What was it all about?

Pyrethrum (Chrysanthemum cinerariifolium (Trev.) Vis.) is a perennial plant that produces daisies that can be harvested over successive years. The variation in the pyrethrum crop growth and canopy over time creates a number of weed management challenges, including weed resistance and spread of weed seeds in harvesting equipment.  

To combat these challenges, this project aimed to improve machine vision technology to inform and enhance weed management strategies. Machine vision is a viable methodology to identify weeds in images, combining a camera, computer and software.

Researchers developed a test rig and data collection unit to capture image data of pyrethrum under a range of weed conditions to ‘teach’ the system to identify weeds.

As approved in a project variation, a pre-production spot spray prototype was developed. It incorporated a precision spray delivery system, a guidance system and a precision weed sensing system.

The proof of concept and pre-production prototype were evaluated on 60cm and 45cm row spacings to enable weed management between the row and on-the-row. The system discriminated between pyrethrum and weeds 100 mm in diameter with 98 per cent accuracy, with 1.2 per cent overspray on pyrethrum for 60 cm row spacings, traveling at speeds up to 8 km/h.

Ongoing commercialisation of the technology is expected to result in a commercial ready sensor that can be bought from a commercial company and used in pyrethrum production and other crop industries.