Automated classification of log quality using airborne sound

Assessing the quality of wood is an essential task within the primary wood processing industry. Precise and rapid classification is crucial, not only for optimizing the economics of sawmills but also for the sustainable use of wood as a resource. While expensive solutions are available, they are only profitable at high-throughput sawmills. In smaller companies, logs are currently classified as rotten or healthy through visual inspection by an operator. With a limited number of highly skilled personnel in the sector and the increasing importance of maximizing the material and value obtained from harvested logs, there is a need for new cost-effective solutions. Over the years, considerable research has been conducted on acoustic techniques for classifying logs. The anisotropic structure of wood and non-standardized log sorting at early stages pose significant challenges for using conventional audio-based algorithms. Therefore, this paper explores an approach that uses airborne sound signals to train a machine learning model capable of determining the internal decay of logs that may not be identified visually.