Innovative Methodology for Determining Basic Wood Density Using Multispectral Images and MAPIR RGNIR Camera

Basic wood density (BWD) is a critical parameter in the forestry sector, directly influencing yield, operational costs, and final product quality. Conventional methods for BWD determination are often time-consuming, destructive, and costly, necessitating the development of faster and more economical alternatives. In this study, multispectral imaging was employed to estimate the BWD of <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Eucalyptus grandis</i> and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Pinus oocarpa</i>. A MAPIR Survey3 (RGNIR) camera was used to capture images of 142 wood samples at four drying stages: saturated, naturally drying, equilibrium moisture content, and oven-dry. Reflectance data were extracted using ENVI Classic software, while statistical correlations were analyzed in Minitab 16. Machine learning models—Linear Regression, Multilayer Perceptron (MLP), Random Forest (RF), and Support Vector Machine for Regression (SMOreg)—were implemented in Weka 3.8.6. Reflectance in the cross-sectional plane, particularly in the red spectral band, exhibited the highest correlation with BWD. The Random Forest model outperformed other approaches, achieving correlation coefficients exceeding 71% (percentage split) and 65% (cross-validation), with root mean square errors (RMSE) below 14% and 15%, respectively. The MLP, SMOreg, and Linear Regression models also demonstrated satisfactory performance, particularly for saturated wood samples. These findings confirm the viability of multispectral imaging as a rapid, non-destructive, cost-effective, and easily applicable method for BWD estimation in forestry applications.