PRELIMINARY EVALUATION OF SHAPE AND COLOUR IMAGE SENSING FOR AUTOMATED WEED IDENTIFICATION IN SUGARCANE

AUTOMATED WITHIN-ROW WEED spot spraying is expected to provide a mechanism for controlling difficult weeds in the sugarcane industry while reducing herbicide usage and the labour of manual weed spot spraying. However, technologies need to be developed that enable robust and automated in-field crop/weed discrimination. Weed identification is potentially achievable using machine vision, a technology that enables low-cost sensing and analysis of colour, shape, texture and depth (i.e. plant and leaf height) information. The National Centre for Engineering in Agriculture (NCEA) has evaluated machine vision and image analysis approaches for colour and shape sensing for the purpose of automatic discrimination of sugarcane from in-field weeds. The approach involves application of line detection techniques to high quality in-field colour camera images. This follows on from research in which NCEA developed a colour-based image analysis system that was effective at discriminating in-field mature Panicum spp. (Guinea grass) from sugarcane at night time. The current research has demonstrated that shape sensing in addition to colour sensing enhances sugarcane/weed discrimination. Preliminary image analysis results are presented for the evaluated machine vision approach on a range of weed species.