Новий підхід до дистанційного оцінювання стану рослин
The problem of remote estimation of chlorophyll content in vegetation is considered. Some reflectance spectra have been recorded for winter wheat leaves with various level of chlorophyll concentration. To reduce the level of noise in the measurement data produced by measuring system and possible inf...
Збережено в:
| Дата: | 2019 |
|---|---|
| Автори: | , , |
| Формат: | Стаття |
| Мова: | Англійська |
| Опубліковано: |
The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute"
2019
|
| Онлайн доступ: | https://journal.iasa.kpi.ua/article/view/171358 |
| Теги: |
Додати тег
Немає тегів, Будьте першим, хто поставить тег для цього запису!
|
| Назва журналу: | System research and information technologies |
| Завантажити файл: |  |
Репозитарії
System research and information technologies| _version_ | 1867334367478546432 |
|---|---|
| author | Bidyuk, P. I. Litvinenko, V. I. Ponomarenko, S. O. |
| author_facet | Bidyuk, P. I. Litvinenko, V. I. Ponomarenko, S. O. |
| author_institution_txt_mv | [
{
"author": "P. I. Bidyuk",
"institution": null
},
{
"author": "V. I. Litvinenko",
"institution": null
},
{
"author": "S. O. Ponomarenko",
"institution": null
}
] |
| author_sort | Bidyuk, P. I. |
| baseUrl_str | http://journal.iasa.kpi.ua/oai |
| collection | OJS |
| datestamp_date | 2019-06-24T16:34:27Z |
| description | The problem of remote estimation of chlorophyll content in vegetation is considered. Some reflectance spectra have been recorded for winter wheat leaves with various level of chlorophyll concentration. To reduce the level of noise in the measurement data produced by measuring system and possible influence of soil surface smoothing procedure proposed by Savitzky and Golay was applied. The 1st derivative of reflectance spectra curves had been computed and analyzed with respect to correlation with pigment content. To compute an estimate of chlorophyll content multiple regression as well as neural net approach have been applied and both proved to be successful. |
| first_indexed | 2025-07-17T10:25:18Z |
| format | Article |
| fulltext |
© P.I. Bidyuk, V.I. Litvinenko, S.O. Ponomarenko, 2005
Системні дослідження та інформаційні технології, 2005, № 1 119
UDC 62-50
A NOVEL APPROACH TO REMOTE SENSING OF VEGETATION
P.I. BIDYUK, V.I. LITVINENKO, S.O. PONOMARENKO
The problem of remote estimation of chlorophyll content in vegetation is considered.
Some reflectance spectra have been recorded for winter wheat leaves with various
level of chlorophyll concentration. To reduce the level of noise in the measurement
data produced by measuring system and possible influence of soil surface smoothing
procedure proposed by Savitzky and Golay was applied. The 1st derivative of reflec-
tance spectra curves had been computed and analyzed with respect to correlation
with pigment content. To compute an estimate of chlorophyll content multiple re-
gression as well as neural net approach have been applied and both proved to be
successful.
Introduction. Chlorophyll content in plant leaves is an important characteristic
showing on the plant status. For example, pigment level correlates with power of
photosynthetic apparatus and may be used for yield prognosis. Influence of vari-
ous stresses on photosynthetic apparatus first of all induces changes in chloro-
phyll content. This phenomenon allows to develop effective approaches to fast
revealing stresses in phytocenoses of various types, using remote estimates of
chlorophyll content and its dynamics. It is well known that reflectance spectra of
plant leaves in an optical range are the most informative with respect to chloro-
phyll content due to the fact that they are correlated with specific spectrum of the
pigment. Moreover, these spectra may serve as reliable information for remote
measurements.
Therefore, an idea to apply this characteristic to monitoring of agrocenosis
and other types of plant canopies seems to be quite acceptable. There are numer-
ous relations based on regression equations used to calculate chlorophyll concen-
tration in leaves by making use of reflectance spectral coefficients or their combi-
nations [1–6]. Practically all of them are effective enough when chlorophyll
concentration is measured under laboratory conditions. When estimation of chlo-
rophyll concentration is performed for «soil-vegetation» system then estimation
results may be distorted. The main cause is in contribution of soil reflectance.
There are several approaches to minimize this type of noise by using complex
vegetation indices including components for correction of soil interference [7–9].
Their application demands measurements of additional parameters of the system,
such as leaf index and spectral coefficients of soil reflectance. This approach
makes the procedures low effective and moreover, the problem is not solved
completely because of variability of soil reflectance depending on its moisture.
Therefore, it is very important to develop effective methods acceptable for chlo-
rophyll measurement in the «soil-vegetation» system.
Application of high resolution spectroscopy to remote sensing of vegetation
[10] displayed new characteristics correlating with chlorophyll content. It ap-
peared that a position of the so called red edge in spectral curves of reflectance
spectra and the shape of this spectral region were dependent on pigment concen-
tration. Various characteristics of red edge (680–800 nanometers) have been used.
P.I. Bidyuk, V.I. Litvinenko, S.O. Ponomarenko
ISSN 1681–6048 System Research & Information Technologies, 2005, № 1 120
Position of maximum in the 1st derivative plot of reflectance spectral curve [11]
and position of red edge were found to be in regression relation with chlorophyll
content [12–14]. It was also revealed that there exists correlation between the ra-
tio of two maxima in the 1-st derivative plot and chlorophyll [15–16]. Some resis-
tance of this characteristic to soil contribution has been shown as well [17]. A
shortage of this approach is in its sensitivity to instrumental noises, which may
considerably distort the results influencing, for example, intensity of maxima and
their position. Besides, one of the two maxima is very often difficult to distin-
guish what creates extra computing burden.
Thus, to develop a device for remote measurements of chlorophyll content in
vegetation using the 1st derivative characteristics it is necessary to solve the fol-
lowing problems: 1) to reduce the level of noise in the measurement data; 2) to
propose appropriate variables and computing procedures for estimating chloro-
phyll content that would exhibit acceptable quality.
The methodology of spectral and biochemical measurements. Leaves of
two varieties of winter wheat were studied during the experiments, and appropri-
ate measurements had been collected. Variations of chlorophyll content were
caused by differences in mineral nutrition and plant vegetation age. Reflectance
spectra were measured with spectrophotometer SF-10 (USSR) equipped with in-
tegrating sphere [17]. The recording system allowed to obtain the spectra in digi-
tal form. The spectra were recorded with the increment of 1 nm in the range of
680–750 nm. The pieces of the wheat leaves used for recording of reflectance
spectra, and the data were used for chlorophyll estimation by chemical method
proposed by Arnon [18].
Preliminary processing of experimental data. Preliminary data processing is
based on application of smoothing procedure proposed in [19, 20]. To perform
data smoothing we used several moving windows of different size: 5, 7 and 9
point windows. After data smoothing the 1st derivative was computed using 2-d
order polynomial estimated by making use of LS procedure. For this purpose an-
other moving window was applied to smoothed data. Next, the resulting first de-
rivative curve was used to determine estimates of chlorophyll content.
Estimation of chlorophyll content in vegetation. To determine estimates of
chlorophyll content in vegetation the following approaches were applied: regres-
sion equations of various structure (order) and neural nets.
Preliminary smoothing of measurement data was performed with 7 and 9-
point windows that revealed their acceptable effectiveness. After smoothing the
data we computed the 1st derivative of reflectance spectra recorded for winter
wheat leaves with various chlorophyll concentration. Some examples of the 1st
derivative plots are shown in figure. Visual estimation of the plots shows that the rip-
ples are insignificant. In this study we did not consider the two maxima approach
reported earlier [16, 17]:
)1(
max
)2(
max
10 x
xkky += , (1)
where )2(
max
)1(
max , xx are two maxima of the first derivative of reflectance spectra.
Though ratio of the maxima intensity is dependent on chlorophyll content in
leaves. To determine coefficients 0k and 1k two reflectance spectra are required.
A Novel approach to remote sensing of vegetation
Системні дослідження та інформаційні технології, 2005, № 1 121
The approach had been studied in [16, 17], and proved to be useful to some ex-
tent. The drawback of this approach is that it uses only two points of the first de-
rivative plot what creates a problem with reaching desirable quality of estimates
of chlorophyll content in vegetation. Also, computing experiments with locating
«lower» maximum showed that it was not always possible to find it. That is why
we propose to use multiple regression equation considered below.
Regression equation. The regression equation proposed has the following
form
)()(...)()()( 22110 llxalxalxaalz mm ε+++++= , (2)
where mixi ...,,1, = are values of the first derivative of smoothed reflectance
spectrum taken with appropriate interval that could be varied depending on the
Plots of first derivatives of reflectance spectra for winter wheat leaves versus wavelength, nm
2
2
dL
Rd
2
2
dL
Rd
2
2
dL
Rd
2
2
dL
Rd
680 680
680 680
P.I. Bidyuk, V.I. Litvinenko, S.O. Ponomarenko
ISSN 1681–6048 System Research & Information Technologies, 2005, № 1 122
regression order and number of experiments available; l is a number of an ex-
periment, with Nl ...,,2,1= ; ε is equation error that is practically inevitable in
estimating regression. Here one experiment is referred to one session of spectrum
measurement using spectrometer, and to determining one estimate of chlorophyll
concentration by chemical test. Values of the first derivative were taken in the
interval 680–740 nm. Thus, theoretically order of the regression could be ex-
panded to about 60. However, in practice order of linear regression could be re-
duced to the value of 10–15. Simplicity of the equation allows to use it even in
real time with signal processor implementation.
Number of independent variables in equation (2) is much higher than in
equation (1) what makes it logically more suitable for description of the first de-
rivative of measurement data. The larger is the number of independent variables
the higher should be degree of adequacy of equation (2), but there are computa-
tional limits to expanding the right-hand side, and shortage of an interval between
samples of spectra may result in linearly dependent vectors of measurement ma-
trix and, consequently, in the so called co-linearity problem. To find estimates of
coefficients miai ...,,1,0, = we need at least 1+m measurements of reflected
spectra and respective estimates of chlorophyll content.
Some results of applying equation (2) to estimation of chlorophyll content
using experimental reflectance spectra are given in Tabl. 1.
T a b l e 1 . Results of chlorophyll content estimation using multiple regression.
№ of
exp.
Regres.
coeff.
Chlorof.
estimate
Error of
estimate
x 1E-08
Regres.
coeff.
Chlorof.
estimate
Error of
estimate
x 1E-06
Regres.
coeff.
Chlorof.
estimate
Error of
Estimate
x 1E-06
1 2 3 4 5 6 7 8 9
1 –3.346 4.76 0.4912 –35.935 4.76 0.1134 –21.779 4.76 0.0498
2 –2.756 5.67 0.5575 30.508 5.67 0.1093 0.952 5.67 0.0168
3 9.865 3.07 0.3703 –14.511 3.07 0.1057 –10.075 3.07 0.1829
4 –12.996 3.58 0.5072 –32.932 3.58 0.0867 34.203 3.58 0.1630
5 19.376 6.82 0.5391 5.490 6.82 0.1154 –42.364 6.82 0.0542
6 –28.401 6.57 0.5695 56.401 6.57 0.1372 58.494 6.57 0.0178
7 30.747 8.02 0.4117 –127.459 8.02 0.1091 –79.216 8.02 0.0965
8 –28.317 3.75 0.5485 215.172 3.75 0.0913 69.327 3.75 0.1008
9 21.174 7.65 0.5663 –134.240 7.65 0.1378 –32.189 7.65 0.1013
10 –29.104 3.25 0.3816 57.537 3.25 0.0719 –46.969 3.25 0.2364
11 26.135 2.43 0.4817 –159.463 2.43 0.1160 61.573 2.43 0.0170
12 –4.337 5.04 0.5806 122.207 5.04 0.1079 –9.275 5.04 0.0958
According to Tabl. 1, estimates of chlorophyll, computed via equation (2),
exhibit rather high quality that is acceptable for field studies of vegetation state.
Columns 1, 4, and 7 contain regression coefficients computed for different values
of samples that belong to the range of wavelength between 680–740 nm. Substan-
tial difference between values of the coefficients indicates that the values of the
first derivative, used for computing regression, should be carefully selected. Col-
A Novel approach to remote sensing of vegetation
Системні дослідження та інформаційні технології, 2005, № 1 123
umns 2, 5, and 8 show estimates of chlorophyll content computed with regression
estimated via the first derivatives of reflectance spectra. The estimates have the
same values due to the fact that the estimate errors (columns 3, 6, and 9) are very
low for each of the three regressions tested in this study. The smallest values of
estimate errors are presented in column 3 for the first regression what indicates
that left edge of the first derivative curve is more informative than the right one.
Thus, this method seems to be promising for the field estimation of chloro-
phyll concentration in vegetation. Further study of the regression approach to es-
timation is required in the following directions: (1) vary the number of independ-
ent variables; (2) vary the range of wavelength used for estimation; (3) try various
structures of regression including nonlinear one.
Neural net. As an alternative way to regression model we have examined
possibility of application of a neural net to determining estimates of chlorophyll
content. In many cases neural nets exhibit robustness with respect to noise of
various origin, and provide acceptable estimates when regression works on-
satisfactorily. A multilayer neural net approach was applied. It includes one input
layer; two, three or more hidden layers, and one output layer. The input layer is a
distributor of input sequence }...,,...,,1{ Jj for 1n neurons of the first hidden
layer. Each neuron has its weighting coefficient 1,njK which form a matrix of
weighting coefficients. The products formed by each neuron are summed and ap-
plied to the input of non-linear sigmoid type element. Outputs of the sigmoid
elements are applied to inputs of the second hidden layer of the network and so
on. The output layer of the net consists of a number of neurons, which corre-
sponds to the number of classes each of which includes definite selected values of
chlorophyll estimates. The large is the number of classes the lower is error of es-
timate of chlorophyll content. Number of neurons in each layer, number of layers,
coefficients 1,njK , functions of non-linear elements are adjusted in the process of
constructing specific net. There are several approaches to determine minimum
and maximum number of neurons for a net. To estimate a number of neurons in
hidden layers we used the expression for estimating necessary number of synaptic
weights [21, 22]:
mmn
m
NmL
N
Nm
w +++⎟
⎠
⎞
⎜
⎝
⎛ +≤≤
+
)1(1
log1 2
, (3)
where n is a dimension of input signal; m is dimension of output signal; N
number of elements of training sample. Then necessary number of neurons in
hidden layers for a two later net is:
)( mnLL w += . (4)
Other parameters are determined during the training process when the net is
fed with representative spectral data and estimates of chlorophyll content pro-
duced by chemical analysis. To determine the coefficient values and parameters
of non-linear functions were used the training procedures given in Tabl. 2.
As an effectiveness indicator of the neural net functioning were selected sta-
tistical characteristics of net response to training and control sequences. The ap-
P.I. Bidyuk, V.I. Litvinenko, S.O. Ponomarenko
ISSN 1681–6048 System Research & Information Technologies, 2005, № 1 124
proach, based on a neural net, does not require a priory knowledge of decision
rules. Though, depending on a net structure it could be constructed robust with
respect to disturbances (noise) of various nature. It is also important to stress that
they do not require estimating of a model structure what is usually done when
regression approach is applied.
T a b l e 2 . The training algorithms used
№ Type of training algorithm Matlab function
1 BFGS quasi-Newton backpropagation Trainbfg
2 Powell-Beale conjugate gradient backpropagation Traincgb
3 Fletcher-Powell conjugate gradient backpropagation Traincgf
4 Polak-Ribiere conjugate gradient backpropagation Traincgp
5 Gradient descent backpropagation Traingd
6 Gradient descent with adaptive of learning backpropagation Traingda
7 Gradient descent with momentum backpropagation Traingdm
8 Levenberg-Marcquardt backpropagation Trainlm
9 One step secant backpropagation Trainoss
10 Scaled conjugate gradients backpropagation Trainscg
After training quality of operation of each of the networks studied was esti-
mated by making use of mean squared error (MSE) computed for p samples of
data:
∑
=
−=
p
i
ii yd
p
MSE
1
2)(1 ,
where id is desirable value of output, and iy is a actual estimate found for i th
sample. MSE was computed separately for a sample with known, cMSE , and un-
known (test sample), tMSE , concentrations of chlorophyll. The results of estima-
tion and number of training cycles are given in Tabl. 3. It follows from columns 3
and 4 that cMSE and tMSE are close to each other.
T a b l e 3 . Results of estimation and neural nets training
№ Training function tMSE cMSE № of training
cycles
1 Trainbfg 0.201 0.198 3120
2 Traincgb 1.263 1.351 6320
3 Traincgf 0.351 0.345 4550
4 Traincgp 1.725 1.801 5150
5 Traingd 2.101 2.214 21000
6 Traingda 2.478 2.541 7014
7 Traindm 3.001 3.124 19950
8 Trainlm 1.145 1.102 6651
9 Trainoss 1.996 1.897 3125
10 Trainscg 0.471 0.456 4100
A Novel approach to remote sensing of vegetation
Системні дослідження та інформаційні технології, 2005, № 1 125
Application of neural nets shows that generally they provide different results
for different types of training algorithms. The best results in our experiments were
achieved with application of BFGS quasi-Newton backpropagation and Fletcher-
Powell conjugate gradient backpropagation optimization training procedure with
MSE (Mean Squared Error) of estimate 0.201, and 0.471 for scaled conjugate
gradient algorithm. In some cases (#5, and #7) number of iterative training cycles
is very high what proves again that the problem of training time reduction is also
present in this study. Also it is clear that application of neural nets still require
further research with respect to refining net structure aiming to improve the esti-
mates.
Conclusions. Application of multiple regression to estimating concentration
of chlorophyll in vegetation led to positive results. Multiple linear regression was
quite effective, and increasing the number of independent variables improved
quality of estimates substantially. In a case of eleven independent variables the
estimate errors turned out to be very small, what proves the possibility of using
the equation in a field device for preliminary estimation of vegetation state. A
neural net approach could be used for chlorophyll estimation as well, and in some
cases it produced even better estimates than regression. However, in general it
provided in this study somewhat lower quality of estimates than regression. This
result could be explained by the fact that we used conventional net structures.
Perhaps, special types of nets could be adjusted for this specific problem to obtain
better results, and further research in this direction is required. Another promising
approach to approximation of the first derivative curves could be the use of or-
thogonal polynomials, and application of image recognition techniques for esti-
mation of chlorophyll content in vegetation. We got positive preliminary results
of application of such approaches to estimation of chlorophyll concentration, and
previously excellent results had been received in performing medical and engi-
neering diagnosis.
REFERENCES
1. Benedict H.M., Swidler R. Nondestructive method for estimating chlorophyll content
of leaves // Science. — 1961. — 133. — P. 2015–2016.
2. Hardwik K., Baker N.R. In vivo measurement of chlorophyll content of leaves // New
Phytol. — 72. — P. 51–54, 1973.
3. Carter G.A. Ratios of leaf reflectance in narrow wavebands as indicators of plant
stresses // Int. J. Rem. Sens.— 1994. — 15. — P. 697–703.
4. Yoder B.J., Pettigrew-Crosby R.E. Predicting nitrogen and chlorophyll content and
concentration from reflectance spectra (400–2500 nm) at leaf and canopy scales
// Rem. Sens. Envir. — 1995. — 53. — P. 199–211.
5. Gitelson A.A., Merzlyak M.N. Signature analysis of leaf reflectance spectra: algo-
rithm development for remote sensing of chlorophyll // J. Plant Physiol. — 1996.
— 148. — P. 494–500.
6. Datt B. A new reflectance index for remote sensing of chlorophyll content in higher
plants: test using Eucaliptus leaves // J. Plant Physiol. — 1999. — 154. — P. 30–
36.
7. Huete A.R. A soil adjusted vegetation index (SAVI) // Rem. Sens. Envir. — 1988. —
25. — P. 295–309.
P.I. Bidyuk, V.I. Litvinenko, S.O. Ponomarenko
ISSN 1681–6048 System Research & Information Technologies, 2005, № 1 126
8. Qi J., Huete A.R. Modified soil adjusted vegetation index (MSAVI) // Rem. Sens.
Envir. — 1994. — 48. — P. 119–126.
9. Li Xia. A two-axis adjusted vegetation index (TWVI) // Int. J. Rem. Sens. — 1994.
— 14. — P. 1447–1458.
10. Collins W. Remote sensing of crop type and maturity // Photogr. Eng. Rem. Sens. —
1978. — 44. — P. 43–55.
11. Horler D.N.H., Dokrey M., Barber J. The red edge of plant leaf reflection // Int. J.
Rem. Sens. — 1983. — № 4. — P. 273–288.
12. Miller J.R., Hare E.W., Wu J. Quantitative characterization of the vegetation red
edge reflectance. 1. An inverted-Gaussian reflectance model // Int. J. Rem. Sens.
— 1990. — № 11. — P. 1755–1773.
13. Joumar A.Yu. Research of influence of parameters of vegetation canopy on the posi-
tion of maximal declination of red edge // Research Earth from Space. — 1992.
— № 4. — P. 15–20.
14. Demetriodes-Shan T.H., Steven M.D., Clark J.A. High resolution derivative spectra
in remote Sensing // Rem. Sens. Envir. — 1993. — 33. — P. 55–64.
15. Kochubey S.M., Kobets N.I., Shadchina T.M. The quantitative analysis of shape of
spectral reflectance curves of plant leaves as a way for testing their status //
Physiology and biochemistry of cultivated plants. — 1988. — 20. — P. 535–539.
16. Кочубей С.М., Кобец Н.И., Шадчина Т.М. Спектральные свойства растений как
базис для дистанционных методов диагностики. — Киев: Наук. думка,
1990. — 260 с.
17. Kochubey S.M. Comparative analysis of information power of multispectral imaging
and high-resolution spectrometry in the remote sensing of vegetation cover //
Space Sciences and Technology. — 1999. — 5. — P. 41–48.
18. Arnon D.I. Copper enzymes in isolated chloroplasts // Polyphenoloxidase in Beta
vulgaris Plant Physiol. — 1949. — 24. — P. 1–15.
19. Savitzky A., Golay M.G.E. Smoothing and differentiation of data by simplified least
squares Procedures // Analyt. Chem. — 1964. —36. — P. 1627–1639.
20. Max J. Methodes et techniques de traitment du signal et applications aux mesures
physiques. — Paris: Masson, 1981. — 375 p.
21. Дьяконов В., Круглов В. Математический инструментарий Matlab. — Санкт-
Петeрбург: Питер, 2001. — 280 с.
22. Применение искусственных нейронных сетей для прогнозирования длины вол-
ны в максимуме генерации лазерных красителей класса имитринов /
В.И. Литвиненко, Ю.А. Бюргер, В.П. Кругленко, Ю.Ю. Стойлов // 8-я Все-
российская конф. по нейрокомпьютерам, НКП-2002, Москва, 21–22 марта
2002. — М., 2002. — С. 232–238.
Received 11.04.2003
_________________________
From the Editorial board: The article corresponds completely to submitted manuscript.
|
| id | journaliasakpiua-article-171358 |
| institution | System research and information technologies |
| keywords_txt_mv | keywords |
| language | English |
| last_indexed | 2025-07-17T10:25:18Z |
| publishDate | 2019 |
| publisher | The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute" |
| record_format | ojs |
| resource_txt_mv | journaliasakpiua/92/8c8349b08afd46c193a72bded7b86c92.pdf |
| spelling | journaliasakpiua-article-1713582019-06-24T16:34:27Z A novel approach to remote sensing of vegetation Новый подход к дистанционному оцениванию состояния растений Новий підхід до дистанційного оцінювання стану рослин Bidyuk, P. I. Litvinenko, V. I. Ponomarenko, S. O. The problem of remote estimation of chlorophyll content in vegetation is considered. Some reflectance spectra have been recorded for winter wheat leaves with various level of chlorophyll concentration. To reduce the level of noise in the measurement data produced by measuring system and possible influence of soil surface smoothing procedure proposed by Savitzky and Golay was applied. The 1st derivative of reflectance spectra curves had been computed and analyzed with respect to correlation with pigment content. To compute an estimate of chlorophyll content multiple regression as well as neural net approach have been applied and both proved to be successful. Рассмотрена проблема оценивания концентрации хлорофилла в растениях. Проведены эксперименты с целью записи отраженных спектров для озимой пшеницы с различными значениями содержания хлорофилла. Построены графики первых производных и выполнен анализ содержания пигмента. Для уменьшения уровня шумов экспериментальных данных использована процедура сглаживания, предложенная Савицким и Голеем, которая основывается на 9-точечной свертке. Для оценивания содержания хлорофилла успешно применена множественная регрессия и нейронные сети. Розглянуто проблему оцінювання концентрації хлорофілу в рослинах. Проведено експерименти з метою запису відбитих спектрів для озимої пшениці з різними значеннями вмісту хлорофілу. Побудовано графіки перших похідних та виконано аналіз вмісту пігменту. Для зменшення рівня шумів експериментальних даних використано процедуру згладжування, запропоновану Савицьким і Голеєм, яка базується та 9-точковій згортці. Для оцінювання вмісту хлорофілу успішно застосовано лінійну множинну регресію та нейронні мережі. The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute" 2019-06-24 Article Article application/pdf https://journal.iasa.kpi.ua/article/view/171358 System research and information technologies; No. 1 (2005); 119-126 Системные исследования и информационные технологии; № 1 (2005); 119-126 Системні дослідження та інформаційні технології; № 1 (2005); 119-126 2308-8893 1681-6048 en https://journal.iasa.kpi.ua/article/view/171358/171013 Copyright (c) 2021 System research and information technologies |
| spellingShingle | Bidyuk, P. I. Litvinenko, V. I. Ponomarenko, S. O. Новий підхід до дистанційного оцінювання стану рослин |
| title | Новий підхід до дистанційного оцінювання стану рослин |
| title_alt | A novel approach to remote sensing of vegetation Новый подход к дистанционному оцениванию состояния растений |
| title_full | Новий підхід до дистанційного оцінювання стану рослин |
| title_fullStr | Новий підхід до дистанційного оцінювання стану рослин |
| title_full_unstemmed | Новий підхід до дистанційного оцінювання стану рослин |
| title_short | Новий підхід до дистанційного оцінювання стану рослин |
| title_sort | новий підхід до дистанційного оцінювання стану рослин |
| url | https://journal.iasa.kpi.ua/article/view/171358 |
| work_keys_str_mv | AT bidyukpi anovelapproachtoremotesensingofvegetation AT litvinenkovi anovelapproachtoremotesensingofvegetation AT ponomarenkoso anovelapproachtoremotesensingofvegetation AT bidyukpi novyjpodhodkdistancionnomuocenivaniûsostoâniârastenij AT litvinenkovi novyjpodhodkdistancionnomuocenivaniûsostoâniârastenij AT ponomarenkoso novyjpodhodkdistancionnomuocenivaniûsostoâniârastenij AT bidyukpi novijpídhíddodistancíjnogoocínûvannâstanuroslin AT litvinenkovi novijpídhíddodistancíjnogoocínûvannâstanuroslin AT ponomarenkoso novijpídhíddodistancíjnogoocínûvannâstanuroslin AT bidyukpi novelapproachtoremotesensingofvegetation AT litvinenkovi novelapproachtoremotesensingofvegetation AT ponomarenkoso novelapproachtoremotesensingofvegetation |