Comparison of different approaches for automatic bark detection on log images

The sawmilling industry stores and measures logs with bark in order to maximize efficiency, quality conservation and preservation. However since billing is based on the diameter under bark, it is necessary to differentiate between bark areas and wood areas automatically or via manual assignment. This paper compares different methodologies to automatically differentiate between these areas based on colour images of the log surface of two species, spruce and pine. Additionally, the performance of the different methodologies is evaluated using the proportion of correctly detected bark areas, correctly detected wood areas and the total amount of detected bark and wood areas. In the end, an algorithm taking into account colour and texture information was found to perform well on both species. Based on a larger dataset, this methodology has the potential to detect the diameter under bark based on measurements of logs with bark.