Detection of Wood Cross-Section Regions with Growcut for Measurement of Wood Diameter Grade

The Japanese forestry industry as a whole has been stagnant and needs to be revitalized. In this study, the timber yard is focused on and image processing is used to measure wood diameter grade to improve the efficiency of measurement work. To detect the cross-section regions of wood, GrowCut, which uses not only pixel color information but also positional information is employed. The detection accuracy is improved by changing the seed pixel detection method and the initial confidence level of the seed pixel to automate the process. In computer experiments, the maximum IoU of the detection result in the cross-section regions was 0.946.