Low cost high precision seedling inventories

SKYLAB in Hamburg, Germany have just presented at this year’s ForestTECH 2019 series. Two practical workshops yesterday finished the series of conferences, exhibitions and workshops that have been run for local forestry companies over the last two weeks.

Local foresters have heard how SKYLAB has just developed a machine learning algorithm for high precision seedling inventories. Manual surveys for seedling survival monitoring could become obsolete for most common plantation types. The Seedling Counter (see photo of a Seedling Counter density map) has already been tested and successfully applied in seedling inventories in Pine, Eucalyptus and Spruce plantations around the world – so far in Chile, Spain, UK and Australia – with accuracies usually between 90 and 99%.

Surveys are traditionally carried out as representative spot samples at some stage after planting to ensure the right number of trees have been planted and survive to reach the target for each site. Sampling carries the risk of missing areas of low density or even complete failure. Aerial surveys can capture the complete stand and thus ensure that all such underperforming areas are identified.

According to SKYLAB, the RGB images from a standard DJI drone are of sufficient quality to be analysed by the Seedling Counter as long as the flight planning is optimized according to their instructions. In addition to seedlings, the system can also identify and quantify the density of weeds in any stand. This can help in more targeted weed management and to improve seedling survival.

All results are presented in geo-referenced density maps (pictured) highlighting the low-density seedling and high-density weed hotspots. This enables the forest manager to focus on inspecting those areas where targeted actions are required. Naturally the analysis can also be used as a basis for performance-based payment arrangements with planting contractors.

Working with growers across the world, but especially in Australia, the algorithm is continually trained and its performance increases with each new data set providing more reference data. This, combined with the very significant time savings and complete coverage could mean, that the need for time consuming and sometimes dangerous terrestrial seedling surveying may become a thing of the past. For those not attending the recent series, further information can be found at www.skylabmaps.com.

Source: SKYLAB