Owens, Frank C.
Date of Degree
Dissertation - Open Access
Doctor of Philosophy (Ph.D)
College of Forest Resources
Department of Sustainable Bioproducts
Globally, illegal logging poses a significant threat. This results in environmental damage as well as lost profits for legitimate wood product producers and taxes for governments. A global value of $30 to $100 billion is estimated to be associated with illegal logging and processing. Field identification of wood species is fundamental to combating species fraud and misrepresentation in global wood trade. Using computer vision wood identification (CVWID) systems, wood can be identified without the need for time-consuming and costly offsite visual inspections by trained wood anatomists. While CVWID research has received significant attention, most studies have not considered the generalization capabilities of the models by testing them on a field sample, and only report overall accuracy without considering misclassifications. The aim of this dissertation is to advance the design and development of CVWID systems by addressing three objectives: 1) to develop functional, field-deployable CVWID models for Peruvian and North American hardwoods, 2) test the ability of CVWID to solve increasingly challenging problems (e.g., larger class sizes, lower anatomical diversity, and spatial heterogeneity in the context of porosity), and 3) to evaluate the generalization capabilities by testing models on independent specimens not included in training and analyzing misclassifications. This research features four main sections: 1) an introduction summarizing each chapter, 2) a chapter (Chapter 2) developing a 24-class model for Peruvian hardwoods and testing its generalization capabilities with independent specimens not used in training, 3) a chapter (Chapter 3) on the design and implementation of a continental scale 22-class model for North American diffuse-porous hardwoods using wood anatomy-driven model performance evaluation, and 3) a chapter (Chapter 4) on the development of a 17-class models for North American ring-porous hardwoods, in particular examining the model's effectiveness in dealing with the greater spatial heterogeneity of ring-porous hardwoods.
Wade, Adam Carter, "Advancement of field-deployable, computer-vision wood identification technology" (2022). Theses and Dissertations. 5573.