Comparison of two forensic wood identification technologies for ten Meliaceae woods: computer vision versus mass spectro
- PDF / 860,949 Bytes
- 12 Pages / 439.37 x 666.142 pts Page_size
- 75 Downloads / 146 Views
Comparison of two forensic wood identification technologies for ten Meliaceae woods: computer vision versus mass spectrometry Prabu Ravindran1 · Alex C. Wiedenhoeft1,2,3,4 Received: 31 December 2019 © This is a U.S. Government work and not under copyright protection in the US; foreign copyright protection may apply 2020
Abstract A wealth of forensic wood identification technologies has been developed or improved in recent years, with many attempts to compare results between technologies. The utility of such comparisons is greatly reduced when the species tested with each technology are different and when performance metrics are not calculated or presented in the same way. Here, a species-level XyloTron computer vision model is presented along with a side-by-side comparison for species- and genus-level identification of the 10 species of Meliaceae studied by Deklerck et al. using mass spectrometry. The species-level accuracies of the XyloTron model and the mass spectrometry models are comparable, while the genus-level accuracy of the XyloTron model is higher than that of the mass spectrometry model. The paper concludes with a call for better practices to compare disparate forensic wood identification technologies from a performance driven perspective.
* Prabu Ravindran [email protected] Alex C. Wiedenhoeft [email protected] 1
Department of Botany, University of Wisconsin, Madison, WI 53706, USA
2
Forest Products Laboratory, Center for Wood Anatomy Research, USDA Forest Service, Madison, WI 53726, USA
3
Department of Forestry and Natural Resources, Purdue University, West Lafayette, IN 47907, USA
4
Ciências Biológicas (Botânica), Universidade Estadual Paulista–Botucatu, São Paulo, Brazil
13
Vol.:(0123456789)
Wood Science and Technology
Introduction A wide range of contexts drives the need for scientifically rigorous forensic wood identification technologies, including the identification of cultural property (Ruffinatto et al. 2010; Ostapkowicz et al. 2017; Guo et al. 2019), determination of structure–property wood technology relationships (Wiedenhoeft and Kretschmann 2014), analysis of evidence from criminal forensic contexts (Graham 1997), and investigation of forest products supply chains (Wiedenhoeft et al. 2019). With the sustained interest in enforcing national and international laws and treaties to control trade in endangered species and ensure legal timber trade, evaluating the validity, reliability, and contextual applicability of various wood forensic techniques is more important than ever. The primary scientific questions for any forensic wood identification technology are geographic origin and botanical identification. The former question has been primarily addressed by stable isotope methods (Kagawa and Leavitt (2010) and reviewed in Meier-Augenstein (2019)], DNA-based methods (e.g., Degen et al. 2013; Vlam et al. 2018), and chemometric methods (Bergo et al. 2016; Finch et al. 2017; Ma et al. 2018), with recent work showing newfound applicability of wood anatomy for this que
Data Loading...
