Assessment of surface crack- ratio and length of weathered Norway spruce using a new region-based image segmentation method

Abstract Assessing the total crack formation on wood surfaces is important to further our understanding of the wood degradation and aging process and to examine both untreated wood as well as different modification systems. Manually assessing cracks can be both time consuming and difficult to perform accurately without bias and is therefore usually performed according to a rating scheme based on visual assessment. To improve accuracy and generate more possible output parameters, a model to detect and classify surface cracks on naturally- and artificially exposed wood surfaces was developed using region-based image segmentation. The model uses an iterative thresholding approach aimed to provide better results on complex, multi-coloured wood surfaces. The model was able to detect all cracks with widths greater than 0.6 mm in both exposure classes and calculated the total length of the cracks close to those that were measured manually (R 2 = 0.990–0.995). In its current state, the model is best used to calculate the total ratio of the surface that is cracked which allows for analysis of the crack development over time.