Has anyone ever had to figure out which side of a 2D sprite a collision occurred on?

Im using pixel-perfect collision detection and I do need to determine it. My best idea at the moment is during the pixel collision routine, upon a successfully collision, save the x,y value of the pixel in the sprite to a variable, then caclulate where that pixel is on the sprite and go from there.

not sure if it will work, and also not sure if theres a simpler way?

(bearing in mind its not a sidescrolling platformer, it's a top down with all manner of rotations and possible 'approach vectors' from other sprites)

Thanks!

_________________
"None are more hopelessly enslaved than those who falsely believe they are free."
"It is no measure of health to be well adjusted to a profoundly sick society."
"Hope is the first step on the road to dissapointment." -Jonah Orion
http://tankzgame.blogspot.com

Not sure of your requirement, but I would look at the relationship between the centres of your objects (or whatever 'handle' you are using to position the sprites).

_________________
10 PRINT "Bad Monkey ";
20 GOTO 10

This depends on the step size of course. A big step could cause the centres to cross, and yield incorrect results.

My preferred method is to look at the relative velocity between the two objects. That vector can be used to determine which side, or it can be used to directly to calculate what happens next.

_________________
I ain't pushing no moon buttons.

I could certainly calculate the vector between the 2 sprites at the time of collision, but the problem is that either of them could be rotated at any angle during the collision. I've thought about a method something like this:

1) Save the position of each sprite at the time of impact
2) Use that to calculate the relative vectors of the 2 objects
3) Figure based on the rotation of the one sprite, and the vector of the other, calculate side of impact

Kindof like, "If Im here, and he's there, and im facing this way, it could only have collided with surface X"..
I got thinking about that kindof logic last night, and im not sure how accurate it would be. I'l give it a try, and see how it works out.

_________________
"None are more hopelessly enslaved than those who falsely believe they are free."
"It is no measure of health to be well adjusted to a profoundly sick society."
"Hope is the first step on the road to dissapointment." -Jonah Orion
http://tankzgame.blogspot.com

I think if your objects aren't convex you're going to struggle, because all sorts of weirdness might occur.

What are you going to use the collision direction for? If it's to model a reaction force on the two objects I'd say that's probably nigh-impossible given only pixel data.

_________________
Whatever the mind can conceive and believe, it can achieve