Background
The Precision Silviculture Programme (PSP) is a partnership between the Ministry for Primary Industries and the forest industry, led by Forest Growers Research Ltd in New Zealand. PSP aims to transform forest management practices by applying sensing technologies, mechanisation, and automation across forest management operations. By doing so, the Programme aims to address labour shortages, improve safety across all forestry operations, reduce costs in the forestry value chain, maintain a pruned log supply, enhance productivity, and maintain a licence to operate through sustainable practices, including reduced chemical inputs. The Scion Group within the Bioeconomy Science Institute for Forest Growers Research Ltd (FGR) recently released a report on Vegetative precision propagation – 2.1 imaging technologies (final report of indoor and outdoor radiata pine seedling base models authored by D Pont, D. Siqueira, G. Salvatierra, M Palmer, P Massam.
Key highlights from the report
1. Indoor Seedling Monitoring (Containerized Growing Zones):
o AI Model Development: A bespoke AI model named "Pine-Eye" was developed in collaboration with Corvus Drones and Track32 to detect and characterize conifer seedlings from RGB imagery. Initial manual image training showed promising results, with the model improving rapidly to achieve 87.5% germination detection accuracy (matching actual counts) by June 2025, up from 8% in May 2025.
o Drone-as-a-Service (DaaS) System: A mini-drone system was set up and tested for autonomous flights within greenhouses. This "drone-in-a-box" solution integrates IoT technologies for automated data transfer and cloud-based AI processing.
o Safety Protocols: Comprehensive safety protocols, compliant with CAA and WorkSafe NZ guidelines, were established for mini-drone operations.
o Future Integration: Once "Pine-Eye" reaches consistent reliability, it will be integrated into the DaaS dashboard to provide real-time germination monitoring, plant counts, and facilitate early interventions in containerized nurseries.
2. Outdoor Stool Bed Production Optimization:
o Data Capture: UAVs equipped with RGB cameras and LiDAR were used to capture high-resolution imagery and topographical data of outdoor radiata pine stool beds. This provided a baseline for understanding stool bed growth and productivity.
o Outdoor AI Seedling Detector: A deep learning-based AI model (Mask R-CNN architecture) was developed to detect and assess radiata pine mother plants in the outdoor nursery environment.
o Performance: The outdoor model achieved high accuracy, with an average precision (AP) score of 0.96 for February imagery (sunny conditions), demonstrating reliable detection and segmentation of seedlings. Performance was lower in cloudy conditions (May imagery) due to challenges with background vegetation, highlighting areas for further training.
o Baseline Data: The collected data and developed AI model provide a valuable reference for future comparisons and research aimed at optimizing stool bed production.
Summary and future plans
The project successfully demonstrated the positive application of novel UAV and AI technologies for seedling detection in both indoor and outdoor environments. For indoor environments, the "Pine-Eye" model and DaaS system offer a powerful tool for efficient, real-time monitoring and decision-making. For outdoor stool beds, the developed AI model and data collection methods lay the groundwork for future optimization efforts. Next Steps (next 12 months): Further refine the "Pine-Eye" model for indoor use by gathering more diverse training data (various seedling stages, lighting, and presence of weeds), conduct regular mini-drone flights, and explore industry trials. For outdoor stool beds, augment remote sensing data with industry operational data to optimize production systems.
For more information, visit:
The full report can be accessed from Precision Silviculture Programme
