are. When we have 1 array of array of string for each profile (totally 125 of them) and we know which profile our current minimap view represents, we can easily load all zones that have that matching profile. This means that one area can be in multiple arrays.

The lenght of the array would be 444, because as profiles are presented as single integers, 444 is the highest integer. This means that there are 444 - 125 = 319 empty subarrays, but we don't care about that.

When we analyze our minimap and get the previous example profile 023 = 23, we have all the areas matching our current profile listed in SPS_ProfiledAreas[23], the 2D string array I just talked about.
The content of SPS_ProfiledAreas[23] could be, for example ['1_12', '4_5', '7_2']

[DaWu]

Lunar island got messed recently, needs a new map.

[Kasi]

Lunar island got messed recently, needs a new map.

one way to put it =D how so did it get messed if you dont mind me asking ;D

[DaWu]

one way to put it =D how so did it get messed if you dont mind me asking ;D

First of all, sps walks mega slow on the whole island. Second, tiles are off near the altar as they changed the altar peninsula and the whole island slightly.

Compare current worldmap to one you got in your SPS image folder and notice it.

[Coh3n]

How recent? I've had new maps made for a while I just haven't had enough time to properly test. Plus the fonts thing set me back a day. Also, the "slow walking" you're talking about will be fixed.

[Kasi]

First of all, sps walks mega slow on the whole island. Second, tiles are off near the altar as they changed the altar peninsula and the whole island slightly.

Compare current worldmap to one you got in your SPS image folder and notice it.

lol no i understand i was just questioning the way you said it in the first place lol

[DaWu]

How recent? I've had new maps made for a while I just haven't had enough time to properly test. Plus the fonts thing set me back a day. Also, the "slow walking" you're talking about will be fixed.

Well it worked absolutely perfect before the graphics update.

So very recent.