Detection of rice sheath blight using an unmanned aerial system with high-resolution color and multispectral imaging
Detection of rice sheath blight using an unmanned aerial system with high-resolution color and multispectral imaging
Dongyan Zhang 1 2 3
Xingen Zhou 1 3
Jian Zhang 0 1 3
Yubin Lan 1 3 4
Chao Xu 1 2 3
Dong Liang 1 2 3
☯ These authors contributed equally to this work. 1 3
0 College of Resources and Environment, Huazhong Agricultural University , Wuhan, Hubei , China
1 Funding: This research was financially supported by the National Natural Science Foundation of China (Grant No. 41771463, 41771469), and the Science and Technology Development Program of Anhui Province , 1604A0702016
2 Anhui Engineering Laboratory of Agro-Ecological Big Data, Anhui University , Hefei, Anhui , China , 2 Texas A&M AgriLife Research Center, Texas A&M University System , Beaumont, Texas , United States of America
3 Editor: Zonghua Wang, Fujian Agriculture and Forestry University , CHINA
4 College of Engineering, South China Agricultural University , Guangzhou, Guangdong , China
Detection and monitoring are the first essential step for effective management of sheath blight (ShB), a major disease in rice worldwide. Unmanned aerial systems have a high potential of being utilized to improve this detection process since they can reduce the time needed for scouting for the disease at a field scale, and are affordable and user-friendly in operation. In this study, a commercialized quadrotor unmanned aerial vehicle (UAV), equipped with digital and multispectral cameras, was used to capture imagery data of research plots with 67 rice cultivars and elite lines. Collected imagery data were then processed and analyzed to characterize the development of ShB and quantify different levels of the disease in the field. Through color features extraction and color space transformation of images, it was found that the color transformation could qualitatively detect the infected areas of ShB in the field plots. However, it was less effective to detect different levels of the disease. Five vegetation indices were then calculated from the multispectral images, and ground truths of disease severity and GreenSeeker measured NDVI (Normalized Difference Vegetation Index) were collected. The results of relationship analyses indicate that there was a strong correlation between ground-measured NDVIs and image-extracted NDVIs with the R2 of 0.907 and the root mean square error (RMSE) of 0.0854, and a good correlation between image-extracted NDVIs and disease severity with the R2 of 0.627 and the RMSE of 0.0852. Use of image-based NDVIs extracted from multispectral images could quantify different levels of ShB in the field plots with an accuracy of 63%. These results demonstrate that a customer-grade UAV integrated with digital and multispectral cameras can be an effective tool to detect the ShB disease at a field scale.
Data Availability Statement: All relevant data are
within the paper and its Supporting Information
Competing interests: The authors have declared
that no competing interests exist.
Rice is one of the most important food crops in the world, providing a major source of
nourishment for over half the world population. Rice diseases, especially sheath blight (ShB) caused
by Rhizoctonia solani AG1-1A, are among the most important factors limiting rice production
]. The ShB disease usually develops in the later tillering or early internode
elongation stage of rice. Infected tissue looks brown to yellow or even bleached in color. Symptoms
of the disease first appear on the lower sheaths of plants and then develop onto upper sheaths
and leaves. ShB spreads from plant to plant through the growth of the fungus and usually
forms in a circular pattern in the field [
]. Under favorable conditions, this disease spreads
quickly to top plant parts, causing lodgings of the plants. The disease occurs every year, causing
significant losses in grain yield and quality in the US and other rice-producing counties of the
]. ShB has become the second most economically-important disease in rice in the
Cultivar resistance can be the most effective, economic means for management of rice ShB
and other crop diseases. Plant phenotyping plays a critical role in breeding programs for the
development of cultivars of rice and other field crops with improved disease resistance. Plant
phenotyping also has been widely used in crop breeding programs to assess plant traits such as
plant growth, development, abiotic stress tolerance, architecture, physiology, ecology, and
]. Recently, optical sensor technologies have been developed and used to assist plant
phenotyping and disease detection and diagnosis . Digital (red, green, and blue)-,
multispectral-, hyperspectral-, flrorescence imaging- and thermal infrared-based sensors have been
utilized to characterize plant disease symptoms, detect different diseases, and even quantify
severity of several diseases in the laboratory and field [
]. These studies have been conducted
on wheat, barley and vegetables including tomatoes and cucumbers . However, no
significant advances in high-throughput phenotyping have been made on rice. This is probably due
to the flooding conditions under which rice plants grow, limiting the ability to use this remote
Unmanned Aerial Systems, one of new remote sensing platforms, has played an important
role in application in precision agriculture. This technology has the advantages of high-spatial
resolution, high efficiency, low costs, and flexibility in use [9±11]. Use of the technology can
facilitate the quick and accurate detection of plant diseases at a field scale and thus improve
disease management efficacy through on-time applications and/or site-specific applications of
In this study, a quadrotor unmanned aerial vehicle (UAV), equipped with digital (Red,
Green, and Blue bands) or multispectral (Near-infrared, Red edge, Red, Green to Blue bands)
camera, was utilized to capture high-spatial resolution imagery data of different ShB-resistant
rice cultivars and lines in research plots to characterize and detect the ShB disease in rice at a
field scale. The objectives of this work were to: 1) quantify the changes in color, texture and
structure associated with the symptoms of ShB at the level of canopy by high-resolution digital
and multispectral images, 2) determine the correlations between color features or vegetation
indices and ShB severity, and 3) evaluate the advantages and disadvantages of vegetation
indices-based detection of ShB.
Materials and methods
A field trial was conducted at Texas A&M AgriLife Research Center (30Ê03'53.4"N, 94Ê
17'38.7"W), Beaumont, Texas, USA. Sixty-seven rice cultivars and elite lines with different
levels of resistance to ShB (Table 1) were selected to characterize and detect their differences in
the development of the disease in research plots. Rice was drill seeded in 67 plots consisting of
seven 2.4-m rows, spaced 18 cm between rows on May 8, 2015. Each plot was divided into two
equal-length sections. One end section of the plot was inoculated with the ShB pathogen by
2 / 14
VS = Very susceptible, S = Susceptible, and MS = Moderately susceptible.
Original field data in Table 1 are included in the files S1 Table, S1 Fig and S2 Fig in the Support Information.
manually broadcasting 100 ml of the R. solani inoculum on July 10; the other end section was
left with no pathogen inoculation, serving as the disease-free control. Our previous
observations showed that the ShB pathogen has very limited ability to spread from one section to
]. This experiment was specifically designed to allow for observing and comparing
the development of the disease at different growth stages of rice. Fertility, irrigation, and weed
and insect pest control followed local recommendations [
]. On August 23 and 30, the severity
of ShB was rated on a scale of 0 to 9 where 0 represents no symptoms and 9 represents most
severe in symptoms and damage (leaves dead or plants collapsed). The resistance of selected
3 / 14
cultivars to ShB was indicated as very susceptible (VS), susceptible (S), and moderately
susceptible (MS) (Table 1). Both disease assessment dates corresponded approximately 3 and 2 weeks
from maturity of most of the rice cultivars and elite lines evaluated in this study.
Unmanned aerial system
The 4-rotor UAV equipped with high-resolution digital or multispectral camera was used to
collect imagery data in the research plots. The UAV is the Phantom 2 Vision+ (Da-Jiang
Innovations Science and Technology Co., Ltd, Shenzhen, China) with the advantages of stable
3-axis gimbal, automatically GPS (Global Position System) data recording, and user-friendly
flight control. The digital camera offers the image quality of 14 Megapixels with the size of the
4384×3288 pixel array for Blue, Green and Red bands. The multispectral camera Micasense
RedEdgeTM (MicaSense, Inc., Seattle, WA, USA) can acquire 12-bit raw image in five narrow
bands from blue, green, red, red edge to near-infrared (NIR). The 5-channel camera could
measure plant reflectance to capture subtle information about crop stress more accurately
than the regular 3-channel (blue, green, and red) camera [
Data collection and processing
Image capture. The digital and multispectral cameras were used to capture high-spatial
resolution images over the field plots. Before imagery data were collected, optimal exposure
time for different cameras was selected based on weather conditions; the actual parameters
were set according to the instructions of the software [
]. During the flights, the cameras
were positioned at the nadir at two altitudes, 27 m above the ground to cover all 67
experimental plots and 5.5 m to cover four plots in each image. The appropriate flight overlaps were
adjusted by different flying heights. In order to stably capture the images, the UAV was
instructed to fly along the experimental plots with the wind direction and the flights were set
at a speed between 0 and 10 m/s depended on wind speed. The flight experiments were
conducted from 12:30 pm to 2:00 pm on Aug 23rd and 30th. The weather conditions were favorable
for the flights, with partly cloudy and breeze condition during the flights.
These imagery data are included in the S1 File in the Support Information File.
Image processing. The ENVI (Exelis Visual Information Solutions, Boulder, CO, USA)
software was used to extract color features from the acquired images and convert them to
different color spaces. Previous studies demonstrate that color space transformation can improve
the presentation of information, and the transformed images can be interpreted more easily
than the original ones [
]. In this study, the digital images (Red, Green and Blue bands) of
different levels of ShB severity were transformed to hue, lightness and saturation (HLS), and then
the mean values of HLS were extracted.
Vegetation indices calculation. Five kinds of vegetation indices (VIs), including
Normalized Difference Vegetation Index (NDVI), Ration Vegetation Index (RVI), Difference
Vegetation Index (DVI), Normalized Difference Water Index (NDWI) and Red Edge (RE), were
calculated from the acquired multispectral images. And then, the NDVIs-change maps of
different levels of disease severity were generated so as to illustrate that multispectral imagery
data could detect the symptoms and development of the ShB disease at a field scale.
Collection of ground truths. Ground-based NDVI values of rice cultivars and lines were
measured by GreenSeeker handheld crop sensor (Trimble Navigation Limited, California,
USA). The operating mechanism of the sensor is based on the fact that green plants absorb
most of the red light and reflect most of the infrared light. The relative strength of the detected
light is a direct indicator of the density of the foliage within the sensor view. The denser and
more vigorous the plant, the greater the difference is observed between the reflected light
4 / 14
signals. When taking the ground NDVI readings in this study, the GreenSeeker handheld crop
sensor was held 100 cm above the canopy of rice plants, with an ovalfield of view covering the
area of 42 cm2. Multiple measurements were taken in each plot to increase the accuracy of the
NDVI values that were representative of the levels of ShB severity. A total of 134 NDVI average
readings (the relative ground truths) were collected from the pathogen-inoculated and
uninoculated control areas of 67 plots in the trial. On the same assessment dates, severity of ShB
was rated at a scale of 0±9 based on the symptoms of the disease as described before.
Data analyses and drawing. The Pix4D mapper (Pix4D Inc, Lausanne, Switzerland) was
used for UAV image processing. ArcGIS 9.1 (Esri, Redlands, CA, USA) was used for geospatial
data analysis and mapping. PASW Statistic 18 (SPSS Inc., Chicago, IL, USA) software was used
for statistical analyses. Determination coefficient and root mean square error (RMSE) were
utilized to evaluate the accuracy of correlation model [
Results and analyses
Qualitative ShB detection based on high-resolution images
Characterizing ShB at two development stages. In this study, color was selected as the
most important characteristic to detect the ShB disease. This is because infected plant tissue
usually changes its color from green (healthy tissue) to brown-to-yellow (diseased tissue) with
the development of the disease. Color has been demonstrated as the most effective means to
distinguish different image targets and achieve object identification among the morphological
features, such as color, texture, size, etc. extracted from images in previous studies [
As shown in Fig 1A and 1C, the color of rice canopy in the RBG images of field plots
changed significantly with the growth of rice plants. However, no significant differences in
color change were observed between ShB-infected and uninfected control areas in most of 67
plots at either assessment date. Previous studies demonstrate that the illumination factor
produces color features difference of the images, but color space transformation can eliminate
illumination difference and strengthen color features such as hue, lightness, and saturation for
different targets [
]. Therefore, RGB images were transformed to HLS images to differentiate
the ShB-infected areas. The images in Fig 1B and 1D clearly showed yellow ShB-infected areas
and green or blue healthy areas for all plots at either observation date. Moreover, these
differences in images were more apparent at the Aug 30th assessment date (Fig 1D) than at the Aug
23rd assessment date (Fig 1B) although they were only seven days in apart. Meanwhile, these
differences were more obvious in plots with susceptible cultivars and lines than in plots with
resistant or partially resistant ones. These results illustrate that HLS-transformed color space is
a useful means to qualitatively detect rice ShB in the field.
Comparison of ShB detection using RGB and multispectral images. It is well known
that multispectral cameras can provide good service to precision agriculture management such
as disease and pest detection, drought monitoring, nutrition diagnosis, and spray drift
evaluation [19±22]. In this study, high-resolution RGB and 5-bands multispectral images were
analyzed to detect ShB-infected areas in the field plots. As shown in Fig 2A and 2B, the imagery
data collected from multispectral camera could reflect field environments (green weed plants,
ground earth, plots shadow, etc.) and canopy characteristics (color, texture, and structure
information, etc.) more accurately. When the false color image (Fig 2B) was transformed into
HLS combination (Fig 2C), it resulted in more apparent display of the ShB-affected areas with
yellow to white in color in the 67 experimental plots. In addition, the NDVIs map of rice
cultivars and lines in the field plots was also developed (Fig 2D) after NDVI values were calculated.
The darker the image color, the more severe the ShB disease. In Fig 2D, the diseased areas
were clearly differentiated from the healthy areas in each of the plots. These differentiations
5 / 14
Fig 1. Original RGB and HLS transformation images of 67 field plots on Aug 23rd (A and B) and Aug 30th(C and
D). The ratings of resistance to ShB with very susceptible (VS), susceptible (S), and moderately susceptible (MS) were
indicated in selected plots (cultivars or lines).
were more effective compared to the differences made by RGB, False, or HLS images (Fig 2A,
2B and 2C). It can be explained that red and near-infrared lights are more sensitive to the
changes in canopy color from green (healthy) to yellow (diseased) and the changes in canopy
structure from dense to sparse in density caused by the development of ShB [
the vegetation index NDVI is a good indicator of different levels of ShB observed in this study.
Quantitative ShB detection based on color feature parameters
Correlations between color features and ShB severity. The regular RGB has been used
to quantitatively detect disease and insect pests, and other crop stresses caused by drought and
nitrogen deficiency in agricultural production systems in previous studies [23±25]. In this
study, quantitative detection of ShB was also evaluated based on high-resolution RGB images.
The relationships between color features and disease severity were analyzed for both disease
assessment dates (Table 2). The color features were extracted from red, green, blue bands, and
color space transformed hue value. It is noted that the relationships between other color space
transformed values (light and saturation) and ShB severity were insignificant, so we only
presented the results of hue values in Table 2.
6 / 14
Fig 2. RGB, False, HLS, and NDVI images of 67 field plots on Aug 30th (A, B, C and D). The ratings of resistance to
ShB with very susceptible (VS), susceptible (S), and moderately susceptible (MS) were indicated in selected plots
(cultivars or lines).
As shown in Table 2, the determination coefficients (R2) were significantly low with a value
ranging from 0.038 to 0.251 for color features of RGB image at both disease assessment dates.
However, R2 value for color space hue on Aug 30th was 0.554, which was obviously higher than
other color features. Therefore, RGB-based color features are less effective to quantitatively
detect different levels of ShB, but color space transformation can improve its ability to quantify
the severity of the disease.
y = 0.017x+3.3356
y = 0.0185x+2.9465
y = 0.0223x+3.4965
y = -0.004x+5.7393
y = 0.0522x+1.3841
y = 0.0168x+4.3868
y = 0.0342x+2.9826
y = 0.0143x+4.6117
7 / 14
Correlations between VIs and ShB severity. Vegetation index is considered as a simple
and effective quantitative parameter to monitor the growth status and coverage of green
vegetation on the earth surface [
]. It has been extensively used to determine disease severity,
nutrition status, drought stress, and yield of several crops [27±29]. In this study, five kinds of
vegetation indices, NDVI, RVI, DVI, NDWI and RE, were chosen to determine their ability to
quantify ShB severity. NDVI had the highest R2 value (0.627) with the lower RMSE (0.0854)
compared to those of other VIs (Table 3). NDVI was also superior to all the color features
evaluated in this study (Table 2). Thus, these results indicate that NDVI has the best performance
on the detection of different levels of ShB in the field plots. This is because NDVI is sensitive to
the changes in the density of ground vegetation from high to low due to the damage caused by
the disease, and the changes in canopy color from green to yellow or white caused by plant
growth toward maturity [
]. For the other VIs, RVI is able to monitor the growth status of
green vegetation but its sensitivity decreases significantly when canopy coverage reduces to
less than 50% [
]. DVI is a good vegetation index to detect low-to-moderate levels of
vegetation coverage but it is more likely to be affected by backgrounds such as soil, water and shadow
]. NDWI is sensitive to the changes in liquid water content of vegetation canopies.
However, it can be affected by residual irrigation water in field plots . RE is able to detect
accurately the changes in leaf chlorophyll content and biochemical components of plant. Its
sensitivity, however, decreases significantly when the crop approaches the late stages of
Influence of spectral bands and their combination on the detection of ShB
Previous studies indicate that digital image can be a good data resource to detect crop diseases
[27±29]. In this study, high-resolution RGB and multispectral imagery data were acquired and
analyzed with the aim of qualitatively and quantitatively detecting the rice ShB in research
plots. The results demonstrate that color features extracted from RGB and multispectral
images could partially distinguish the canopy changes caused by the disease. However, it does
not provide sufficient information to differentiate different levels of ShB because of less
spectral wavelength and broad bands associated with the RGB camera used in this study. This
limits the potential application of UAV equipped with RGB camera to the detection of the disease
at a commercial scale and to the screening of ShB-resistant cultivars and elite lines in research
plots for the rice breeding program.
On the contrary, NDVI, one of five kinds of VIs calculated from multispectral imagery data
in this study, had a strong relationship with ShB severity. The R2 and RMSE values are,
respectively, 0.627 and 0.0852 for the correlation between image-based NDVIs and ShB severity
(Table 3), and 0.635 and 0.0854 for the correlation between ground-measured NDVIs and ShB
severity (Fig 3A). Meanwhile, we also analyzed the correlation between ground-measured
NDVIs and image-based NDVIs, the R2 and RMSE values are, respectively, 0.907 and 0.0854
(Fig 4). These results demonstrated that NDVIs calculated from multispectral imagery data
could provide better spectral information to differentiate different levels of ShB in the field
In addition, the correlations between ground-measured NDVIs and ShB severity for
different growth stages were compared. The R2 and RMSE values are, respectively, 0.635 and 0.0854
on Aug 30th (Fig 3A), and 0.346 and 0.0608 on Aug 23rd (Fig 3B). These results indicate that it
is insufficient to utilize NDVIs to quantify ShB severity at an earlier stage of disease
development. This is in agreement with the results obtained from previous studies [
]. More works
is needed to improve the differentiation accuracy of rice ShB at early growth stages. Using
8 / 14
y = -0.0513x+0.479
y = -2.0782x+9.6484
y = -7E-05x+7.4872
y = 17.929x+7.5417
y = 2.5935x+4.8018
hyperspectral sensor might be a way to characterize subtle details of crop diseases. Yang
demonstrated that VIs calculated from nano-level hyperspectral bands can assess the severity of
bacterial leaf blight in rice precisely . Mahlein pointed out that hyperspectral camera and
VIs derived from sensitive bands have data benefits to diagnose plant diseases and phenotypes
]. Our previous research also confirmed that hyperspectral sensor with broad wavelengths
and narrow bands can capture subtle changes in the symptoms of ShB in rice at an earlier
stage of disease development [
]. Therefore, additional investigations are needed to explore
the potential use of narrow-bands multispectral and hyperspectral sensors to effectively detect
ShB at early stages of development.
Influence of sampling area selection on the detection of ShB. Selection of sampling
areas in the field can affect the accuracy of detecting ShB severity due to the nature of cluster
distribution of the disease in infected fields [
]. In this study, three methods of selecting a
sample area, circular sampling, rectangle sampling, and manual selection by a plant
pathologist, were evaluated to determine their efficacy of detecting ShB severity (Fig 5). These three
sampling methods were all acceptable in efficacy with no significant differences in R2 and
RMSE. The R2 and RMSE values for the correlation between image-extracted NDVIs and ShB
severity are, respectively, 0.627 and 0.0852 for the circular area sampling method (Fig 6A),
0.598 and 0.0825 for the rectangle area sampling method (Fig 6B), and 0.635 and 0.0854 for
the selection of sampling area by an investigator (Fig 3A). The reason for the method of
Fig 3. Correlations between ground-based NDVIs and ShB severity on Aug 30th (A) and Aug 23rd (B).
9 / 14
Fig 4. Correlation between image-based NDVIs and ground-based NDVIs on Aug 30th.
selecting sample area by an investigator having a highest value of R2 is because such manual
selection can effectively target diseased areas for disease assessment. The R2 value is relatively
greater for the circular area sampling method than for the rectangle area sampling method.
This is due to fact that the circular area sampling method could cover more percentage of
diseased areas in the field plots than the later one. The rice ShB tends to spread more likely in
circular pattern than in rectangle pattern from a point focus of inoculation in the field . This
resulted in a decreased error in disease detection made by the circular area sampling method.
Therefore, selecting an optimal sampling method based on the spread pattern of a disease is
also a factor to improve disease detection efficacy when using image-based UAV platforms.
Influence of flying altitude on the detection of ShB
Flying altitude can be an important factor affecting the ability of UAVs to detect the details of
ShB development and plant canopy in the field. In this study, UAV imagery data at an altitude
of 5.5 and 27 m above the ground were collected and evaluated on two assessment dates. At
the altitude of 5.5 m, detailed information about the lodging caused by ShB, and canopy
components such as leaf color, rice ears and other canopy structures were clearly displayed in
10 / 14
Fig 5. Three sampling methods. They were assigned as rectangle area sampling (black rectangle), circular area
sampling (red circle), and manual sampling by a plant pathologist (yellow dashed lines).
original RGB (Fig 7A and 7C), but there was no significant difference in the transformed HLS
images (Fig 7B and 7D) on Aug 23rd and Aug 30th. At this altitude, however, quantitative
analyses were not conducted for multispectral images since the flight height was too low to have the
high quality of imagery data. At the altitude of 27 m (Figs 1 and 2), we were able to not only
Fig 6. Correlations between image-based NDVIs and ShB severity using circular sampling (A) and rectangle sampling (B) methods on Aug 30th.
11 / 14
Fig 7. RGB and HLS images at 5.5-m flying altitude on Aug 23rd (A and C) and Aug 30th (B and D).
qualitatively analyze the color characteristics of rice canopy obtained from RGB- and
multispectral cameras but also quantitatively evaluate the performance of NDVIs extracted from
multispectral imagery on the detection of the rice ShB in the field. The results of this study indicate
that such flying altitude was acceptable but further research is still needed to explore the
optimum flying altitudes that can maximize the ability of UAV to detect ShB in rice.
A commercial UAV equipped with a high-resolution RGB and multispectral cameras was used
to capture imagery data. Collected imagery data were then processed and analyzed to
characterize the development of ShB and quantify different levels of the disease in the field. Ground
truth data of ShB severity and NDVIs were also measured for comparisons. Hue value of color
features obtained from RGB images can clearly differentiate the infected areas from the
healthy, uninfected areas in field plots on August 30th. NDVIs calculated from the
multispectral images could quantify different levels of the disease in the plots with an accuracy of 63%.
Image-based NDVI values were strongly correlated with ground-NDVI values with R2 of 0.91.
There was a good relationship (R2 = 0.64) between ground-NDVI values and disease severity.
These results demonstrate for the first time that image-based NDVI is an effective means to
detect ShB and quantify the severity of the disease at a field scale.
Combined use of an UAV with high-spatial resolution camera is an innovation that has the high
potential for quick and accurate detection of ShB, one of the most important diseases in rice in the
world. This technology can aid in the scouting and monitoring process of this disease and reduce
the costs in time and effort associated with this process. This UAV system in the current form can
also assist crop breeders in breeding for rice cultivars with resistance to ShB. In addition, such new
UAV system developed from this research also has provided a basis to develop site-specific
precision fungicide application technology for control of this important disease in rice in the future.
S1 Table. Disease assessment data.
12 / 14
S1 Fig. Ground truth data (0823).
S2 Fig. Ground truth data (0830).
S1 File. Imagery data (0823±0830).
We greatly thank Dr. Chenghai Yang at the USDA ARS-AATU for providing experimental
devices, and Drs. Guo-Zhong Zhang and Huai-Bo Song for data collection in the field. This
research was financially supported by the National Natural Science Foundation of China
(Grant No. 41771463, 41771469), and the Science and Technology Development Program of
Anhui Province (1604A0702016).
Formal analysis: Dongyan Zhang, Jian Zhang.
Funding acquisition: Chao Xu, Dong Liang.
Investigation: Dongyan Zhang, Xingen Zhou, Jian Zhang.
Methodology: Dongyan Zhang, Yubin Lan, Chao Xu.
Supervision: Dong Liang.
Writing ± review & editing: Xingen Zhou.
13 / 14
1. RPH. Rice Production Handbook . 2014 ; Available from: https://beaumont.tamu.edu/eLibrary/ RiceResource/Rice_Production_Handbook.pdf.
2. Zhou XG , Jo Y . - K. Disease management. The Texas Rice Production Guidelines. Texas AgriLife Research and Texas AgriLife Extension. B-6131 . 2015 . pp. 44 ± 56 .
3. Zhang DY , Lan YB , Zhou XG , Murray SC , Chen LP . Research imagery and spectral characteristics of rice sheath blight using three portable sensors . ASABE International Meeting , New Orleans, Louisiana. 2015 , ID: 152190801 .
4. Yang WN , Duan LF , Chen GX , Xiong LZ , Liu Q. Plant phenomics and high-throughput phenotyping: accelerating rice functional genomics using multidisciplinary technologies . Curr Opin Plant Biol . 2013 ; 16 ( 2 ): 180 ± 187 . https://doi.org/10.1016/j.pbi. 2013 . 03 .005 PMID: 23578473
5. Granier C , Vile D . Phenotyping and beyond: modelling the relationships between traits . Curr Opin Plant Biol . 2014 ; 18 : 96 ± 102 . https://doi.org/10.1016/j.pbi. 2014 . 02 .009 PMID: 24637194
6. Mahlein AK . Detection by imaging sensors±Parallels and specific demands for precision agriculture and plant phenotyping . Plant Dis . 2016 ; 1 ± 11 .
7. Kuska M , Wahabzada M , Leucker M , Dehne HW , Kersting K , Oerke EC , et al. Hyperspectral phenotyping on the microscopic scale: Towards automated characterization of plant-pathogen interactions . Plant Methods . 2015 ; 11 ( 1 ): 28 .
8. Bauriegel E , Herppich W. Hyperspectral and chlorophyll fluorescence imaging for early detection of plant diseases, with special reference to Fusarium spec. infections on wheat . Agriculture . 2014 ; 4 : 32 ± 57 .
9. Mulla DJ . Twenty five years of remote sensing in precision agriculture: Key advances and remaining knowledge gaps . Biosyst Eng . 2013 ; 114 ( 4 ): 358 ± 371 .
10. Ballesteros R , Ortega JF , Hernandez D , Moreno MA . Applications of georeferenced high-resolution images obtained with unmanned aerial vehicles. Part I: Description of image acquisition and processing . Precis Agric . 2014 ; 15 ( 5 ): 79 ± 592 .
11. Peña JM , Torres-SaÂnchez J , Castro AID , Kelly M , LoÂpez-Granados F . Weed mapping in early-season maize fields using object-based analysis of unmanned aerial vehicle (UAV) images . PLoS ONE . 2013 ; 8 ( 10 ): e77151. https://doi.org/10.1371/journal.pone. 0077151 PMID: 24146963
12. Zhou XG , Liu G , Tabien RE , Vawter J . Evaluation of rice cultivars and elite lines for resistance to diseases in Texas . PDMR . 2010 ; 5:FC052: 1±2 .
14. http://www.dji.com/cn/phantom-2 - vision-plus
15. Kruse FA , Raines GL . Technique for enhancing digital color images by contrast stretching in Munsell color space . Center for Integrated Data Analytics Wisconsin Science Center . 1984 ; 755 ± 760 .
16. Yang CM . Assessment of the severity of bacterial leaf blight in rice using canopy hyperspectral reflectance . Precis Agric . 2010 ; 11 ( 1 ): 61 ± 81 .
17. Arora A , Gupta A , Bagmar N , Mishra S , Bhattacharya A . A plant identification system using shape and morphological features on segmented leaflets: Team IITK . CLEF. 2012 ; 2012 .
18. Hemmin J , Rath T. PA-Precision Agriculture : Computer-vision-based weed identification under field conditions using controlled lighting . J Agr Eng Res . 2001 ; 78 ( 3 ): 233 ± 243 .
19. Qin ZH , Zhang MH . Detection of rice sheath blight for in-season disease management using multispectral remote sensing . Int J Appl Earth Obs . 2005 ; 7 ( 2 ): 115 ± 128 .
20. Berni JAJ , Zarco-Tejada PJ , Suarez L , Fereres E . Thermal and narrowband multispectral remote sensing for vegetation monitoring from an unmanned aerial vehicle . IEEE T Geosci Remote . 2009 ; 47 : 722 ± 738 .
21. Lan YB , Thomson SJ , Huang YB , Hoffmann CW , Zhang HH . Current status and future directions of precision aerial application for site-specific crop management in the USA . Comput Electron Agr . 2010 ; 74 ( 1 ): 34 ± 38 .
22. Huang YB , Thomson SJ , Ortiz BV , Reddy KN , Ding W , Zablotowicz RM , et al. Airborne remote sensing assessment of the damage to cotton caused by spray drift from aerially applied glyphosate through spray deposition measurements . Biosyst Eng . 2010 ; 107 ( 3 ): 212 ± 220 .
23. Mewes T , Franke J , Menz G . Spectral requirements on airborne hyperspectral remote sensing data for wheat disease detection . Precis Agric . 2011 ; 12 ( 6 ): 795 ± 812 .
24. Bausch WC , Khosla R. QuickBird satellite versus ground-based multi-spectral data for estimating nitrogen status of irrigated maize . Precis Agric . 2010 ; 11 : 274 ± 290 .
25. Yengoh GT , Dent D , Olsson L , Tengberg AE , Tucker CJ . The use of the Normalized Difference Vegetation Index (NDVI) to assess land degradation at multiplescales: A review of the current status, future trends, and practical considerations. Lund University Center for Sustainability Studies (LUCSUS), and The Scientific and Technical Advisory Panel of the Global Environment Facility (STAP/GEF ). 2015 .
26. Haboudane D , Miller JR , Pattey E , Zarco-Tejada PJ , Strachan IB . Hyperspectral vegetation indices and novel algorithms for predicting green LAI of crop canopies: modeling and validation in the context of precision agriculture . Remote Sens Environ . 2004 ; 90 ( 3 ): 337 ± 352 .
27. Devadas R , Lamb DW , Simpfendorfer S , Backhouse D. Evaluating ten spectral vegetation indices for identifying rust infection in individual wheat leaves . Precis Agric . 2009 ; 10 : 459 ± 470 .
28. Robertson N A , Gitelson A , Peng Y , Viña A , Arkebauer T , Rundquist D . Green Leaf Area Index Estimation in Maize and Soybean: Combining Vegetation Indices to Achieve Maximal Sensitivity . Agron J . 2012 ; 104 ( 5 ): 1336 .
29. Gao BC . NDWIÐA normalized difference water index for remote sensing of vegetation liquid water from space . Remote Sens Environ . 1996 ; 58 ( 3 ): 257 ± 266 .
Wu W , Huang J , Cui KH , Nie LX , Wang Q , Yang F , et al. Sheath blight reduces stem breaking resistance and increases lodging susceptibility of rice plants . Field Crop Res . 2012 ; 128 ( 2 ): 101 ± 108 .