3



So, im building an architectural visualization with Three.js, and one of the things the user should be able to do is to click on things and orbit around them. The problem is that the camera is able to clip through wall. I fixed that by assigning each clickable object its own limiting azimuth and polar angles.

Now the Problem is that azimuth angles go from -PI to +PI and its impossible to limit between for example 1.5, and -2.4 because its limiting the "wrong" way. I hope this graphic explains that a little better:

enter image description here

Heres a link to the live version: (You control by clicking on the ground)

https://jim-fx.com/modern/

As you can see, on objects on the right side of the room the limiting works flawless, but on the cabinet and the vases the camera clips through the wall.

If anyone could help me that would be amazing. And any other tipps are welcome aswell.

Greetings, Max

gman
  • 100,619
  • 31
  • 269
  • 393
Max Richter
  • 33
  • 4

1 Answers1

0

There is several solution to your problem. One is to implements a kind of collision detection with some real or virtual wall for your camera, wich stops the rotation. However, I guess your are looking for something simpler to implement.

As i don't know Three.js very well, I will provides you a generic solution, but which should be easily adaptable to Three.js.

The first thing is to do not use the built-in Three.js orbit control, but to implement your own, where you control all your transformations. And, this is in fact very easy.

To create an orbitable camera, you simply have to crate:

  • A "null" transformable object, which mean a simple transformable entity that does not embed any shape (is not rendered, is invisible, but exists). I hope Three.js provides such elementary thing.

  • A camera, which should be itself another transformable.

Once you have this, you simply parent the camera to the "null" object. Now parented to the "null" object, if you rotate the "null" object, you rotate the camera with. Then to orbits, you now have to move back the camera from the parent object:

     Null                Camera

       + - - - - - - - - - |>

Like this, the "null" object becomes the camera "look at point", and if you rotate the "null" object around Y (I believe Three.js use Y up), you controls the camera azimuth. If you rotate the "null" object in X or Z (depending coordinate system), you will control the camera altitude. Then, you even can control the camera forward-backward to close up to the "look at point" by moving your camera in its local Z axis..

Well, you now have an orbit-camera easy to control. But your problem is not yet solved: How to make this control Pi / -Pi possible in every camera initial orientation ?

Simple: You create second "null" transform object, name it "the socle", and you parent the first one to this last one: Like this, the rotation of the camera "look at point" is always local, and you can now rotate "the socle" to give your "Orbital camera" group, an initial orientation in the world space.

In fact, it is pretty like creating virtual gimbals. I hope I was clear, with pictures this would be more easy to visualize...

  • Okay, wow, thank you very much for the in depth answer. I think implementing the controls myself is a really good idea, especially as i had to do some "hacky" things to get the first person camera to work. And that would probably cut about 1000 Lines of OrbitControls code from my project. – Max Richter Oct 25 '17 at 20:17
  • ""UPDATE"" Okay, i wrote my own controls and it was a lot of fun, and works very well, they are already live at the link posted above. I have learned a lot about OOP and using vector math and programming animations, so again, thanks a lot. – Max Richter Dec 01 '17 at 09:31