Evaluating the accuracy of mapping weeds in seedling crops using airborne digital imaging: Avena spp. in seedling triticale

Airborne multispectral imaging has been used to map patches of Avena spp. (wild-oats) in a field of seedling triticale (X Triticosecale , Wittmack). Images of the target field were acquired using a four-camera airborne digital imaging system, recording in the infrared, red, green and blue wave-bands. Spectral information derived from images of 0.5-, 1.0-, 1.5- and 2.0-m spatial resolution were correlated with detailed on-ground weed density measurements to investigate the effect of image resolution on mapping accuracy. Comparisons between normalized-difference vegetation index (NDVI) or soil-adjusted vegetation index (SAVI) images and weed data achieved correlations of up to 71%. The highest correlation was achieved with the 0.5-m-resolution images and the lowest with the 2.0-m-resolution images. At 0.5-m resolution, NDVI images could not reliably discriminate weed populations of less than 28 weeds m^–2 from weed-free regions, while SAVI images could not discriminate populations of less than 17 weeds m^–2. At 1.0-, 1.5- and 2.0-m resolution, SAVI images could not discriminate populations of less than 23 weeds m^–2, while NDVI images again demonstrated a higher discrimination threshold. Results suggest that airborne multispectral imaging could be used as part of a stratified weed sampling system.