mjrb4

2009/10/19

Zerg

2009/10/21

Does this work for other 3D solids besides spheres?
Builderboy2005

2009/10/21

It could be done. The egning would have to be changed a bit, but it wouldn't be that hard. :) Unfortunately, GreenfootImage is a bit slow, and i need to figure out how to write to the awt image directly XD
Zerg

2009/10/21

Speaking from experience, you might find this a bit harder than it first appears. I believe that your raytracing algorithm works perfectly for any solid defined by a mathematical function, but I think you could have problems when you define your solid through polygons instead of a function, as you are doing here. Your code is a bit tricky to read (consider intrinsic variable naming?), but I must say I am impressed, particularly with the spherical texture mapping! Do you intend to design lighting shaders as well for your engine?
mjrb4

2009/10/21

"Unfortunately, GreenfootImage is a bit slow, and i need to figure out how to write to the awt image directly XD" GreenfootImage.getAwtImage() ..?
Builderboy2005

2009/10/21

@Zerg: A polygon could just be an equation for a plane with some bounds on it. For example, if you wanted to map a triangle, you would first find the plan that it lies on, and find the point at which the ray collides with the plane. If the point is inside the triangle, go to linear interpolation for the mapping. And yeah, my code is a bit messy and hard to read right now, i hope to clean that up in the future. And normal shading might not be out of the question, after all it only takes a bit of multiplication. @Mjrb I am aware of that, i am just experimenting with Graphics and Graphics2D to see how I can maximize speed.
Zerg

2009/10/22

Using triangles results in slow-down issues--not only do you have to find where it intersects, you then have to make sure that it is contained within the triangle. Then you have all the fun of iterating through all the triangles. The biggest issue I think is that Greenfoot/Java was not intended to be able to solve multiple simultaneous equations quickly while iterating. I believe I have read though on some OpenGL for Java... not sure where you would find it, but that would be the way to go for 3D anything.
Builderboy2005

2009/10/22

The slowdowns you talk about are always pressent, even in this scenario. Finding whether or not a point is inside a polygon is trivial using normals and dot products. The multiplication is Trivial compared to the trig needed for rotation, and the slowdowns that come from reading and writing from greenfootimages. I think I may have discovered a way to make at least the graphics faster though. Converting an image to 3 arrays of color bytes could speed up reading from images, and writing directly to the writable raster of an image drasticaly decreases writing times, and gives more control over color manipulation.
Zerg

2009/10/22

I would disagree. Finding whether or not a point resides within a triangle requires solving a simultaneous linear equation set in 3 variables--not trivial for a computer!
Builderboy2005

2009/10/22

There is more than one way to find whether or not a point lies inside a triangle. The way you said works, as does computing the normals of the sides, then finding the dot product of the point against the normal. The dot product returns a positive number if the point is inside, and a negative if outside. Calculating the normal is trivial given the coordinates of the vertexes of the triangle, and the dot product is merely multiplication and addition. Do this for all 3 sides and your good :/ I do this in my rigid body physics simulator to detect collisions, and let me assure you, it is quite fast.
Zerg

2009/10/23

There is also the linear equation set for determining where the point intersects the plane. This would normally be no problem, but you have to do this for each polygon in the scene to determine which on is "on top" relative to the camera (unless you want to disregard all shading). Like I said, I am speaking from experience; my rasterizer posted previously used this method(minus Z-buffer), and was indeed quite slow. Just want to point out as well that your rigid body simulator is not running 40000(200px by 200px image)+ iterations per act cycle per object, and is also in 2D, not 3D. ... Of course, there really is no better method than yours. :P
A new version of this scenario was uploaded on Mon Oct 26 04:06:38 UTC 2009
Zerg

2009/10/27

mmm... I lied to you... There is a way to do intersection with polygons with only a single linear equation--still not trivial, but also faster. It requires a form of vector parametrization (which may not be trivial) along with the equation for the plane.
Builderboy2005

2009/10/27

Yes I know. http://local.wasp.uwa.edu.au/~pbourke/geometry/planeline/ Simpler than you might think, and requires stunningly less math than a sphere, and even less to transform the image coordinates.
A new version of this scenario was uploaded on Tue Oct 27 22:39:00 UTC 2009
bourne

2009/10/28

it seems to break when worsening the quality with the plane too far (like 100). not sure about with the earth, it just kept going.
Builderboy2005

2009/10/28

I suppose I could put a cap on the quality settings :P I hadn't even thought of lowering the quality that much! XD
talmantel

2011/1/14

you have bag when do zoom and bad quality! the game stopped!! in quality 9.00+.... and the best zoom(most big) its a bag
MathManiac

2011/5/1

I can see the Rocky Mountians!
Cocky

2012/8/26

schön schön