Mcguinness, B. J., & Duke, M. (2017). Machine vision system grading of pine tree seedlings. Presented at the PA17 International Tri-Conference for Precision Agriculture, Claudelands, Hamilton, New Zealand.
Permanent Research Commons link: https://hdl.handle.net/10289/11802
A PC-based machine vision system for grading pine tree seedlings has been tested at a forestry nursery. The machine has been designed to be implemented in the field at the point of harvesting, removing the need for extra handling steps. The machine measures the height, RCD and root quadrants and makes a decision whether to reject or accept the tree. The grading specification for pine tree cuttings and seedlings appears to be black and white, with clear rules defining whether a tree is acceptable or not; however, the organic nature of the product introduces ambiguity into the decision. Three experts were gathered and asked to independently grade a raw lift of 200 trees with no knowledge of the other experts’ decisions. A consensus was not reached on one in every four trees. The same set of trees was graded by placing them in the machine one by one. The machine achieved 97% agreement with the group of experts, ignoring the decisions on trees where they did not all agree. The machine has been proven to be capable of making decisions on pine seedling quality comparable to that of an expert. It performed well in a shed under controlled conditions; however, the effect of an outdoor environment has yet to be determined.
© 2017 copyright with the authors.