Greyscale-density calibration of an industrial CT scanner for wood microdensitometry

Abstract

ENGLISH ABSTRACT: The microdensitometry of wood, that is the quantification of density in the micrometre scale, is important for different scientific disciplines such as wood science, dendrochronology, dendroecology, dendroclimatology, and biomass determination and in the industrial applications of engineered wood. With the advent of industrial computed tomography scanners, powerful equipment for density measurement is available. However, the methodological framework for wood density determination with those machines is not fully established, because possible error sources have not so far been analysed and described sufficiently. Thus, the objective of this study was the development of a methodology, in order to quantify wood density with an industrial computed tomography scanner. In this study, industrial cone-beam computed tomography was used for the microdensitometry of wood. Fifty stacks of stem discs from an equal number of Pinus radiata trees were used as the sample dataset. Firstly, the fifty stacks were scanned along with selected reference materials, which were measured for density in a conventional way. Then, fifty linear regressions were performed between the conventionally-determined density and the grey value of the reference materials. A strong correlation (mean R2=0.998) between conventionally-determined density and grey value was observed. The regressions also provided calibration equations, which could translate a given grey value to a certain density. For validation purposes, the fifty calibration equations were tested on extra reference materials. The density from the computed tomography scanner had a mean absolute error of 0.008 g/cm3 (σ=0.004) and a mean percent error of 1.4% (σ=0.7). Additionally, one of the calibration equations was applied to the creation of tree ring density profiles. An important result of the study was that it was not possible to deduce a sole calibration equation, as typically done with densitometers and medical computed tomography scanners. Thus, the calibration of each scan was inevitable, due to the various degrees of freedom of the industrial computed tomography scanner.

AFRIKAANSE OPSOMMING: Die mikrodensitometrie van hout, dws die kwantifisering van die digtheid op die mikrometer skaal is belangrik vir verskillende wetenskaplike dissiplines soos houtwetenskap, dendrochronologie, dendro-ekologie, dendro-klimatologie, biomassa bepaling en vir die industriële gebruik van verwerkte hout. Met die koms van die industriële tomografie skandeerders is kragtige toerusting vir die meting van houtdigtheid beskikbaar. Die metodologiese raamwerk vir die bepaling van hout digtheid met die masjiene is egter nog nie ten volle ontwikkel nie. Dit kan aanleiding gee tot metingsfoute wat nie voldoende geanaliseer en beskryf word nie. Die doel van hierdie studie is dus om ‘n metode te ontwikkel om houtdigtheid te kwantifiseer deur die gebruik van ‘n industriële tomografie-skandeerder. In hierdie studie is industriële laser ligstraal tomografie gebruik om die mikrodensitometrie van hout te bepaal. Vir die monster datastel is vyftig stapels van die stam skywe van 'n gelyke aantal Pinus radiata bome gebruik. Eerstens is die vyftig stapels geskandeer saam met geselekteerde verwysingsmateriaal wat vir digtheid gemeet is op 'n konvensionele manier. Daarna is vyftig lineêre regressies ontwikkel tussen die digtheid wat bepaal is deur die konvensionele digtheidsmetode en die grys waarde van die verwysing materiaal. Met die studie is waargeneem dat daar ‘n sterk korrelasie (R² = 0.998) is tussen die digtheid gemeet deur die gebruik van die konvensionele metode en die gryswaarde. Die regressies voorsien ook kalibrasie vergelykings, wat 'n gegewe grys waarde kan assosieer met 'n sekere digtheid. Vir validerings doeleindes, is die vyftig kalibrasie vergelykings getoets op ekstra verwysingsmateriaal. Die digtheid gemeet deur die tomografie-skandeerder het 'n gemiddelde absolute afwyking van 0.008 g/cm³ (σ=0.004) en 'n gemiddelde 1.4% (σ=0.7) fout getoon. Verder is een van die ontwikkelde kalibrasie vergelykings toegepas om boomring digtheidsprofiele te skep. Die studie het bevind dat ‘n enkele kallibrasie vergelyking nie gebruik kan word soos tipies met die gebruik van densitometers en mediese tomografie skandeerders nie. Die industriële tomografie-skandeerder moes dus gekalibreer word met elke skandering as gevolg van die verskillende grade van vryheid wat bevind is in die studie.