Lari Melander: Machine learning helps forest machines to understand forest environment
The forest resource data in Finland has very good coverage through the country, is accurate and mostly publicly available. Furthermore, the forest machines in Finland, logging machines in particular, have very high degree of automation. In his dissertation, Melander investigated how these two assets could be used together for producing additional value for forestry applications.
“Forest resource data has been utilized for planning in forest management for many years, but not so much for optimizing the work of the forest machine in the forest. However, it would be beneficial to understand how changing forest environment affects the performance of the forest machine and its operator”, Melander says.
Melander proposes a method for the data fusion of the forest resource data and the fieldbus data of the forest machines. Machine learning methods can then be easily applied for the resulting dataset for searching interesting associations between the performance of the machine, its operator, and the forest environment. The results in the dissertation study suggest that Finnish forests should be clustered based on the forest resource data for representing different combinations of tree and soil attributes. Such clusters enable comparisons of the forest machines and their operators to each other, as every cluster has a national reference.
In addition, Melander suggests methods for directly perceiving the forest environment. His dissertation proposes techniques for measuring the wheel rut after a vehicle passage and stoniness of the forest soil while the forest machines are conducting their main tasks. Currently, forests are being surveyed or measured from the airplanes for example, but according to Melander, forests machines should be considered as moving sensors in the forest.
“Due to their high purchase price, forest machines spend considerable time in the forests. Thus, measuring the forest environment while working is both cheap and efficient way of producing new or updating existing information about the forest”, says Melander.
Both improved sensing capabilities and fusion of existing datasets enable better planning of the forest operations. The aim is that right things are done at the right time, and for example biggest forest damages can be avoided.
The methods presented in the doctoral thesis promote so called precision forestry. Precision forestry refers to forest planning that utilizes novel technologies and very detailed data collected from forests. Precision forestry can improve the performance of the logging operations, but can also help preventing global climate change and improving forest biodiversity.
The doctoral dissertation of M.Sc. Lari Melander in the field of automation technology titled Towards Precision Forestry: Methods for Environmental Perception and Data Fusion in Forest Operations will be publicly examined at the Faculty of Engineering and Natural Sciences of Tampere University at 12 o’clock on Friday 5 March 2021. The opponents will be Professor Ola Lindroos from Swedish University of Agricultural Sciences and Research Professor Annika Kangas from Natural Resources Institute Finland. The custos will be Professor Risto Ritala from the Faculty of Engineering and Natural Sciences at Tampere University.
Due to the corona virus situation, the public can follow the event via remote connection only.
The dissertation is available online at https://trepo.tuni.fi/handle/10024/124745
Photo: Kaarina Melander