No label defined (Q4351273)

The canonical way of estimating the intensity of a stationary \(k\)-flat process is to measure the \(k\)-content of the intersections of the flats of the process with a compact convex observation window. For planar line processes that means to sum the lengths of all segments of lines of the process being visible inside the window. This work was inspired by an -- at first glance -- paradox concerning the estimation of the intensity of stationary isotropic Poisson line processes in the plane. For this class of processes an unbiased estimator for the intensity can be obtained by counting the intersection points of the lines with the boundary of the observation window. Ohser, Baddeley and Cruz-Orive showed that this new estimator is better than the canonical estimator in the sense that its variance is lower. This is astonishing in so far as this point count estimator only uses the (lower-dimensional) information provided by sections with the boundary of the observation window whereas the canonical estimator measures the length of the line segments. In his lecture in Valencia, Cruz-Orive posed the problem, whether there exist similar estimators for more general classes of line processes. Here an answer is given in the setting of \(k\)-flats in \(\mathbb{R}^d\). Starting from stationary isotropic Poisson flat processes the considered class of flat processes can be changed in several ways. The first generalization is the step to stationary Poisson flat processes with a known directional distribution, which is not necessarily the isotropic one. Next only some information about invariance properties of the directional distribution is assumed. This includes no knowledge at all, i.e. all stationary Poisson flat processes are considered. Mixing leads to the corresponding classes of Cox processes. To drop the Poisson assumption is another way of generalization. It leads to stationary ergodic and general stationary flat processes.