Towards automatic tree rings detection in images of scanned wood samples

•Fully automatic approach for three-ring boundaries detection is proposed.•The method is dedicated to scanned images of wood cores.•The method was tested on wood scans of 12 tree species.•For some tree species the method performs almost faultlessly.•The dataset used in the study is made accessible for other researchers.

In this paper, the problem of automation of dendrochronological measurements is considered. In particular, a fully automatic, image-based approach for detecting tree-ring boundaries in wood core images is introduced. The method is based on image gradient peak detection and linking. Unlike other existing approaches, the proposed method performs well on a variety of wood types and corresponding tree species. The results of applying the proposed approach to scanned images of wood cores representing 12 tree species (4 conifer and 8 angiosperm) are presented and discussed. Analysis of the results shows that the method performs almost faultlessly for conifer species, detecting almost 100% of tree-ring boundaries. In the case of diffuse-porous species, the results are not quite as good, but still at a level of 85% of properly detected tree-ring boundaries. For ring-porous species, the method has some problems, which are described in the paper.