Researchers have investigated whether it is possible to use AI to identify tree species based on images of the bark. After collecting and analysing over 14,000 photos of spruce and pine trunks, a first AI model has been trained.
The images were collected at a sawmill in northern Sweden and linked to tree data from the sawmill’s log files. After a quality check, the dataset amounted to over 9,000 images of spruce and 5,000 images of pine trees. AI has since been trained to determine tree species based solely on the appearance of the bark. The project was carried out by Skogforsk, with support from Mistra Digital Forest and Norra Skogs Research Foundation.
Morgan Rossander, Skogforsk.
– The initial AI model shows promising results. More studies are needed to evaluate how the model performs in different situations, for example on trees from other geographical areas and under different conditions. This could involve different photo angles, backgrounds, light and weather conditions, says Morgan Rossander, a researcher at Skogforsk.
One possible future area of application would be to integrate similar technology into the harvester head, in order to automate tree species identification during harvesting. But first, more understanding is needed about the extent to which this is technically and financially feasible.
– To promote further development and innovation, Skogforsk is making this extensive image data available to forestry companies, machine manufacturers, researchers and other forestry stakeholders, says Morgan Rossander.