Machine vision-based weed spot spraying: a review and where next for sugarcane?

摘要

Automated precision weed spot spraying in the sugarcane industry has potential to increase production while reducing herbicide usage. However, commercially-available technologies based on sensing of weed optical properties are typically restricted to detecting weeds on a soil background (i.e. detection of green on brown) and are not suited to detecting weeds amongst a growing crop. Machine vision and image analysis technology potentially enables leaf colour, shape and texture to achieve discrimination between vegetation species. ududThe National Centre for Engineering in Agriculture (NCEA) has developed a machine vision-based weed spot spraying demonstration unit to target the weed Panicum spp. (Guinea Grass) in a sugarcane crop, which requires discrimination of a green grass weed from a green grass crop. The system operated effectively at night time for mature Guinea Grass but further work is required for the system to operate under a greater range of conditions (e.g. different times of day and crop growth stages). Techniques such as multispectral imaging and shape analysis may potentially be required to achieve more robust weed identification. The implications for machine vision detection of Guinea Grass and other weed species in sugarcane crops are considered.ud