1. Introduction
Weed control is an important element of crop protection in modern intensive farming. The productivity of agricultural crops all over the world depends on it. According to, the Food and Agriculture Organization (FAO) annual crop losses from weeds on a global scale are about 10% [1].
Many weeds are considered extremely dangerous. They are most damaging to the agricultural sector, in addition to affecting human health and ecosystem stability in general [2,3,4,5,6]. Successful weed control is only possible with timely monitoring of their distribution in field agrocenoses. At the same time, the suppression of weed development is carried out by introducing expensive herbicides that are harmful to the environment. Traditional methods of phytosanitary monitoring, based on visual ground surveys of cultivated crops, are laborious and do not grant prompt and comprehensive data on the spread of weeds.
The accurate recognition of weeds using remote monitoring tools and its localization in space allows for the spot application of herbicides. Therefore, it is possible to increase the effectiveness of protective measures, obtain a high economic effect and reduce the pesticide pressure on agrocenoses [7,8].
The physical basis for the recognition of plant objects using remote monitoring tools is the specifics of their reflectivity in the visible and infrared ranges of the electromagnetic radiation spectrum. It is closely related to the nature and direction of the physiological processes occurring in plants [9,10]. Features of the spectral characteristics of plant objects can become the key to their identification using remote sensing of the Earth [11].
The approach based on the use of quantitative spectral characteristics of plant objects imposes fewer restrictions on the response time. It is most appropriate to develop methods for the spot application of herbicides in real time [12,13].
Hyperspectral data from next-generation sensors provide highly informative data [14]. In this connection, there is a tendency among researchers to search for optimal spectral features that help to classify different types of vegetation for specialized monitoring tasks.
Francisco-Fernández et al. [15], using the method of nonparametric linear regression, revealed patterns of spatial distribution of field bindweed (Convolvulus arvensis L.) foci in the cultivated crop rotation of wheat and sunflower crops and established the boundaries of territories within which weediness exceeded the economic threshold of damage.
There are some monitoring systems based on high quality images that allow the use of spectral, spatial and texture information for the recognition of the following plants: creeping thistle (Cirsium arvense (L.) Scop.) in wheat crops [16], as well as couch grass (Agropyron repens (L.) Beauv.), spiny sowthistle (Sonchus asper L.) Hill.), white marrow (Chenopodium album L.), dog’s mercury (Mercurialis perennis L.), wild cabbage (Brassica campestris L.), and wild chamomile (Matricaria inodora L.) on carrot plantations [17]. A similar example is research on the development of weed recognition methods of field bindweed (C. arvensis L.), sorrels (Rumex crispus L.) and creeping thistle (C. arvense (L.) Scop.) on corn crops using machine learning based on the processing of hyperspectral mosaic images obtained in the near infrared range of the spectrum [18].
Herrmann et al. [19], from the analysis of ground-based spectroscopy data and using classification models of discriminant analysis, revealed the separability of wheat from broadleaf plants of the genera Chenopodium, Malva, Knapweed and Chrysanthemum as well as grassy weeds of Lolium rigidum (Lolium rigidum Gaudin.) and little barley (Hordeum plaucum L.). They also established the scalability of this method for aerospace imaging systems.
A method for classifying crops and weeds based on the use of the spectral G and R components of the RGB color image has been studied [20]. The results of the statistical analysis showed the effectiveness of this method for the detection of white marrow (C. al-bum L.), annual mercury (Mercurialis annua L.) and bluegrass (Poa praténsis L.) weeds on corn and sugar beet crops.
The efficiency of deciphering agricultural crops is largely determined by the complexity of combinations of plant backgrounds in a particular agricultural landscape, the availability of standards for the reflectivity of plant objects, as well as the choice of classification methods and the method of preliminary processing of the data obtained [21].
One of the most effective ways to study the vegetation parameters of plant objects is to convert their reflectance spectra into vegetation indices using elementary mathematical operations [22]. Hyperspectral vegetation indices help to decipher subtle changes in the nature and direction of biochemical processes occurring in plants [23].
In this paper, we aim to study correlation structures of spectral patterns of cultivated and weed plants based on factor analysis methods, converting them into spectral vegetation indices. We also aim to evaluate the potential possibility of classifying objects under study using Earth remote sensing tools. We chose 15 generally accepted vegetation indices, which are widely used in Earth remote sensing to study natural and anthropogenic landscapes [24,25,26]. Our primary objectives are sunflower crops (Helianthus annuus L.), corn (Zea mais L.) and soybean (Glycine max (L.)), as well as ragweed (Ambrosia artemisiifolia L.), California-bur (Xanthium strumarium L.), red-root amaranth (Amaranthus retroflexus L.), white marrow (C. album L.) and field milk thistle (Sonchus arvensis L.). These types of cultivated and weed plants form the basis of Krasnodar Krai agrocenoses and are actual components of artificial plant communities in other geographical regions of the world [27,28,29,30,31].
2. Materials and Methods
2.1. Arrangement of Test Plots
We made test plots with different species composition of cultivated and weed plants on the experimental fields of the scientific crop rotation of the Federal Research Center of Biological Plant Protection. These plants form the basis of the agrocenoses of Krasnodar Krai.
Our primary subjects are sunflower crops (H. annuus L.), corn (Z. mais L.) and soybean (G. max (L.)) Besides on the test plots, pure and mixed backgrounds of weeds were identified, represented by the following species: ragweed (A. artemisiifolia L.), Cali-fornia-bur (X. strumarium L.), red-root amaranth (A. retroflexus L.), white marrow (C. al-bum L.) and field milk thistle (S. arvensis L.) (Table 1). The total area under crops of test plots within which experimental plots with different vegetation backgrounds were allocated was about 3 ha. The area of each individual experimental plot was 8 m2.
2.2. Ground Spectrometric Measurements
We used the Ocean Optics Maya 2000 PRO automated spectrometer calibrated by the absolute radiation intensity. This spectrometer allows on-line measurements of the spectral brightness density of objects in the spectral range from 350 nm to 1000 nm with a high spectral resolution ~1 nm of reflected solar radiation.
The Maya 2000 PRO Spectrometer is part of Ocean Optics’ high-sensitivity spectrometer series, featuring frame-by-frame CCD array detectors with back-lit illumination. It is an uncooled spectrometer primarily designed for low light measurements, UV spectrum analysis and other scientific applications.
Ground-based spectrometric measurements of individual plants were carried out in a perpendicular direction to the Earth’s surface at a height of ~1 m. In this case, the lowest degree of specular reflection is provided for most surfaces. In this case, the solid angle of the light flux was 250, and the area of the visual field of the spectrometer was about 0.25 m2. Vegetation areas with a protective cover of at least 60% were selected for surveying. This minimized the soil influence on the reflective spectral characteristics of the vegetation background.
Fifteen ground-based spectrometric measurements of the canopy were carried out at each designated experimental plot. They were supplemented with measurements of the white panel (reference reflector). SRM-990 Spectralon white reference was used as ideal white (reflection coefficient-0.95–0.99). Further, based on the results of statistical processing of the obtained data, we obtained the average reflection spectra.
During the measurements, we considered the state of atmosphere with respect to cloudiness, aerosols, etc.
2.3. Data Analysis and Processing
The reflectance spectra were converted into 15 generally accepted vegetation indices to study the spectral features of plant objects [24,25,26]. The total number of index dependencies was divided into separate groups of indices that evaluate certain vegetation parameters of plant objects (Table 2).
Further, the calculated indicators of index values were processed using factor analysis methods. Its purpose was to determine the nature of the relationships between the variables under study and to combine strongly correlated values into separate factors.
We used a posteriori comparisons of variable index values based on the methods of analysis of variance to assess the information content of the spectral features of the studied plant objects.
All of the above statistical methods were carried out using Statistica 10.0.1011 software package.
3. Results
In the experimental part of the research, we conducted ground spectrometric measurements of sunflower (H. annuus L.), corn (Z. mais L.) and soybean (G. max (L.)), as well as pure and mixed phonons of ragweed (A. artemisiifolia L.), California-bur (X. strumarium L.), red-root amaranth (A. retroflexus L.), white marrow (C. album L.) and field milk thistle (S. arvensis L.) (Figure 1).
The reflectance spectra were converted into 15 generally accepted vegetation indices [24,25,26], and then divided into separate groups that evaluated certain vegetation parameters of plant objects:
Indices for assessing xanthophylls and changes in chlorophyll content (NDVI705, mSR705, mNDVI705, VOG1, VOG2, VOG3, REPI);
Indices for assessing the content of carotenoids and anthocyanins in plants (CRI1, CRI2, ARI1, ARI2);
Light Efficiency Indices (PRI, SIPI);
Narrow-band indices of plant stress assessment (REP, RVSI).
Factor analysis revealed three factors that fully described the variances of the compared variables by 50.3%, 27.3% and 16.1, respectively (Table 3).
Within the first factor, the indices for assessing the content of carotenoids and anthocyanins in plants CRI1, CRI2, ARI1 and ARI2 are characterized by a negative correlation with the narrow-band indices of plant stress assessment REP and RVSI, and have a positive correlation with Vogelmann indices for the near infrared slope region VOG2 and VOG3 (Table 4).
The second factor consists of a positive correlation of the modified relative index mSR705 with the photochemical wave reflection index PRI, having a negative correlation with Vogelmann index for the near infrared slope region VOG1.
The third factor is determined by the mutual positive correlation of the normalized difference index NDVI705, the modified relative index mNPVI, and the position index of the near infrared slope REPI. All of them are included in the group of vegetation indices for assessing xanthophylls and changes in the content of chlorophyll, calculated from the data on the values of the reflection coefficients in the spectral region 690–750 nm.
Analysis of variance confirms the influence of the plant species factor at a high level of statistical significance for all variations of dependent variable values of vegetation in-dices.
A posteriori analysis established a pronounced statistically significant difference between sunflower and the totality of compared cultivated and weed plants in terms of the values of all vegetation indices that make up the first factor (Table 5). Corn is characterized by an absolute difference from other plant objects in terms of vegetation indices, which reflect the content of carotenoids and anthocyanins CRI1, CRI2, ARI1, ARI2, as well as in plant stress indices REP and RVSI. However, in terms of the Vogelmann VOG2 and VOG3 indices, it coincides with the soybean values.
We noted a significant similarity of the factor patterns of soybean and field milk thistle. It was statistically confirmed only by the values of the vegetation indices CRI2, ARI1 and RVSI.
A remarkable fact is the complete statistically significant and strongly pronounced identity of ragweed values and the vegetation background of the experimental site No. 11 where the quantitative predominance of ragweed was observed.
Vegetation indices CRI1, CRI2, ARI1 and REP confirmed the patterns of California-bur and red-root amaranth. According to the RVSI vegetation stress index, California-bur is similar to soybean and red-root amaranth is similar to white marrow.
White marrow (No. 7) is completely different from other plant objects in terms of the values of carotenoid indices CRI1 and CRI2, but in other respects it is similar to the indicators of experimental plots with mixed plant backgrounds No. 9 and 10.
Experimental plots with mixed plant backgrounds No. 9 and 10 form a single homogeneous group within the common factor values. This group, according to the CRI1, CRI2 and RVSI vegetation indices, is similar to California-bur and red-root amaranth, and, according to the values of the ARI1 and REP vegetation indices, to white marrow plants.
Statistically significant isolation of values was revealed by the values of generalized factorial coefficients. It helps to identify sunflower, corn, soybean and field milk thistle plants, as well as experimental plots with mixed plant backgrounds No. 9 and 10. Statistically significant identity of the values of ragweed and plant background of experimental plot No. 11 with a predominance of ragweed was confirmed. California-bur, red-root amaranth and white marrow separated into a single homogeneous group.
According to the values of the vegetation indices MSR705, PRI and VOG1 that make up the second factor, a complete significant difference between corn and the totality of compared cultivated and weed plants was revealed (Table 5). As in the previous case, there is a statistically significant identity of the values of ragweed and the vegetation background of the experimental site 11, with a predominance of ragweed for all signs of the factor.
We should note the statistically significant similarity between sunflower and California-bur plants in terms of the combination of MSR705 and PRI vegetation indices. For individual values of the MSR705 vegetation index, there is a similarity between the indicators of sunflower and California-bur with the experimental site No. 10, where the quantitative predominance of California-bur was observed. According to the PRI photochemical activity index, sunflower and California-bur are similar to red-root amaranth. Sunflower is combined into one similarity group with soybean and field milk thistle plants according to Vogelmann index VOG1. California-bur on this index is similar to experimental sites No. 9 and 10.
Field milk thistle is completely different from other plants in MSR705 and PRI indices.
Soybean plants are similar to white marrow according to the MSR705 vegetation index and similar to experimental plot No. 10 according to the PRI index.
Red-root amaranth, according to the MSR705 vegetation index, is similar to the experimental sites 10 and 11, but it is similar to California-bur and white marrow according to the PRI index.
Red-root amaranth and white marrow are completely different from other plants in terms of the VOG1 index.
We determined a statistically significant isolation of the patterns of sunflower, corn, red-root amaranth, white marrow and field milk thistle according to factor coefficients. The statistically significant identity of the values of ragweed and the vegetation back-ground of experimental site No. 11 with the predominance of ragweed was confirmed. Soybean, California-bur and experimental plots with mixed plant backgrounds No. 9 and 10 stood out in a single homogeneous group.
We revealed a statistically significant difference between soybean plants and all studied plant backgrounds according to the third factor vegetation indices NDVI705, mNPVI and REPI (Table 5).
Corn is completely different from all plant objects in terms of the NDVI705 and mNPVI vegetation indices, but according to the REPI index, it is included in the same group of similarities as California-bur and experimental plots with mixed plant back-grounds No. 9 and 10.
White marrow and experimental plot No. 9 have a similar identity in terms of the NDVI705 and mNPVI vegetation indices.
Sunflower, in terms of the NDVI705 and mNPVI vegetation indices, completely coincides with the characteristics of field milk thistle, but according to the REPI index, it is similar to white marrow.
Ragweed, California-bur, red-root amaranth, as well as experimental plots with mixed vegetation backgrounds No. 10 and 11 were combined into a single homogeneous group for the NDVI705 and mNPVI vegetation indices.
The REPI index revealed the absolute difference between ragweed and other plant objects, as well as the identity of the indicators of red-root amaranth and experimental site No. 9.
In general, according to factor coefficients, we registered a statistically significant isolation of the patterns of corn, soybean, red-root amaranth, white marrow and experimental site No. 9. In addition, a statistically significant identity of the values of ragweed and the vegetation background of experimental site No. 11 with a predominance of ragweed were reported. The homogeneity of the patterns of sunflower, California-bur and experimental site No. 10 was revealed.
4. Discussion
Our study is devoted to the development of criteria for the separation of different types of vegetation in field agrocenoses based on the analysis of high-resolution ground-based spectrometric measurements. In terms of the general focus and content of the experimental part, this research is similar to other studies [17,18,19,20,21,22].
A distinctive feature of our study is the methodical approach to data processing, aimed at studying the spectral characteristics of plants by vegetation index values characterizing the biophysical parameters of plants. Other researchers show high efficiency of classification of different types of vegetation based on the use of discriminant functions [20,21,22].
Factor analysis identified three factors that fully describe the variances of the compared variables of index values.
As a rule, most plants under consideration have homogeneity of patterns in terms of index indicators that make up the selected factors. However, in almost all cases, we note the possibility of identifying different plant backgrounds, both by the values of individual vegetation indices and by generalized factorial coefficients.
For example, sunflower has a pronounced statistically significant isolation of pat-terns in terms of the values of the first factor, but in terms of the second and third factors it shows similarities with the indicators of California-bur, red-root amaranth, white marrow and field milk thistle.
Clearly expressed identification features of cultivated soybean plants appeared only in terms of the values of the third factor.
California-bur and red-root amaranth, which are identical in terms of the first factor, can be separated by the index values of the second and third factors.
White marrow has a strongly pronounced statistically significant isolation of char-acters for all values of the third factor, and according to the generalized values of the first factor, it is similar to California-bur and red-root amaranth.
Field milk thistle is well identified by the generalized values of the patterns of all three factors; although, in terms of individual index values of the first and third factors, it coincides with the indices of cultivated sunflower and soybean plants.
The mixed plant backgrounds of experimental plots No. 9 and 10 are close or completely identical to those of California-bur and red-root amaranth. Nevertheless, they can be identified in the total set of compared objects according to the generalized features of the first factor.
At the same time, we determined plant backgrounds that have pronounced, statistically-confirmed absolute differences in characteristics from the totality of compared objects in terms of the values of all three factors.
These objects include cultivated corn plants, as well as identical plant backgrounds of ragweed and experimental site No. 11.
The Maya 2000 PRO spectrometer is a highly sensitive instrument capable of detecting subtle differences in the reflectivity of studied plant backgrounds. The use of such equipment is an original approach to conducting ground-based, sub-satellite spectrometric measurements of vegetation cover.
The use of hyperspectral data opens up the possibility of fairly accurate recognition of different types of vegetation, considering their reflectivity in a large number of narrow-band spectral channels [14,15,16]. However, a large number of channels imply the need for volumetric information processing, which, in turn, makes it difficult to apply this method of vegetation classification in practice [19,22]. Therefore, it seems appropriate to create specialized applications focused on solving specific problems using optimal spectral channels [20,21,22].
Another problem is the possibility of a comparative evaluation of the measurement results of ground-based spectrometers, which record reflection spectra from micro-surfaces with data from specialized Earth remote sensing equipment.
A monitoring system seems like a solution to this problem. Such a system should be based on the synchronization of high-precision ground-based spectrometric measurements with satellite and unmanned remote surveys, and comparison of the obtained data with the results of field surveys [10,14,15].
5. Conclusions
As a result of the research, we determined the structure of the relationship of variable values of hyperspectral vegetation indices into separate factor patterns that make it possible to classify cultivated and weed plants.
The potential possibility of identifying different plant backgrounds, both by the values of individual vegetation indices and by generalized factorial coefficients, has been revealed. However, in order to apply the vegetation classification method in practice, it seems appropriate to create specialized applications focused on solving specific problems using optimal spectral channels.
The data obtained contribute to the operational control over agricultural crops and ensure the high efficiency of protective measures. This, in turn, will help to optimize the costs of growing crops and will reduce the pesticide load on agro-ecosystems. What is more, our research results are important for the validation of remote aerospace observations using multispectral and hyperspectral instruments.
Conceptualization, R.D. and O.K.; methodology, R.D. and O.K.; validation, O.K. and A.P.; formal analysis, R.D. and O.K.; data curation, R.D. and O.K.; writing—original draft preparation, R.D.; writing—review and editing, R.D., O.K. and A.P.; visualization, A.P.; project administration, O.K.; funding acquisition, O.K. All authors have read and agreed to the published version of the manuscript.
Not applicable.
Not applicable.
Not applicable.
The authors declare no competing interests.
Footnotes
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.
Enlarge this image.Figure 1. Average graphs of the dependence of spectral brightness on the wavelength of cultivated and weed plants.
Characteristics of the studied plant backgrounds in test plots.
| No. Test Plot | Plant Species | Number of Plants, pcs/m2 | Development Stages |
|---|---|---|---|
| 1 | Sunflower |
5–6 | BBCH 65–78 «budding» |
| 2 | Corn (Z. mais L.) | 5–6 | BBCH 71–73 |
| 3 | Soybean |
7–10 | BBCH 51–53 «budding» |
| 4 | Ragweed |
450–480 | Stage 2-Shooting «lengthening of 4–5 internodes» |
| 5 | Californian cocklebur |
250–300 | Stage 2-Shooting «lengthening of 4–5 internodes» |
| 6 | Pigweed |
250–300 | Stage 2-Shooting «lengthening of 4–5 internodes» |
| 7 | Muchweed |
250–300 | Stage 2-Shooting «lengthening of 4–5 internodes» |
| 8 | Field sow thistle |
100–150 | Stage 4-Flowering |
| 9 | Ragweed |
448–482 | Stage 2-Shooting «lengthening of 4–5 internodes» |
| Field sow thistle |
6–8 | Stage 4-Flowering | |
| Californian cocklebur |
2–3 | Stage 2-Shooting «lengthening of 4–5 internodes» | |
| 10 | Ragweed |
435–448 | Stage 2-Shooting «lengthening of 4–5 internodes» |
| Californian cocklebur |
35–42 | Stage 2-Shooting «lengthening of 4–5 internodes» | |
| Muchweed |
1–2 | Stage 2-Shooting «lengthening of 4–5 internodes» | |
| 11 | Ragweed |
431–450 | Stage 2-Shooting «lengthening of 4–5 internodes» |
| Californian cocklebur |
2–3 | Stage 2-Shooting «lengthening of 4–5 internodes» |
Description of the sample of vegetation indices for assessing the vegetation parameters of plant objects.
| Spectral Index | Calculation Formula | Authors |
|---|---|---|
| Indices for assessing the state of xanthophylls and changes in chlorophyll content | ||
| NDVI705 | P750 * − P705/P750 + P705 | Gitelson A.A. et al., 1994 |
| mSR705 | P750 − P445/P750 + P445 | Datt B., 1999 |
| mNDVI705 | P750 − P705/P750 + P705 − 2P445 | Datt B., 1999 |
| VOG1 | P740/P720 | Vogelmann J.E. et al., 1993 |
| VOG2 | P734 − P747/P715 + P726 | |
| VOG3 | P734 − P747/P715 + P720 | |
| REPI | REPI = NDVI705 + mSR705 + VOG1 + VOG2 + VOG3 | |
| Indices for assessing the content of carotenoids and anthocyanins in plants | ||
| CRI1 | (1/P510) + (1/P550) | Gamon J.A, 1997 |
| CRI2 | (1/P510) + (1/P700) | |
| ARI1 | (1/P550) + (1/P700) | |
| ARI2 | P800 × ((1/P550) + (1/P700)) | |
| Light efficiency indices | ||
| PRI | P531 − P570/P531 + P570 | Gamon J.A, et al., 1990 |
| SIPI | P800 − P845/P800 + P680 | |
| Plant stress assessment narrowband indices | ||
| REP | 700 + 40 × Predegre-P700/P740 + P700 | Merton J.P. and Hunnington S.V., 1999 |
| RVSI | P714 + P752/2 − P733 | |
Note: * PXXX-reflectivity value of the object under study in a specific spectral channel; Predegre—the average reflectivity of the object under study in the area of the “red slope” is 720–800 nm.
Factor analysis results for variable index values.
| Variable | Factor 1 | Factor 2 | Factor 3 |
|---|---|---|---|
| NDVI705 | −0.177411 | 0.574509 | 0.748737 * |
| mSR705 | −0.235497 | −0.806815 * | 0.406489 |
| mNDVI705 | −0.177411 | 0.574509 | 0.748737 * |
| PRI | −0.123305 | −0.819804 * | 0.546125 |
| SIPI | 0.587770 | 0.663273 | 0.098082 |
| CRI1 | −0.797328 * | −0.461330 | 0.276217 |
| CRI2 | −0.990745 * | −0.022931 | 0.003545 |
| ARI1 | −0.961214 * | 0.225332 | −0.149379 |
| ARI2 | −0.957798 * | 0.233990 | −0.158228 |
| REP | 0.861582 * | −0.444344 | 0.231426 |
| RVSI | 0.814935 * | 0.376121 | 0.220696 |
| VOG1 | 0.424311 | 0.859234 * | −0.060536 |
| VOG2 | −0.934131 * | 0.299670 | −0.176047 |
| VOG3 | −0.937214 * | 0.293074 | −0.173965 |
| REPI | −0.522780 | 0.307182 | 0.729474 * |
| Expl.Var | 7.548457 | 4.089205 | 2.417299 |
| Prp.Totl | 0.503230 | 0.272614 | 0.161153 |
Note: * the most significant indicators of factor loadings of variables having an absolute value of more than 0.7.
Correlation structure of dependence of values of vegetation indices.
| NDVI705 | MSR705 | mNDVI705 | PRI | SIPI | CRI1 | CRI2 | ARI1 | ARI2 | REP | RVSI | VOG1 | VOG2 | VOG3 | REPI | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| NDVI705 | 1.00 | −0.21 | 1.00 * | −0.05 | 0.27 | 0.05 | 0.15 | 0.18 | 0.18 | −0.23 | 0.16 | 0.30 | 0.19 | 0.19 | 0.74 * |
| MSR705 | −0.21 | 1.00 | −0.21 | 0.91 * | −0.55 | 0.68 | 0.26 | −0.01 | −0.02 | 0.23 | −0.33 | −0.73 * | −0.07 | −0.07 | 0.28 |
| MNPVI | 1.00 * | −0.21 | 1.00 | −0.05 | 0.27 | 0.05 | 0.15 | 0.18 | 0.18 | −0.23 | 0.16 | 0.30 | 0.19 | 0.19 | 0.74 * |
| PRI | −0.05 | 0.91 * | −0.05 | 1.00 | −0.54 | 0.63 | 0.15 | −0.14 | −0.16 | 0.39 | −0.29 | −0.79 * | −0.22 | −0.22 | 0.21 |
| SIPI | 0.27 | −0.55 | 0.27 | −0.54 | 1.00 | −0.69 | −0.56 | −0.41 | −0.40 | 0.25 | 0.78 * | 0.85 * | −0.36 | −0.36 | 0.01 |
| CRI1 | 0.05 | 0.68 | 0.05 | 0.63 | −0.69 | 1.00 | 0.84 * | 0.63 | 0.62 | −0.41 | −0.70 * | −0.73 * | 0.56 | 0.56 | 0.48 |
| CRI2 | 0.15 | 0.26 | 0.15 | 0.15 | −0.56 | 0.84 * | 1.00 | 0.95 * | 0.95 * | −0.84 * | −0.79 * | −0.43 | 0.92 * | 0.92 * | 0.52 |
| ARI1 | 0.18 | −0.01 | 0.18 | −0.14 | −0.41 | 0.63 | 0.95 * | 1.00 | 1.00 * | −0.96 * | −0.72 * | −0.20 | 0.99 * | 0.99 * | 0.47 |
| ARI2 | 0.18 | −0.02 | 0.18 | −0.16 | −0.40 | 0.62 | 0.95 * | 1.00 * | 1.00 | −0.97 * | −0.72 * | −0.19 | 0.99 * | 1.00 * | 0.46 |
| REP | −0.23 | 0.23 | −0.23 | 0.39 | 0.25 | −0.41 | −0.84 * | −0.96 * | −0.97 * | 1.00 | 0.57 | −0.04 | −0.98 * | −0.98 * | −0.43 |
| RVSI | 0.16 | −0.33 | 0.16 | −0.29 | 0.78 * | −0.70 * | −0.79 * | −0.72 * | −0.72 * | 0.57 | 1.00 | 0.73 * | −0.67 | −0.67 | −0.08 |
| VOG1 | 0.30 | −0.73 | 0.30 | −0.79 | 0.85 | −0.73 | −0.43 | −0.20 | −0.19 | −0.04 | 0.73 | 1.00 | −0.11 | −0.12 | 0.08 |
| VOG2 | 0.19 | −0.07 | 0.19 | −0.22 | −0.36 | 0.56 | 0.92 * | 0.99 * | 0.99 * | −0.98 * | −0.67 | −0.11 | 1.00 | 1.00 * | 0.47 |
| VOG3 | 0.19 | −0.07 | 0.19 | −0.22 | −0.36 | 0.56 | 0.92 * | 0.99 * | 1.00 * | −0.98 * | −0.67 | −0.12 | 1.00 * | 1.00 | 0.47 |
| REPI | 0.74 * | 0.28 | 0.74 * | 0.21 | 0.01 | 0.48 | 0.52 | 0.47 | 0.46 | −0.43 | −0.08 | 0.08 | 0.47 | 0.47 | 1.00 |
Note: * the most significant indicators of the correlation coefficient having an absolute value of more than 0.7.
Results of a posteriori comparison of variable index values (according to Duncan’s test).
| CRI1 | CRI2 | ARI1 | ARI2 | REP | RVSI | Factor 1 | MSR705 | PRI | VOG1 | Factor 2 | NDVI705 | mNDVI705 | REPI | Factor 3 | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | 1.3 a | 0.9 a | 0.4 a | −1.1127 a | 714.24 g | −0.079 f | 2.9731 i | 0.6878 d | 0.0777 f | −0.0061 a | −0.769 b | 0.5343 g | 0.2039 c | 1.31 b | 0.49 g |
| 2 | 1.7. b | 1.6 b | 1.3 b | −0.0063 b | 695.47 a | −0.106 e | 0.2917 h | 0.5482 a | 0.0019 a | −0.0011 h | 2.538 h | 0.6107 i | 0.2948 h | 1.46 h | −0.08 e |
| 3 | 8.0 d | 8.4 c | 4.2 c | 0.0099 c | 697.68 b | −0.161 cd | −0.1623 f | 0.7151 e | 0.0526 c | 0.0002 g | −0.033 e | 0.5277 f | 0.1792 b | 1.43 f | −0.84 c |
| 4 | 18.2 h | 19.2 g | 9.8 f | 0.0123 c | 699.73 f | −0.156 d | −0.6255 a | 0.7619 f | 0.0972 g | −0.0026 de | −0.456 c | 0.4889 c | 0.2612 f | 1.51 i | 1.36 h |
| 5 | 11.4 e | 11.9 d | 5.2 cd | 0.0103 c | 698.68 d | −0.172 bc | −0.4219 c | 0.6933 d | 0.0764 f | −0.0019 f | 0.017 ef | 0.4985 d | 0.2631 fg | 1.45 gh | 0.59 g |
| 6 | 11.3 e | 11.8 d | 5.3 cd | 0.0104 c | 698.69 d | −0.198 a | −0.5046 b | 0.6596 c | 0.0720 ef | −0.0028 cd | −0.251 d | 0.4684 b | 0.2644 g | 1.38 d | 0.13 f |
| 7 | 15.0 g | 15.8 f | 7.9 e | 0.0129 c | 699.20 e | −0.193 a | −0.4637 bc | 0.7132 e | 0.0692 de | −0.0023 ef | −1.550 a | 0.4585 a | 0.1295 a | 1.29 a | −1.84 a |
| 8 | 6.7 c | 7.2 c | 4.3 c | 0.0117 c | 697.99 c | −0.149 d | 0.1857 g | 0.6005 b | 0.0332 b | −0.0021 f | 0.795 g | 0.5391 h | 0.2018 c | 1.33 c | −1.30 b |
| 9 | 13.0 f | 13.7 e | 7.5 e | 0.0130 c | 699.19 e | −0.173 bc | −0.2533 e | 0.6722 c | 0.0576 c | −0.0031 c | 0.035 ef | 0.5072 e | 0.2223 d | 1.39 e | −0.33 d |
| 10 | 12.4 ef | 13.0 de | 6.3 de | 0.0121 c | 699.13 e | −0.176 b | −0.2533 d | 0.6890 d | 0.0650 d | −0.0036 b | 0.161 f | 0.5071 e | 0.2576 e | 1.44 g | 0.43 g |
| 11 | 19.9 i | 20.9 h | 9.8 f | 0.0122 c | 699.73 f | −0.157 d | −0.6683 a | 0.7619 f | 0.0972 g | −0.0026 de | −0.487 c | 0.4889 c | 0.2612 f | 1.51 i | 1.39 h |
Note: Different letters indicate significant differences (p < 0.05) according to Duncan’s test.
References
1. FAO (Food and Agriculture Organization of the United Nations). Available online: https://www.fao.org/home/en/ (accessed on 12 June 2022).
2. Martinez, B.; Reaser, J.K.; Dehgan, A.; Zamft, B.D.; McCormick, C.; Giordano, A.J.; Aicher, R.; Selbe, S. Technology innovation: Advancing capacities for the early detection of and rapid response to invasive species. Biol. Invasions; 2020; 22, pp. 75-100. [DOI: https://dx.doi.org/10.1007/s10530-019-02146-y]
3. Enders, M.; Hűtt, M.-T.; Jeschke, J.M. Drawing a map of invasion biology based on a networkof hypotheses. Ecosphere; 2018; 9, e02146. [DOI: https://dx.doi.org/10.1002/ecs2.2146]
4. Nie, S.; Li, W. How spatial structure of species and disturbance influence the ecological invasion. Ecol. Model.; 2020; 431, 109199. [DOI: https://dx.doi.org/10.1016/j.ecolmodel.2020.109199]
5. Chung, H.I.; Choi, Y.; Ryu, J.; Jeon, S.W. Validating management strategies for invasive species from a spatial perspective: Common ragweed in the Republic of Korea. Environ. Sci. Policy; 2020; 114, pp. 52-63. [DOI: https://dx.doi.org/10.1016/j.envsci.2020.07.018]
6. Reaser, J.K.; Burgiel, S.W.; Kirkey, J.; Brantley, K.A.; Veatch, S.D.; Burgos-Rodríguez, J. The early detection of and rapid response (EDRR) to invasive species: A conceptual framework and federal capacities assessment. Biol. Invasions; 2020; 22, pp. 1-19. [DOI: https://dx.doi.org/10.1007/s10530-019-02156-w]
7. Tataridas, A.; Kanatas, P.; Chatzigeorgiou, A.; Zannopoulos, S.; Travlos, I. Sustainable Crop and Weed Management in the Era of the EU Green Deal: A Survival Guide. Agronomy; 2022; 12, 5897. [DOI: https://dx.doi.org/10.3390/agronomy12030589]
8. Allmendinger, A.; Spaeth, M.; Saile, M.; Peteinatos, G.; Gerhards, R. Precision Chemical Weed Management Strategies: A Review and a Design of a New CNN-Based Modular Spot Sprayer. Agronomy; 2022; 12, 1620. [DOI: https://dx.doi.org/10.3390/agronomy12071620]
9. Ngom, R.; Gosselin, P. Development of a Remote Sensing-Based Method to Map Likelihood of Common Ragweed (Ambrosia artemisiifolia) Presence in Urban Areas. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens.; 2013; 7, pp. 126-139. [DOI: https://dx.doi.org/10.1109/JSTARS.2013.2254469]
10. Danilov, R.Y.; Ismailov, V.Y.; Tret’yakov, V.A.; Kremneva, O.Y.; Shumilov, Y.V.; Rizvanov, A.A.; Krivoshein, V.V.; Kostenko, I.A. Development of precision technologies of agroecosystems phytosanitary monitoring based on the use of data of remote hyperspecral sensing of the Earth. Dostizheniya Nauk. Tekhniki APK; 2018; 10, pp. 82-86. [DOI: https://dx.doi.org/10.24411/0235-2451-2018-11019]
11. Danilov, R.Y.; Kremneva, O.Y.; Ismailov, V.Y.; Tretyakov, V.A.; Rizvanov, A.A.; Krivoshein, V.; Pachkin, A.A. General methods and results of ground hyperspectral studies of seasonal changes in the reflective properties of crops and certain types of weeds. Curr. Probl. Remote Sens. Earth Space; 2020; 1, pp. 113-127. [DOI: https://dx.doi.org/10.21046/2070-7401-2020-17-1-113-127]
12. Kawamura, K.; Asai, H.; Yasuda, T.; Soisouvanh, P.; Phongchanmixay, S. Discriminating crops/weeds in an upland rice field from UAV images with the SLIC-RF algorithm. Plant Prod. Sci.; 2020; 24, pp. 73-82. [DOI: https://dx.doi.org/10.1080/1343943X.2020.1829490]
13. Rakhmatulin, I.; Kamilaris, A.; Andreasen, C. Deep Neural Networks to Detect Weeds from Crops in Agricultural Environments in Real-Time: A Review. Remote Sens.; 2021; 13, 4486. [DOI: https://dx.doi.org/10.3390/rs13214486]
14. Mishra, P.; Asaari, M.S.M.; Herrero-Langreo, A.; Lohumi, S.; Diezma, B.; Scheunders, P. Close range hyperspectral imaging of plants: A review. Biosyst. Eng.; 2017; 164, pp. 4-9. [DOI: https://dx.doi.org/10.1016/j.biosystemseng.2017.09.009]
15. Francisco-Fernández, M.; Jurado-Expósito, M.; Opsomer, J.D.; López-Granados, F. A nonparametric analysis of the spatial distribution of Convolvulus arvensis in wheat-sunflower rotations. Environmetrics; 2006; 17, pp. 849-860. [DOI: https://dx.doi.org/10.1002/env.803]
16. Sørensen, R.A.; Rasmussen, J.; Nielsen, J.; Jørgensen, R.N. Thistle detection using convolutional neural networks. Proceedings of the EFITA WCCA 2017 Conference; Montpellier Supagro, Montpellier, France, 2–6 July 2017; 75.
17. Piron, A.; Heijden, F.; Destain, M.F. Weed detection in 3D images. Precis. Agric.; 2011; 12, pp. 607-622. [DOI: https://dx.doi.org/10.1007/s11119-010-9205-2]
18. Gao, J.; Nuyttens, D.; Lootens, P.; He, Y.; Pieters, J.G. Recognising weeds in a maize crop using a random forest machine-learning algorithm and near-infrared snapshot mosaic hyperspectral imagery. Biosyst. Eng.; 2018; 170, pp. 39-50. [DOI: https://dx.doi.org/10.1016/j.biosystemseng.2018.03.006]
19. Herrmann, I.; Shapira, U.; Kinast, S.; Karnieli, A.; Bonfil, D.J. Ground-level hyperspectral imagery for detecting weeds in wheat fields. Precis. Agric.; 2013; 14, pp. 637-659. [DOI: https://dx.doi.org/10.1007/s11119-013-9321-x]
20. Mao, W.; Hu, X.; Zhang, X. Weed detection based on the optimized segmentation line of crop and weed. Comput. Comput. Technol. Agric.; 2008; 2, pp. 959-967. [DOI: https://dx.doi.org/10.1007/978-0-387-77253-0_27]
21. Terekhin, E.A. Assessment the spatial-temporal changes in green phytomass of agricultural vegetation using spectral response. Curr. Probl. Remote Sens. Earth Space; 2021; 1, pp. 138-148. [DOI: https://dx.doi.org/10.21046/2070-7401-2021-18-1-138-148]
22. Xue, J.; Su, B. Significant Remote Sensing Vegetation Indices: A Review of Developments and Application. J. Sens.; 2017; 1, 1353691. [DOI: https://dx.doi.org/10.1155/2017/1353691]
23. Gewali, U.B.; Monteiro, S.T.; Saber, E. Gaussian Processes for Vegetation Parameter Estimation from Hyperspectral Data with Limited Ground Truth. Remote Sens.; 2019; 11, 1614. [DOI: https://dx.doi.org/10.3390/rs11131614]
24. Sims, D.A.; Gamon, J.A. Relationships between leaf pigment content and spectral reflectance acr oss a wide range of species. leaf structures and developmental stages. Remote Sens. Environ.; 2002; 81, pp. 337-354. [DOI: https://dx.doi.org/10.1016/S0034-4257(02)00010-X]
25. Sims, J.; James, C.; Harrison, D.A. Monitoring the contribution of general practice to population health activities. Health. Promot. J. Aust.; 2002; 13, pp. 189-192. [DOI: https://dx.doi.org/10.1071/HE02189]
26. Vogelmann, J.E.; Rock, B.N.; Moss, D.M. Red edge spectral measurements from sugar maple leaves. Remote Sens. Environ.; 1993; 14, pp. 1563-1575. [DOI: https://dx.doi.org/10.1080/01431169308953986]
27. Tretyakova, A.S.; Baranova, O.G.; Luneva, N.N.; Terekhina, T.A.; Yamalov, S.M.; Lebedeva, M.V.; Khasanova, G.R.; Grudanov, N.Y. Segetal flora of some regions of Russia: Characteristics of the taxonomic structure. Proc. Appl. Botany. Genet. Breed.; 2020; 181, pp. 123-133. [DOI: https://dx.doi.org/10.30901/2227-8834-2020-2-123-133]
28. Olsen, J.; Kristensen, L.; Weiner, J. Influence of sowing density and spatial pattern of spring wheat (Triticum aestivum) on the suppression of different weed species. Weed Biol. Manag.; 2006; 6, pp. 165-173. [DOI: https://dx.doi.org/10.1111/j.1445-6664.2006.00210.x]
29. José-María, L.; Blanco-Moreno, J.M.; Armengot, L.A.; Sans, F.X. How does agricultural intensification modulate changes in plant community composition?. Agric. Ecosyst. Environ.; 2011; 145, pp. 77-84. [DOI: https://dx.doi.org/10.1016/j.agee.2010.12.020]
30. Pyšek, P.; Jarošík, V.; Chytrý, M.; Kropác, Z.; Tichý, L.; Wild, J. Alien plants in temperate weed communities: Prehistoric and recent invaders occupy different habitats. Ecology; 2005; 86, pp. 772-785. [DOI: https://dx.doi.org/10.1890/04-0012]
31. Rauber, R.B.; Demaría, M.R.; Jobbágy, E.G.; Arroyo, D.N.; Poggio, S.L. Weed communities in semiarid rainfed croplands of Central Argentina: Comparison between corn (Zea mays) and soybean (Glycine max) crops. Weed Sci.; 2018; 66, pp. 368-378. [DOI: https://dx.doi.org/10.1017/wsc.2017.76]
© 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
Details
; Kremneva, Oksana
; Pachkin, Alexey