Cutting-edge weed-ID tool to cut herbicide use Down Under

SUBSCRIBE to our fortnightly e-newsletter to receive more stories like this. Raphanus raphanistrum sepal, aka wild radish, is an introduced weed that's been declared noxious in some parts of southern NSW.

Weeds and the herbicides used to control them, cost Australian growers millions a year. The runoff from chemical applications can harm the environment and seep into surrounding soil and nearby waterways. And potentially toxic herbicide residues can make their way into the resulting produce.

In a bid to help Aussie growers reduce their use of chemical herbicides in controlling weeds, researchers at the Society for Precision Agriculture Australia (SPAA), backed by the South Australian Grain Industry Trust Fund (SAGIT), have been trialling a commercially viable weed identification and mapping system sourced from Germany known as the H-sensor.

How does the H-sensor work?

AgriCon, creators of the H-sensor (in which the H is short for herbicide) claim it’s “the world's first sensor to safely distinguish crop plants from weeds and grasses”. 

Specially-developed cameras in the tool recognise different plant species automatically, switching the corresponding partial spray width in real time according to the weeds it identifies. The H-sensor’s “robust” hardware enables working speeds of up to 12 kilometres per hour, and the tool comes with its own light source so it can be used 24-7.  

The H-sensor is designed to be used with special herbicides that are applied in a precisely targeted fashion only where weeds and grasses are identified – reducing herbicide costs, negative environmental impacts and herbicide resistance. 

In large-surface tests, AgriCon reports, “savings of 50 to 70 percent of weeds and grasses could be detected with the H-sensor”.

Adapting the H-sensor to Aussie conditions

Since 2015, SPAA has been trialling the H-sensor system in Aussie paddocks and producing new, adapted classifiers for identifying important Australian weeds in crops not typically grown in Germany, including all the grain legumes – lentils, field peas, faba beans, chickpeas and lupins.

The goal of the project is to adapt the site-specific weed management (SSWM) tool to conditions and farming practices Down Under.

The researchers hope that, once fully adapted to Aussie conditions, the sensor-mapping SSWM tool will be widely adopted by grain farmers, resulting in more efficient use of herbicides.

Potential benefits are many; they include:

  • cost savings;
  • reduced phytotoxic effects of some herbicides on crops; and
  • environmental and social benefits thanks to reduced herbicide load, notably lessened “off-target impacts” and lower herbicide residue levels in food.

Mapping trial paddocks

A single-camera, “operations-ready” H-sensor, supplied by AgriCon for the project, was boom-mounted on a ute and used to map trial paddocks planted with winter crops.

Mappings were timed to coincide with herbicide applications, with the resulting data used to generate ‘weed density’ maps.

The accuracy of these weed maps was assessed using field observations from specific GPS locations.

The H-Sensor, mounted to a ute for mapping and collecting images of the SPAA trial crop and weeds.
The H-Sensor, mounted to a ute for mapping and collecting images of the SPAA trial crop and weeds.
Grains Research & Development Corporation (GRDC)

Modifying the classification database

Initially, the H-sensor ran with its German weed-classification database but as trial data accumulates, the database is being modified and fine-tuned, improving the sensor’s ability to identify local weeds.

SPAA is building additional databases to enable classification of new crop types and special-interest weed groups, which it has been assessing in the field. These new and refined classification databases will become valuable intellectual property. Weeds of particular interest to the SPAA researchers include annual ryegrass (Lolium rigidum) and wild radish (Raphanus raphanistrum).

The initial focus in winter crops was to:

  • discriminate all broadleaf weeds from cereal crops, ensuring that broadleaf weed species of vastly different shapes are correctly classified together, using similar weed groups as those covered by the GRDC weed ID app – such as ‘carrot-like’ (bifora, fumitory); having multiple leaflets (soursob, medic, clover); and having rosettes (capeweed, sow thistle, Indian hedge mustard, wild radish).
  • correctly distinguish all grass weeds from broadleaf crops, including canola and the grain legumes lentils, peas, beans, chickpeas and lupins.
Wheat and an Indian hedge mustard collected in the red and near-infrared spectrum, and how the sensor has classified these differently.
Wheat and an Indian hedge mustard collected in the red and near-infrared spectrum, and how the sensor has classified these differently.
Grains Research & Development Corporation (GRDC)

Crop and weed distinctions

The project also investigated other valuable crop and weed discriminations, including:

  • classifying annual ryegrass in wheat and barley;
  • classifying weeds with rosettes (particularly wild radish) and distinguishing them from other broadleaf weeds in cereals and legume crops;
  • classifying carrot-like weeds, with particular focus on bifora, and distinguishing them from other broadleaf weeds in cereals and broadleaf crops.

Field testing and training was conducted in 2015 and 2016.

In addition to field testing, the SPAA researchers grew ‘weeds of interest’ in in pots in a glasshouse during the ‘off’ seasons (spring and autumn of 2014-15 and 2015-16) and imaged them using the H-Sensor, using the resulting images to help build the classification database. 

The flower of wild radish, a major weed of winter crops especially canola and turnips, and if grazed, can taint milk or meat.
The flower of wild radish, a major weed of winter crops especially canola and turnips, and if grazed, can taint milk or meat.
Harry Rose, Flickr CC

Assessing the effect of crop residue on classification accuracy

Given that no-till farming techniques have been adopted widely in Australian cropping, whereas tilling is far more prevalent in Germany, the researchers also assessed whether the H-Sensor could correctly identify plants emerging through high levels of crop residue. Large field trials were required to assess the ‘stubble effect’ under differing types and amounts of stubble. Treatments tested included plots with stubble retained standing, plots with ‘stubble removed completely, and different stubble types, including wheat and canola.

Field observations were made from specific GPS locations within each treatment to compare the effect of the various crop residues on the accuracy of the H-sensor.

The project was completed at the end of the 2016-17 financial year, with findings to be released later in 2017.

For further details, check out GRDC’s update paper, ‘New technology for improved herbicide use efficiency’

About SPAA

An independent, membership-based non-profit group, the Society for Precision Agriculture Australia (SPAA) was established in 2002 to “promote the development and adoption of precision agriculture (PA) technologies”.

The association’s stated goal is to be “the leading advocate for PA in Australia and, through this role, improve the profitability and sustainability of agricultural production systems via the adoption of PA”.

Current SPAA members include those involved in the production of grains, horticultural crops and wine grapes, and include growers, consultants, equipment manufacturers, contractors and researchers. The association takes an Australia-wide focus, often partnering with national organisations to form broad industry alliances.

More information

For more information about this project or the association, and/or to participate in SPAA’s research, development and extension activities, phone 0437 422 000.

SUBSCRIBE to our fortnightly e-newsletter to receive more stories like this. 

Project: 
Innovation Central connected tech laboratory

Comments