Investigation of a 'Field Factory' to Harvest and Grade Tree Stock in a Forestry Nursery
McGuinness, B. J. (2018). Investigation of a ‘Field Factory’ to Harvest and Grade Tree Stock in a Forestry Nursery (Thesis, Doctor of Philosophy (PhD)). The University of Waikato, Hamilton, New Zealand. Retrieved from https://hdl.handle.net/10289/12245
Permanent Research Commons link: https://hdl.handle.net/10289/12245
Primary industries are facing an ever increasing labour problem. Major concerns with labour include lack of staff training, high costs, poor efficiency, non-optimal quality control and health and safety issues. While automation is commonplace in factory environments, such technologies have not yet migrated to an outdoor, agricultural environment. Forestry nurseries are no exception, where the most problematic and labour intensive task is lifting and grading tree stock. Mechanical lifting of tree stock is already performed commercially; however, these machines are incapable of performing the additional steps required by this research, particularly root trimming, coupled with a machine vision system that can replicate the human decision making process for selecting ’good’ and ’bad’ tree stock. In particular, there are strict criteria for root structure which must be assessed. Currently, human graders are proving to be poor assessors of this, to such an extent that tree stock is graded up to three times before being shipped to the customer. Additionally, there is the need to remove expensive pack houses. This research investigates a field factory capable of processing forestry tree stock in the field, from lifting through to grading and boxing. The machine vision component of the field factory was tested in controlled conditions, on a sample of 200 trees. There was good agreement between machine vision measurements and manually measured tree features. There is much ambiguity in the grading process, with three experts only reaching a consensus 75% of the time when grading a sample of trees. The machine vision grading system performed very well, showing less bias than human graders. The machine agreed with the specification 96% of the time, significantly higher than the experts’ agreements of between 86 and 90%. While classification systems such as fuzzy logic and artificial neural networks seem to be a good match for this research, they did not outperform the ’crisp’ grading system. A field factory for harvesting and grading forestry tree stock proved to be feasible; however, further development, particularly on mechanical systems, is required to produce a machine reliable enough to be implemented commercially.
The University of Waikato
All items in Research Commons are provided for private study and research purposes and are protected by copyright with all rights reserved unless otherwise indicated.
- Higher Degree Theses