This project was built on progress made in the initial Virtual Reality (VR) work explored in FWPA funded project (PNC464-1718). The project developed VR methods and workflows to support VR visualisation and measurement of plot-level point-cloud data in forest inventory. Additionally, this project was built on progress in the development of automated point cloud processing algorithms for individual tree detection and segmentation developed in FWPA-funded research (PNC377-1516) and National Institute for Forest Product innovation (NIFPI)-funded project (NIF073-1819). This project investigated the concept of the Human-in-the-loop (HITL) framework, where VR users directly collaborate with the automated Machine Learning (ML) point cloud processing algorithms.
The project provides the following key outputs:
Human-in-the-loop (HITL) framework guiding the interaction between VR users and Machine Learning (ML) algorithms.
- User evaluation of the tree assessment accuracy using different visual cues of tree characteristics computed by machine learning algorithms.
- VR point cloud software capable of rendering a high-density plot-level point cloud dataset such as those scanned by HoverMap TLS sensor.
- ML-based VR Software and instructions with a trained knowledge of stem segmentation (using the training dataset) and a knowledge retraining capability.
- Training workshop and presentations to disseminate the uses of the VR-ML tools and knowledge gained during the project.
These outputs help increase the capacity and understanding of forest growers to utilise high-density point cloud data in forest inventory. The HITL provides a framework for VR users to collaborate with the ML engine in tree assessments. The framework allows human field operators to interchange knowledge (of stem segmentation) with machine-learning algorithms. The ML algorithms developed in this project demonstrated the continuing advancement of the ML-based approach, which will significantly enhance and complement how humans assess tree characteristics. Combining VR human operators and ML algorithms could improve forest inventory practices.
