To estimate the densityfof the mean measure of a discrete random measureη, the sequence of the observed random measures is usually supposed to be independent. In this paper, that sequence is associated: this definition is a particular case of association of probability measure on a partially ordered Polish space (Lindqvist [18]). The kernel estimator offconverges in probability and almost surely, pointwise and uniformly, under second-order moment conditions onηnand on the sequence (hn) of the window-widths of the estimator.