A vision-based system for automatic growing ring detection and measurement

Abstract

An automatic visual inspection system for wood growing ring identification and measurement is described. Given the irregular contrast distributions in wood samples, the identification of these elements is not straightforward. Furthermore, it is observed that many rings exhibit very narrow widths. To overcome these problems, several new methods are introduced, namely at the local image scale definition and at the image segmentation levels. First, local scales are automatically computed and applied for ring center identification. This information is then used to reconstruct the noisy image signal, which is segmented using a local minimum distance-minimum cross-entropy principle. The system uses a reduced and intuitive set of user defined parameters; for this reason, it is easy to tune.