nomenator

Well, I used "brown" just as an example, because it seems to me iron ore is going to be the hardest to automine. My idea on using this is as follows: 1. Collect some images of objects of interest from various angles and in various lighting conditions/screen setups. 2. "Codify" the images. Doesn't need to be colour at all, I know the first object-recognition algorithms were black-and-white. 3. Use a library of codified images to determine objects on the display. For example we start off with focusing on the minimap area on the display. Say for now we have our view turned exactly North (again, we can use recognition for that to identify if the letter "N" is in its proper position, if not, continue turning the screen). Then we compare the image on the minimap to the whole map, and actually find the spot which we are in. Then chose the best way to get to the nearest goal point (for example the ladder from the mining guild to Falador). Or maybe use it to see if our pickaxe head has flown away for starters. I imagine it is possible to divide the game screen into areas, such as "chatbox", "inventory", "minimap", and whatever else. With a good way to codify images, a good database for their identification, and a simple script, it could actually be a rather interesting script, far beyond the limitations of mining ores and dropping them. I am not sure if it would be possible to apply matrix transormations to this idea, I always had a rather cold relationship with my algebra.

THis is so amazing, as I was just thinking about this like 5 seconds before I decided to register on the forum and ask about it. =) This thread is going into my favourite links list. =)

Is it possible to "dump" the current picture on the display as a matrix into some array for example? This way it can be used to actually analyze the picture and see, for example, where the player is on the map. I would imagine that any image smaller than the display could be imported as another matrix and compared to the bigger matrix. For example suppose we have a small matrix like this: 1 5 2 6 3 4 4 9 1 2 5 2 5 1 8 7 and smaller matrix like this: 4 4 2 5 Then we can look into the big matrix and start the comparison: 1. Switch through all the values in matrix_1 2. When any value occures to be equal to initial value of matrix_2, find the coordinates of the value, in our case it will be matrix_1[1,1] 3. Get properties of matrix_2, in our case it is [2,2] 4. Compare values in matrix_1[1+i1,1+i2] to values in matrix_2[i1,i2] where i1 is between 0 and the vertical size of matrix_2, and i2 is between 0 and horizontal size of matrix_2. 5. When all the values match, return the key coordinates in matrix_1, in our case they are [1,1] [2,1] [1,2] [2,2], when no matching sub-matrix is found, do something else, for example return false at which point the function which receives this return would know from false that it needs to look in another area of the world. 6. Finally, when the key coordinates are found, calculate the central point inbetween them, move the cursor over it, click, and wait for the result. This would also open up space for more flexible matrix analysis, for example we are comparing to a matrix with values that we don't care for, or with values that might fall into some rage: (0<=i<=4) * * 5 where "*" is a value we don't care for, and (0<=i<=4) is a value that falls between 0 and 4 inclusively. So instead of looking for a pixel that looks brown (posing problems when we're mining iron ore), we are going to be looking for a shape which is filled with a bunch of shades of brown with some variety between them. Then we can see if we are somewhere specific in the world, find the nearest bank, walk to it, deposit all we need to, and return to mining. Or even more complicated, run multiple bots on different machines, which would spend 8 hours every night mining, one collecting ores off the other and depositing them into bank. I have also collected some packets, but was unable to crack the cyphers, if anyone has experience with Java network programming, and decoding network packets, maybe we could try to break the runescape code together and see if there is room for a more sophisticated botting engine? EDIT: Actually I just found a topic on this forum which addresses the issue of looking up images on the screen: hehe, seems like someone's making a runescape traveller bot =)