IEEE Geoscience and Remote Sensing Magazine - March 2021 - 214

early season crop (wheat, cotton, spring maize, sugarcane,
and rice) type classification based on the combination
of optical satellite and SAR data [302], [307]-[309]. Hao
et al. [307] used the improved artificial immune network
and Sentinel data to carry out early season crop mapping
in Hengshui, China, and the result indicated that the overall accuracy for summer crops and winter wheat was 98.55
and 99%, respectively. Jiang et al. [308] combined machine
learning algorithms and Sentinel-1A/2 time series data to
map sugarcane during the early season in Zhanjiang City,
China, and the outcome showed that the value of the kappa
coefficient was 0.902 and that the estimation accuracy of
the sugarcane mapping area was approximately 86.3%.
Compared with satellite platforms, UAVs can provide
ultrahigh spatial resolution for classifying crops during the
early season. UAVs equipped with RGB and multispectral
cameras have been employed to classify weeds and crops
[41], [311]. Zheng et al. [311] used RGB, NIR-G-B, and multispectral images from UAVs to detect rice plants during
the early season, and the results indicated that a spectralfeatures decision tree had a high classification accuracy
during the early growth stages. With the improvement of
the spatial and temporal resolution of images from satellites and UAVs, deep-learning algorithms, e.g., convolutional neural networks (CNNs) [312] and recurrent neural
networks [313], will ideally be employed to map early season crop types, with their advantages for image processing
and analysis.
CROP GROWTH CONDITION MONITORING
NITROGEN STRESS
Nitrogen deficiency in crops results in a decreased chlorophyll content, a slower growth rate, and less photosynthesis
as well as more sensitivity to diseases and pests [314]. These
changes in the chlorophyll content, growth rate, photosynthesis, and disease and pest occurrence can be monitored
using near-infrared and visible spectral reflectance [8], [29],
[34]-[36]. These factors have been estimated using UAVs to
monitor different nitrogen levels. The results demonstrated
that multispectral and hyperspectral images from UAVs
can estimate the biomass, nitrogen content, and grain yield
with multiple nitrogen treatments [34]-[36]. In addition, a
study by Camino et al. [35] showed that solar-induced chlorophyll fluorescence derived from hyperspectral images is
a good indicator of photosynthesis under multiple nitrogen
treatments in crop phenotyping experiments. Similarly,
multispectral and hyperspectral sensors are integrated with
phenomobiles to estimate the nitrogen content under different nitrogen treatments [67], [278]. Additionally, fluorescence sensors combined with UAVs and phenomobiles
have a great potential for the early monitoring of crop traits
under nitrogen stress in the field. Developing specific, lightweight spectral fluorescence sensors in combination with
UAVs and phenomobiles for the early detection of nitrogen
stress is a great opportunity.
214

WATER STRESS
Stomata are typically closed under water-stressed conditions, which reduces crop growth and photosynthesis and
may aggravate crop heat stress because of decreased transpirational cooling [8]. The canopy temperature is a good
indicator of the response of crops to water stress [315], and
scientists have reported that under water stress the measurement can be used to identify drought-adapted cultivars
[114], [123], [136], [268], [316]-[318]. Thermal images need
to be normalized to the ambient temperature and relative
humidity to control the effect of environmental conditions
on measurements. In a study, Jones et al. [268] indicated
that the relative value among genotypes was more important than the absolute value for crop phenotyping evaluations. In addition, the background temperature (soil and
dead leaves) of thermal images ideally should be eliminated from the signal of the green leaves using related imagesegmentation algorithms [121].
Using thermal cameras with UAVs and phenomobile
platforms enables the collection of more crop temperature
images in less time compared to ground-based approaches
[259], and it may assist in selecting water-stress-resistant
genotypes. The limitation of the crop growth status (such
as the biomass) owing to water stress could be estimated
using near-infrared and visible optical sensors [319].
Therefore, hyperspectral and multispectral cameras can
be applied in combination with thermal cameras to select
water-stress-resistant genotypes [66], [123]. Images from
thermal, multispectral, and hyperspectral cameras collected through the whole crop growing season can offer
important quantitative and qualitative data sets that could
be correlated with other ground-truthing data sets. These
methods will promote the uptake of high-throughput,
nondestructive sensors for estimating crop phenotyping
traits in the field and may finally take the place of traditional data collection methods.
DISEASES
Crop yield losses are persistent problems in agriculture due
to the widespread occurrence of pathogens, such as fungi,
viruses, nematodes, and bacteria. Advanced disease-monitoring technologies are required to minimize crop yield
losses [98]. Remote sensing technologies have been applied to estimate diseases [320] and monitor new outbreaks
worldwide [265]. However, the application of UAV-based
optical sensors for detecting crop diseases and the tolerance
of different crop varieties to maladies is less well developed
[8], [99]. Nebiker et al. [99] used a threshold of NDVI values from multispectral images to identify blight infestation in potatoes. Crop diseases are mainly detected using
near-ground optical sensors (such as hyperspectral, RGB,
and thermal cameras) [100], [103], [104]. Recent research
on detecting Huanglongbing (HLB) disease-infected citrus
trees demonstrated that a good classification accuracy of
HLB disease was obtained using the combination of visible,
near infrared, and thermal spectral bands [100].
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
