Machine-Learning-Based Identification Method for <italic style="font-style: italic">Quercus</italic> Wood Species Using Direct Analysis in Real Time Mass Spectrometry

LIU Shoujia; ZHAN Weihui; LI Tianxiao; JIAO Lichao; YIN Yafang

doi:10.12326/j.2096-9694.2026013

RESEARCH PAPERS | Views : 0 下载量: CSCD: 0 CNKI被引量: 0 Crossref: 0 Wos: 0 Scopus: 0 更多指标

PDF
Export
Share
Collection
Album

Machine-Learning-Based Identification Method for Quercus Wood Species Using Direct Analysis in Real Time Mass Spectrometry
Research Institute of Wood Industry，Chinese Academy of Forestry
Wood Collection of Chinese Academy of Forestry
Wood Specimen Resource Center of National Forestry and Grassland Administration
CITES Wildlife Forensic Testing Laboratory （Timber and Timber Products），Beijing 100091，China
Vol. 40, Issue 3, Pages: 20-29(2026)
DOI： 10.12326/j.2096-9694.2026013
稿件说明：

移动端阅览

刘守佳,詹伟辉,李天笑等.实时直接分析质谱结合机器学习的栎属木材树种识别方法研究[J].木材科学与技术,2026,40(03):20-29. DOI： 10.12326/j.2096-9694.2026013.

LIU Shoujia,ZHAN Weihui,LI Tianxiao,et al.Machine-Learning-Based Identification Method for Quercus Wood Species Using Direct Analysis in Real Time Mass Spectrometry[J].Chinese Journal of Wood Science and Technology,2026,40(03):20-29. DOI： 10.12326/j.2096-9694.2026013.

摘要

高效提取木材特征信息从而建立科学准确的木材树种识别方法，对推动生物多样性保护、林木资源可持续利用等具有重要现实意义。然而现有遗传学、计算机视觉等方法仍难以满足快速精准应用场景下的树种鉴定需求。以构造特征高度相似的3种栎属木材为研究对象：美国白栎（

Quercus alba

）、蒙古栎（

Q. mongolica

）和夏栎（

Q. robur

），对其开展树种识别方法研究。采用实时直接分析-飞行时间质谱技术（Direct Analysis in Real Time-Time-of-Flight Mass Spectrometry，DART-TOF-MS）获取木材化学指纹图谱，通过构建逻辑回归、多层感知机、随机森林和K-近邻算法等4种机器学习方法分析木材质谱数据，评估不同机器学习方法在树种识别中的可行性，同时比较质谱数据预处理参数对识别模型性能的影响。研究结果表明：逻辑回归算法具有最高的树种识别准确率（达89.5%），性能优于其他3种模型；不同机器学习方法在处理复杂质谱特征时表现出较明显差异，合理的分箱参数设置对模型性能具有关键影响，在相对峰强度阈值为3%、分箱容差为20 mmu时模型整体表现最优；3种栎属木材在木质素单体及其衍生物组成上具有高度相似性，但在黄酮类化合物、木质素二聚体及鞣花酸的空间分布方面存在差异。DART-MS结合机器学习方法对3种栎属木材呈现出较强的识别能力，可为木材树种的精准识别提供科学依据和技术路径。

Abstract

Establishing scientific and accurate wood species identification methods through the efficient extraction of characteristic information from different wood species holds significant practical importance for promoting biodiversity conservation and the sustainable utilization of forest resources. However

existing techniques such as computer vision and DNA barcoding are still insufficient to meet the demand for rapid and accurate species discrimination

underscoring the need for methodological innovation in wood identification. In this study

three morphologically similar

Quercus

species common in trade:

Quercus alba

Quercus mongolic

and

Quercus robur

which are difficult to differentiate using conventional anatomical methods

were selected as investigated species to acquire spectral fingerprint data using Direct Analysis in Real Time-Time-of-Flight Mass Spectrometry (DART-TOF-MS). Four machine learning algorithms

including logistic regression

multilayer perceptron (MLP)

random forest(RF)

and k-nearest neighbors (KNN)

were then applied to DART-TOF-MS data to assess their feasibility for species classification. We further investigated the influence of different spectral preprocessing parameters on model performance. The results showed that the logistic regression model achieved the highest classification accuracy (89.5%)

outperforming the other three algorit

hms. Distinct differences were observed among the four machine learning approaches in their ability to handle complex spectral features

and appropriate binning strategies-particularly threshold control and mass tolerance selection-were found to be critical for optimal model performance

with the best results obtained at a parameter setting of 3% relative threshold and 20 mmu binning tolerance. Although the three

Quercus

species exhibited similarities in lignin monomers and their derivatives

significant differences were detected in the distribution of flavonoids

lignin dimers

and ellagic acid. The present results demonstrated that DART-TOF-MS combined with machine learning exhibited a strong capability for identifying the three

Quercus

wood species examined in this study

providing a scientific basis and technical pathway for the accurate identification of wood species.

关键词

Keywords

references

LIU S J , HE T , WANG J J , et al . Can quantitative wood anatomy data coupled with machine learning analysis discriminate CITES species from their look-alikes [J ] . Wood Science and Technology , 2022 , 56 ( 5 ): 1567 - 1583 .

JIAO L C , YU M , WIEDENHOEFT A C , et al . DNA barcode authentication and library development for the wood of six commercial Pterocarpus Species: the critical role of xylarium specimens [J/OL ] . Scientific Reports , 2018 , 8 : 1945 . https://doi.org/10.1038/s41598-018-20381-6 https://doi.org/10.1038/s41598-018-20381-6 .

JIAO L C , LU Y , HE T , et al . DNA barcoding for wood identification: global review of the last decade and future perspective [J ] . IAWA Journal , 2020 , 41 ( 4 ): 620 - 643 .

CODY R B , LARAMEE J A , DURST H D . Direct analysis in real time (DART) mass spectrometry of solid materials, surfaces, and liquids [J ] . Journal of the American Society for Mass Spectrometry , 2005 , 16 ( 4 ): 681 - 685 .

张毛毛 , 蒋劲东 , 刘波 , 等 . 实时直接分析质谱技术在木材识别研究中的应用 [J ] . 木材工业 , 2019 , 33 ( 1 ): 29 - 33 .

ZHANG M M , JIANG J D , LIU B , et al . Application of direct analysis in real time mass spectrometry (DART-MS) for wood identification [J ] . Wood Industry , 2019 , 33 ( 1 ): 29 - 33 .

ZHANG M M , ZHAO G J , LIU B , et al . Wood discrimination analyses of Pterocarpus tinctorius and endangered Pterocarpus santalinus using DART-FTICR-MS coupled with multivariate statistics [J ] . IAWA Journal , 2019 , 40 ( 1 ): 58 - 74 .

RAVINDRAN P , WIEDENHOEFT A C . Comparison of two forensic wood identification technologies for ten Meliaceae woods: computer vision versus mass spectrometry [J ] . Wood Science and Technology , 2020 , 54 ( 5 ): 1139 - 1150 .

ESPINOZA E O , LANCASTER C A , KREITALS N M , et al . Distinguishing wild from cultivated agarwood ( Aquilaria spp.) using direct analysis in real time and time of-flight mass spectrometry [J ] . Rapid Communications in Mass Spectrometry , 2014 , 28 ( 3 ): 281 - 289 .

PARK G , LEE Y G , YOON Y S , et al . Machine learning-based species classification methods using DART-TOF-MS data for five coniferous wood species [J/OL ] . Forests , 2022 , 13 ( 10 ): 1688 . https://doi.org/10.3390/f13101688 https://doi.org/10.3390/f13101688 .

DEKLERCK V , FINCH K , GASSON P , et al . Comparison of species classification models of mass spectrometry data: Kernel discriminant analysis vs random fo rest; a case study of Afrormosia ( Pericopsis elata (Harms) Meeuwen) [J ] . Rapid Communications in Mass Spectrometry , 2017 , 31 ( 19 ): 1582 - 1588 .

DEKLERCK V , MORTIER T , GOEDERS N , et al . A protocol for automated timber species identification using metabolome profiling [J ] . Wood Science and Technology , 2019 , 53 ( 4 ): 953 - 965 .

ZHANG M M , ZHAO G J , GUO J , et al . Timber species identification from chemical fingerprints using direct analysis in real time (DART) coupled to Fourier transform ion cyclotron resonance mass spectrometry (FTICR-MS): comparison of wood samples subjected to different treatments [J ] . Holzforschung , 2019 , 73 ( 11 ): 975 - 985 .

路佳佳 . 基于交叉验证的集成学习误差分析 [J ] . 计算机系统应用 , 2023 , 32 ( 1 ): 302 - 309 .

LU J J . Error analysis of ensemble learning based on cross validation [J ] . Computer Systems & Applications , 2023 , 32 ( 1 ): 302 - 309 .

LIU S J , ZHENG C , WANG J J , et al . How to discriminate wood of CITES-listed tree species from their look-alikes: using an attention mechanism with the ResNet model on an enhanced macroscopic image dataset [J/OL ] . Frontiers in Plant Science , 2024 , 15 : 1368885 . https://doi.org/10.3389/fpls.2024.1368885 https://doi.org/10.3389/fpls.2024.1368885 .

IAWA Committee . IAWA list of microscopic features for hardwood identification [J ] . IAWA Bulletin , 1989 , 10 ( 3 ): 219 - 332 .

周志华 . 机器学习 [M ] . 北京 : 清华大学出版社 , 2016 : 56 - 62 .

邱锡鹏 . 神经网络与深度学习 [M ] . 北京 : 机械工业出版社 , 2020 : 89 - 105 .

冯晓荣 , 瞿国庆 . 基于深度学习与随机森林的高维数据特征选择 [J ] . 计算机工程与设计 , 2019 , 40 ( 9 ): 2494 - 2501 .

FENG X R , ZHAI G Q . Feature selection for high-dimensional data based on deep learning and random forest [J ] . Computer Engineering and Design , 2019 , 40 ( 9 ): 2494 - 2501 .

李驰 , 段雨梅 . K近邻算法优化设计策略 [J ] . 电脑知识与技术 , 2019 , 15 ( 31 ): 200 - 202, 211 .

LI C , DUAN Y M . Optimal design strategy of K-nearest neighbor algorithm [J ] . Computer Knowledge and Technology , 2019 , 15 ( 31 ): 200 - 202, 211 .

TRAORÉ M , MARTÍNEZ CORTIZAS A . Color and chemical composition of timber woods ( Daniellia oliveri , Isoberlinia doka , Khaya senegalensis , and Pterocarpus erinaceus ) from different locations in southern Mali [J/OL ] . Forests , 2023 , 14 ( 4 ): 767 . https://doi.org/10.3390/f14040767 https://doi.org/10.3390/f14040767 .

SJOSTROM E . Wood chemistry: fundamentals and applications [M ] . Gulf Professional Publishing , 1993 .

OSAKABE K , TSAO C C , LI L G , et al . Coniferyl aldehyde 5-hydroxylation and methylation direct syringyl lignin biosynthesis in angiosperms [J ] . Proceedings of the National Academy of Sciences of the United States of America , 1999 , 96 ( 16 ): 8955 - 8960 .

PUECH J L . Extraction of phenolic compounds from oak wood in model solutions and evolution of aromatic aldehydes in wines aged in oak barrels [J ] . American Journal of Enology and Viticulture , 1987 , 38 ( 3 ): 236 - 238 .

WEI L M , MA R K , FU Y L . Differences in chemical constituents between Dalbergia oliveri heartwood and sapwood and their effect on wood color [J/OL ] . Molecules , 2022 , 27 ( 22 ): 7978 . https://doi.org/10.3390/molecules27227978 https://doi.org/10.3390/molecules27227978 .

CODY R B , ESPINOZA E O , PRICE E R , et al . Wood from hardwood angiosperms and coniferous gymnosperms shows distinctive lignin peaks in direct analysis in real time (DART) mass spectra [J ] . Journal of the American Society for Mass Spectrometry , 2023 , 34 ( 4 ): 784 - 789 .

CHATONNET P , BOIDRON J N . Identification and determination of volatile compounds released by oak wood used for aging of wines [J ] . American Journal of Enology and Viticulture , 1989 , 40 ( 1 ): 76 - 83 .

QUIDEAU S , DEFFIEUX D , DOUAT-CASASSUS C , et al . Plant polyphenols: chemical properties, biological activities, and synthesis [J ] . Angewandte Chemie International Edition , 2011 , 50 ( 3 ): 586 - 621 .

DEKLERCK V , LANCASTER C A , VAN ACKER J , et al . Chemical fingerprinting of wood sampled along a pith-to-bark gradient for individual comparison and provenance identification [J ] . Forests , 2020 , 11 ( 1 ): 107 .

ZHANG M M , GUO J , LU Y , et al . Similarity network fusion for aggregating headspace GC-MS and direct analysis in real time-mass spectrometry data from solid samples to enhance species identification efficiency of high-temperature heated wood [J/OL ] . Journal of Wood Science , 2022 , 68 : 1 - 13 . https://doi.org/10.1186/s10086-022-02044-3 https://doi.org/10.1186/s10086-022-02044-3 .

Alert me when the article has been cited

提交

Machine Learning Prediction of Mechanical Properties of Larix gmelinii var. principis-rupprechtii Wood by Integrating Multi-Source Features of Anatomy, Physics and Non-Destructive Testing

Identification and Comparison of Seven “Walnut Woods”

Development of Online Retrieval System Based on the IAWA List of Microscopic Features for Softwood Identification

Application of Gas Chromatography-mass Spectrometry for Wood Identification

Preparation and Performance of Room Temperature Phosphorescent Cellulose-based Transparent Wood

Machine-Learning-Based Identification Method for Quercus Wood Species Using Direct Analysis in Real Time Mass Spectrometry

DOI： 10.12326/j.2096-9694.2026013

摘要

Abstract

关键词

Keywords

references