In production forestry, radiata pine is grown at scale using both containerised and bareroot production systems. Seedlings are typically lifted, graded, and selected manually using standard measurements. With labour availability tightening and demand increasing, machine‑based harvesting and selection technologies are being developed to increase throughput and improve data consistency for research and operational decision‑making.
The Precision Silviculture Programme engaged Tui Technology and Contempo Lab to design and build a prototype conveyor system with integrated AI grading capability and graphical operator control software. Validation was performed with Appletons Tree Nursery, and feedback was used to make improvements to the prototype.
The newly developed operator software runs the computer‑vision AI and seedling rules engine. It allows an operator to control conveyor speed and tune parameters for maximising seedling yield under varying shed lighting conditions. The interface displays live computer‑vision output, including object and segment detection, and provides batch reporting, user controls, several operational modes, and a final daily report option.
Building the model involved continuously running seedlings through the system as either “normal” or “abnormal,” with varied spacing and orientation. Three types were included: bareroot seedlings, bareroot cuttings, and containerised seedlings. Height measurements were accurate to within 10 mm, and root‑collar diameter (RCD) to within 0.5 mm when checked against vernier callipers. Root assessments were visually consistent, and colour‑based pass/fail outcomes aligned with nurseryman judgement.
The project succeeded in building a proof-of-concept conveyor with computer vision built into operational software for the operator. Results showed that a single camera should be acceptable. Shed tests and industry review found the performance acceptable with some suggestions for future improvements.
Such enhancements will likely include a singulation mechanism and consideration for inclusion of bunching and packing mechanism within an integrated, mobile solution to improve mobility and ease of demonstration. Further measurement development is required, including crooked stem assessment, bareroot length, and containerised root area. In the future the same model could be implemented into a bareroot lifting machine in the field.
