NIR And QR Code Of Aspen Wood For The Timber Traceability

Populus davidiana wood density and water content prediction model was developed with Near Infrared Spectroscopy(NIRS)and Partial Least Squares(PLS)method.Three wavelet functions of Haar?Daubechies?Symlet were used for NIR denoising.The optimal parameters were determined for the best denoising results.NIR-based wood information procurement was then integrated with Quick Response(QR)Code for wood tracing.The effects of error correction level,number of characters,and pixel size on the QR decoding time and efficiency were discussed for a brand new way of wood tracing.(1)Collecting the NIRS of aspen wood with the bifurcate optical fiber probe,single and combo NIR models for wood density/moisture content prediction were developed with PLS.R2 for calibration was 0.8364,0.8448,and 0.8394 for the density,moisture content,and combo model,respectively.For the single and combo model,R2was 0.8327 and 0.8300 for density prediction while it was 0.8545 and 0.8498 for moisture content prediction.Results showed that combo NIR model could improve the prediction efficiency without sacrificing model accuracy.(2)Denoising pretreament to NIRS was implemented with three wavelet packets.When Haar scale was 3,the denoising effect based on the "SURE" standard entropy and balance sparasity-norm hard thresholding was the best.Using dbN for wavelet decomposition,denoising effect based on "SURE" entropy db9,Hard&Balance sparasity-norm was the best with SNR of 25.7689 and RMSE of 0.0160.For SymN wavelet,when the decomposition level is 4 and entropy standard is "SURE",the denoising effect based on Sym4 balanced sparasity-norm standard hard wavelet threshold is the best.It indicated that the denoising results with Sym4 wavelet function is the best.The NIR-based wood density and moisture content model was then developed with the R2of 0.9109 for moisture content calibration model and 0.8955 for the combo model.Compared to the raw model,the R2 improved 0.0657 for moisture content prediction and 0.0733 for density prediction.It indicates that denoising with wavelet packet could effectively eliminate the redundant information while keeping the typical information and improve model efficiency and accuracy.(3)NIR-based combo model was developed for the prediction of wood density,moisture content simultaneously.In combination with Quick Response(QR)Code technology,the wood related information including aspen collection locations,collection units,density and moisture content,etc.)were integrated into QR code as its "identity label".With the same electronic equipment,decoding was implemented to varied error correction level,varied number of characters,and varied pixel size based QR Code.Results showed that number of characters and error correction level have little effects on decoding time and eficiency.QR Code has low readability with pixel size of 100*100px while the highest readability of 30%was achieved with error correction level of 7%.QR Code has low readability with pixel size of 100*100px.With pixel size of 150-600px,the readability is 100%with except that readability of 90%when pixel size was 300x300px at error correction of 30%.Generally,with the integration of NIR and QR Code technology,it could provide a new and green way for wood quality tracing.