Agriculture Reference
In-Depth Information
RED
CCD Sensor
Infrared
Trim filters
CCD Sensor
Green
FIGURE 2.24
Construction of a multispectrum imaging sensor (MSIS).
Skye Instruments Ltd.), an RTK-GPS, and a PC. The imaging sensor was a three-
CCD camera. Separate optical paths permitted the installation of special optical
filters for each sensor, providing the three video channels of green (G), red (R), and
near-infrared (NIR) illustrated in Figure 2.24. These three channels have a band-
width of approximately 100 nm and center wavelengths of 550, 650, and 800 nm.
To ensure a wide dynamic range in the sensor, computer control of each channel
via an RS232C interface was included in the design for both lens aperture and CCD
gain. An ambient illumination sensor was used to compensate for changes in lighting
power. This sensor also has the three filters (G, R, and NIR), which were almost the
same bandwidth as those of the MSIS. Imaging systems on the vehicle were set to
observe the same field of view.
An estimator of crop height, chlorophyll content, and SPAD value using MSIS
response was developed. The motivation for detecting crop parameters with the
MSIS was to enable remote estimation of these values. Because the MSIS can control
the CCD gains as well as the aperture, the MSIS response relating to light reflectance
can be calculated as follows:
X
response
=
c
,
(2.2)
1
∗
*
AI
*
cg
E
10
2
where
X
is the gray level,
E
is the aperture, AI is the output from the illumination
sensor, and
c
1
and
c
2
are constants. As is evident in expression 2.2, the factor AI
compensates for changes in outdoor illumination. An MSIS response is individually
acquired for each of the three channels: red, green, and NIR. An important issue
in image processing is the method for segmenting fields of interest from various
unnecessary regions in the image. For instance, soil and weed regions have to be seg-
mented from the image before a decision is made on crop status. Even after segment-
ing the image, regions of shadow and areas of direct reflectance from the crop leaves
must be considered by the image processing system. An NIR image was used for
