Flexural behavior of wood in the transverse direction investigated using novel computer vision and machine learning approach

Abstract A deep-learning-based semantic segmentation approach (U-Net) was used to partition the anatomical features in cross-section of hinoki ( Chamaecyparis obtusa ) wood during a micro three-point bending test. Using Crocker–Grier linking algorithm, thousands cells were successfully extracted, and several parameters (area, eccentricity, fitted ellipse aspect ratio, bounding box ratio) evaluate intensity cells’ deformation. Thus, 2D map deformation distribution constructed. By analyzing flat-sawn, quarter-sawn, rift-sawn specimens, it confirmed that annual ring orientation affects flexural behavior transverse direction. The quarter-sawn specimens exhibited largest modulus elasticity (MOE) rupture (MOR). ray tissue aligned against load may have contributed restriction cell smallest MOE MOR, possibly owing loading specimen in-plane off-axial direction, which induced shear wall. For all three types, fracture had high occurrence probability tension part specimen, large Therefore, proposed method can be adapted prediction fractures. With different test species, this great help elucidating relationship between mechanical improve effective utilization resources.