智能化精准获取木材基本信息是木材质量追溯系统信息采集的核心， 同时也是木材后期分流、 加工、 精细化利用的重要依据。 旨在探讨基于小波变换的木材近红外光谱（NIRS）去噪及模型优化， 研究近红外技术用于木材质量追溯的基础理论与方法， 构建基于近红外的木材质量追溯体系。 以山杨木材气干密度为例， 采用小波变换进行光谱去噪及模型优化， 应用偏最小二乘法（PLS）构建了基于近红外光谱技术的山杨木材气干密度定标模型。 在此基础上， 集成二维码技术， 在Matlab环境下， 将近红外定标模型预测结果（以木材密度为例）及其他木材相关信息（树种名称、 产地、 采集单位、 数据获取方式等）生成快速响应矩阵码（QR Code）， 以实现木材信息的有效、 快速追溯。 同时研究分析了不同纠错等级、 字符数、 像素下QR码的可读性及有效性。 结果显示： (1)当db5小波基分解层为5时， 经启发式硬阈值去噪后得到的信噪比（SNR）最大， 均方根误差（RMSE）最小， 基于小波变换的近红外光谱去噪效果最好， 将校正模型决定系数由0774 8提高到了0850 1； (2)字符数一定时（本追溯信息的字符数为217）， 当像素大于100 px×100 px时， QR码的可读性均大于90%； 当纠错等级为7%、 像素大小为200 px×200 px时， 随着字符数由100增加至600， 解码率和可读性均为100%。 研究表明， 基于小波去噪及近红外光谱技术， 可以实现木材密度的准确预测， 并能有效集成QR码技术， 以QR码作为传递木材信息的有效载体， 为木材材性信息的无损获取及有效追溯提供了理论依据与技术支撑。

Wood information intelligent acquiring is the key for timber quality tracing. It is also the prerequisite for timber sorting， processing and fine applications. This study aims to discuss the denoising of aspen wood near infrared spectroscopy (NIRS) with wavelet transform and develop the calibration model for wood density， to analyze the feasibility of NIR-based wood quality tracing. In this study， calibration model was developed for air-dry density of aspen wood based on NIRS and partial least squares (PLS) algorithm. Wavelet transform was used for NIR denoising treatment and model optimization. The best denoising method was determined. Aspen wood density predicted with NIR calibration model together with other wood information (species， locality of growth， measuring unit， ways of data acquisition etc. ) was recorded with QR (Quick Response) code for the quick and effective tracing. The readability and effectiveness of the QR code with varied correction levels， number of characters， and pixel sizes were compared. The results showed that： (1) The best model fitting was achieved with the decomposition layer of 5 (db5 wavelet) under the heuristic hard threshold denoising treatment. The determination coefficient (R2) was increased from 0774 8 to 0850 1 for the PLS calibration model. (2) As the number of coded characters were 217 in this study， the readability of QR code was low with pixel size of 100 px×100 px while the QR code readability was higher than 90% with pixel size greater than 100 px×100 px. The readability could be 100% with pixel size of 200 px×200 px and number of coded characters up to 600 at error correction level of 7%. It can be concluded that QR code could be an effective carrier for timber tracing information acquired with NIRS.