A METHOD TO QUANTIFY WEED DISTRIBUTION FOR RELATING TO PATCH SPRAYING SYSTEMS

Abstract

A method of processing weed maps to quantify the spatial distribution of weeds in a field was developed using distance transform image analysis. Distance transform analysis involved the calculation of the distance of every non-weed pixel from the nearest weed pixel or vice versa. An exponential association function was fitted to the cumulative distance distribution curve describing the distance-transformed image histogram, and parameters were used to group fields with similar weed distribution patterns. Nine significantly different (p < 0.001) classes of weed map were established. Those with large weed patches were evaluated as being most suitable for patch spraying, but a secondary analysis using image dilation and erosion indicated that fields with small, fragmented weed patches could also benefit from patch spraying by up to 30% reduction in wasted chemical. In conjunction with knowledge of sprayer properties and costs, sprayer selection can now be based on machine performance, operational costs, and the type of weed pattern, to give the most appropriate performance for a specific field or farm. The method was developed using maps of grass weeds in cereal fields, but it could be used for any patchy weed problem. It could also be used to specify sprayer design criteria or applied to quantification of field spatial patterns for other precision agriculture applications.