Application of image processing and artificial intelligence techniques for the automatic dendrometry of native and commercial wood species

Abstract

Tree-ring analysis is a cornerstone of dendrometry, providing essential information for dendrochronology, forest dynamics, and growth studies. Traditionally, ring marking is performed manually, a process that is time-consuming, subjective, and difficult to scale to large image datasets. Moreover, annual growth measurements are often taken in a onedimensional manner, which complicates comparisons among samples. This thesis proposes the use of ring area as an alternative to ring width, since the latter is a one-dimensional measure that is difficult to standardize. Image-processing algorithms were developed to accurately delineate annual growth curves and calculate growth area in both trees and shrubs. A graphical interface was implemented to combine automatic detection with manual correction tools when needed. Annotated image databases for several species were also created, enabling systematic evaluation of the algorithms. Finally, a case study in ecology?climatology is presented, comparing ring width and ring area as climate indicators.