Forests provide natural habitats, opportunities for recreation and a sustainable resource base for the forestry sector – one of the largest primary industries in Australia and New Zealand.
We're working with our industry and academic research partners to develop new technologies for measuring and managing forest resources using state-of-the-art sensors, robotics, autonomous systems and intelligent processing algorithms.
Our research focuses on resourcing and inventories, forestry automation and forest health, achieved by utilising our expertise in:
Our expert: Dr Mitch Bryson
Our collaborator: Forest and Wood Products Australia, NSW Department of Primary Industries (Forest Science), SCION New Zealand, University of Tasmania
We’ve developed new 3D pointcloud processing techniques for individual tree detection, segmentation and assessment using aerially-acquired, dense pointcloud datasets from airborne laser scanning.
These techniques can be used to perform plot-level inventories of commercial plantations prior to harvest without the need for sending field crews into remote and potentially hazardous sites.
Our collaborator: NSW Department of Primary Industries (Forest Science and Forest Health)
We've developed new methods for analysing plantation tree health from high-resolution, multi-spectral aerial imaging at a tree-by-tree level. We've developed methods for segmenting individual trees and classifying their health from 3D photogrammetric pointclouds. These methods have the potential to be used as part of early detection system for pest and disease outbreaks such as from Sirex wood wasp.
Our collaborators: National Institute for Forest Products Innovation, NSW Department of Primary Industries (Forest Science), SCION New Zealand, University of South Australia, University of Tasmania
We're collaborating with partners to develop new approaches to forest inventory based on deep learning at an individual tree level using Light Detection and Ranging and photogrammetric pointclouds, collected from aerial platforms such as drones.
We're developing new approaches based on active learning and human-computer interaction in which algorithms intelligently query inventory training examples from a human expert for learning predictive models from pointcloud data. These techniques have the potential to enable efficient workflows for inventory based on deep learning that minimise the time and effort required by foresters.