A final illustration is to be found by Li et al. (2002). They developed
computer image technology for apple surface defect detection
which allows graders to remove image background, to
identify the stem-end and calyx areas, and to ascertain segment
defects. The detection hardware features two CCD monochromatic
cameras mounted above and below a conveyor in a lighting chamber
and two mirrors fixed on both sides of the conveyor, which
provided four simultaneous images of the scanned fruit. The background
is removed by the subtracting method and defects are extracted
by comparing the scanned fruit with normalized reference
apple images. Testing of the single row computer image sorting
system indicated 93% classification accuracy.