Methods of weed patch detection in cereal crops

This paper reviews the merits of real-time and map-based systems of weed assessments and discusses the importance of the precision of the map. 'Traditional' mapping based on weed density counts on a regular grid are useful for research purposes but are too expensive for practical on-farm mapping. Visual weed detection systems can already be used to record weeds but tend to be labour intensive and so need improvement to increase their cost-effectiveness. The potential for automatic weed detection is being explored by a number of research teams. Reflectance techniques can already be used to detect weeds in uncropped areas and between crop rows in crops sown with wide row spacings. However, although progress is being made with experimental systems, commercially viable methods to detect weeds within narrow row crops are not pet available. Systems involving combinations of reflectance, colour and leaf shape are in development and some are being tested in the field. There has to be a compromise between the cost of mapping and the need for the map to reflect the distribution of weeds. Thus fine maps are preferred to the cheaper coarse maps since some of the advantages of the technique will be lost if sample grids are too coarse.