tions
in ambient lighting intensity due to location within a facility
and wavelength dependent differences associated with the type of
ambient lighting used in the facility. The next software iteration
used a single gain array where gains could be calculated and stored
at 5 nm intervals. The use of a single gain profile proved impractical
as gain profiles had to be recalculated frequently primarily due
to changes in the composition of target backgrounds, e.g., HDPE
exhibits an auto fluorescence response to supplemental LED illumination
while SS does not. In the current software, users can select
among five named gain profiles and, if desired, recalculate one of
the three variable profiles by pressing the appropriate icon. The
gain profiles are graphed to allow users to better select the appropriate
gain array for use.
The finding that all types of anomalies of interest could not be
detected using a single wavelength resulted in the addition of
the ability to automatically cycle among a set of preselected wavelengths.
The addition of the ability to cycle was dependent on the
development of automated gain profiles. The number of wavelengths
used for cycling was eventually restricted to three as users
had difficulty following more than three wavelengths. The ability
to tap on the display to start and stop cycling facilitated the acceptance
of the cycling option.
The three wavelengths used for cycling were selected by examining
the wavelengths where ambient illumination was relatively
low to see if a subset of these wavelengths could be used to detect
all anomalies of interest. The 475, 520, and 675 nm wavelengths
were selected based on an understanding of wavelengths appropriate
for detecting organic residues (Wiederoder et al., 2012). When
the device was tested in produce plants, the wavelengths were
found both to be necessary and adequate for detecting all
anomalies of interest, including defects in HDPE and SS surfaces
(Wiederoder et al., accepted for publication). It should be noted