IEEE Geoscience and Remote Sensing Magazine - March 2021 - 207

angles, and it realizes the extraction of the surface information and the reconstruction of the 3D scene [112]. Lidar
includes a single-beam narrowband laser and a receiving
system. The laser generates and emits a light pulse, which
hits the target object and reflects back to a receiver [199].
The receiver accurately measures the propagation time of
the light pulse from emission to reflection [112]. Because
the pulse travels at the speed of light, the receiver always
obtains a reflected pulse before the next one is sent. Given
that the speed of light is known, the time of flight can be
converted into a measurement of distance [200].
The laser itself has a very accurate ranging ability, and its
ranging accuracy can reach the millimeter level. In general,
an airborne lidar system integrates three technologies: laser
scanning, GPS/differential GPS, and an inertial navigation
system. This combination can locate with high accuracy
a laser beam spot hitting an object [255]. The principle of
measuring the ground from a flying carrier concerns the
" two-way " time measurement method [200]. That is to
say, we need to record only the time between a laser signal's emission and its return. Combined with other data,
we can accurately calculate the x-, y-, and z-coordinates of
the ground spot. Lidar does not need to be calibrated, but
known referenced objects are necessary to accurately obtain crop structure information.
Figure 1(h) shows an example image of sunflowers that
was acquired with a Sick LMS400. Lidar can provide a 3D
point cloud to describe the crop canopy structure [129]. The
point cloud can be used to monitor dynamic crop changes
during the growing season [197], [200]. Table 2 describes a
typical lidar model and its parameters, phenotype estimating traits, imaging environment, imaging techniques, and
limitations. Compared with other sensors, lidar is less influenced by environmental conditions, and it has a relatively high estimation accuracy for several crop phenotyping
traits because it is an active sensor [198]. However, the price
of lidar is generally higher than that of passive sensors. Lidar is sensitive to small differences in the path length, although the data processing is more complex (Table 2). Lidar
has been integrated with ground and aerial phenotyping
platforms to estimate crop phenotyping traits [112], [197],
[200], [220], [255]. However, in the future, the density of
the achievable 3D point cloud needs to be increased to
better describe 3D crop structures.
AIRBORNE SENSORS
RGB CAMERAS
RGB cameras record red, green, and blue spectral bands in
the visible spectral regions [39]. An RGB camera produces
digital images and can mimic human visible perception to
obtain information for estimating crop phenotyping traits
in plant breeding [74]. The most common implementation
of RGB cameras is based on silicon sensors with chargedcoupled devices or CMOS arrays that are very sensitive to
visible bands of light and show images in 2D; this is the
MARCH 2021

IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE

simplest imaging technique for crop phenotyping traits
[247]. RGB camera images are typically shown in spatial
matrices of intensity values (from zero to 255) corresponding to photon fluxes in blue, green, and red spectral bands
of the visible region [46]. To obtain better high-quality RGB
images, a color calibration plate is commonly used to adjust
images using algorithms to reduce the effect of different illumination conditions on the photo quality [256], [257].
Figure 1(a) gives an example image of maize taken with
a Sony Alpha 7 camera. The RGB camera can export images
that include gray value, color, and texture information [39].
The applications of RGB cameras for crop phenotyping traits
are shown in Tables 1 and 2, including GAI, plant density,
ear density, FAPAR, canopy coverage, canopy height, leaf rolling, leaf angle, lodging, and so on, during the whole crop
growing season. In addition, a typical camera model and its
parameters, phenotype estimating traits, imaging environment, imaging techniques, and limitations are presented in
Table 2. RGB cameras have been widely applied to evaluate
crop phenotyping traits in the field to provide measurements
at an affordable price [46], [47], [162], [164], [258].
Compared with other sensors, the advantage of the RGB
camera relates to its affordable price and high-spatial-resolution imagery [237]. However, it remains a huge challenge
to get good image segmentation results when adjacent plant
leaves overlap [39]. In addition, the gray values and color
information are limited to three visual spectral bands. New
technologies are required to solve these shortcomings and
improve the application of the RGB camera for crop phenotyping trait estimation [7]. Currently, a modified RGB
camera [the so-called color-infrared (CIR) camera] with
near-infrared, red, and green bands may be applied to obtain spectral vegetation images when it is radiometrically
calibrated [259], [260]. The pseudonormalized-difference
vegetation index (pseudo-NDVI) from a CIR camera has
been applied for crop phenotyping [7], [258], [261], [262].
CIR cameras can be used to obtain high-spatial-resolution
imagery with a limited number of wide spectral bands at a
relatively low cost. However, CIR cameras may be unstable,
and they have a limited application, as they are not designed for precise radiometric measurements [263].
MULTISPECTRAL AND HYPERSPECTRAL CAMERAS
Multispectral and hyperspectral cameras depend on the
interaction between solar radiation and crops [71]. The reflectance of single leaves or canopies is low in the visible
spectral ranges (400-700 nm) because solar radiation is
absorbed by leaf pigments (such as chlorophyll), with a
peak of reflectance in the green spectral region of roughly
550 nm [45]. The reflectance (so-called red-edge region)
is sharply increased with the transition from the visible to
near-infrared spectral region [32]. A large fraction of the
incident solar radiation is reflected by leaves because of
scattering within the leaf structure and mesophyll in the
near infrared (700-1,200 nm) region [24]. Furthermore,
near-infrared radiation is transmitted from the upper to the
207
IEEE Geoscience and Remote Sensing Magazine - March 2021

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - March 2021
