<p>’Tracheid effect’, or laser light scattering in wood is an important phenomenon in the study of wood, but not well researched for Malaysian timber species. Sixty (60) common commercial timber species of Malaysia, as defined by the Forest Research Institute Malaysia (FRIM), were tested for ’tracheid effect’ response using 650 nm (red) laser. 5 samples each of 11 heavy hardwood, 15 medium hardwood, 32 light hardwood, and 2 softwood species were tested. Medium density fibreboard (MDF) was used as a reference. The grain angle performance metric, which is the root mean squared error (RMSE) between the observed and actual angles of ellipse generated by the dot laser and imaged by the camera, was determined and tabulated. To predict the grain angle performance for unknown species using colour and density features, two machine learning (ML) classification approaches were tested, namely k-Nearest Neighbour (k-NN), and shallow feed-forward artificial neural network (ANN), as well as one function-fitting ANN. For predicting RMSE <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="226_2025_1653_Article_IEq1.gif" Format="GIF" Height="14" Rendition="HTML" Resolution="72" Type="Linedraw" Width="35" /> </InlineMediaObject> <EquationSource Format="TEX">$<5^\circ$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mo><</mo> <msup> <mn>5</mn> <mo>∘</mo> </msup> </mrow> </math></EquationSource> </InlineEquation>, the function-fitting ANN performed the best at 82.7%, while k-NN scored the highest overall performance of 89.3% when predicting RMSE <InlineEquation ID="IEq2"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="226_2025_1653_Article_IEq2.gif" Format="GIF" Height="14" Rendition="HTML" Resolution="72" Type="Linedraw" Width="43" /> </InlineMediaObject> <EquationSource Format="TEX">$<10^\circ$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mo><</mo> <msup> <mn>10</mn> <mo>∘</mo> </msup> </mrow> </math></EquationSource> </InlineEquation>. The density of wood did not directly correlate with the grain angle performance, but its inclusion as a feature together with the colour features improved the accuracy of the ML predictions. The colour features related to brightness were dominant features that affected performance. In summary, this study confirmed that wood colour as well as density plays an important role in the ability to determine grain angle by means of the tracheid effect using 650 nm lasers.</p>

‘Tracheid effect’ light scattering response of common commercial species of Malaysia using red laser (650 nm) for grain angle detection, and machine learning based performance prediction

期刊论文
出版模式：

开放获取
发表日期：

2025年5月7日

查看全文

Chiat Oon Tan,
Shigenobu Ogata,
Hwa Jen Yap,
Jin Xue Soo,
Zuriani Usop,
Mohd ’Akashah Fauthan,
Shaer Jin Liew,
Siew-Cheok Ng

摘要

’Tracheid effect’, or laser light scattering in wood is an important phenomenon in the study of wood, but not well researched for Malaysian timber species. Sixty (60) common commercial timber species of Malaysia, as defined by the Forest Research Institute Malaysia (FRIM), were tested for ’tracheid effect’ response using 650 nm (red) laser. 5 samples each of 11 heavy hardwood, 15 medium hardwood, 32 light hardwood, and 2 softwood species were tested. Medium density fibreboard (MDF) was used as a reference. The grain angle performance metric, which is the root mean squared error (RMSE) between the observed and actual angles of ellipse generated by the dot laser and imaged by the camera, was determined and tabulated. To predict the grain angle performance for unknown species using colour and density features, two machine learning (ML) classification approaches were tested, namely k-Nearest Neighbour (k-NN), and shallow feed-forward artificial neural network (ANN), as well as one function-fitting ANN. For predicting RMSE $<5^\circ$ $< 5^{\circ}$ , the function-fitting ANN performed the best at 82.7%, while k-NN scored the highest overall performance of 89.3% when predicting RMSE $<10^\circ$ $< 10^{\circ}$ . The density of wood did not directly correlate with the grain angle performance, but its inclusion as a feature together with the colour features improved the accuracy of the ML predictions. The colour features related to brightness were dominant features that affected performance. In summary, this study confirmed that wood colour as well as density plays an important role in the ability to determine grain angle by means of the tracheid effect using 650 nm lasers.