Development of a wood damage monitoring system for mechanized harvesting

Abstract

Cut-to-length harvesting is a cost-efficient method of the wood supply chain. However, it risks causing stem damage in the mechanized process of thinning forest stands, thereby reducing the growth and technical quality of the remaining trees, which would then be exposed on the increased vulnerability to fungal diseases. For these reasons, it is critical to support quality monitoring of harvesting machines. One way to support quality monitoring is through the application of machine vision solutions. In this study, the damaged stems were photographed systematically from a strip road. The success of the stem-damage detection was analyzed to determine the relationships between successful detection, stand condition, and the image-processing technique. Statistically meaningful relationships were identified via logistic regression analysis, which can be used in selection of tailored image processing technique. The study indicated that the quality-monitoring system of mechanized harvesting could be improved by an increased focus on developing the multi-view photogrammetry of stem damages according to different stand conditions. Further, refining the machine learning system would support the need to determine accurate image-processing thresholds of the texture of stem damages. Then, the overall proportion of successful stem-damage detections will be 89%. These improvements of the quality monitoring system will provide the efficient thinning process in the sustainable wood supply from forests to forest industry. The implementation of such a system could be much broader, initially under Nordic conditions and then in other countries as well, given that its development takes into considerations the significant calibration factors of local conditions.