Cheat Engine

thenewguy

Trying to find an easier way to location the location of a set of known pixels. Pixel range is about 5x15 and can be easily converted to a bitmap to use as a compare. Looking for a way to easily find the pixels and return the location of.

2 methods I've used to do this:
1) Take the window and cut off some of the top/the bottom and the sides to reduce the area I have to search in. Then starting in the top left of what's left go 1 pixel to the right and test. Repeat till you reach the right side then go down 1 pixel and start again from the left. When I find 3 pixels of the same color in a row I then send that location. I've later changed it to search by 4(1, 5, 9, 13,,,,) and if I find 1 pixel that matches I then check the right/left of it. If either the right/left are the same as well I check right/left once more to be sure then send that location. Very involved and considering I'm only trying to do this to make a bot more simple, this is far from simple.

2)Kind of the same method as described above except using recursion instead. It's really no different except it does seems to find a valid target quicker since it's covering more of the top left of the screen and spanning out where the other one just scans all of the top then goes down to repeat.

Any ideas, thoughts or pointers would be greatly appreciated.

TerryDonahugh · Posted: Mon Oct 08, 2007 8:38 am Post subject:

What is this for?

In case you can do exact matching this problem is not that hard. If your template and source images are small enough a naive exhaustive search should be fast enough. There are a lot of tricks to speed this up though.

Here is an idea I whipped up:

I <- 0
J <- 0

Take a N pixel horizontal segment from your source image starting at position I,J.

Compare this segment against all rows in your template image. If row K of your template matches, you have a high probability that the whole template matches. Set J <- J -K and compare the rest of the scanlines to validate this.

If there is no match set I <- I+1.

If you reached the end of the row set J <- J + M

thenewguy · Posted: Mon Oct 08, 2007 10:43 am Post subject:

well, the pixel constant that I'm comparing is 7x25,, lets say. On the screen there maybe be many, there may be none. Of the 7x25,, all the pixels aren't the same so taking a horizontal segment sounds like a good idea but if there is the 7x25 near the segment and it gets cut in half, that would then defeat the purpose since it would never match. of the 7x25 there was 1 pixel row that was the exact same color, at least in the middle like 1x7. I was using that as my basis to search but still had to go pixel by pixel, skipping some to increase speed. sounds like there really isn't an easy way to do this in C++ then is there? I was hoping to find something like pixelsearch for autoit.

TerryDonahugh · Posted: Mon Oct 08, 2007 12:19 pm Post subject:

DeltaFlyer · Posted: Mon Oct 08, 2007 1:21 pm Post subject:

Is your problem basically searching for a pattern match in a 2D array? If so, 4 nested loops can do the job:

thenewguy · Posted: Mon Oct 08, 2007 2:41 pm Post subject:

atom0s · Posted: Mon Oct 08, 2007 2:43 pm Post subject:

AutoIT does Pixel detection by pixel CRC's. You could look at the source of AutoIT (parts are open source, not sure if the pixel stuff is or not) and see if that would be of any help.

http://www.autoitscript.com/autoit3/files/archive/autoit/