>We need to calculate the volume of a box defined by n point where all the

>coordinates (Xi, Yi, Zi) of all the surfaces are known.

>the technic of calculating the volume of a polyhedron with the signed

>volume of the tetrahedron (FAQ 5.1.9) shouldn't work on convex surfaces.

>So we would like to find an other algorythm (like the verification of tthe

>belonging of a moving cube to this volume ...)

If your surface is closed and if you can also compute (outer) unit normal

vectors at each of the sample points, then the code

http://www.cs.unc.edu/~eberly/misc.htm , file meas3d.cpp computes

approximations to the surface area and volume for such data sets. If

your samples (Xi,Yi,Zi) were obtained by thresholding from a 3D image,

then normals can be computed from that image by estimating the

gradient at the sample by using central finite differences for each of

the three partial derivatives.

The algorithm uses the Divergence Theorem from calculus. It also

does not require computing a polygonal mesh and does not require

any sorting. You simply feed it the list of surface points and normal

vectors.

Dave Eberly