Camera help
Author
I'm trying to make a camera which will allow me to look around my heightmap that I have created. I've looked for tutorials and such on the web but I can only seem to find camera systems that involve moving the object, I dont want this, I want to draw the image and then look around it by moving the camera's position.
Is this possible just by using gluookAt()? To start with I want to keep the camera looking at the centre of my heightmap the entire time, so I would have thought I only need a function for each of the first 3 parameters of the gluLookAt function (eyeX, eyeY, eyeZ) to move the camera round. Is this correct? And if so could you point me in the direction of a guide that would help me do this, I understand the level of maths I would need to use but it's linking this maths with the 3D coordinate system that gets me in a muddle!
Sorry if its a stupid question but I'm a bit of a newbie to openGL and programming in general so I would appreciate any help. Thanks.
What you are asking for is a little like a way to take a picture of a beach and then magically walking around inside the picture. You can't. It's a 2D snapshot of the scene. If you want to change what you see on the picture, you obviously have to take another one from a different angle (ie draw your stuff again). All you can do is move the picture around in front of your eyes (basically rendering to texture and move a textured rectangle around the screen), which most likely isn't what you want.
The reason people do cameras by "moving the objects" is because it's the same. Not just because if you move to the left, everything else seems to be moving to the right, but because the resulting matrix for moving the camera left is virtually the same as the matrix to move the world to the right. gluLookAt does nothing but setting up the modelview matrix with the inverse of the cameras matrix, effectively moving the world the opposite way.
The reason people do cameras by "moving the objects" is because it's the same. Not just because if you move to the left, everything else seems to be moving to the right, but because the resulting matrix for moving the camera left is virtually the same as the matrix to move the world to the right. gluLookAt does nothing but setting up the modelview matrix with the inverse of the cameras matrix, effectively moving the world the opposite way.
Author
Dammit!! OK thanks for the help Trienco. Is the any guide that has sample code and goes through this in detail because I'm a little confused?!
Thanks again.
Thanks again.
For a full description you have to dive into matrix math with transformations. But for the basic understanding it is sufficient to look at translational transformations only. Notice please that the following is the way of thinking that the camera should be part of the world.
Assume your camera is located at a position given by the point c. Here the co-ordinates of c are assumed to be given w.r.t. the "world", i.e. in the global co-ordinate system. Assume furthur you have a mesh located at the point m, also given w.r.t. the global co-ordinate system.
When it comes to rendering, you have to make the camera the "center of the world". This is because rendering happens in the viewing rectangle on your monitor, and the graphics hardware defines that rectangle to be located at zero. To get this, you subtract the camera's position from everything related to the world. This becomes clear when looking at the camera itself: Its new position will become
c' := c - c == 0
what is exactly what we want. Notice please that subtraction here is to be interpreted as "applying the inverse transformation"!
Similarly, the new position of the mesh becomes
m' := m - c
That is, looking at the structure of the formula, nothing else than a translational transformation of the mesh! And due to the subtraction the translation is along the opposite direction. That is what Trienco means: E.g. shifting the camera to the right has the same effect as shifting the world (the mesh in this case) to the left.
Extending this to matrices looks as follows. The local co-ordinate frame (from which the position is a part of) of the camera w.r.t. the global frame is
CG
and the same of the mesh is
MG
Applying the inverse transformation of CG to the mesh hence means
CG-1 * MG
(I'm using column vectors here, as OpenGL does).
The left part is called the "view matrix" (at least OpengGl does so), and the right part is called the "model matrix". In total it is the MODELVIEW matrix. It means in fact that 1st a mesh, given in its local co-ordinate frame, is 1st transformed into the global frame (done by the model matrix) and the result is then transformed into the camera's local frame (done by the view matrix).
Assume your camera is located at a position given by the point c. Here the co-ordinates of c are assumed to be given w.r.t. the "world", i.e. in the global co-ordinate system. Assume furthur you have a mesh located at the point m, also given w.r.t. the global co-ordinate system.
When it comes to rendering, you have to make the camera the "center of the world". This is because rendering happens in the viewing rectangle on your monitor, and the graphics hardware defines that rectangle to be located at zero. To get this, you subtract the camera's position from everything related to the world. This becomes clear when looking at the camera itself: Its new position will become
c' := c - c == 0
what is exactly what we want. Notice please that subtraction here is to be interpreted as "applying the inverse transformation"!
Similarly, the new position of the mesh becomes
m' := m - c
That is, looking at the structure of the formula, nothing else than a translational transformation of the mesh! And due to the subtraction the translation is along the opposite direction. That is what Trienco means: E.g. shifting the camera to the right has the same effect as shifting the world (the mesh in this case) to the left.
Extending this to matrices looks as follows. The local co-ordinate frame (from which the position is a part of) of the camera w.r.t. the global frame is
CG
and the same of the mesh is
MG
Applying the inverse transformation of CG to the mesh hence means
CG-1 * MG
(I'm using column vectors here, as OpenGL does).
The left part is called the "view matrix" (at least OpengGl does so), and the right part is called the "model matrix". In total it is the MODELVIEW matrix. It means in fact that 1st a mesh, given in its local co-ordinate frame, is 1st transformed into the global frame (done by the model matrix) and the result is then transformed into the camera's local frame (done by the view matrix).
Author
What about if I were to redraw the image every time the camera position was changed, would it be possible to just use the gluLookAt() function to view it from a different perspective, rather than changing the objects position?
I just seems more logical to me to have a couple of functions to control the position of the camera rather than translating the object.
Cheers.
I just seems more logical to me to have a couple of functions to control the position of the camera rather than translating the object.
Cheers.
Manipulating the camera as it were an object in the world is usual. That is exactly why OpenGL deals with both the VIEW and the MODEL matrix (although it deals with them in a combined manner). gluLookAt can be used for that purpose.
However, perhaps here a misinterpretation of terms causes some confusion. When you say "draw an image" you perhaps mean instead "render the scene". You're, of course, able to re-render the scene each time the camera's position and/or orientation has changed, yielding in an adapted view into the scene (in a texture or on the screen). You're _not_ able to render the scene into an image (e.g. a texture) and navigate in a 3D manner inside the image. That is what Trienco has meant above.
However, perhaps here a misinterpretation of terms causes some confusion. When you say "draw an image" you perhaps mean instead "render the scene". You're, of course, able to re-render the scene each time the camera's position and/or orientation has changed, yielding in an adapted view into the scene (in a texture or on the screen). You're _not_ able to render the scene into an image (e.g. a texture) and navigate in a 3D manner inside the image. That is what Trienco has meant above.
Author
This is what i'm after: http://www.lighthouse3d.com/opengl/glut/index.php?6
There's the two functions orientMe and moveMeFlat which call gluLookAt after making adjustments to the camera position. I think all I need to do is add more similar functions so I can rotate around different axis and move in different directions but I dont understand how it works. I understand the x, y, and z values for the camera position but its the lx, ly, lz values defining the line of sight that I dont understand. Why do you need to change the line of sight when we are just moving backwards and forwards for example?
Is there anywhere that explains this in more depth?
Thanks again :)
There's the two functions orientMe and moveMeFlat which call gluLookAt after making adjustments to the camera position. I think all I need to do is add more similar functions so I can rotate around different axis and move in different directions but I dont understand how it works. I understand the x, y, and z values for the camera position but its the lx, ly, lz values defining the line of sight that I dont understand. Why do you need to change the line of sight when we are just moving backwards and forwards for example?
Is there anywhere that explains this in more depth?
Thanks again :)
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement
