=Paper=
{{Paper
|id=Vol-3126/paper50
|storemode=property
|title=Validation of data obtained after field sensing using UAV for management of future crops
|pdfUrl=https://ceur-ws.org/Vol-3126/paper50.pdf
|volume=Vol-3126
|authors=Natalia Pasichnyk,Dmytro Komarchuk,Oleksiy Opryshko,Yurii Gunchenko,Sergey Shvorov,Oksana Zui
}}
==Validation of data obtained after field sensing using UAV for management of future crops==
https://ceur-ws.org/Vol-3126/paper50.pdf
Validation of Data Obtained After Field Sensing Using UAV for
Management of Future Crops
Natalia Pasichnyk1, Dmytro Komarchuk2, Oleksiy Opryshko3, Yurii Gunchenko4, Sergey
Shvorov5, Oksana Zui6
1,2,3,5
National University of Life and Environmental Sciences of Ukraine, Heroyiv Oborony st., 15, Kyiv, 03041,
Ukraine
4,6
Odessa I.I.Mechnikov National University, Dvoryanskaya str., 2, Odessa, 65082, Ukraine
Abstract
UAVs are innovative equipment for monitoring fields that are free from a lot of the
disadvantages of satellites such as availability, low cost, and high image resolution. However,
the issues of quality, reproducibility, and suitability for crop management processes remain
relevant. Now, the issue of assessing the suitability of the results of spectral monitoring of
plantations in relation to the condition of plants has not been resolved. Since spectral monitoring
is a necessary component in the concept of crop management, the development of a
methodology for assessing the suitability of remote monitoring spectral data for the calculation
of agrochemical practices was the purpose of the work. According to the publications, the
dependence of the number of pixels on the values of the intensity of color components for plants
and soil is described by the Gaussian distribution. Deviation from such distribution is caused
by the imposing of distributions from various objects fixed on a photo. The experimental test
was carried out on the basis of wheat, using the results obtained during 2017-2020 when
considering the stresses of nutrient deficiency and technological nature. The investigation found
experimental evidence that the pixel distribution of plantations on the example of the wheat
crop is described by the Gaussian distribution. It was found that the analysis of the
correspondence of the nature of the distribution on the spectral channels, namely the presence
of several max peaks that affects the value of the distribution maximum may indicate the
presence of foreign inclusions or a transitional stage of vegetation. The suitability of the data
can be assessed on the basis of the reference values of the width of the distribution on the
spectral channels. Vegetation indices GNDVI and RNDVI were unsuitable for assessing the
suitability of the data based on the parameters of the pixel distribution of the image in the
experimental plots. This determines the feasibility of introducing in the sets of regular
vegetation indices of geographic information systems additional packages that reflect the
spectral channels.
Keywords 1
UAV, spectral monitoring, crop management, data validation
1. Introduction of fundamental shortcomings of satellites in terms
of availability, cost, image resolution. However,
the issues of quality, reproducibility, and
UAVs are innovative equipment for
suitability for crop management processes remain
monitoring fields, which are deprived of a number
ISIT 2021: II International Scientific and Practical Conference
«Intellectual Systems and Information Technologies», September
13–19, 2021, Odesa, Ukraine
EMAIL: N.Pasichnyk@nubip.edu.ua (A. 1);
dmitruyk@gmail.com (A. 2); sosdok@nubip.edu.ua (A. 3);
gunchenko@onu.edu.ua (A. 4); ozon.kiev@nubip.edu.ua (A. 5);
oks.zuj@gmail.com(A. 6)
ORCID: 0000-0002-2120-1552 (A. 1); 0000-0003-3811-6183 (A.
2); 0000-0003-3358-1297 (A. 3); 0000-0003-4423-8267 (A. 4);
0000-0001-6433-3566 (A. 5); 0000-0001-9520-4441 (A. 6)
©️ 2021 Copyright for this paper by its authors. Use permitted under Creative
Commons License Attribution 4.0 International (CC BY 4.0).
CEUR Workshop Proceedings (CEUR-WS.org)
relevant. More often, designers focus on the others (2021) in [2]. Information on plant
improvement of spectral equipment, but there are dimensions is useful for determining stress
also methodological problems in the perception conditions, but in the early stages of the growing
and interpretation of information from devices of season, accurate image resolution is required for
technical vision. Thus, most of the vegetation accurate identification, which can only be
indices currently used to interpret UAV data, such obtained from low altitudes, which will not
as NDVI, were developed for satellite platforms contribute to the scalability of technology on an
with their inherent low image resolution when industrial scale. An alternative technical means
each pixel had a group of plants. The indices for estimating plant dimensions are LiDARs
developed on the basis of the soil line concept described in the review article by Yue Pan and
were primarily intended to assess the availability others (2019) in [3]. However, such innovative
of biomass, and crop management issues require equipment for small plants, with a leaf width of
other methodological approaches to crop several millimeters, according to Tai Guoa and
monitoring. It should be borne in mind that the others (2019) in [4].
implementation of agrochemical measures, in Another approach is based on the use of
particular fertilization should be carried out only reference values of plant spectral indicators to
at certain stages of the growing season. However, identify the spread of forest pests described in Per-
the state of plant development is determined by Ola Olsson and others (2016) in [5]. The estimate
many factors, including the state of mineral is based on recording the deviation from the
nutrition, water supply, etc., so within one field seasonal changes of the NDVI index is designed
there may be a situation when the plants are at for different stages of the growing season because
different stages of the growing season. satellite imagery is carried out at high intervals
Accordingly, in such situations, the calculation of and you can select data for uniquely the same
the mean value over the site, which is inherent in stage of the growing season. A similar approach
satellite solutions, is erroneous. At present, the to the selection of spectral data from an existing
issue of assessing the suitability of the results of array of rapidly changing data is shown in the
spectral monitoring of plantations in relation to work of Ameer Shakayb Arsalaan and others
the condition of plants has not been resolved. (2016) in [6] on the example of forest fires.
Since spectral monitoring is a necessary However, under normal conditions, farms in crop
component in the concept of crop management, management should be able to decide on the basis
the development of a methodology for assessing of a single departure on the need for additional
the suitability of remote monitoring spectral data flights that require free equipment.
for the calculation of agrochemical practices was An original approach to the identification of
the purpose of our work. plants in terms of changes in their dimensions on
the example of sugar beet is shown in the work of
2. The state of the issue Yang Cao Liu and others (2020) in [7].
Researchers have proposed a new wide-dynamic-
range vegetation index (WDRVI) where an
The spectral performance of objects critically additional coefficient is introduced for the
depends on the state of illumination, and the infrared channel. However, in production, the
reproducibility of data is tried to ensure by a achieved accuracy increase of up to 5% should
combination of technical and organizational still recoup the cost of determining the
measures. The work of Helge Aasen and others dynamically changing coefficients for the infrared
(2015) in [1] considered the construction of 3D channel. That is, the most promising approach is
models of plants, where to ensure accuracy, they based on the comparison of spectral indices with
proposed a method of combining data from certain reference samples.
several flights. Despite the interesting and Spectral indicators of plants, even those that
encouraging results, such a technique will require are in the same stage of the growing season have
several flights in a row from different directions, some differences. To obtain the average value for
which is unsuitable for industrial-scale in plants when fixing the soil in a photograph,
conditions of time shortages. An approach to Yaokai Liu et al. (2012) in [8] proposed the use of
determine the features of the dome of plants in the Gaussian distribution combinations where the
mass phenotyping of plants using UAVs based on ranges belonging separately to plants and soil
a comparison of the obtained portraits with were recorded. Positive results were obtained, but
reference templates is shown in Fusang Liu and
the resolution of images from a height of 3 m was
very high, which is difficult to implement on an the color components for plants and soil is
industrial scale. According to the data presented described by the Gaussian distribution. Deviation
in the work of Guangjian Yan and others (2019) from such distribution is caused by the imposing
in [9], when the resolution of the images is of distributions from various objects fixed on a
reduced, the ability to select individual ranges photo. However, experiments were performed in
corresponding to the soil and plants is lost. hospitals where the plants were in one phase of the
Improving identification by estimating the growing season in the air-dry state of the soil,
intensity distribution of color components is respectively, it is advisable to check the suitability
shown in André Coy et al. (2016) in [10] where of the method and in moist soil.
the CIE L * a * b * space model was used instead
of the RGB color model. The authors have 3. Materials and research software
proposed threshold values to determine the area of
the dome, but this approach will be effective only and hardware
in the initial stages of the growing season when in
particular the shade on the lower tiers of plant The research was carried out on the basis of
leaves can be neglected. The method was wheat, using the results obtained during 2017-
improved in the work of Linyuan Li et al. (2018) 2020. Stresses due to lack of nutrients were
in [11], when the identification of soil and plants studied in the fields of the long-term stationary
was attempted on the basis of the Gaussian half- experiment of the Department of Agrochemistry
distribution. This approach allows you to identify and Plant Quality of NULES of Ukraine, where
2 components, but in the case of 3 components, its fertilizer application systems are studied.
efficiency is questionable. Technological stresses were studied on and in the
Thus, based on the analysis of the literature, production fields of farms in the Kyiv region. In
we can conclude that the dependence of the fig. Green Chlorophyll index distribution maps
number of pixels on the values of the intensity of are presented (Fig. 1).
Figure 1: Green Chlorophyll index distribution maps on the research hospital on the left and
production fields (on the right) are created by Slantrange software. Blue intersections highlight
checkpoints for accurate positioning of pixels of different spectral channels and index distribution
maps
The experiments were performed in the optical Dolia and others (2019) in [13] (2019).
range using a standard UAV camera DJI Phantom Multispectral studies using the infrared range
3+. A description of the methodology of were performed using the Slantrange 3p system
experimental research was covered in the work of and Slantview software (version 2.13.1.2304)
V. Lysenko and others (2017) in [12] and M. designed specifically for this sensor equipment. A
feature of Slantview software is the ability to
quickly and autonomously create vegetation As can be seen from the above data when using
distribution maps directly in the field. Slantview the proposed method, it was found that the value
software compiles a general orthophoto from of the maximum distribution shifted by 2 units,
images, corrects for lighting, and provides the while reducing the width w by 3 units. The
user with ready-made maps of the distribution of presence of the Max2 distribution can be
vegetation indices such as various NDVI variants. explained both by the presence of shadow on the
Slantview software can export data to geotiff lower and upper leaves and by the fixation of the
format. Areas of rapeseed with and without signs soil.
of technological stress were considered for The proposed approach to the processing of
research. Data on individual spectral channels and experimental results will be effective if the
vegetation indices calculated by the Slantview condition Max1≫ Max2 is satisfied. In practice, a
program were considered. The research situation is possible when plants of the same crop
methodology is described in the work of S. are in the field at the same time, but at different
Shvorov and others (2020) in [14]. Maximum stages of the growing season or in a fundamentally
detail (GSD 0.04 m / pixel) was obtained from the different physiological state, such as the
Slantview software image window (available appearance of a flag leaf, which was recorded on
NDVI index variants - Green, Red, and RedEdge). 06.08.2018. According to the presented in fig. 3
Monochrome images were used to study the data Max1≅Max2, so the approach was used
results on separate spectral channels (image when at the first stage separately determined
window), which were stored in BMP format to separately 2 Gaussian distributions, after which
ensure the completeness of the information. To do the calculations were carried out according to the
this, a copy of the screen was saved in Paint method proposed in section 3.
(Microsoft Windows 7.0 Sp.1).
600
4. The results and discussions were Wheat
All xc=170, w=28;
obtained 500
Max1 xc=156, w=16;
Max2 xc=193, w=12;
Number of points
400 Max3 xc=118, w=17;
In fig. 2 shows the results of calculations for Max1+Max2+Max3
the red component for experimental data obtained
300
on 2017.05.05 in studies of the impact on the
spectral indicators of the state of mineral nutrition
200
using a universal camera FC200 (a standard tool
for UAV DJI Phantom 3). 100
0
1000 20 40 60 80 100 120 140 160 180 200 220 240 260
Wheat
All xc=116, w=25; Red color intensity
Max1 xc=118, w=22;
800
Max2 xc=64, w=23.
Figure 3: The results of approximation of the
dependence of the number of pixels on the
Number of points
600 intensity of the red component of the color for
winter wheat (2018.06.08 - there is a flag sheet)
400
Detection of the presence of several individual
200 maxima can be done based on the magnitude of
the distribution when using to approximate the
0
experimental data. For the presented data, the
0 20 40 60 80 100 120 140 160 180 200 value was 28 while in the remaining sections was
Red color intensity 18… 23.
Figure 2: The results of the approximation of the Based on the obtained results, the results
dependence of the number of pixels on the obtained by approximating all the data by a single
intensity of the red component of color Gaussian dependence (All) are incorrect because
(05.05.2017) they do not correspond to any of the distribution
maxima. That is, monitoring was performed when
the plants were in a transitional state and carried out in production fields near the village of
monitoring should be repeated after a few days Gvardiyske with the coordinates of lat. 50,0347
when the vast majority of plants in the field are in long. 30,0286 is presented in fig. 5.
a single stage of vegetation. For automatic According to the results obtained under stress
processing of monitoring results, reference values conditions, the width of distribution on both the
for distribution parameters can be obtained in green and red channels is 1.5≤ times greater than
stationary experiments, etc. in healthy plants. On the red channel, regardless
For universal digital cameras in the optical of the presence of technological stresses, 2
range, such as FC200, strict compliance with the pronounced maxima of distribution were not
selectivity of light filters is not required, so to observed, in contrast to the results in Figs. 4,
verify the results, a study was conducted using a regardless of the channel, the coefficient of
specialized spectral complex Slantrange 3. The determination is 0.98≤R2. According to the
results of mineral nutrition studies are presented authors, the difference in plant development is
in Fig. 4. explained by the difference in climatic factors due
to the location of the plots, so the production field
is protected by dense forest strips in contrast to the
2
3000 G Fertilizers MIN w=7.1, R =0,98; used area of the experimental hospital.
2
G Fertilizers MAX w=3.6, R =1;
2
R Fertilizers MIN w=18, R =0.84;
2500 2
R Fertilizers MAX w=9.8, R =0.98;
260
Number of pixels
2000 G Teh.str. +, w=8.4;
240
G w=5.5;
220 R Teh.str. +, w=9;
1500
200 R w=5,6.
180
1000
Number of pixels
160
140
500
120
100
0
20 40 60 80 100 80
Spectral channel 60
Figure 4: Dependence of the number of pixels on 40
20
the value of the intensity of the green (G) and red
0
(R) components of the color and the wall of 50 100
mineral nutrition at a dose of mineral fertilizers Spectral channel
(Fertilizers MAX) and without fertilizers Figure 5: Dependence of the number of pixels on
(Fertilizers MAX). Date of research 2020.04.27 the intensity of the green (G) and red (R)
components of color and the presence of
When approximating the experimental data by technological stresses (The.str). Date of research
the GaussAmp dependence, the distribution width
2020.04.27
for the green channel was 7.1 for plants under
stress and 3.6 for healthy plants, respectively, at
0,98≤R2. For the red component, regardless of the 5. Wheat (distribution maps of
state of mineral nutrition, the imposition of 2 vegetation indices)
maxima will be recorded, which were more
pronounced in the absence of nutrients. Similarly Since the experimental plots with different
to the green channel, the calculated distribution fertilizer contents of the stationary experiment
width in healthy plants was approximately twice have a relatively small width of 5 meters for
less than in stress plants 9, 8 and 18, respectively. remote sensing using a UAV, the results obtained
The coefficient of determination at 1.5 doses of from the Slantview software map window were
mineral fertilizers was 0.98 and for affected plants used for the research. The obtained results are
0.84. shown in Fig. 6 for stresses caused by the state of
The results of research on the technological mineral nutrition and technological stresses,
stress caused by the action and aftereffect of respectively.
herbicides from the predecessor culture were
Based on the data obtained for the distribution Data suitability can be assessed on the basis of
of the NDVI index, there is a difference in the spectral channel width reference values.
distribution of spectral channels. Thus, the width Vegetation indices GNDVI and RNDVI were
of the distribution regardless of the nature of stress unsuitable for assessing the suitability of data
in stress plants was similar or even smaller than in based on the parameters of the pixel distribution
healthy plants. The coefficient of determination of the image in the experimental plots. This
was 0.85-0.95, it was much lower than in the determines the feasibility of introducing in the
green and red spectral channels. sets of regular vegetation indices of geographic
information systems additional packages that
reflect the spectral channels.
800
GNDVI Fertilizers MIN, w=0,005;
GNDVI Fertilizers MAX, w=0,010;
700
RNDVI Fertilizers MIN, w=0,01; 7. References
600 RNDVI Fertilizers MAX, w=0,01.
Number of pixels
500 [1] Helge Aasen, Andreas Burkart, Andreas
400
Bolten and Georg Bareth 2015 Generating
3D hyperspectral information with
300
lightweight UAV snapshot cameras for
200 vegetation monitoring From camera
100
calibration to quality assurance (ISPRS
Journal of Photogrammetry and Remote
0
0,64 0,66 0,68 0,70 0,72 0,74 0,76 0,78 0,80 0,82 0,84 0,86 Sensing Vol.108) рр. 245–259
Vegetation indexes (NDVI) https://doi.org/10.1016/j.isprsjprs.2015.08.0
80
RNDVI RNDVI Teh.stress +
GNDVI GNDVI Teh.stress + 02
70 [2] Fusang Liu, Pengcheng Hu, Bangyou Zheng,
60
Tao Duan, Binglin Zhu and Yan Guo 2021 A
field-based high-throughput method for
Number of pixels
50
acquiring canopy architecture using
40 unmanned aerial vehicle images Agricultural
30 and Forest Meteorology vol.296, 108231
20
https://doi.org/10.1016/j.agrformet.2020.10
8231
10
[3] Yue Pan, Yu Han, Lin Wang, Jian Chen, Hao
0
0,65 0,70 0,75 0,80 0,85 0,90
Meng, Guangqi Wang, Zichao Zhang and
Vegetation indexes (NDVI) Shubo Wang 2019 Reconstruction of Ground
Figure 6: Dependence of the number of pixels on Crops Based on Airborne LiDAR
the value of the variant of the vegetation index Technology IFAC-PapersOnLine vol.52 (24)
GrennNDVI (GNDVI) and RedNDVI (RNDVI) at pp.35–40
stresses caused by lack of mineral nutrition https://doi.org/10.1016/j.ifacol.2019.12.376
(upper) and technological nature (lower) [4] Tai Guoa, Yuan Fanga, Tao Chenga,
Yongchao Tiana, Yan Zhua, Qi Chenb,
Xiaolei Qiua and Xia Yaoa 2019 Detection
6. Conclusions of wheat height using optimized multi-scan
mode of LiDAR during the entire growth
The study found experimental evidence that stages Computers and Electronics in
the pixel distribution of plantations on the Agriculture vol. 165, 104959
example of wheat crops is described by the https://doi.org/10.1016/j.compag.2019.1049
Gaussian distribution. 59
Analysis of the conformity of the nature of the [5] Per-Ola Olsson, Johan Lindström and Lars
distribution along the spectral channels, namely Eklundh 2016 Near real-time monitoring of
the presence of several maxima that affect the insect induced defoliation in subalpine birch
value of the maximum distribution may indicate forests with MODIS derived NDVI Remote
the presence of foreign inclusions or a transitional Sensing of Environment vol.181 рр. 42–53
stage of vegetation. https://doi.org/10.1016/j.rse.2016.03.040
[6] Ameer Shakayb Arsalaan, Hung Nguyen, Infocommunications Science and
Andrew Coyle and Mahrukh Fida 2021 Technology, PIC S and T 2017 Proceedings
Quality of information with minimum рр. 265-267
requirements for emergency https://doi.org/10.1109/INFOCOMMST.201
communications Ad Hoc Networks vol.111 7.8246394
https://doi.org/10.1016/j.adhoc.2020.102331 [13] M.Dolia, V.Lysenko, N.Pasichnyk,
[7] Yang Cao, Guo Long Li, Yuan Kai Luo, Qi O.Opryshko, D.Komarchuk, V.Miroshnyk,
Pan, and Shao Ying Zhang 2020 Monitoring T.Lendiel and A.Martsyfei 2019 Information
of sugar beet growth indicators using wide- Technology for Remote Evaluation of after
dynamic-range vegetation index (WDRVI) Effects of Residues of Herbicides on Winter
derived from UAV multispectral images Crop Rape 2019 3rd International
Computers and Electronics in Agriculture Conference on Advanced Information and
vol.171, 105331 Communications Technologies, AICT 2019
https://doi.org/10.1016/j.compag.2020.1053 Proceedings рр. 469–473,
31 https://doi.org/10.1109/AIACT.2019.88478
[8] Yaokai Liu, Xihan Mu, Haoxing Wang and 50
Guangjian Yan 2012 A novel method for [14] S.Shvorov, V.Lysenko, N.Pasichnyk,
extracting green fractional vegetation cover Y.Rosamakha, А.Rudenskyi, V.Lukin and
fromdigital images Journal of Vegetation A.Martsyfei 2020 The method of
Science vol. 23 рр. 406–418 determining the amount of yield based on the
https://doi.org/10.1111/j.1654- results of remote sensing obtained using
1103.2011.01373.x UAV on the example of wheat 2020 IEEE
[9] Guangjian Yan, Linyuan Li, André Coy, 15th International Conference on Advanced
Xihan Mu, Shengbo Chen, Donghui Xie, Trends in Radioelectronics,
Wuming Zhang, Qingfeng Shen and Telecommunications and Computer
Hongmin Zhou 2019 Improving the Engineering (TCSET) рр. 245–248
estimation of fractional vegetation cover http://dx.doi.org/10.1109/PICST47496.2019
from UAV RGB imagery by colour .9061238
unmixing ISPRS Journal of
Photogrammetry and Remote Sensing
vol.158 рр. 23–34
https://doi.org/10.1016/j.isprsjprs.2019.09.0
17
[10] André Coy, Dale Rankine, Michael Taylor,
David C. Nielsen and Jane Cohen 2016
Increasing the Accuracy and Automation of
Fractional Vegetation Cover Estimation
from Digital Photographs Remote Sensing
vol.8, 474
https://doi.org/10.3390/rs8070474
[11] Linyuan Li, Xihan Mu, Craig Macfarlane,
Wanjuan Song, Jun Chen, Kai Yan and
Guangjian Yan 2018 A half-Gaussian fitting
method for estimating fractional vegetation
cover of corn crops using unmanned aerial
vehicle images Agricultural and Forest
Meteorology vol. 262 рр. 379-390
https://doi.org/10.1016/j.agrformet.2018.07.
028
[12] V.Lysenko, D.Komarchuk, O. Opryshko,
N.Pasichnyk, N.Zaets 2017 Determination of
the not uniformity of illumination in process
monitoring of wheat crops by UAVs 2017
4th International Scientific-Practical
Conference Problems of