gluUnProject stopped working...
Author
I'm sure the problem is with the intersector, since the line always points directly from the camera to the triangle, but it's a completely different triangle that gets selected. Unfortunately I don't understand any of that horribly messy code.
You mean I'M MESSY!?!?!? :P
If you don't understand what its doing, here's the basic idea:
1) First, you need to get the line's direction.
Subtract point B on the line from point A. This give the vector from A pointing
to where B is relative to A (treating A as the origin), I'll call it D.
2) Next, you need to measure the line w.r.t. one of the triangle's veritces.
This is done by picking one (any one) of the vertices to act as the origin.
Now subtract point A from that vertex and store it in a temp (call it
C).
3) Now, you need the triangle's normal. Optimally these are best stored with
each triangle or vertex (depending on your representation) simply to save a lot
of time.
4) To get a normal to a triangle, take the cross product of any of the two
edges. I.E: I would say edge 1 = vert[1]-vert[0] and edge 2 = vert[0]-vert[2].
Now cross edge 1 and edge 2 (in that order) to create a normal, then normalize
it.
5) Next, intersect the line with the plane represented by that normal.
Use the following formula:
t = -Dot(Normal,C)/Dot(Normal,D)
Intersection = t*D + C.
Note that this is still relative to the vertex that was picked in step 2.
6) Next we check to see if the point is actually in the triangle:
We need vectors from each of the triangle's vertices to the point of intersection.
We already have the one from the vertex we chose in step 2 to the intersection,
just need the other two.
Why is this necessary? It is a heuristical intersection test which uses the
cross product of each edge versus each vector form the triangle's vertices to
the point of intersection dotted against the traingle's normal. The fact is,
for ANY n-sided convex polygon, those dot products will all have the same sign
if the point is inside.
The piled if()s in the algorithm I posted prevent unneeded computation. In the
way I was performing the dot products, they would all be positive if the point
was inside. If one happened to be negative, the point is obviously outside the
triangle.
Here's a picture of what I'm saying:
If the dot product of each of the three red vectors versus the triangle's
normal are all the same sign, the point of intersection (blue dot) lies within
the triangle.
Edit: To get the actual point of intersection, add the vertex picked in step 2
to the intersection obtained from step 5.
[Edited by - i-photon on August 5, 2006 9:40:09 PM]
If you don't understand what its doing, here's the basic idea:
1) First, you need to get the line's direction.
Subtract point B on the line from point A. This give the vector from A pointing
to where B is relative to A (treating A as the origin), I'll call it D.
2) Next, you need to measure the line w.r.t. one of the triangle's veritces.
This is done by picking one (any one) of the vertices to act as the origin.
Now subtract point A from that vertex and store it in a temp (call it
C).
3) Now, you need the triangle's normal. Optimally these are best stored with
each triangle or vertex (depending on your representation) simply to save a lot
of time.
4) To get a normal to a triangle, take the cross product of any of the two
edges. I.E: I would say edge 1 = vert[1]-vert[0] and edge 2 = vert[0]-vert[2].
Now cross edge 1 and edge 2 (in that order) to create a normal, then normalize
it.
5) Next, intersect the line with the plane represented by that normal.
Use the following formula:
t = -Dot(Normal,C)/Dot(Normal,D)
Intersection = t*D + C.
Note that this is still relative to the vertex that was picked in step 2.
6) Next we check to see if the point is actually in the triangle:
We need vectors from each of the triangle's vertices to the point of intersection.
We already have the one from the vertex we chose in step 2 to the intersection,
just need the other two.
Why is this necessary? It is a heuristical intersection test which uses the
cross product of each edge versus each vector form the triangle's vertices to
the point of intersection dotted against the traingle's normal. The fact is,
for ANY n-sided convex polygon, those dot products will all have the same sign
if the point is inside.
The piled if()s in the algorithm I posted prevent unneeded computation. In the
way I was performing the dot products, they would all be positive if the point
was inside. If one happened to be negative, the point is obviously outside the
triangle.
Here's a picture of what I'm saying:
If the dot product of each of the three red vectors versus the triangle's
normal are all the same sign, the point of intersection (blue dot) lies within
the triangle.
Edit: To get the actual point of intersection, add the vertex picked in step 2
to the intersection obtained from step 5.
[Edited by - i-photon on August 5, 2006 9:40:09 PM]
Author
OK, so like this?
With that, no matter where I click it thinks I clicked a polygon at (62.625, -6.250, 63.875). Maybe there's something wrong with the vector class itself? Are the operators in the right order and all that?
[edit] I investigated a bit and it seems to actually be interseting many triangles. Something must be horribly, horribly wrong because when I changed < 0.0 to <= 0.0 it suddenly hit them all - somehow the result is less than zero, but neither less than or equal to zero? O_o
[edit again] I realized one flaw in my code is that I was calling this function too many times per shape, comparing it to nonexistant points. Having fixed that, I changed the comparisons to <= 0.0 again. As it is, this results in no triangles being hit, but if I change this:
GLfloat T = -Normal.DotProduct(Measure) / Normal.DotProduct(Line);
to this:
GLfloat T = -(Normal.DotProduct(Measure) / Normal.DotProduct(Line));
it almost works. The wrong triangles still get hit but they're all in a line (as if it's working correctly, but with the line in the wrong place).
[more edits] Well, after screwing around with it for an hour, I've managed to make it...
-Hit every polygon.
-Hit every polygon that has the same Y coordinate for all points.
-Hit every polygon that has the same X/Z coordinates for all points.
-Hit random polygons.
-On rare occasions, hit all the polygons in a straight line, but not the line being tested.
I've tried everything and no longer have any idea what I'm doing. >_<
[Edited by - HyperHacker on August 8, 2006 8:57:40 PM]
bool LineIntersectsTriangle(Vector LineStart, Vector LineEnd, SHAPE3D* Tri, unsigned int Point1, unsigned int Point2, unsigned int Point3){ Vector Line = LineEnd - LineStart; Vector Measure = LineStart - Tri->Point[Point1].Position; //Calculate normal Vector Edge1 = Tri->Point[Point2].Position - Tri->Point[Point1].Position; Vector Edge2 = Tri->Point[Point1].Position - Tri->Point[Point3].Position; Vector Normal = Edge1.CrossProduct(Edge2); Normal.Normalize(); GLfloat T = -Normal.DotProduct(Measure) / Normal.DotProduct(Line); Vector Intersect = (Measure + Line) * T; if((Intersect - Tri->Point[Point1].Position).CrossProduct(Edge1).DotProduct(Normal) < 0.0) return false; if((Intersect - Tri->Point[Point2].Position).CrossProduct(Edge2).DotProduct(Normal) < 0.0) return false; if((Intersect - Tri->Point[Point3].Position).CrossProduct(Tri->Point[Point3].Position - Tri->Point[Point2].Position).DotProduct(Normal) < 0.0) return false; return true;}With that, no matter where I click it thinks I clicked a polygon at (62.625, -6.250, 63.875). Maybe there's something wrong with the vector class itself? Are the operators in the right order and all that?
class Vector { private: public: GLfloat X, Y, Z; //Position Vector() //Constructor { X = 0.0; Y = 0.0; Z = 0.0; } Vector(GLfloat InX, GLfloat InY, GLfloat InZ) //Constructor { X = InX; Y = InY; Z = InZ; } inline bool operator== (const Vector& V2) const { return ((X == V2.X) && (Y == V2.Y) && (Z == V2.Z)); } inline Vector operator+ (const Vector& V2) const { return Vector(X + V2.X, Y + V2.Y, Z + V2.Z); } inline Vector operator- (const Vector& V2) const { return Vector(X - V2.X, Y - V2.Y, Z - V2.Z); } inline Vector operator- () const //Negative { return Vector(-X, -Y, -Z); } inline Vector operator/ (const Vector& V2) const { return Vector(X / V2.X, Y / V2.Y, Z / V2.Z); } inline Vector operator/ (GLfloat S) const { GLfloat F = 1.0 / S; return Vector(X * S, Y * S, Z * S); } inline Vector operator* (const Vector& V2) const { return Vector(X * V2.X, Y * V2.Y, Z * V2.Z); } inline Vector operator* (GLfloat S) const { return Vector(X * S, Y * S, Z * S); } inline Vector CrossProduct(const Vector& V2) { return Vector((Y * V2.Z) - (Z * V2.Y), (Z * V2.X) - (X * V2.Z), (X * V2.Y) - (Y * V2.X)); } inline GLfloat DotProduct(const Vector& V2) { return (X * V2.X) + (Y * V2.Y) + (Z * V2.Z); } inline void Normalize() //Normalizes the vector { GLfloat t = sqrtf((X * X) + (Y * Y) + (Z * Z)); X /= t; Y /= t; Z /= t; }};[edit] I investigated a bit and it seems to actually be interseting many triangles. Something must be horribly, horribly wrong because when I changed < 0.0 to <= 0.0 it suddenly hit them all - somehow the result is less than zero, but neither less than or equal to zero? O_o
[edit again] I realized one flaw in my code is that I was calling this function too many times per shape, comparing it to nonexistant points. Having fixed that, I changed the comparisons to <= 0.0 again. As it is, this results in no triangles being hit, but if I change this:
GLfloat T = -Normal.DotProduct(Measure) / Normal.DotProduct(Line);
to this:
GLfloat T = -(Normal.DotProduct(Measure) / Normal.DotProduct(Line));
it almost works. The wrong triangles still get hit but they're all in a line (as if it's working correctly, but with the line in the wrong place).
[more edits] Well, after screwing around with it for an hour, I've managed to make it...
-Hit every polygon.
-Hit every polygon that has the same Y coordinate for all points.
-Hit every polygon that has the same X/Z coordinates for all points.
-Hit random polygons.
-On rare occasions, hit all the polygons in a straight line, but not the line being tested.
I've tried everything and no longer have any idea what I'm doing. >_<
[Edited by - HyperHacker on August 8, 2006 8:57:40 PM]
Author
W00t! I've got it THIIIIIIIS close to working, by writing a whole new app that just renders some triangles (using as few classes as possible in case of memory corruption), analyzing what the function is doing (specifically, where it's returning false and why), and tweaking it until it worked 100% in the test program. I copied that back into the main program, and it's working in almost every case. I have yet to see it claim to hit a polygon that it shouldn't be hitting, the only issue is it's not hitting a few that it should.
A few examples of places I can't hit:
From this angle, I cannot hit the brick wall. By facing toward it, no problem.
I can hit the left Yoshi sign no problem, but not the right. There is at least one other such sign I can't seem to hit.
I can hit the top of the brick wall, but not the sides.
The signs are actually composed of four shapes like this:
Two make up the front of the sign, and two make up the back which has a different texture. There's some Z-fighting here which I can't seem to get rid of (anyone have any ideas?) so I suspect that might contribute to the problem there. The wall has me stumped though. Whatever the case, these spots all have something in common - all the triangles face straight up. I can hit other such triangles fine, though.
What bugs me the most is I can hit all of these - indeed, any triangle at all - in wireframe mode. O_o Also, I can't hit the back of any triangle (hence why I suspect Z-fighting is part of the problem) even when culling is disabled. (I never have managed to.)
[edit] Wrong image tags. ^_^;
[another edit] Alright, I figured out why I can only select certain polygons in wireframe mode. In normal mode, the line only extends out to the polygon, and at some angles (or with culled polygons) does not actually intersect it. In wireframe mode it goes right on through. Now to figure out how to fix that...
[edit again] Of course, I do that by changing a line in ScreenToWorldCoords:
gluUnProject((GLdouble)XP, (GLdouble)YP, (GLdouble)ZP, ModelView, Projection, Viewport, &WX, &WY, &WZ);
to
gluUnProject((GLdouble)XP, (GLdouble)YP, (GLdouble)1.0, ModelView, Projection, Viewport, &WX, &WY, &WZ);
99% working now. The intersection code is working perfectly. The only problem I'm having is pretty minor, but baffling nonetheless. I use ScreenToWorldCoords() to get the line coordinates, then LineIntersectsTriangle() to see which polygons fall under the cursor. I then do some simple comparisons to see which of these polygons' Z position is closest to the cursor position, so I select the object under the cursor rather than one behind it. This works most of the time, but if I click a certain position on the polygon, it seems to get omitted from the depth test, and one behind it gets selected. In fact it's usually the farthest one that gets selected in this case, no matter how many are actually being tested. Again, the intersector is working and correctly telling me that these polygons are being hit; the problem lies in the depth test not correctly testing them.
Here is the updated code that does all of this. (I might end up removing WorldPos3 from ScreenToWorldCoords(). I thought it might give me the coordinates of every pixel that the line intersects, but it only gives the closest one, which would be useful except it doesn't work in wireframe mode.)
Again, depending where I click, the depth test will just ignore one or more polygons or in some cases work in reverse. (Possibly just ignoring all but the farthest polygon.) The pixel coords returned are correct.
Here's a picture demonstrating the problem:
The yellow circle is where the camera was; the line indicates where I clicked and what direction the camera was facing. The line intersects a polygon on either side of the wall, as well as the grass. The red polygon is the one that ends up being selected, when it should be one on the other side of the wall.
[Edited by - HyperHacker on August 10, 2006 12:03:11 AM]
bool LineIntersectsTriangle(Vector LineStart, Vector LineEnd, SHAPE3D* Tri, unsigned int Point0, unsigned int Point1, unsigned int Point2){ Vector V, Intersect, T1, T2, T3, Norm; GLfloat D1, D2; V = Tri->Point[Point1].Position - Tri->Point[Point0].Position; Norm = V.CrossProduct(Tri->Point[Point2].Position - Tri->Point[Point0].Position); Norm.Normalize(); D1 = (LineStart - Tri->Point[Point0].Position).DotProduct(Norm); D2 = (LineEnd - Tri->Point[Point0].Position).DotProduct(Norm); if((D1 * D2) > 0.0) return false; //Doesn't cross if(D1 == D2) return false; //Parallel Intersect = Vector(LineStart + (LineEnd - LineStart) * (-D1 / (D2 - D1))); T1 = Norm.CrossProduct(Tri->Point[Point1].Position - Tri->Point[Point0].Position); T2 = Norm.CrossProduct(Tri->Point[Point2].Position - Tri->Point[Point1].Position); T3 = Norm.CrossProduct(Tri->Point[Point0].Position - Tri->Point[Point2].Position); if((T1.DotProduct(Intersect - Tri->Point[Point0].Position) >= 0.0) && (T2.DotProduct(Intersect - Tri->Point[Point1].Position) >= 0.0) && (T3.DotProduct(Intersect - Tri->Point[Point0].Position) >= 0.0)) return true; return false;}A few examples of places I can't hit:
From this angle, I cannot hit the brick wall. By facing toward it, no problem.
I can hit the left Yoshi sign no problem, but not the right. There is at least one other such sign I can't seem to hit.
I can hit the top of the brick wall, but not the sides.
The signs are actually composed of four shapes like this:
_ _|\|\| ¯ ¯Two make up the front of the sign, and two make up the back which has a different texture. There's some Z-fighting here which I can't seem to get rid of (anyone have any ideas?) so I suspect that might contribute to the problem there. The wall has me stumped though. Whatever the case, these spots all have something in common - all the triangles face straight up. I can hit other such triangles fine, though.
What bugs me the most is I can hit all of these - indeed, any triangle at all - in wireframe mode. O_o Also, I can't hit the back of any triangle (hence why I suspect Z-fighting is part of the problem) even when culling is disabled. (I never have managed to.)
[edit] Wrong image tags. ^_^;
[another edit] Alright, I figured out why I can only select certain polygons in wireframe mode. In normal mode, the line only extends out to the polygon, and at some angles (or with culled polygons) does not actually intersect it. In wireframe mode it goes right on through. Now to figure out how to fix that...
[edit again] Of course, I do that by changing a line in ScreenToWorldCoords:
gluUnProject((GLdouble)XP, (GLdouble)YP, (GLdouble)ZP, ModelView, Projection, Viewport, &WX, &WY, &WZ);
to
gluUnProject((GLdouble)XP, (GLdouble)YP, (GLdouble)1.0, ModelView, Projection, Viewport, &WX, &WY, &WZ);
99% working now. The intersection code is working perfectly. The only problem I'm having is pretty minor, but baffling nonetheless. I use ScreenToWorldCoords() to get the line coordinates, then LineIntersectsTriangle() to see which polygons fall under the cursor. I then do some simple comparisons to see which of these polygons' Z position is closest to the cursor position, so I select the object under the cursor rather than one behind it. This works most of the time, but if I click a certain position on the polygon, it seems to get omitted from the depth test, and one behind it gets selected. In fact it's usually the farthest one that gets selected in this case, no matter how many are actually being tested. Again, the intersector is working and correctly telling me that these polygons are being hit; the problem lies in the depth test not correctly testing them.
Here is the updated code that does all of this. (I might end up removing WorldPos3 from ScreenToWorldCoords(). I thought it might give me the coordinates of every pixel that the line intersects, but it only gives the closest one, which would be useful except it doesn't work in wireframe mode.)
/*Converts screen coords to world coordsInputs: -XPos, YPos: Screen coords. -WorldPos1: [Out] A vector containing the actual world coordinates. -WorldPos2: [Out] A vector containing the world coordinates, but at maximum depth. -WorldPos3: [Out] A vector containing the world coordinates of the pixel.Notes: -Automatically flips the Y coord - 0,0 is the top left like it should be. -WorldPos2 is provided so that you can easily determine if a given screen coordinate is overtop of a triangle by forming a line between the two (from the clicked position to the farthest visible point). WorldPos3 is provided so that you can do manual depth testing, in case you want to filter out polygons behind others. WorldPos1, WorldPos2 and WorldPos3 can all be NULL if you don't need them.*/void ScreenToWorldCoords(int XPos, int YPos, Vector* WorldPos1, Vector* WorldPos2, Vector* WorldPos3){ GLint Viewport[4]; GLdouble ModelView[16], Projection[16]; GLfloat XP, YP, ZP; GLdouble WX, WY, WZ; //World coords this translates to glLoadIdentity(); glRotatef(Camera.Rotation.X, 1.0, 0.0, 0.0); glRotatef(Camera.Rotation.Y, 0.0, 1.0, 0.0); glRotatef(Camera.Rotation.Z, 0.0, 0.0, 1.0); glTranslatef(Camera.Position.X, Camera.Position.Y, Camera.Position.Z); /*glRotatef(Camera.Rotation.X, 1.0, 0.0, 0.0); glRotatef(Camera.Rotation.Y, 0.0, 1.0, 0.0); glRotatef(Camera.Rotation.Z, 0.0, 0.0, 1.0); glTranslatef(-Camera.Position.X, -Camera.Position.Y, -Camera.Position.Z);*/ glGetIntegerv(GL_VIEWPORT, Viewport); //Get the viewport coords - Viewport[] = x, y, width, height //XP = (GLfloat)XPos / Viewport[2]; /*XP = ((GLfloat)XPos / (Viewport[2] / 2.0) - 1.0) * ((GLfloat)Viewport[2] / Viewport[3]); YP = (GLfloat)((Viewport[3] / 2.0) - YPos) / (Viewport[3] / 2.0);*/ XP = XPos; YP = Viewport[3] - YPos; glGetDoublev(GL_MODELVIEW_MATRIX, ModelView); //Get the matrices glGetDoublev(GL_PROJECTION_MATRIX, Projection); glReadPixels((GLint)XP, (GLint)YP, 1, 1, GL_DEPTH_COMPONENT, GL_FLOAT, &ZP); //Get the depth of the pixel //DebugOut(DO_ALL, "Mouse: %d, %d = %2.3f, %2.3f, %2.3f\n", XPos, Viewport[3] - YPos, XP, YP, ZP); if(WorldPos1) { gluUnProject((GLdouble)XP, (GLdouble)YP, (GLdouble)0.0, ModelView, Projection, Viewport, &WX, &WY, &WZ); //This gives us the near end of the line (at the cursor) (*WorldPos1) = Vector((GLfloat)WX, (GLfloat)WY, (GLfloat)WZ); } if(WorldPos2) { gluUnProject((GLdouble)XP, (GLdouble)YP, (GLdouble)1.0, ModelView, Projection, Viewport, &WX, &WY, &WZ); //the far end, (*WorldPos2) = Vector((GLfloat)WX, (GLfloat)WY, (GLfloat)WZ); } if(WorldPos3) { gluUnProject((GLdouble)XP, (GLdouble)YP, (GLdouble)ZP, ModelView, Projection, Viewport, &WX, &WY, &WZ); //and the coords of the actual pixel at this location. (*WorldPos3) = Vector((GLfloat)WX, (GLfloat)WY, (GLfloat)WZ); }}/*Checks if a line intersects a triangle.Inputs: -LineStart: Position the line starts at. -LineEnd: Position the line ends at. -Tri: Shape to check. -Point1, Point2, Point3: Points of Tri that form the triangle to test.Returns: True if the line intersects the triangle, false otherwise.*/bool LineIntersectsTriangle(Vector LineStart, Vector LineEnd, SHAPE3D* Tri, unsigned int Point0, unsigned int Point1, unsigned int Point2){ Vector V, Intersect, T1, T2, T3, Norm; GLfloat D1, D2; V = Tri->Point[Point1].Position - Tri->Point[Point0].Position; Norm = V.CrossProduct(Tri->Point[Point2].Position - Tri->Point[Point0].Position); Norm.Normalize(); D1 = (LineStart - Tri->Point[Point0].Position).DotProduct(Norm); D2 = (LineEnd - Tri->Point[Point0].Position).DotProduct(Norm); if((D1 * D2) > 0.0) return false; //Doesn't cross if(D1 == D2) return false; //Parallel Intersect = Vector(LineStart + (LineEnd - LineStart) * (-D1 / (D2 - D1))); T1 = Norm.CrossProduct(Tri->Point[Point1].Position - Tri->Point[Point0].Position); T2 = Norm.CrossProduct(Tri->Point[Point2].Position - Tri->Point[Point1].Position); T3 = Norm.CrossProduct(Tri->Point[Point0].Position - Tri->Point[Point2].Position); if((T1.DotProduct(Intersect - Tri->Point[Point0].Position) >= 0.0) && (T2.DotProduct(Intersect - Tri->Point[Point1].Position) >= 0.0) && (T3.DotProduct(Intersect - Tri->Point[Point0].Position) >= 0.0)) return true; return false;}//-------code that checks for intersections and selects the closest polygon that was hit#define Difference(x, y) ((x > y) ? x - y : y - x)int i, j, k;int ClickedObj=-1, ClickedShape=-1, ClickedPoint=-1;Vector ClickNear, ClickFar, ClickPixel;Vector TDepth;if(MouseMoved){ ClickedObj = -1; ClickedShape = -1; ClickedPoint = -1; TDepth.X = 10000.0; TDepth.Y = 10000.0; TDepth.Z = FRUSTUM_DEPTH; //Get the world position we clicked ScreenToWorldCoords(MouseX, MouseY, &ClickNear, &ClickFar, &ClickPixel); //DebugOut(DO_ALL, "Click at %d, %d = %2.3f, %2.3f, %2.3f\n", MouseX, MouseY, ClickPos1.X, ClickPos1.Y, ClickPos1.Z); int Hits = 0, R; //todo: remove for(i=0; i<NumLevelObjects; i++) { for(j=0; j<LevelObject->NumShapes; j++) { for(k=0; k<(LevelObject->Shape[j].NumPoints - 2); k++) { /* See if the line between the cursor and the edge of the world intersects any polygons. There's a depth test here to ensure only the closest polygon gets selected, but it doesn't work very well. */ LevelObject->Shape[j].Point[k].Highlighted = false; LevelObject->Shape[j].Point[k+1].Highlighted = false; LevelObject->Shape[j].Point[k+2].Highlighted = false; R = LineIntersectsTriangle(ClickNear, ClickFar, &LevelObject->Shape[j], k, k+1, k+2); if(R) { Hits++; //LevelObject->Shape[j].Point[k].Highlighted = true; //if(Difference(LevelObject->Shape[j].Point[k].Position.Z, ClickPixel.Z) < TDepth) //Closer than previous check if( (Difference(LevelObject->Shape[j].Point[k].Position.X, ClickPixel.X) <= TDepth.X) && (Difference(LevelObject->Shape[j].Point[k].Position.Y, ClickPixel.Y) <= TDepth.Y) && (Difference(LevelObject->Shape[j].Point[k].Position.Z, ClickPixel.Z) <= TDepth.Z)) { ClickedObj = i; ClickedShape = j; ClickedPoint = k; TDepth.X = Difference(LevelObject->Shape[j].Point[k].Position.X, ClickPixel.X); TDepth.Y = Difference(LevelObject->Shape[j].Point[k].Position.Y, ClickPixel.Y); TDepth.Z = Difference(LevelObject->Shape[j].Point[k].Position.Z, ClickPixel.Z); //DebugOut(DO_ALL, "Hit %d: Pixel @ %2.3f, %2.3f, %2.3f\n", Hits, ClickPixel.X, ClickPixel.Y, ClickPixel.Z); //break; } } //if(R) } //for k //if(Hits) break; } //for j //if(Hits) break; } //for i DebugOut(DO_ALL, "%d hits\n", Hits); if((ClickedObj >= 0) && (ClickedShape >= 0) && (ClickedPoint >= 0)) { //LevelObject[ClickedObj]->Shape[ClickedShape].Point[ClickedPoint].Highlighted = true; LevelObject[ClickedObj]->Shape[ClickedShape].Point[0].Highlighted = true; LevelObject[ClickedObj]->Shape[ClickedShape].Point[1].Highlighted = true; LevelObject[ClickedObj]->Shape[ClickedShape].Point[2].Highlighted = true; DebugOut(DO_ALL, "Clicked obj %d, shp %d, pt %d (%2.3f, %2.3f, %2.3f; px@ %2.3f, %2.3f, %2.3f)\n", ClickedObj, ClickedShape, ClickedPoint, LevelObject[ClickedObj]->Shape[ClickedShape].Point[2].Position.X, LevelObject[ClickedObj]->Shape[ClickedShape].Point[2].Position.Y, LevelObject[ClickedObj]->Shape[ClickedShape].Point[2].Position.Z, ClickPixel.X, ClickPixel.Y, ClickPixel.Z); } MouseMoved = false;}Again, depending where I click, the depth test will just ignore one or more polygons or in some cases work in reverse. (Possibly just ignoring all but the farthest polygon.) The pixel coords returned are correct.
Here's a picture demonstrating the problem:
The yellow circle is where the camera was; the line indicates where I clicked and what direction the camera was facing. The line intersects a polygon on either side of the wall, as well as the grass. The red polygon is the one that ends up being selected, when it should be one on the other side of the wall.
[Edited by - HyperHacker on August 10, 2006 12:03:11 AM]
It sounds like you've got this all worked out.
As for that weird triangle on the other side of the wall,
just dot its normal against the direction of the pick ray
(from the camera to the point). If its positive, reject the triangle.
However, I wasn't able to read the whole post
(really, really big on my laptop).
Well, good luck with that one!
As for that weird triangle on the other side of the wall,
just dot its normal against the direction of the pick ray
(from the camera to the point). If its positive, reject the triangle.
However, I wasn't able to read the whole post
(really, really big on my laptop).
Well, good luck with that one!
Author
Thanks, but I'm not sure you understood the problem. I want triangles to be selected from the back like that. In this case, however, there is another triangle on the other side of the wall that should be getting selected instead. The depth testing routine seems to have ignored that one and taken the next, which is the one shown in the picture.
Author
Well, I think I managed to hammer out a solution. The depth testing is working quite nicely now. :)
ScreenToWorldCoords() now returns ClickPixel, the coordinates of the exact pixel I clicked on. I figured the best way to find the correct triangle would be a point-in-triangle algorithm. However I couldn't find any and don't know much about writing my own, so I cheated a bit. This actually uses the modelview matrix to create a very small line segment going through the pixel, and then simply uses the LineIntersectsTriangle() function I already had to determine if that line is within any triangles. Of course if the two are really close together it still can't tell them apart, but usually at that point they're Z-fighting anyway. (POLY_HIT_LENGTH is defined as 0.5. Reducing it makes the segment smaller, working around this problem in many cases, but using too small a value makes it impossible to hit triangles at near-parallel angles.)
int Hits = 0, R; //todo: removeGLdouble M[16];ScreenToWorldCoords(MouseX, MouseY, &ClickNear, &ClickFar, &ClickPixel);glGetDoublev(GL_MODELVIEW_MATRIX, M);for(i=0; i<NumLevelObjects; i++){ for(j=0; j<LevelObject->NumShapes; j++) { for(k=0; k<(LevelObject->Shape[j].NumPoints - 2); k++) { /* See if the line between the cursor and the edge of the world intersects any polygons. There's a depth test here to ensure only the closest polygon gets selected, but it doesn't work very well. */ LevelObject->Shape[j].Point[k].Highlighted = false; LevelObject->Shape[j].Point[k+1].Highlighted = false; LevelObject->Shape[j].Point[k+2].Highlighted = false; Vector N = Vector( ClickPixel.X + (POLY_HIT_LENGTH * M[0]) + (POLY_HIT_LENGTH * M[1]) + (POLY_HIT_LENGTH * M[2]), ClickPixel.Y + (POLY_HIT_LENGTH * M[4]) + (POLY_HIT_LENGTH * M[5]) + (POLY_HIT_LENGTH * M[6]), ClickPixel.Z + (POLY_HIT_LENGTH * M[8]) + (POLY_HIT_LENGTH * M[9]) + (POLY_HIT_LENGTH * M[10]) ); Vector F = Vector( ClickPixel.X + (-POLY_HIT_LENGTH * M[0]) + (-POLY_HIT_LENGTH * M[1]) + (-POLY_HIT_LENGTH * M[2]), ClickPixel.Y + (-POLY_HIT_LENGTH * M[4]) + (-POLY_HIT_LENGTH * M[5]) + (-POLY_HIT_LENGTH * M[6]), ClickPixel.Z + (-POLY_HIT_LENGTH * M[8]) + (-POLY_HIT_LENGTH * M[9]) + (-POLY_HIT_LENGTH * M[10]) ); R = LineIntersectsTriangle(N, F, &LevelObject->Shape[j], k, k+1, k+2); if(R) { Hits++; ClickedObj = i; ClickedShape = j; ClickedPoint = k; TDepth.X = Difference(LevelObject->Shape[j].Point[k].Position.X, ClickPixel.X); TDepth.Y = Difference(LevelObject->Shape[j].Point[k].Position.Y, ClickPixel.Y); TDepth.Z = Difference(LevelObject->Shape[j].Point[k].Position.Z, ClickPixel.Z); } //if(R) } //for k //if(Hits) break; } //for j //if(Hits) break;} //for iScreenToWorldCoords() now returns ClickPixel, the coordinates of the exact pixel I clicked on. I figured the best way to find the correct triangle would be a point-in-triangle algorithm. However I couldn't find any and don't know much about writing my own, so I cheated a bit. This actually uses the modelview matrix to create a very small line segment going through the pixel, and then simply uses the LineIntersectsTriangle() function I already had to determine if that line is within any triangles. Of course if the two are really close together it still can't tell them apart, but usually at that point they're Z-fighting anyway. (POLY_HIT_LENGTH is defined as 0.5. Reducing it makes the segment smaller, working around this problem in many cases, but using too small a value makes it impossible to hit triangles at near-parallel angles.)
Quote:Original post by i-photon
Well, as far as GLfloat goes, its just a typedef of a float (so it's treated the same way), the compiler should make automatic conversions, albeit while complaining at the same time.
This is not quite correct.
a GLfloat has a fixed size on ALL platforms, the size of float is platform dependant.
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement
