PDA

View Full Version : Critique my normal calculations, they're backwards but work!



dirk103
11-01-2012, 12:12 AM
Hi, I just recently was able to render coherent terrains after quite a long time of trying. I'm not sure why this works but it does, I simply switched the equation around:

instead of: < what every example i've seen says to do
normal = (v1 - v2) * (v2 - v3)



i tried
normal = (v1 * v2) - (v2 * v3)

and it looks good

why is this so? I don't understand this and it baffles me. did the forces of OpenGL come together and i have a working bug, i've been experimenting with light positioning, and it doesn't appear as if the position of my light matches what I feed it. for example. If i set my light position to:
x = 2571.15
y = 6000
z = 3064.18
??? = 45

it appears as if the center of the light source is closer to
x = z = 60

at least that is where the center of my spot_light_cutoff point is in relation to the terrain that i am drawing.

- I am calling glLightfv after a switch to the modelview matrix, and after i translate to my camera position.


__________________________________________________
Another question i have is:


GLfloat poslight[4] = {x, y, z, ??? };
glLightfv(GL_LIGHT0, GL_POSITION, poslight );


what is the fourth parameter for?
__________________________________________________




* Note, this normal list jumps by four because i'm rendering quads


void ModelObject::CalculateNormals()
{
for(int i=0;i<normal_count;i+=12) {
float v1_x = vertexes[i+0];
float v1_y = vertexes[i+1];
float v1_z = vertexes[i+2];


float v2_x = vertexes[i+3];
float v2_y = vertexes[i+4];
float v2_z = vertexes[i+5];


float v3_x = vertexes[i+6];
float v3_y = vertexes[i+7];
float v3_z = vertexes[i+8];



normals[i+0] = (v1_x * v2_x) - (v2_x * v3_x);
normals[i+1] = (v1_y * v2_y) - (v2_y * v3_y);
normals[i+2] = (v1_z * v2_z) - (v2_z * v3_z);


normals[i+3] = (v1_x * v2_x) - (v2_x * v3_x);
normals[i+4] = (v1_y * v2_y) - (v2_y * v3_y);
normals[i+5] = (v1_z * v2_z) - (v2_z * v3_z);


normals[i+6] = (v1_x * v2_x) - (v2_x * v3_x);
normals[i+7] = (v1_y * v2_y) - (v2_y * v3_y);
normals[i+8] = (v1_z * v2_z) - (v2_z * v3_z);


normals[i+9] = (v1_x * v2_x) - (v2_x * v3_x);
normals[i+10] = (v1_y * v2_y) - (v2_y * v3_y);
normals[i+11] = (v1_z * v2_z) - (v2_z * v3_z);

}


// I believe i should normalize to unit length here

// Smooth the normals (for heightmap only)
for(int i=0;i<normal_count;i+=12) {
float n1_x = normals[i+0];
float n1_y = normals[i+1];
float n1_z = normals[i+2];


float n2_x = normals[i+3];
float n2_y = normals[i+4];
float n2_z = normals[i+5];


float n3_x = normals[i+6];
float n3_y = normals[i+7];
float n3_z = normals[i+8];


float n4_x = normals[i+9];
float n4_y = normals[i+10];
float n4_z = normals[i+11];

float n1 = n1_x + n2_x + n3_x + n4_x;
n1 /= 4;
normals[i+0] = normals[i+3] = normals[i+6] = normals[i+9] = n1;

float n2 = n1_y + n2_y + n3_y + n4_y;
n2 /= 4;
normals[i+1] = normals[i+4] = normals[i+7] = normals[i+10] = n2;

float n3 = n1_z + n2_z + n3_z + n4_z;
n3 /= 4;
normals[i+2] = normals[i+5] = normals[i+8] = normals[i+11] = n3;



}
}


thank you!

Aliii
11-01-2012, 03:01 AM
//************************************************** ******************
bool CVector3f::crossP(CVector3f* v1, CVector3f* v2){

coord[0]= v1->coord[1]*v2->coord[2] - v1->coord[2]*v2->coord[1];
coord[1]= v1->coord[2]*v2->coord[0] - v1->coord[0]*v2->coord[2];
coord[2]= v1->coord[0]*v2->coord[1] - v1->coord[1]*v2->coord[0];

if(coord[0] || coord[1] || coord[2]) return true;
else return false; //if they are paralell.
}
//************************************************** ******************

^^^^this is the function Im using

normal = (v1 - v2) * (v2 - v3) = v2*(v1 - v2) + v3*(v2 - v1) //I never seen this one
normal = (v1 * v2) - (v2 * v3)

You need vectors if you want to calculate cross product not vertexes. Except if the "vertex" is a vector that starts from the origo.

If A,B,C,D are the points of a quad, then you should do:
AB= B - A
AC= C - A

normal = AB x AC

If you dont need it to be 1.0 long then dont normalize it.

This seems like some magic:)

float n1 = n1_x + n2_x + n3_x + n4_x;
n1 /= 4;
normals[i+0] = normals[i+3] = normals[i+6] = normals[i+9] = n1;

Try this:

//************************************************** ******************
float CVector3f::normalize(){


float temp;

if( !(temp= sqrt(coord[0]*coord[0] + coord[1]*coord[1] + coord[2]*coord[2])) ) return 0.0;


coord[0]/= temp;
coord[1]/= temp;
coord[2]/= temp;

return temp;
}
//************************************************** ******************

Dark Photon
11-01-2012, 05:15 AM
Please use
... tags to mark your source code (or ...). Inserted them for you.

Dan Bartlett
11-01-2012, 06:17 AM
In this equation:

normal = (v1 - v2) * (v2 - v3)
The "-" is a vector subtraction operation, and the "*" is a cross product (http://en.wikipedia.org/wiki/Cross_product), it doesn't mean you multiply each term (x,y,z) in that way.

Check out this article (http://en.wikipedia.org/wiki/Euclidean_vector#Basic_properties) for more info about vectors.

If you write a vector class and overload certain operators, eg. "+", "-", "+=", "-=", "*", you could use the original equation directly.

dirk103
11-07-2012, 04:18 PM
Ah, very good sir. Thank you!