When turning a camera around or changing the point of view position you generally experience an apparent shifting of foreground objects relative to background objects which is called parallax. When taking the pictures for a virtual tour scene you want to rotate the camera around a line that runs through (or as close as possible to) a certain point such that the parallax error is completely eliminated. Failing to do so may result in problems with the stitching process and will require a lot more work with the possibe need for image retouching.
So first of all, in order to set the terminology to rights I will provide you with the proper definition of the nodal point. In fact, it consists of two points - the front and rear nodal point. In order to understand this issue let us first define:
- The principal planes (front and rear) which have the property that a ray emerging from the lens appears to have crossed the rear principal plane at the same distance from the axis that that ray appeared to cross the front principal plane, as viewed from the front of the lens.
- The principal points which are the points where the principal planes cross the optical axis.
If the medium on both sides of the optical system is the same (e.g. air), which is indeed our case, then the front and rear nodal points coincide with the front and rear principal planes, respectively (see the diagram). So there is nothing like a single nodal point to which so many pano tutorials are wrongly referring.
On the other hand, the most common explanation of nodal (or principal) point as being usually heard from pano photography folks says that it is the point inside a lens where light rays converge and flip over before being focused onto a film plane or a digital sensor. It should also be the place where the entrance pupil of the camera is situated. So far so good, this explains where the point to which one should set the center of rotation for shooting panorama is situated. However, using the term "nodal point" is absolutely improper here and one shoud rather think of a different term to avoid confusion. I personaly prefer naming it "pivot point". By the way, it is also worth noting that the nodal point is not the same as the film/sensor plane (sometimes marked at the bottom of cameras) as some photographers think.
In practice determining the pivot point of a lens is quite easy to do visually. The only thing you need is two vertical features that you can use as reference lines, for instance a flag or light pole, corner of a wall etc. Set one of them very close to the camera (1 meter or so) while the other one stays far away. Take two shots at different angles so that the vertical features you chose are close to the right or left edge of the frame, respectively. Most likely you will see the parallax shift of the objects. Adjust the corresponding rail on your panoramic head (the one that changes the vertical axis of camera rotation) so that you reduce the parallax. Eventually you should be able to eliminate it entirely which usually requires few iterations depending on how accurate your judgement was in each step. By checking the test images down to a pixel resolution a sub-milimeter accuracy in the setting of pivot point can be achieved.