Precision FarmingThe future of weed control

Cameras, algo­rithms, drones, and other new ideas: Intel­li­gent tech­nology can make plant protec­tion much more effi­cient. We explain how it works, what John Deere is working on and why the farmer still plays the deci­sive role.

Every farmer knows that weeds occur in patches but due to a lack of appro­priate tech­nology the entire field is currently treated through broad­cast spraying. An effi­cient tech­no­log­ical solu­tion has to address three areas: Detect weeds, decide on the type and quan­tity of treat­ment and finally treat them.

This can lead to signif­i­cant bene­fits for growers through increased yields, a signif­i­cant reduc­tion in input costs as well as bene­fits for the envi­ron­ment through lower chem­ical appli­ca­tion and reduc­tion in the like­li­hood of resis­tances devel­oping. See & Act tech­nolo­gies are key to over­coming the prob­lems described above and unlocking earning poten­tial for growers within multiple produc­tion systems.

Detect weeds reli­ably

Although it’s rela­tively easy for tech­nology to distin­guish between green plants and brown soil, that doesn’t really help. This can be enhanced to iden­tify weeds in between rows in tradi­tional row crops (eg maize). However, weeds are partic­u­larly harmful within the row. To unlock the full value of this tech­nology both weeds between and within the rows need to be detected and treated.

The greater the simi­lar­i­ties, the more diffi­cult it is to iden­tify the weed.

But that is also the greatest chal­lenge. Some weeds have the same morpho­log­ical appear­ance as the crops them­selves, making them diffi­cult to iden­tify. For example, in early growth stages it’s hard to sepa­rate grass weeds from a cereal crop. The greater the simi­lar­i­ties, the more diffi­cult the detec­tion of weeds. It is even more demanding to develop a tech­nical solu­tion that meets the very complex require­ments of camera, sensor and algo­rithm.

For this, John Deere has devel­oped an industry first – computing power the same as IBM’s “Blue Gene” super­com­puter, that revo­lu­tionised the IT space in 2008. In a rugged casing to meet the needs of agri­cul­ture it will be mounted on spray booms of the future. It will enable capturing an image, processing the data, taking a deci­sion, and initi­ating execu­tion within millisec­onds to ensure produc­tivity is main­tained.

Is it crop or weed? The devel­op­ment of modern systems allows auto­matic plant clas­si­fi­ca­tion in order to decide how to proceed with them.

Machine mounted or remote sensing

Complexity also deter­mines whether you choose to use online or offlin e tech­nology. The so-called real-time recog­ni­tion or ‘online tech­nology’ has the great advan­tage of seeing and treating within one pass. The short time window and thus high-speed oper­a­tion plus the influ­ences that affect the camera and sensor system – eg dusty field condi­tion, dew, wind, and sun shading, makes execu­tion of real-time sensing chal­lenging.

In the offlin e approac h th e are a is scanned first, for example by a drone or maybe satel­lites in future. A chal­lenge is the required reso­lu­tion which demands low alti­tude flying for drones. The two-stage approach offers the poten­tial to analyse the data with cloud-based services which doesn’t require high processing power on the machine. However, offline data analysis results in time delays which may lead to changes in the weed popu­la­tions and densi­ties in the mean­time.

Deci­sion making process

If weeds are detected the farmer needs to decide how to control them. In conven­tional produc­tion systems, the farmer decides on the herbi­cide used and the amount sprayed. However, the more detailed the analysis of the field, including the type of weed, the crop, the crop rota­tion, expected yield, the weather etc, the higher the poten­tial savings can be. Handling this enor­mous amount of data is crit­ical for success. There­fore, farmers will continue to play a key role in sustain­able weed manage­ment. John Deere provides the full solu­tion from farm data handling, scouting infor­ma­tion, and machine inte­gra­tion to fully support the future line-up of weed manage­ment.

Future weed manage­ment equip­ment

Today, John Deere is the world’s leading sprayer manu­fac­turer and already has tech­nology devel­oped to apply sprays to a very precise level. Tech­nology like Exac­tAp­plyTM enables highly accu­rate appli­ca­tion and is already leading in nozzle tech­nology. With this current tech­nology customers can reduce input costs by 5% while improving the quality of appli­ca­tion. Soon, the See & Act tech­nology supported by deci­sion-making processes and highly precise and site-specific appli­ca­tion will allow 50% to 90% input reduc­tion.

The green spots show where chem­i­cals have been sprayed in the field.

Research into effi­cient solu­tions

John Deere is working on various solu­tions to actively shape the future of plant protec­tion. With the acqui­si­tion of Blue River Tech­nology in 2017, the company has demon­strated its commit­ment to invest in preci­sion equip­ment that will allow customers to unlock signif­i­cant input savings in multiple produc­tion systems. Blue River continues its healthy devel­op­ment in the area of Arti­fi­cial Intel­li­gence (AI) – with this tech­nology the machine will learn as it goes – addi­tional cameras will verify that job execu­tion was as intended and self-opti­mise to improve its perfor­mance contin­u­ously.

The journey started with plant detec­tion on bare soil. Further devel­op­ments will grad­u­ally increase the full poten­tial of indi­vidual, loca­tion-based weed treat­ment.

Further infor­ma­tion