Machine vision at a plant scale: methodologies and applications

Abstract

Machine vision at a plant scale, which involves an on-field image sensor, allows specific applications, such as plant counting, automatic weeding, or vegetation detailed characterisation. Plant characterisation first addresses agronomical researchers, e.g. for growth modelling or seed selection. But overall, complete technical solutions should then allow the agricultural producers to optimise their cultural operations in terms of product quality, chemical inputs and environmental impact, by taking into account in-field measurements. However, the kind of machine vision approach brings out various difficulties, from outdoors lighting control to leaf overlapping problems. We present here the main techniques that can be used to overcome these difficulties - some of them still being at a research stage - as well as various application examples.