Determination of log moisture content using early-time ground penetrating radar signal

AbstractFast, reliable and non-destructive measurement of log moisture content (MC) is important to optimize the forest value chain. We investigated the use of early-time ground penetrating radar signals to determine MC of stacked logs of black spruce, quaking aspen and balsam poplar in mill yards. Two approaches are presented: a linear fitting between the average envelope amplitude (AEA) and MC and a partial least square (PLS) regression between the signal amplitude and MC. We show that PLS regression enable us to greatly improve the prediction of MC in comparison with the AEA method. Moreover, the PLS technique allows us building models which integrate the signal variability due to the different species and log states (thawed and frozen). Models acquired with measurements collected on the logs ends produced usually higher accuracies (with ranging from 0.67 to 0.95 and root mean square error of prediction [RMSEv] ranging from 6% to 13%) than models acquired with measurements collected through the bark of the logs (with ranging from 0.18 to 0.83 and RMSEv ranging from 7% to 22%). The depth of influence of the ground wave was estimated to be between 8.3 cm and 21.7 cm, with higher penetration in the frozen wood.Keywords: Ground penetrating radarimpulse radarmoisture contentwoodearly-timedirect waveground wavepartial least square (PLS) regressionmultivariate analysis AcknowledgementsThe authors wish to thank FPInnovations and Sensors & Software Inc. for their support. We thank A. Haddadi and C. Lamason from the University of New Brunswick and K. Phung from the University of British Columbia for their help during the experiments. This work was supported by the Natural Sciences and Engineering Research Council of Canada (NSERC) [Strategic Grant STPGP 396789] and the New Brunswick Innovation Foundation (NBIF) [Research Technician Initiative Grant 2012–018].