THE use of drones to find skeleton weed is set to amp up with new technology aiming to greatly increase the amount of hectares that can be covered.
New remote sensing data collection technology to complement traditional surveillance methods is now being employed by program funder, the Grains, Seeds and Hay Industry Funding Scheme (IFS).
IFS chair Rohan Day said a project began in 2015 to investigate the potential of remotely piloted aircraft, or drones, as a more efficient way to search for skeleton weed.
"There were two key areas of the project, to test the abilities of various drones and camera technologies, and to promote the development of machine learning software to identify and map skeleton weed from images captured by drones," Mr Day said.
This has resulted in Precise AI, a company specialising in developing software applications to process large amounts of remote sensing data, to apply its Optiweed program, a unique weed mapping and detection platform, to the detection of skeleton weed.
"This platform has progressed from detecting clumps of skeleton weed to being able to detect individual skeleton weed plants and their precise GPS location within the paddock," Mr Day said.
Department of Primary Industries and Regional Development (DPIRD) Skeleton Weed Program manager Martin Atwell said the skeleton weed surveillance activity aimed to cover up to 40,000 hectares of farmland per year.
"Drone technology was used in the 2018/19 search season to undertake targeted surveillance of uninfested areas where skeleton weed plants were most likely to spread," Mr Atwell said.
Previously all surveillance was undertaken by farmers and program staff in vehicles.
While this has worked well, there are some potential advantages in using drones.
"This first large-scale surveillance test of the drone and software application system worked well," Mr Atwell said.
"The work is currently delivering a good skeleton weed detection rate at a similar cost to the current methods of surveillance.
"This work is still new, and many important procedures and workflows were developed in the 2018/19 search season which will greatly improve the drone surveillance undertaken in future."
Mr Day said the future potential offered by the technology included improved cost-effectiveness of skeleton weed surveillance.
"This means reduced costs to the industry for funding for the program," Mr Day said.
"It also minimises the impact of vehicle traffic on farmland."
The company behind the drone surveillance, Precise AI said it was invited to assist with the project in 2015.
Its general manager Ed Boxall said the company saw the use of algorithm-driven precision agriculture as a likely efficient and cost-effective tool for locating and eliminating declared pests like skeleton weed.
"Precise AI's Optiweed software application utilises cutting-edge aerial and optical systems to collect large amounts of data, which was then processed using in-house machine learning algorithms that have been developed to discriminate between the target (skeleton weed) and the surrounding terrain," Mr Boxall said.
"Machine learning in this context means that the technology builds up this ability incrementally with each new batch of data.
"Optiweed has the ability to store aerial data, categorise different weed species and, most importantly in the context of this project, recognise skeleton weed.
"This provides DPIRD with precise GPS co-ordinates for weeds within the paddocks included in the project.
"In the 2018/2019 Skeleton Weed season, Precise AI moved from collecting and processing aerial data on a trial basis, and commenced a contract to map a portion of the total hectares allocated to the skeleton weed surveillance program."
Mr Boxall said from 2015 to 2018, approximately 6000ha were mapped.
"This was increased to 9000ha for the 2018/19 period and the 2019/20 period will see this area expanded further, to 20,000ha," he said.
"Despite the impressive leaps and bounds made by drone technology in recent years, as far as markets go, we are only at the early stages so progress is likely to be comparatively rapid.
"Even within a single project timeline we have seen leading edge technology become superseded by the subsequent iteration of the very same product.
"Concurrent to Optiweed's process of machine learning has been a process of regular upgrades to our aerial fleet and optical systems.
"In the precision agriculture space, small improvements can have significant benefits.
"For example, improvements to drone battery technology allows us to stay airborne for longer each day, which, over the course of a single project timeline can significantly increase the total hectares that can be covered."
Mr Boxall said the drone work was currently yielding 97 per cent accuracy rates.
"To achieve this percentage has been a fantastic achievement given the relativity small window available each year to collect data," he said.
"Precise AI has a team of industry leading software developers who are committed to undertaking a process of continual optimisation.
"Our team of drone pilots and technicians are continuously researching and experimenting with aerial systems and optical sensors to ensure that we have the highest quality data being processed through the Optiweed software application."
