Can someone please link a good article/tutorial/anything or maybe even explain how to correctly cast a ray from the mouse coordinates to pick objects in 3D?
I already have the Ray and intersection works, now I only need to create the ray from the mouse click.
I would just like have something which I know actually should work, thats why I ask the professionals here, not something where I am unsure if it is even correct in the first place.
I have a ray class, which actually works and detects intersection if I set the origin and direction to be the same as the camera, so when I move the camera it actually selects the right thing.
Now I would like to actually have 3D picking with the mouse, not camera movement.
I have read so many other questions about this, 2 tutorials, and especially so much different math stuff, since I am really not good at it.
But that didn't help me much, because the people there often use some "unproject" functions, which seem to actually be deprecated and which I have no idea how to use and also don't have access to.
Right now I set the ray origin to the camera position and then try to get the direction of the ray from the calculations in this tutorial.
And it works a little bit, meaning the selection works when the camera is pointed at the object and also sometimes along the whole y-axis, I have no idea what is happening.
If someone wants to take a look at my code right now:
public Ray2(Camera cam, float mouseX, float mouseY) {
origin = cam.getEye();
float height = 600;
float width = 600;
float aspect = (float) width / (float) height;
float x = (2.0f * mouseX) / width - 1.0f;
float y = 1.0f - (2.0f * mouseX) / height;
float z = 1.0f;
Vector ray_nds = vecmath.vector(x, y, z);
Vector4f clip = new Vector4f(ray_nds.x(), ray_nds.y(), -1.0f, 1.0f);
Matrix proj = vecmath.perspectiveMatrix(60f, aspect, 0.1f, 100f);
proj = proj.invertRigid();
float tempX = proj.get(0, 0) * clip.x + proj.get(1, 0) * clip.y
+ proj.get(2, 0) * clip.z + proj.get(3, 0) * clip.w;
float tempY = proj.get(0, 1) * clip.x + proj.get(1, 1) * clip.y
+ proj.get(2, 1) * clip.z + proj.get(3, 1) * clip.w;
float tempZ = proj.get(0, 2) * clip.x + proj.get(1, 2) * clip.y
+ proj.get(2, 2) * clip.z + proj.get(3, 2) * clip.w;
float tempW = proj.get(0, 3) * clip.x + proj.get(1, 3) * clip.y
+ proj.get(2, 3) * clip.z + proj.get(3, 3) * clip.w;
Vector4f ray_eye = new Vector4f(tempX, tempY, tempZ, tempW);
ray_eye = new Vector4f(ray_eye.x, ray_eye.y, -1.0f, 0.0f);
Matrix view = cam.getTransformation();
view = view.invertRigid();
tempX = view.get(0, 0) * ray_eye.x + view.get(1, 0) * ray_eye.y
+ view.get(2, 0) * ray_eye.z + view.get(3, 0) * ray_eye.w;
tempY = view.get(0, 1) * ray_eye.x + view.get(1, 1) * ray_eye.y
+ view.get(2, 1) * ray_eye.z + view.get(3, 1) * ray_eye.w;
tempZ = view.get(0, 2) * ray_eye.x + view.get(1, 2) * ray_eye.y
+ view.get(2, 2) * ray_eye.z + view.get(3, 2) * ray_eye.w;
tempW = view.get(0, 3) * ray_eye.x + view.get(1, 3) * ray_eye.y
+ view.get(2, 3) * ray_eye.z + view.get(3, 3) * ray_eye.w;
Vector ray_wor = vecmath.vector(tempX, tempY, tempZ);
// don't forget to normalise the vector at some point
ray_wor = ray_wor.normalize();
direction = ray_wor;
}
First,unproject() method is the way to go.It is not deprecated at all.You can find it implemented in GLM math library for example.Here is my implementation of Ray based 3D picking:
// let's check if this renderable's AABB is clicked:
const glm::ivec2& mCoords = _inputManager->GetMouseCoords();
int mouseY = _viewportHeight - mCoords.y;
//unproject twice to build a ray from near to far plane"
glm::vec3 v0 = glm::unProject(glm::vec3(float(mCoords.x), float(mouseY), 0.0f),_camera->Transform().GetView(),_camera->Transform().GetProjection(), _viewport);
glm::vec3 v1 = glm::unProject(glm::vec3(float(mCoords.x), float(mouseY), 1.0f),_camera->Transform().GetView(),_camera->Transform().GetProjection(), _viewport);
glm::vec3 dir = (v1 - v0);
Ray r(_camera->Transform().GetPosition(),dir);
float ishit ;
//construct AABB:
glm::mat4 aabbMatr = glm::translate(glm::mat4(1.0),renderable->Transform().GetPosition());
aabbMatr = glm::scale(aabbMatr,renderable->Transform().GetScale());
//transforms AABB vertices(need it if the origianl bbox is not axis aligned as in this case)
renderable->GetBoundBox()->RecalcVertices(aabbMatr);
//this method makes typical Ray-AABB intersection test:
if(r.CheckIntersectAABB(*renderable->GetBoundBox().get(),&ishit)){
printf("HIT!\n");
}
But I would suggest you also to take a look at color based 3d picking which is pixel perfect and even easier to implement.