Monitoring canola emergence from the air
GroundCover™ Issue: 120 | Author: Alistair Lawson
Researchers in southern Victoria are taking to the skies to see if unmanned aerial vehicles (UAVs) can be used to help growers make timely interventions in canola establishment.
Traditionally, canola establishment can be challenging for some growers in the high-rainfall zone (HRZ) of south-eastern Australia due to pressure from pests – especially slugs – feeding on the seedlings.
Three sites were set up across southern Victoria in 2015 at Rokewood, Inverleigh and Winchelsea. At each site three different baiting treatments were applied: one was no bait, another had a baiting regime managed by Neil Hives from IPM Technologies and the third was managed by the grower on whose property the trial was taking place.
Four flights with the UAV were conducted at each site throughout the project at pre-sowing, cotyledon stage, two-leaf stage and four-leaf stage. There were six stages involved in the image processing:
- collection of colour infrared images;
- colour, texture and shape used to identify plants and create a binary image;
- self-learning algorithms used to detect larger and average-sized plants;
- the same algorithms then modified to detect small plants;
- the two images put together to calculate totals; and
- calculation of ground cover.
Southern Farming Systems researcher Megan Beveridge carried out ground-truthing of plant counts at each of the sites by using quadrats on the ground.
“We conducted manual plant counts to improve and calibrate the algorithms,” Ms Beveridge says.
PEAQ Management consultant Alex Russell says the team encountered some barriers when transferring the processed UAV images to Alberto Dri at Luminis Analytics in Perth.
“The UAVs picked up approximately 30 gigabytes per flight, so 90GB in total over the three sites, and an internet transfer test revealed it takes 90 minutes to transfer 1GB of data, so it was going to take us 135 hours to transfer 90GB,” he says.
“However, using express post transfer between the FarmingIT office in Meredith, Victoria, and Perth we can deliver 90GB of data in 24 hours … so postal transfer is currently preferable.”
Mr Russell says accurate plant counts can be gathered at the two-leaf stage using UAV imagery; however, before this stage current camera specifications limit plant detection at cotyledon stage.
“Detection much later than the two-leaf stage is likely to be too late for re-sowing as a result of slug damage but we are still going through the image analysis,” he says.
“We are continuing to work with Alberto to see if we can get detection at cotyledon stage, which would be in time for a re-sowing decision based on that four-day data turnaround.
“We will continue to work on improving the plant counts at those later stages because once you get a certain percentage of ground cover you have to make some assumptions on what those counts are.”
More information:Alex Russell,
0467 089 843,
Megan Beveridge, Southern Farming Systems,
0488 600 116,
GRDC Project Code SFS00030