Determination of log moisture content using ground penetrating radar (GPR). Part 1. Partial least squares (PLS) method

Ground penetrating radar (GPR) is a handheld system showing good potential for the real-time and nondestructive characterization of wood moisture content (MC). However, measurements performed over logs can be challenging because of their curved surface that can affect the GPR signal. In this study, the MC of thawed and frozen logs was estimated for three species (quaking aspen, balsam poplar, and black spruce) using the full GPR signals and the partial least squares (PLS) regression method. The signal was acquired from the cross-section (CS) and through the bark (TB) of the logs with and without an aluminum plate placed under the log. The full GPR signal does not provide better log MC prediction accuracy for small logs compared with the early-time GPR signal. The information about the shape and diameter of the log is contained in the direct and reflected waves of the GPR signal. CS models provided more accurate log MC prediction (RMSE =7–25%) than TB models (RMSE =6–40%) for the hardwood species. Thawed and frozen log models showed similar performances. This study demonstrates that GPR in combination with PLS regression is suitable for predicting log MC in the field.