OSG can define many cameras as well as many (rendering) views. Then one or many views can be bound to a single camera. Views can then be tied to a QT view...
In order to compute the intersection between a scene and plane (by order of recommendation) :
at the OSG level and within the OSG formalism one can weed out the scene with boolean operation (cut the scene with an OSG plane).
at the OSG level one can set a camera with a depth field reduced to epsilon (this is cheating though)
use the same trick (forward plane, backward plane) at the OpenGL level (where only simple geometric objects are available). Note that doing so at the OSG level enables some optimizations.
The user now needs to use the GUI to define the cutting plane spatial position. In order to do so one can bind, at the QT level, some keyboard event with an OSG method. Expect some technical difficulties in order to trap events: refer to the "event section" of SimpleAdapterVisu code
Cities underground geological data becomes available
References
Technical notes concerning the cutting planes
According to Jeremy Levalois: