A normalized plant community remote sensing index, its construction method and its application

Abstract

The present invention discloses a normalized plant community remote sensing index, a construction method and an application thereof. The construction method of the remote sensing index NCRSI has the following steps: S1: plant community characteristic investigation and classification; S2: environmental temperature and humidity data collection; S3: multi-source remote sensing data collection; S4: extracting spectral information, texture information and temperature information from the multi-source remote sensing data obtained in S3; S5: performing correlation analysis on the data of S2 and S4, and screening out remote sensing variables highly correlated with the temperature and humidity data; S6: using principal component analysis to construct the remote sensing variables obtained in S5 into a normalized plant community remote sensing index NCRSI. The present invention proposes for the first time to construct a normalized plant community remote sensing index for detecting plant community variability using multi-source remote sensing data; and provides an objective, rapid and effective new method for plant community monitoring, which has important guiding significance for carrying out plant community construction and improving ecological benefits.

Description

Technical Field

[0001] The present invention relates to the field of plant community remote sensing management and smart agricultural technology, and in particular to a plant community remote sensing index, a construction method and applications thereof. Background Art

[0002] Ecological tea garden is a complex ecosystem model with tea trees as the main species. It not only conforms to the biological characteristics of tea trees that like warmth, humidity and diffuse light, but also follows the natural law of ecological balance. The structure and function of plant communities play a leading role in regulating ecological stability. In order to create a stable tea garden ecological environment, more and more ecological tea gardens with different plant community structures have emerged in northern China. In fact, people often build or transform ecological tea gardens based on the general model of ecological tea gardens and their own subjective ideas. This has become a good way to alleviate the low temperature in winter and high temperature and drought in summer in the Jiangbei tea area of ​​China. However, the structure and stability of these plant communities are difficult to quantify.

[0003] Monitoring of plant community structure and its response to climate change is an important basis for evaluating the ecological structure and function of tea gardens. Traditional methods of monitoring vegetation changes usually involve field surveys of plant types, crown width, plant height, leaf area and other indicators, and then calculating differences and coefficients of variation to reflect their responses to climate change. This method indirectly reflects the stability of the ecosystem. However, ground surveys not only require a lot of manpower and material resources, but also have low accuracy, making it difficult to achieve large-scale and rapid surveys. In addition, environmental data is also crucial for ecological monitoring. Studies have shown that climate change, as the main driving factor of terrestrial ecosystems, determines the survival and transformation of surface vegetation.

[0004] At present, remote sensing technology, with its efficient, unbiased, dynamic and quantitative monitoring characteristics, has gradually become the most effective and practical method for people to better evaluate regional or global ecology. However, it is currently usually applicable to ecosystems such as forests and cities, and it is difficult to fully reflect the changes in small-scale tea garden ecology, especially the impact of multiple plant communities on the ecosystem. In addition, satellite remote sensing data has the characteristics of low temporal and spatial resolution, which makes it difficult to fully capture detailed information on small-scale plant communities. And no research has focused on the monitoring of farmland and tea garden ecosystems.

[0005] Therefore, the existing technology needs to be further improved. How to reveal the differences in the community structures of different tea gardens and their impact on the stability of the tea garden ecosystem has become an important issue that needs to be urgently addressed at this stage. Summary of the invention

[0006] In view of the problems existing in the prior art, the present invention provides a normalized plant community remote sensing index, a construction method and its application based on coordinated multi-source remote sensing data and environmental data, so as to realize timely, accurate and continuous detection of plant community stability. The present invention uses the meteorological station environmental monitoring system to continuously obtain the temperature and humidity data of air and soil, and uses unmanned aerial vehicles to obtain multi-spectral, RGB and thermal infrared remote sensing data; coordinates environmental data and remote sensing data to determine remote sensing variables sensitive to tea garden ecological changes; uses multi-source remote sensing variables to construct a normalized plant community remote sensing index (NCRSI); uses Fourier first-order functions to establish a mechanism model of NCRSI, quantifies the change curve of NCRSI, and uses CNN-GRU hybrid network, CNN, GRU and SVM to establish a regression model between different plant community environmental data and their NCRSI data to predict future NCRSI.

[0007] To solve the above problems, this application provides the following technical solutions:

[0008] In a first aspect, the present invention provides a normalized plant community remote sensing index NCRSI, wherein the formula of the remote sensing index NCRSI is as follows:

[0009] NCRSI=0.095X1+0.096X2+0.093X3+0.099X4+0.091X5+0.092X6+0.092X7+0.075X8+0.078X9+0.073X 10 +0.047X 11 +0.070X 12

[0010] Where X1 is NDVI, X2 is GNDVI, X3 is RECI, X4 is RENDVI, X5 is OSAVI, X6 is RVI, X7 is SR, X8 is BCor, X9 is GCor, X 10 It's RCor, X 11 It's Min, X 12 It's AVG.

[0011] Furthermore, the steps of constructing the remote sensing index NCRSI are as follows:

[0012] S1: Investigation and classification of plant community characteristics;

[0013] S2: Environmental temperature and humidity data collection;

[0014] S3: Multi-source remote sensing data acquisition;

[0015] S4: Extract spectral information, texture information and temperature information from the multi-source remote sensing data obtained in S3;

[0016] S5: Perform correlation analysis on the data of S2 and S4 to screen out remote sensing variables that are highly correlated with temperature and humidity data;

[0017] S6: The remote sensing variables obtained in S5 are constructed into a normalized plant community remote sensing index NCRSI using a principal component analysis method; the normalized plant community remote sensing index NCRSI is used for the ecological status of the plant community.

[0018] Furthermore, the normalized plant community remote sensing index NCRSI obtained in S6 is fitted with the Fourier function; the Fourier function is expressed as: f(x)=a0+acos(wx)+bsin(wx),

[0019] Where x is the number of days from the start of the experiment, a0, a, b and w are all constants; the fitting curve of NCRSI ( Figure 5 ) shows that the variation trend of the NCRSI fitting curve based on Fourier function is consistent with the actual value and can accurately fit the annual variation trend of NCRSI.

[0020] On the basis of the above technical solution, S1 investigates the characteristics of plant communities, and the plant communities are divided into three categories.

[0021] Category 1, the plant community structure is relatively simple, mainly composed of bare ground and herbaceous plants, with low plant density, 0≤the number of observed species≤3; Category 2, mainly composed of herbs and shrubs, 4≤the number of observed species≤7; Category 3, the plant community is relatively complex, mainly composed of shrubs and trees, with high plant density, 8≤the number of observed species≤10.

[0022] Based on the above technical solution, the data of S2 is collected from the temperature and humidity sensor of the weather station. Preferably, the sensor is powered by solar energy, and its data collection cycle is once every 30 minutes.

[0023] Further preferably, the air temperature and humidity data in S2 are taken from 100 cm above the ground; the soil temperature and humidity data are taken from 0 cm on the surface, 20 cm underground, and 40 cm underground, respectively.

[0024] Based on the above technical solution, the S3 uses an unmanned aerial vehicle (UAV) flight platform to collect multi-source remote sensing data in the research area. Preferably, the UAV flight platform is a DJI Matrice M300 equipped with a dual gimbal bracket, equipped with MS600PRO and Zenmuse H20T.

[0025] Furthermore, the flight altitude of the UAV flight platform is 30m, and the flight speed is 2.5m / s. Preferably, the flight time is 11:00-14:00 noon, the weather is clear, the lighting conditions are good, and the wind speed is low.

[0026] Based on the above technical solution, ENVI is used in S4 to extract spectral information and texture information; DJI Thermal Analysis Tool is used to extract temperature information.

[0027] Furthermore, the spectral information, texture information and temperature information are extracted from the multi-source remote sensing data obtained in S3; the steps are as follows:

[0028] Yusense Map (V1.0) software was used to preprocess the original MS, RGB, and TIR images.

[0029] Specifically, the original, single MS image is output as a complete MS image of the study area through processes such as band registration, radiometric calibration, and image stitching. The original RGB and TIR images are output as RGB and TIR images of the study area through processes such as camera parameter generation and temperature calibration.

[0030] Spectral information, texture information and temperature information are extracted from MS, RGB and TIR images respectively.

[0031] Specifically, the region of interest tool in ENVI (V5.3) software was used to extract the average spectral reflectance of each community from the MS image, and 15 common vegetation indices were calculated using the 6-band MS data.

[0032] The three-channel texture information of each community was extracted from the RGB image using the second-order probability statistics tool in ENVI (V5.3) software. Furthermore, the texture information type was an 8-dimensional gray-level co-occurrence matrix, including mean (Mean), variance (Var), homogeneity (Hom), contrast (Con), dissimilarity (Dis), entropy (Ent), second-order moment (SM) and correlation (Cor).

[0033] The maximum temperature (Tmax), minimum temperature (Tmin) and mean temperature (Tmean) of each community were extracted from the TIR images using the rectangular temperature measurement tool in the DJI Thermal Analysis Tool (V3.0) software.

[0034] Based on the above technical solution, SPSS is used in S5 to analyze the correlation between remote sensing variables and temperature and humidity data. The analysis shows that the correlation between temperature and humidity data and TIR variables is the highest, followed by MS variables and RGB variables. Thus, a total of 12 remote sensing variables highly correlated with temperature and humidity data were screened out, including TIR variables (Min, AVG), MS variables (NDVI, GNDVI, RENDVI, OSAVI, RECI, RVI, SR) and RGB variables (BCor, GCor, RCor).

[0035] In the second aspect, based on the same invention, the present invention also provides the application of the normalized plant community remote sensing index NCRSI in plant community assessment.

[0036] Furthermore, the above application used CNN-GRU hybrid network, CNN, GRU and SVM to establish a regression model between the environmental data of different plant communities and their NCRSI data to predict the future NCRSI.

[0037] Furthermore, the plant community is any one of a tea garden, an orchard, an artificial forest, an ecological garden, a flower garden, and a vegetable garden.

[0038] The present invention has the following advantages:

[0039] 1. This invention proposes for the first time to construct a Normalized Plant Community Remote Sensing Index (NCRSI) for detecting plant community stability using multi-source remote sensing data;

[0040] 2. This invention, for the first time, combines environmental data with remote sensing data to determine remote sensing variables that are sensitive to temperature and humidity changes. This not only overcomes the defects of meteorological sensors that are easily disturbed by the external environment and have a limited data acquisition range, but also makes up for the problem of discontinuity in UAV remote sensing data.

[0041] 3. The present invention quantifies the normalized plant community remote sensing index NCRSI based on the Fourier fitting function;

[0042] 4. The present invention uses a CNN-GRU hybrid network, CNN, GRU and SVM to establish a regression model between different community environmental data and their NCRSI data to predict future NCRSI.

[0043] 5. The present invention provides an objective, rapid and effective new method for plant community monitoring, which has important guiding significance for the construction of plant communities and the improvement of ecological benefits in northern tea gardens.

[0044] 6. The normalized plant community remote sensing index NCRSI of the present invention is also suitable for evaluating tea gardens, orchards, plantations, ecological gardens, flower gardens, vegetable gardens, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only one embodiment of the present invention, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0046] Figure 1: A flow chart of a plant community monitoring method of the present invention;

[0047] Figure 2 : A distribution map of the locations of 9 plant communities in Example 1 of the present invention; wherein a is a location marked with Shandong Province, China and Chunxi Tea Garden, ba is a location marked with 9 plant communities, and cb is a picture showing the actual locations of the 9 plant communities;

[0048] Figure 3 : Pearson correlation coefficient analysis diagram between MS, RGB and TIR variables and temperature and humidity in Example 5 of the present invention;

[0049] Figure 4 : NCRSI variation diagram of 9 plant communities in Example 6 of the present invention;

[0050] Figure 5 : NCRSI fitting curves of 9 plant communities in Example 7 of the present invention;

[0051] Figure 6 : Scatter plot of actual and predicted NCRSI values; a is the CNN-GRU model, b is the CNN model, c is the GRU model, and d is the SVM model;

[0052] Figure 7 : Analysis of temperature and NCRSI characteristics of three types of plant communities; a is the response of the highest / lowest temperature at 0 cm to external high or low temperatures, b is the descriptive statistics of NCRSI, and c is the comparison of the coefficient of variation of NCRSI. DETAILED DESCRIPTION

[0053] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention. In the present invention, unless otherwise specified, the equipment and raw materials used can be purchased from the market or are commonly used in the art. The methods in the following embodiments are conventional methods in the art unless otherwise specified.

[0054] This study took 9 plant communities in Chunxi Tea Garden, Shandong Province, China as the research objects, extracted MS, RGB and TIR variables from multi-source remote sensing images of UAVs, selected 12 remote sensing variables that were highly correlated with environmental parameters to construct the Normalized Plant Community Remote Sensing Index (NCRSI); then, the Fourier function was used to fit the NCRSI change curve, and the deep learning network was used to predict the NCRSI. A normalized plant community remote sensing index and its application method were obtained.

[0055] A normalized plant community remote sensing index NCRSI, the formula of the ecological index NCRSI is as follows:

[0056] NCRSI=0.095X1+0.096X2+0.093X3+0.099X4+0.091X5+0.092X6+0.092X7+0.075X8+0.078X9+0.073X 10 +0.047X 11 +0.070X 12

[0057] Where X1 is NDVI, X2 is GNDVI, X3 is RECI, X4 is RENDVI, X5 is OSAVI, X6 is RVI, X7 is SR, X8 is BCor, X9 is GCor, X 10 It's RCor, X 11 It's Min, X 12 It's AVG.

[0058] Furthermore, the steps of constructing the remote sensing index NCRSI are as follows:

[0059] S1: Investigation and classification of plant community characteristics in tea gardens;

[0060] S2: Environmental temperature and humidity data collection in tea garden;

[0061] S3: Multi-source remote sensing data collection in tea gardens;

[0062] S4: Extract spectral information, texture information and temperature information from the multi-source remote sensing data obtained in S3;

[0063] S5: Perform correlation analysis on the data of S2 and S4 to screen out remote sensing variables that are highly correlated with temperature and humidity data;

[0064] S6: The remote sensing variables obtained in S5 are constructed into a normalized plant community remote sensing index NCRSI using a principal component analysis method; the normalized plant community remote sensing index NCRSI can be used to evaluate the ecological status of plant communities.

[0065] Furthermore, the normalized plant community remote sensing index NCRSI obtained in S6 is fitted with the Fourier function; the Fourier function is expressed as: f(x)=a0+acos(wx)+bsin(wx),

[0066] Where x is the number of days from the start of the experiment, a0, a, b and w are all constants; the fitting curve of NCRSI ( Figure 5 ) shows that the variation trend of the NCRSI fitting curve based on Fourier function is consistent with the actual value and can accurately fit the annual variation trend of NCRSI.

[0067] The present invention is described by taking a tea garden as an example, and the specific investigation is taken as an example of Chunxi Tea Garden in Linyi City, Shandong Province (117.45°E, 35.11°N). Normalized Plant Community Remote Sensing Index (NCRSI) is referred to in the present invention by the abbreviation NCRSI.

[0068] Example 1: Investigation of plant community characteristics

[0069] In order to study the distribution and structural characteristics of plant communities in tea gardens, representative plant communities should be selected during the period of vigorous plant growth (May-August) and the square sampling method (5 m side length, 25 m area) should be used. 2 ) to conduct a comprehensive survey of its basic characteristics (plant name, plant quantity and tree height, etc.).

[0070] In this example, the characteristics of plant communities in Chunxi Tea Garden, Linyi City, Shandong Province (117.45°E, 35.11°N) were investigated from May to August 2022. Nine different types of plant communities were selected as the basic types of plant communities in Chunxi Tea Garden ( Figure 2 ). And number the 9 plant communities as NO.1, NO.2, ..., NO.9.

[0071] According to the number of observed species, the nine plant communities were divided into three categories. Category 1 (C1): NO.1, NO.6 and NO.8, the plant community structure is relatively simple, mainly composed of bare ground and herbaceous plants, with low plant density, 0≤number of observed species≤3. Category 2 (C1): NO.2, NO.3 and NO.9, mainly composed of herbs and shrubs, 4≤number of observed species≤7. Category 3 (C3): NO.4, NO.5 and NO.7, the plant community is relatively complex, mainly composed of shrubs and trees, with high plant density, 8≤number of observed species≤10.

[0072] Example 2: Collection of ambient temperature and humidity data

[0073] The temperature and humidity sensors at the meteorological station are used to obtain long-term (≥1 year) and continuous environmental data to characterize the actual climate change and stability of the ecological tea garden.

[0074] In this embodiment, a weather station (Shandong Kerun Information Technology Co., Ltd.) was arranged at the center of 9 plant communities in Chunxi Tea Garden on August 9, 2022. The sensors installed on the weather station are powered by solar energy, and their data collection cycle is once every 30 minutes. The network layer is then responsible for receiving the data collected by the sensor and transmitting it to the server via the 4G network. Finally, the temperature and humidity data are queried through the web page (https: / / kr.miots.cn / login). Based on this, environmental data from August 9, 2022 to September 30, 2023 (a total of 418 days) were obtained, including air temperature and humidity (100 cm above the ground (100 cm)) and soil temperature and humidity (surface (0 cm), 20 cm underground (-20 cm) and 40 cm underground (-40 cm)), and they were statistically analyzed by excel, including daily averages, maximum and minimum values.

[0075] Example 3: Collection of multi-source remote sensing data

[0076] Multi-source and multi-temporal remote sensing data were obtained using a low-altitude UAV platform. The UAV flight platform was a DJI Matrice M300 (DJI Co., Ltd, Shenzhen, China), equipped with a dual gimbal, and equipped with MS600 PRO (Yusense Co., Ltd, Qingdao, China) and Zenmuse H20T (DJI Co., Ltd, Shenzhen, China). MS600 PRO is a customized multispectral (MS) sensor used to obtain MS data in 6 bands (B.450, G.555, R.666, RE.720, RE.750, NIR.840). Zenmuse H20T is equipped with a visible light camera and a thermal imaging camera, which are used to obtain three-channel (RGB) data and thermal infrared (TIR) ​​data, respectively.

[0077] Choose clear and windless weather and perform the flight mission between 11:00-14:00 at noon. Select the automatic planning aerial photography mode on the drone remote controller, set the flight altitude to 30-100m, and the flight speed to 2-10m / s; the shooting mode is triggered by the overlap rate, the heading overlap rate is set to 60%-80%, and the lateral overlap is set to 65%-75%. Save the mapping aerial photography mission and upload the route to perform the flight mission. During the one-year research period, carry out 10-15 flight missions, with an interval of about 30 days.

[0078] In this embodiment, the tea garden drone operation is carried out when the weather is clear, the lighting conditions are good, and the wind speed is low. The heading overlap rate is 80%, and the lateral overlap rate is 70%. The flight time is 11:00-14:00 at noon, the flight altitude is 30m, and the flight speed is 2.5m / s. The specific flight test dates are: June 11, 2022 (x1, indicating the number of days the test is carried out, this date is the first day), June 24 (x14), August 9 (x60), September 28 (x110), November 14 (x157), January 14, 2023 (x218), March 1 (x264), April 15 (x309), May 25 (x349), August 111 (x427), August 30 (x446) and September 27 (x474). A total of 12 flight tests were carried out, with intervals of 13d, 46d, 50d, 47d, 61d, 46d, 45d, 40d, 78d, 19d and 28d respectively.

[0079] Example 4: Extraction of remote sensing variables

[0080] First, the original MS, RGB and TIR images were preprocessed using Yusense Map (V1.0) software. The original, single MS image was output as a complete MS image of the study area through processes such as band registration, radiometric calibration, and image stitching. The original RGB and TIR images were output as RGB and TIR images of the study area through processes such as camera parameter generation and temperature calibration.

[0081] Then, the spectral information, texture information and temperature information are extracted from the MS, RGB and TIR images respectively.

[0082] The average spectral reflectance of each community was extracted from the MS image using the region of interest tool in ENVI (V5.3) software, and 15 common vegetation indices were calculated using MS data in six bands.

[0083] The second-order probability statistics tool in ENVI (V5.3) software was used to extract the R, G, and B texture information of each community from the RGB image. The texture information type is an 8-dimensional gray-level co-occurrence matrix, including mean (Mean), variance (Var), homogeneity (Hom), contrast (Con), dissimilarity (Dis), entropy (Ent), second-order moment (SM), and correlation (Cor).

[0084] The maximum temperature (Tmax), minimum temperature (Tmin) and mean temperature (Tmean) of each community were extracted from the TIR images using the rectangular temperature measurement tool in the DJI Thermal Analysis Tool (V3.0) software.

[0085] In summary, a total of 48 remote sensing variables (Table 1) were extracted in the experiment of the present invention, including 21 spectral variables, 24 texture variables and 3 temperature variables.

[0086] Table 1 Variables extracted from MS, RGB and TIR images

[0087]

[0088]

[0089] Example 5: Screening remote sensing variables that are sensitive to temperature and humidity changes

[0090] In order to better characterize the plant community status of the tea garden and determine the remote sensing variables that are sensitive to changes in temperature and humidity of the plant community, the analysis tools in SPSS software were used to conduct correlation analysis on 48 remote sensing variables and tea garden temperature and humidity data. Specifically, the bivariate correlation analysis tool was used to determine the statistic, the Pearson correlation coefficient was selected, the two-tailed significance test was selected, and the significance level was selected as 0.01.

[0091] The Pearson correlation coefficient matrix shows that ( Figure 3 ): The correlation between temperature and humidity data and TIR variables is the highest, followed by MS variables and RGB variables. A total of 25% (12) remote sensing variables highly correlated with temperature and humidity data were screened out, including TIR variables (Min, AVG), MS variables (NDVI, GNDVI, RENDVI, OSAVI, RECI, RVI, SR) and RGB variables (BCor, GCor, RCor).

[0092] Example 6: Construction and analysis of normalized plant community remote sensing index

[0093] Using principal component analysis (PCA), 12 remote sensing variables highly correlated with temperature and humidity were constructed into a Normalized Differential Plant Community Remote Sensing Index (NCRSI) that was sensitive to the status of plant communities in tea gardens.

[0094] PCA was performed in SPSS software. First, to avoid weight imbalance caused by inconsistent dimensions, the 12 remote sensing variables were standardized using descriptive statistics tools before PCA operation. Next, the dimensionality reduction tool was used to perform PCA operation, and based on the principle that the eigenvalue was greater than 1, three principal components were obtained. The cumulative contribution rate of the principal component was acceptable at more than 80%. Then, the linear combination coefficient (LCC) of each principal component was calculated using formula (1) using the eigenvalue and loading coefficient of the correlation matrix. Finally, the comprehensive score coefficient (CSC) was calculated according to formula (2) and standardized.

[0095]

[0096]

[0097] In formula (1), LCij is the load coefficient of the j-th principal component and the i-th index, and λ is the initial eigenvalue of the j-th. In formula (2), VC is the variance contribution rate, and CVC is the cumulative variance contribution rate.

[0098] In this embodiment, the PCA operation results are shown in Tables 2 and 3. The data in Table 3 show that the eigenvalues ​​of the first three principal components (Y1, Y2, and Y3) are all greater than 1, and their cumulative contribution rate is as high as 95.10%, which can reflect most of the information of the 12 remote sensing variables. Therefore, Y1, Y2, and Y3 are used to construct NCRSI. According to the data in Tables 3 and 4, LCC is calculated using formula (1) to obtain the linear composite expression of Y1, Y2, and Y3:

[0099] Y1=0.366X1+0.367X2+0.370X3+0.362X4+0.366X5+0.351X6+0.355X7+0.003X a +0.016X9-0.007X 10 +0.195X 11 +0.205X 12

[0100] Y2=0.009X1+0.032X2-0.006X3+0.053X4-0.002X5+0.056X6+0.062X7+0.549X8+0.548X9+0.546X 10 -0.267X 11 -0.133X 12

[0101] Y3=-0.047X1-0.090X2-0.072X3-0.059X4-0.101X5-0.1524X6-0.197X7+0.163X8+0.158X9+0.184X 10 +0.593X 11 +0.687X 12

[0102] Then, CSC is calculated using formula (2) to obtain the calculation formula of the principal component (Y):

[0103] Y=0.221X1+0.223X2+0.216X3+0.229X4+0.212X5+0.214X6+0.213X7+0.174X8+0.181X9+0.169X 10 +0.108X 11 +0.162X12

[0104] Finally, each coefficient of the Y equation is normalized based on the percentage method. The final formula for NCRSI is:

[0105] NCRSI=0.095X1+0.096X2+0.093X3+0.099X4+0.091X5+0.092X6+0.092X7+0.075X8+0.078X9+0.073X 10 +0.047X 11 +0.070X 12

[0106] Where X1 is NDVI, X2 is GNDVI, X3 is RECI, X4 is RENDVI, X5 is OSAVI, X6 is RVI, X7 is SR, X8 is BCor, X9 is GCor, X 10 It's RCor, X 11 It's Min, X 12 It's AVG.

[0107] To analyze the ecological significance of NCRSI, we analyzed the seasonal variation characteristics of NCRSI in different plant communities. Figure 4 The variation curve of NCRSI is shown, ranging from -1 to 6, and reaching the lowest peak in winter. The variation of NCRSI is consistent with the actual seasonal growth of deciduous trees in the north. NCRSI gradually decreases with the transition from midsummer to severe winter, and increases with the alternation of winter and summer, showing a cyclical change. Therefore, the change of NCRSI is closely related to the change of environment and vegetation, and can be used to characterize the status of plant communities in tea gardens. The higher the vegetation coverage rate of the tea garden plant community, the larger the NCRSI value; the lower the vegetation coverage rate of the tea garden plant community, the smaller the NCRSI value.

[0108] Table 2 Eigenvalues ​​and variance contribution rates of remote sensing variable correlation matrix

[0109]

[0110] Table 3 Principal component loadings of 12 remote sensing variables

[0111]

[0112]

[0113] Example 7: Fitting of normalized plant community remote sensing index

[0114] Fourier function is a very effective analytical method for fitting periodic data. In order to make up for the lack of remote sensing data during the experiment and promote the use of remote sensing data for multi-time series ecological analysis, we used Fourier function to fit NCRSI data and quantified the change curve of NCRSI. Specifically, using the curve fitter tool in MATLAB 2020 software, selecting Fourier function, setting the number of terms 1-8, the fitting can be completed to obtain the coefficients of the Fourier equation and its confidence interval.

[0115] In this embodiment, in order to improve the calculation efficiency, the number of Fourier terms selected is 1. The initial date of the NCRSI quantization curve is June 11, 2022, and the end date is September 30, 2023, a total of 477 days. The Fourier equation is expressed as:

[0116] f(x)=a0+acos(wx)+bsin(wx)

[0117] Where x is the number of days from the start date of the test (e.g., June 11, 2022, x = 1), and a0, a, b, and w are all constants.

[0118] The sum of squared errors (SSE), coefficient of determination (R 2 ), adjusted coefficient of determination (Adj.-R 2 ) and root mean square error (RMSE) were used to evaluate the fitting effect of the mechanism model on NCRSI. The smaller the SSE and RMSE, the closer they are to 0, indicating that the performance of the model is better; R 2 and Adj.R 2 The larger it is and the closer it is to 1, the better the data fitting effect is.

[0119] The fitting curve parameters in Table 4 show that the R 2 Greater than 0.844, the goodness of fit is in line with research expectations. The fitting curve of NCRSI ( Figure 5 ) shows that the variation trend of the NCRSI fitting curve based on the Fourier function is consistent with the actual value, and can accurately fit the annual variation trend of NCRSI. In addition, the Fourier function is periodic, which is very consistent with the seasonal periodic variation of NCRSI.

[0120] Table 4 NCRSI fitting curve parameters of 9 plant communities

[0121]

[0122] Example 8: Prediction of Normalized Plant Community Remote Sensing Index

[0123] Due to the large amount of environmental data, the environmental data obtained through long-term monitoring can accurately reflect the dynamic characteristics of vegetation growth changes in the form of external factors. Using temperature and humidity data as intermediate data, and using CNN-GRU, CNN, GRU and SVM networks to achieve collaborative conversion between environmental data and NCRSI data, and finally achieve the prediction of NCRSI data.

[0124] Specifically, in this embodiment, NCRSI data from August 9, 2022 to September 30, 2023 were selected, totaling 418 days. They are used in the NCRSI modeling process to better match the environmental data. Therefore, the NCRSI data sets and temperature and humidity data sets of the 9 plant communities each contain 418×9=3762 data. At the same time, the total data set is divided into a training set (2821 samples) and a test set (941 samples) according to the 3:1 rule, which are used for training and testing of the NCRSI regression model, respectively. Finally, a regression model was established between the environmental data of different communities and their NCRSI data using a CNN-GRU hybrid network, CNN, GRU, and SVM to predict future NCRSI (operating environment: MATLAB 2020).

[0125] Six indicators were selected to evaluate the prediction effect of the model on NCRSI, including RMSE of validation (RMSECV), calibration (RMSEC) and prediction (RMSEP), calibration (Rc 2 ) and forecast (Rp 2 ) 2 Among them, the smaller RMSECV, RMSEC and RMSEP are, the greater the Rc 2 and Rp 2 The larger it is, the better the performance of the model.

[0126] Table 5 shows the prediction performance of the model. The data show that CNN-GRU, CNN, GRU and SVM can achieve the expected results in predicting NCRSI (Rp 2 >0.888). Specifically, compared with single models CNN, GRU and machine learning model SVM, CNN-GRU has the best prediction accuracy, Rp 2 and RMSEP were 0.955 and 0.314, respectively. Figure 6 The black diagonal line represents the 1:1 line, and the orange dotted line represents the regression line between the predicted and actual values. The predicted values ​​of NCRSI are all closely distributed around the regression line, indicating that the four models have good predictive performance.

[0127] Table 5 Prediction performance of CNN-GRU, CNN, GRU and SVM

[0128]

[0129] Example 9: Analysis of the time series characteristics of temperature data and NCRSI

[0130] In order to evaluate the response characteristics of different plant community categories to temperature, attention should be paid to periods of extreme high temperature or extreme low temperature. In this example, special attention was paid to four time periods with drastic temperature changes. Summer high temperature stage: 11 / 08 / 2022-18 / 08 / 2022 and 20 / 06 / 2023-27 / 06 / 2023; winter low temperature stage: 28 / 11 / 2022-05 / 12 / 2023 and 11 / 03 / 2023-18 / 03 / 2023. The maximum (Max-) and minimum (Min-) temperature changes (0 cm) of C1, C2 and C3 plant communities were analyzed. Figure 7 a). It can be seen that in the high temperature stage, the temperature of C1 is always higher than that of C2 and C3, and C2 and C3 show good buffering capacity. In the low temperature stage, the temperature of C1 is always lower than that of C2 and C3, and C2 and C3 show good insulation effect. In summary, compared with C1, the rich vegetation of C2 and C3 creates a local climate environment for tea trees that is warm in winter and cool in summer, which can maintain a lower temperature when the temperature rises in summer and maintain a higher temperature when the temperature drops in winter, and has good stability.

[0131] Figure 7 (b) shows the statistical analysis of NCRSI for three types of plant communities, including maximum, minimum, standard deviation, and coefficient of variation (CV). The data show that the CVs of C1, C2, and C3 are 0.518, 0.461, and 0.458, respectively. The three types of plant communities are sorted in ascending order of coefficient of variation: C3<C2<C1. Using the CV value of the plant community helps to clarify the degree and speed of data change and reflects the discreteness and volatility of the NCRSI distribution. If the vegetation characteristics of the plant community and the microclimate tend to be stable, then the smaller the fluctuation of NCRSI, the smaller the CV value; conversely, if the vegetation characteristics of the plant community and the microclimate change dramatically, then the NCRSI will fluctuate greatly and have a larger CV value. Therefore, among the three types of plant communities, C3 has the best stability, while C1 has the worst stability.

[0132] In order to further verify the feasibility of NCRSI of three types of plant communities to represent the stability of plant communities, we analyzed the coefficient of variation of NCRSI in four remote sensing flight survey periods, which were 09 / 08 / 2022-28 / 09 / 2022, 14 / 11 / 2022-14 / 01 / 2023, 01 / 03 / 2023-15 / 04 / 2023 and 25 / 05 / 2023-11 / 08 / 2023, covering the four periods of drastic temperature changes mentioned above. Figure 7(c) shows the comparison results of the coefficient of variation of NCRSI of the three types of plant communities. It can be seen that the CV value of C1 is always the largest, the CV value of C2 is the second, and the CV value of C3 is always the smallest. This is consistent with the differences in the responses of the three types of plant communities to extreme temperatures ( Figure 7 a) is consistent. The results show that NCRSI can effectively characterize the buffering effect of tea garden plant communities on temperature and the stability of vegetation phenology.

[0133] It is understandable that those skilled in the art can make equivalent substitutions or changes based on the technical solution and concept of the present invention, and all these changes or substitutions should fall within the protection scope of the claims attached to the present invention.

Claims

1. A normalized plant community remote sensing index NCRSI, characterized in that: The formula of the remote sensing index NCRSI is as follows: Where, X1 is NDVI, X2 is GNDVI, X3 is RECI, X4 is RENDVI, X5 is OSAVI, X6 is RVI, X7 is SR, X8 is BCor, X9 is GCor, X10 is RCor, X11 is Min, and X12 is AVG; The steps of constructing the remote sensing index NCRSI are as follows: S1: Investigation and classification of plant community characteristics; The S1 investigated the characteristics of plant communities, and the plant communities were divided into three categories; Category 1, plant community structure is relatively simple, mainly composed of bare ground and herbaceous plants, with low plant density, 0≤ number of observed species≤3; Category 2, mainly composed of herbs and shrubs, 4≤ number of observed species≤7; Category 3, plant community is relatively complex, mainly composed of shrubs and trees, with high plant density, 8≤ number of observed species≤10; S2: Environmental temperature and humidity data collection; The data of S2 is collected from the temperature and humidity sensor of the weather station. The sensor is powered by solar energy and the data collection cycle is once every 30 minutes. The air temperature and humidity data in S2 are taken from 100 cm above the ground; the soil temperature and humidity data are taken from 0 cm on the surface, 20 cm underground, and 40 cm underground respectively. S3: Multi-source remote sensing data acquisition; The S3 uses an unmanned aerial vehicle (UAV) flight platform to collect multi-source remote sensing data in the research area. The UAV flight platform is a DJI Matrice M300 equipped with a dual gimbal bracket, equipped with MS600 PRO and Zenmuse H20T. The flight altitude of the UAV flight platform is 30m, the flight speed is 2.5m / s, and the flight time is 11:00-14:00 noon. The weather is clear, the lighting conditions are good, and the wind speed is low. S4: Extract spectral information, texture information and temperature information from the multi-source remote sensing data obtained in S3; In S4, ENVI is used to extract spectral information and texture information; DJI Thermal Analysis Tool is used to extract temperature information; and spectral information, texture information and temperature information are extracted from the multi-source remote sensing data obtained in S3; the steps are as follows: The original MS, RGB, and TIR images were preprocessed using Yusense Map software; Through band registration, radiometric calibration, and image stitching, the original, single MS image is output as a complete MS image of the study area; through camera parameter generation and temperature calibration, the original RGB and TIR images are output as RGB and TIR images of the study area; Extract spectral information, texture information and temperature information from MS, RGB and TIR images respectively; The mean spectral reflectance of each community was extracted from the MS images using the region of interest tool in ENVI software. The three-channel texture information of each community was extracted from the RGB image using the second-order probability statistics tool in the ENVI software; the texture information type was an 8-dimensional gray-level co-occurrence matrix, including mean Mean, variance Var, homogeneity Hom, contrast Con, dissimilarity Dis, entropy Ent, second-order moment SM, and correlation Cor; The maximum temperature Tmax, minimum temperature Tmin, and average temperature Tmean of each community were extracted from the TIR image using the rectangular temperature measurement tool in the DJI Thermal Analysis Tool software; S5: Perform correlation analysis on the data of S2 and S4 to screen out remote sensing variables that are highly correlated with temperature and humidity data; In S5, the analysis tools in SPSS software are used to perform correlation analysis on the 48 remote sensing variables obtained in S4 and the temperature and humidity data of the tea garden; the bivariate correlation analysis tool is used to determine the statistic, the correlation coefficient is selected as Pearson, the significance test is selected as two-tailed, and the significance level is selected as 0.01; the Pearson correlation coefficient matrix shows that the correlation between the temperature and humidity data and the TIR variable is the highest, followed by the MS variable and the RGB variable, thereby screening out a total of 12 remote sensing variables that are highly correlated with the temperature and humidity data, including TIR variables Min, AVG; MS variables NDVI, GNDVI, RENDVI, OSAVI, RECI, RVI, SR and RGB variables BCor, GCor, RCor; S6: constructing the remote sensing variables obtained in S5 into a normalized plant community remote sensing index NCRSI using a principal component analysis method; the normalized plant community remote sensing index NCRSI is used to evaluate the ecological status of the plant community; The normalized plant community remote sensing index NCRSI obtained in S6 is fitted with the Fourier function; the Fourier function is expressed as: , In the formula, x is the number of days from the initial date of the experiment, and a0, a, b and w are all constants. The fitting curve of NCRSI shows that the change trend of the NCRSI fitting curve based on Fourier function is consistent with the actual value, and can accurately fit the annual change trend of NCRSI. The plant community remote sensing index NCRSI is applicable to tea gardens.

2. Application of the Normalized Plant Community Remote Sensing Index (NCRSI) as claimed in claim 1 in the assessment of plant communities in tea gardens.

3. The use according to claim 2, characterized in that: A regression model was established between the environmental data of different plant communities and their NCRSI data using a CNN-GRU hybrid network, CNN, GRU, and SVM to predict future NCRSI.

Images