|
i wish this was my nine to five, its actually a side home project, in my normal life i build ERP business software for a logistics company.... dead boring...
|
|
|
|
|
|
I have worked on several tracking algorithm. In many algorithm we need to give the boundary of the object to track. So if its not possible in your case, first you need to do some motion segmentation, then you need to track.
Tracking is definitely complicated, so if you have time i would suggest you to look into active contours and SIFT feature tracking , otherwise try to implement mean shift algorithm, the problem with mean shift is it is not scale invariant. You also need to predict the camera object movement ( using kallman filter )
Hopes this will give you a very high level introduction. 
If u can Dream... U can do it
|
|
|
|
|
Thanks JK, the Kallman Filter approach seems to be the best option, you have definatly given me some good direction, thanks again.
|
|
|
|
|
Tell me If you got it working.am just curious.
If u can Dream... U can do it
|
|
|
|
|
|
You saw what happened yesterday when you didn't follow the guidelines. Why do you think it will be any different today? 
Albert Einstein once said "The definition of insanity is doing the same thing over and over again and expecting different results". refer here
Software rusts. Simon Stephenson, ca 1994.
|
|
|
|
|
forum, u identify the installshield application send me the name i don't know?
|
|
|
|
|
Wow, you really are stupid. I just saw the email from you saying sorry for multiple posts and now you immediately post your question for a seventh time. That should guarantee that nobody will respond to you.
I must get a clever new signature for 2011.
|
|
|
|
|
You should post your question at the InstallShield forums[^] in the forum specific to the product/version you are using.
Jack of all trades ~ Master of none.
|
|
|
|
|
I agree!
"Program testing can be used to show the presence of bugs, but never to show their absence."
<< please vote!! >>
|
|
|
|
|
SOLVED
For anyone interested here is the solution to my problem:
public IEnumerable<Tile> PathFind(Tile startNode, Tile endNode)
{
List<Tile> openNodes = new List<Tile>() { startNode };
List<Tile> closedNodes = new List<Tile>();
Tile lowestFScore = null;
while (openNodes.Count > 0 && lowestFScore != endNode)
{
lowestFScore = GetLowestScoringTile(openNodes, endNode);
openNodes.Remove(lowestFScore);
if (lowestFScore == null)
return null;
SwitchFromOpenToClosedList(lowestFScore, openNodes, closedNodes);
IEnumerable<Tile> neighbouringTile = GetNeighbouringTiles(lowestFScore);
foreach (Tile n in neighbouringTile)
{
if (IgnoreTile(n))
{
SwitchFromOpenToClosedList(n, openNodes, closedNodes);
continue;
}
if (closedNodes.Contains(n)) continue;
if (!openNodes.Contains(n))
{
openNodes.Add(n);
n.Parent = lowestFScore;
}
}
}
return RenderToGrid(endNode);
}
I have done my research and attempted several times to create my version of the infamous A* path-finding algorithm.
The algorithm I have made does seem to work though!...(unsurprisingly)
If anyone can help me it would be great, I am sure it is just a logic problem in the main part of the method:
public void PathFind(Tile startNode, Tile endNode)
{
List<Tile> openNodes = new List<Tile>() { startNode };
List<Tile> closedNodes = new List<Tile>();
Tile currentNode = startNode;
while (true)
{
foreach (Tile n in GetNeighbouringTiles(currentNode))
{
if (closedNodes.Contains(n)) continue;
if (IgnoreTile(n))
{
if (!closedNodes.Contains(n))
{
closedNodes.Add(n);
continue;
}
}
n.Parent = currentNode;
if (!openNodes.Contains(n))
openNodes.Add(n);
}
if (openNodes.Contains(currentNode))
openNodes.Remove(currentNode);
Tile lowestScoring = GetLowestScoringTile(openNodes, endNode);
if (lowestScoring == null) lowestScoring = currentNode;
if (openNodes.Contains(lowestScoring))
{
openNodes.Remove(lowestScoring);
}
foreach (Tile n in GetNeighbouringTiles(lowestScoring))
{
if (closedNodes.Contains(n)) continue;
if (IgnoreTile(n))
{
if (!closedNodes.Contains(n))
{
closedNodes.Add(n);
continue;
}
}
n.Parent = lowestScoring;
if (!openNodes.Contains(n))
openNodes.Add(n);
else
{
if (Tile.GetCost(lowestScoring, endNode) > Tile.GetCost(n, endNode))
{
n.Parent = lowestScoring;
lowestScoring = n;
}
}
}
currentNode = lowestScoring;
if (openNodes.Count == 0 || currentNode == endNode)
return;
}
}
I am expecting the method to return when there are no more nodes remaining (which tells me that there is no way of getting from a to b), or the current node is the end node (which should mean that I can just re-curse through each "Parent" attribute of the endNode to cycle through the path.
Edit1: I noticed that the first " n.Parent = currentNode;" is never reached!
Edit2: The "GetNeighbouringTiles" method was incorrect, it didn't search for moving down... The pathfinding method hits an infinite loop now...
It only ever does the former result 
modified on Tuesday, January 4, 2011 1:40 PM
|
|
|
|
|
|
Thank Espen,
I was trying to do it based on the concept, I didn't want to use other code and instead build from the concept 
|
|
|
|
|
|
.. A few comments on this, if your using XNA and want to use this code on the XBOX, you should get rid of the Foreach loop, it is a performance killer due to GC on the XBOX....
Also you should look into the weighting Heuristics, because this approach albeit closed and complete is not as efficient as it could be....
Also your sucking up memory as you use this algo... have you tested it against a large number of bots representative to the normal in game load...., I'd wadger it won't run inside your game loop propperly for more then 125 or so bots (using a 256x256 tile map).... just trying to give you the heads up...
I'd blame it on the Brain farts.. But let's be honest, it really is more like a Methane factory between my ears some days then it is anything else...
-----
"The conversations he was having with himself were becoming ominous."-.. On the radio...
|
|
|
|
|
I used to be pretty good at this stuff, but I just can't get a handle on how to express this properly. Thirty years takes a toll on remembered math, so I hope someone more recently experienced with probability can steer me back on course.
In the game of Keno there is a field of 80 numbers, from which 20 are selected for each round. Players pick from n = 3 to 20 numbers and hope that n of them will come up in the next round. In reality, far fewer than n usually pop up, and I'd like to know the probability of each combination occurring. For instance, suppose I pick 6 numbers. How can I calculate the probability that 3 of my numbers will be among the 20 drawn? 5?
The simple approach of calculating draws doesn't work, since the classic method I remember determines the probability of all 3 numbers coming up in 3 draws. But given that 20 draws are made regardless of how many I pick changes things considerably. I just can't figure out how to express it.
Anyone care to take a crack at it?
Will Rogers never met me.
|
|
|
|
|
DISCLAIMER: This could be totally wrong...
Ok, so I start by saying there are 80 squares - 20 of which are picked each round. Intuition says now:
20/80 = 1/4 = 25%
A colleague said "It's not that simple, because not ALL the 80 are picked each round." I justified my answer by saying replace this with a coin, there are two possible outcomes. If you flip it once, not ALL the outcomes are going to be achieved, but it's still a 50% chance (statistically speaking, none of this "the Queen's head is heavier BS").
Anyway, back on track, this would mean that for each individual number on the board, there is a 25% chance it will be picked in a given round. Therefore, for the three numbers YOU selected to ALL come up:
(if you say 'and' it means multiply, 'or' is addition (off the odds))
for outcomes: N1 and N2 and N3
prob = prob(N1) x prob(N2) x prob(N3)
= (1/4) ^ 3
= 1/64
So, to generalise this, as no two numbers have different probabilities:
for outcomes: N1, N2 ... Nz
prob = prob(N1) x prob(N2) x ... x prob(Nz)
= (1/4) ^ z
= 1/ (4 ^ z)
So, for any selection of Keno numbers of length z, the probability of getting all of these is:
1 / (4 ^ z)
For more complex things like "if I pick 10 numbers, what's the chance of getting 8 of them" look up Permutations and Combinations - I cannot for the life of me remember the details, I just know one is concerned with the order of the outcomes and the other is not.
Anyway, statisticians, am I even close???
|
|
|
|
|
Well, you're close, but not all numbers have the same probability of coming up.
The first time you draw a number there is a 1/80 chance of it being a particular number. But the second time, the chance is 1/79, since the first number is not replaced.
I looked at it as the red ball/ black ball problem, where the number I picked are red balls. If I pick 10 numbers, then there are 10 red balls and 70 black balls. The probability of one of my numbers coming up ( order isn't important) on the first draw is then 10/80. The probability of of the second ball also being one of my numbers, given that the first was a good one, is 9/79, since I only have 9 numbers remaining in the bag, and a total of 79 balls altogether. So the total probability of the first two balls being ones I picked becomes (10/80)*(9/79). That's easy enough to calculate. But how do I calculate the probability of perhaps, the 2nd, 9th, 11th, 12th, and 17th ball being good ones? Even if I make this tedious calculation for all the possible draw/hit combinations and orders, how do I determine the total probability of any particular number of hits in all the different orders in which they may occur? That's where I get stumped!
I think this may need to enter the domain of set theory, and calculating the probability that n of my numbers will be in a set of 20 numbers drawn from a field of 80.
My reason is quite pragmatic. There are different payouts for each combination of numbers picked and number hit in a round. And each game has different payouts. And to make it more interesting, each casino nmay have different payouts for the same games. If I can figure the odds on each combination, then multiply each by the payout for each machine and casino, I can learn which combination is optimal for each game and casino. I know that none of them is going to be a paying proposition, but I see no reason not to optimize my chances on a game I'm going to play anyway.
Will Rogers never met me.
|
|
|
|
|
Roger Wright wrote: The first time you draw a number there is a 1/80 chance of it being a particular number. But the second time, the chance is 1/79, since the first number is not replaced.
Damn it! I knew I'd forgotten something, oh well... never mind.
Roger Wright wrote: Even if I make this tedious calculation for all the possible draw/hit combinations and orders, how do I determine the total probability of any particular number of hits in all the different orders in which they may occur? That's where I get stumped!
That smells like Permutations and Combinations - in fact, I'd be surprised if this was not exactly what you were looking for: Permuations[^] and Combinations[^].
I believe you want Combinations, as (forgive me if I'm wrong) I don't remember the order of numbers mattering in Keno.
|
|
|
|
|
Excellent links, and I will look at them closely. But in the meantime, take a look at Luc's answer. It's quite a nice approach to the problem!
Will Rogers never met me.
|
|
|
|
|
Hi Roger,
the easiest way to tackle this is by reversing the chronology:
assume there are 80 numbers (1 to 80). Now let them first pick 20 "good" numbers out of those 80; it could be they pick numbers 1 to 20, and leave 60 "bad" numbers numbered 21 to 80; or any other combination, it does not really matter.
Then you pick N arbitrary numbers out of the range 1-80, each of them obviously has probability 20/80 or 1/4 of being good and 60/80 or 3/4 of being bad.
What you pick is represented by the expression (g+b)^N with g=1/4, b=3/4, and N the number of numbers you pick.
The expression equals 1 because g+b=1; however expanding the power series (using binomial coefficients), you get separate terms each representing a specific number G of good picks and B bad picks.
The probability that corresponds with a specific value of G and B equals COMB(G,N)*(g^G)*(b^B)
where COMB(G,N)=N!/G!/(N-G)!
Note: COMB is symmetric, i.e. COMB(G,N) = COMB(B,N)
As an example:
we continue with g=1/4, b=3/4
assume you pick 2, hence N=2
three possible outcomes:
2 good, 0 bad; hence G=2, B=0; probability = COMB(2,2)*(0.25^2)*(0.75^0) = 1 * 1/16 * 1 = 1/16
1 good, 1 bad: G=1, B=1; prob= COMB(1,2)*(0.25^1)*(0.75^1) = 2 * 1/4 * 3/4 = 6/16
0 good, 2 bad: G=0, B=2; prob= COMB(0,2)*(0.25^0)*(0.75^2) = 1 * 1 * 9/16 = 9/16
check: the sum of all three is 16/16 or 1.
PS: by looking at it in this way, there is no order, hence no 80,79,78... concerns!
|
|
|
|
|
The elegance of your solution is exquisite, but my recall of theory is too inadequate to judge the validity. I will take it as a working hypothesis and test it with real money. If all works out, there will be something nice in your stocking next Christmas.
Thanks, Luc - yet again!
Will Rogers never met me.
|
|
|
|
|
Luc's answer is the combination formula I linked you to
Just between you, me and the gatepost rest of CP, you're not actually going to use this with real money are you? 
|
|
|
|
|
GlobX wrote: you're not actually going to use this with real money are you?
Sure! I play a little Keno in any case, so why not approach it scientifically?
Will Rogers never met me.
|
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
