Data from a photoelectric diode sensor equipped with 670 and 780 nm interfe
rence filters, were analyzed to determine the maximum field-of-view and min
imum percentage of soil covered by bindweed (Convolvulus arvensis) that cou
ld be distinguished from bare soil. Images containing one 150-mm bindweed c
ould be distinguished from images of adjacent bare soil for all fields-of-v
iew implying that adaptive thresholding would enable use of a field-of-view
as large as 0.71 m(2). Detection was 1000% for nine of the II fields-of-vi
ew evaluated and 98% for the other two. When bare soil images were not pair
ed with adjacent bindweed images, a single NDVI threshold could be used to
distinguish between the two images with error Results indicated that the re
quired percentage of bindweed cover within art image was 12% for a 5% error
in falsely detecting or failing to detect the presence of bindweed. Error
for a fixed NDVI threshold decreased when classified by apparent soil moist
ure when compared with unclassified data.