New semi-automated survivability count method
Posted 17 October 2023
PFT trial Esk Spatial's new semi-automated survivability count method testing technology capabilities against more manual survivability count methods.
Round 1 of the Integrated Farm Forestry Demonstration Sites program is now in its third year.
Seven properties received funding to assist with plantation establishment costs, and to develop landscape scale best practice forestry plantings dispersed throughout the state including a range of species, and a mix of shelterbelts and woodlots totalling 201 hectares.
With seedlings now approaching 2 years of age, PFT sought the opportunity to conduct a test trial using ESK Spatial’s new semi-automated survivability count.
PFT trialled this new method on two properties located near Dunalley in the south, and Evandale in the north to gain an understanding of the capabilities of this new technology and to learn how results compared to survival counts done in the more manual ‘traditional’ method. Findings are yet to be determined with PFT exploring the technology further once seedlings are at a greater height and weed competition is at a minimum.
Esk Spatial have provided the following information to explain their exciting new process.
When establishing forests it is critical to ensure that the initial establishment operations achieved the required survival rates and planting stocking rates to fully occupy and maximise the growing potential of the site. This ensures the forest will achieve the goals of the grower, whether they be for environmental or production use. A ‘survival count’ assessment is typically undertaken within the first 6 to 12 months after planting to sign off that the forest has achieved the required standards.
Esk Spatial have developed a semi-automated method for undertaking a 100% survival count by way of capturing and analysing drone imagery, vastly improving on old techniques typically undertaken by foot which were only based on a small sample of seedlings (say 10%) which would not have taken stock of the whole area.
Esk Spatial captured high-resolution imagery (3cm resolution) of the trial sites using a drone, allowing us to see individual seedlings/stems clearly and distinguish them from their often ‘noisy’ surroundings.
A total of 24,343 stems were detected and pinpointed across the two trial sites. Manually extracting these from the imagery would be highly laborious and expensive, taking more than 70 hours.
Using state of the art neural network based object detection models, Esk Spatial trained up algorithms to detect the 1,000s of stems in a matter of minutes. Esk Spatial has adopted and developed these object detection methods for several years successfully locating 100,000s of seedlings and trees for a range of species, environments and age classes.
As part of the process a 3D model is also generated. Using the locations of the detected stems, Esk Spatial can use the 3D model to calculate the individual height for each and every stem which is extremely valuable information for tracking the growth rates and health of the trees.
The richness and accuracy of forest information that can now be collected remotely with semi- and fully automated means across 100% of the area by Esk Spatial provides invaluable information to tree growers, and in a manner that is truly cost effective.