Kernel estimation of the diameter distribution in boolean models with spherical grains (Q4485012)

Consider an i.i.d. sequence \((K_j)\) of regular convex compact random sets in \(\mathbb{R}^d\), an independent Poisson process \((X_j)\) in \(\mathbb{R}^d\), and \(K:= \bigcup_{j\in \mathbb{N}} (K_j+ X_j)\). For fixed \(u\in\mathbb{S}^{d- 1}\), let \(\ell_j(u)\) denote the contact point between \(K_j+ X_j\) and the hyperplane which is orthogonal to \(u\) and lower tangent to \(K_j+ X_j\). Consider then \[ \rho_u(x):= \lim_{\varepsilon\downarrow 0} \varepsilon^{-2} \mathbb{E} \Biggl[\Biggl( \sum_{j\in \mathbb{N}} 1_{K^c\cap B(0;\varepsilon)}(\ell_j(u))\Biggr)\times \Biggl(\sum_{j\in \mathbb{N}} 1_{K^c\cap B(x;\varepsilon)} (\ell_j(u))\Biggr)\Biggr], \] the ``second-order product density'' of the ``process of exposed tangent points''. It is proved that the square root of some kernel estimator of \(\{\rho_u(x), x\in\mathbb{R}^d\}\) satisfies a central limit theorem with constant diagonal covariance. When \(K_j\) is a ball, an analoguous result concerning the distribution function of the radius of \(K_j\) is deduced.