Edit 2: I found what I was looking for here. The near and far planes compound the bounding box of the camera.
I have a Perspective Camera and multiple set of points (SphereGeometry) in the space and I want to define a bounding box according to what is visible in the field of view, so that I only need to load the points that are currently visible in the space according to the camera settings.
My problem when defining the bounding box is that I don't have a reference point between the camera position and what I'm currently looking at. For example, if the camera is at the position v1(20, 20, 20), and let's say the mid point of my set of points is set at v2(5, 5, 5). Then, I can define the initial point of my bounding box as this mid point v2(5, 5, 5) , and I could create a bounding box (square) with size (v1-v2)/2. But once I start moving the camera I lose this reference initial point, and I don't know how to obtain a distance parameter to define a proper bounding box of "what is currently visible".
Every time I move the camera I need to know the size of the bounding box and its position so that it represents as accurate as possible "what is currently visible" in the field of view.
One possible solution could be to translate the initial point v2(5, 5, 5) following every camera movement, is there any way to do this or something similar?
Edit:
So far I've been able to replicate @TheJim01 's code into my component, but I think this solution is not what I was looking for or I'm not able to understand it properly. I'm adding some images next to better explain what is going on.
In the next image I have a set of points of various colors, the 8 brown points represent each of the worldBoxCorners before performing a worldToLocal transformation. Which fits the whole space as it should.
Next, I render worldBoxCorners after the transformation. I do not understand how the bounding box is moved or how can I use it.
What I'm looking for is, for example if I zoom in into the space in the greener zone, to obtain a bounding box alike the last image. The point is to only load the points that are within the field of view, but for that I need to define a bounding box. When I say loading the points, I mean that only the points that are visible should be loaded from the backend (not talking about rendering here).
This solution provides a perfect bounding box for the current loaded space. What I'm missing on is how to define a new bounding box for a given loaded space according to camera parameters change (zoom in/out or rotations).