IEEE Geoscience and Remote Sensing Magazine - March 2021 - 212

guarantees continuous electricity through a regulator. The
cooperation between each system can safeguard the normal operation of the automatic phenomobile.
Currently, the automatic phenomobile can autonomously obtain RGB, multi/hyperspectral, and thermal
images and point clouds using different optical sensors for
each plot in a field. These images and point clouds can be
employed to estimate the canopy cover, GAI, canopy height
and temperature, chlorophyll and nitrogen content, water
content, yield, grain quality, relative values of the in-season
biomass, and so forth. The dynamic changes of crop phenotyping traits can be monitored through the automatic
phenomobile without much human intervention. The advantage of the automatic phenomobile lies in the ability to
simultaneously collect different types of images; however,
such platforms are expensive. Current automatic phenomobile examples include Ladybird [277] at the University of
Sydney, Australia, and Phenomobile [276] at the Institut
National de la Recherche Agronomique, Paris, France.
Human phenomobiles are moved by manpower through
a field. Examples include Phenomobile Lite [200], which
contains data acquisition systems controlled by a person
with a laptop. The system can incorporate RGB, multi/hyperspectral, and thermal sensors to obtain the related crop
phenotyping traits. The price of human phenomobiles is
low, but the data acquisition quality and efficiency of the
sensors cannot be guaranteed.
Tractor phenomobiles consist of integrated sensor and
GPS systems, a data acquisition system, and a power supply [278]. The basic integrated-sensor-and-GPS system
combines non-image-based sensors (such as GreenSeeker
and CropCircle, which are hyperspectral), the the outputs
of which are recorded with GPS coordinates; similarly, the
data acquisition system maintains the brightness of the
flashlight and the consistency of the sensor data to ensure
the quality the gathered information. The tractor provides
the power supply. The PhenoTrac4 could be used to obtain
VIs, nitrogen uptake, biomass, and water status. Its price
is lower than other phenomobiles because its sensors are
non-image-based [278]. Phenomobiles are relatively flexible in terms of sensors.
Finally, the sensors of stationary platforms include RGB
cameras, multi/hyperspectral cameras, and thermal cameras and laser sensors. Stationary platforms are simultaneous
and fully automated, fixed-site phenotyping equipment
[279]. They can carry sensors for the noninvasive estimation of crop growth, physiology, morphology, and health.
Their advantage is that they can simultaneously monitor
different crop traits, although only in a defined area of, for
example, roughly 1 hectare, when they are mounted on
trackways [279].
AERIAL PHENOTYPING PLATFORMS
Aerial phenotyping platforms may be classified as MAVs,
UAVs, and micro/nanosatellites (Figure 3). Recent advances
in technology have promoted the development of UAVs,
212

which have been transformed into crop phenotyping platforms that provide high-spatial-resolution images for crop
phenotyping trait estimation in the field [259], [260]. They
include fixed-wing and multirotor configurations. Fixedwing UAVs have the advantages of fast flight, high flight
efficiency, long endurance time, large payload capacity, and
a high flight altitude. However, they have certain requirements for taking off and landing, they cannot hover, and
they will cause blurred images because of their high-speed
shooting [281]. Multirotor UAVs have a simple structure,
the ability to hover, and modest requirements for taking
off and landing, but they possess a slow flight speed, short
endurance time, a low flight altitude, and small payload
volume [281].
Sensor images from UAVs are processed using Agisoft
Photoscan Professional (Version 1.2.2, Agisoft, Saint Petersburg, Russia). The overlap between images should be
larger than 60% for the software to compute the position
of the camera corresponding to every acquired image [39].
The software can automatically recognize targets applied as
ground control points (GCPs) for the absolute geopositioning of the images, and it can precisely locate the objects' center on the images [39]. The specific processing is conducted
as follows:
1) check the camera calibration and optimizing the parameters
2) align the photos
3) build dense point cloud
4) construct the mesh
5) create the texture
6) build digital elevation models (DEMs)
7) develop the orthomosaic
8) export the orthomosaic and the DEMs.
The lidar point cloud from UAVs can be processed by
CloudCompare or by writing code as needed [112], [199].
Currently, UAVs can carry different sensors (RGB cameras,
multi/hyperspectral cameras, thermal cameras, and lidar)
to estimate different crop traits at varying spatial scales. Images from UAVs can be employed to estimate grain yield
[99], canopy temperature [266] and NDVI [259], [282],
plant height [108], [282], [283], biomass [162], GAI [109],
lodging [134], plant density [39], fluorescence [266], and
nitrogen status [36], mostly through proxies for plant traits.
The advantage of UAVs concerns the relatively high-resolution images that are obtainable in a relatively short time;
however, it is difficult to cover very large areas due to the
vehicles' limited range and speed. The onboard sensors of
MAVs are comparable with those of UAVs. MAVs can cover
larger regions in a relatively short time, albeit with a lower
image resolution, and they have a greater payload capability. Images from MAVs are mainly used to estimate the
GAI, biomass, chlorophyll content, nitrogen content, plant
height, and biomass [8].
Finally, micro/nanosatellites can provide data in several
spectral bands (which are sensitive to the crop structure, leaf
pigment, and water content), and they enable meter-levelresolution images for estimating crop phenotyping traits at
IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE

MARCH 2021

IEEE Geoscience and Remote Sensing Magazine - March 2021

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