A winter wheat phenology monitoring method for heterogeneous regions

Abstract

The application discloses a winter wheat phenological stage monitoring method for heterogeneous regions and relates to the technical field of remote sensing monitoring. The method extracts band reflection information by performing smoothing processing on remote sensing image data of a phenological stage monitoring region. EVI2 is calculated through the band reflection information, and an EVI2 time sequence curve is obtained. An SMF model is constructed to perform curve fitting on the EVI2 time sequence curve, the model phenological stage matching algorithm is optimized in combination with the spatial variation law of the phenological stage monitoring region, and a GSMF model is constructed. The phenological stage difference is calculated in view of the growth period difference caused by geographical factors of the phenological monitoring region, and the monitoring of the winter wheat phenological stage of the heterogeneous region is realized. The method combines remote sensing information and geographical information of the phenological stage monitoring region for the monitoring of the winter wheat phenological stage, has strong spatial expansibility and universality, and can realize the accurate prediction of the winter wheat phenological stage of the heterogeneous region.

Description

Technical Field

[0001] This invention relates to the field of remote sensing monitoring technology, and mainly to a method for monitoring the phenological stages of winter wheat in heterogeneous regions. Background Technology

[0002] Crop phenology refers to the changes in external morphological characteristics during crop growth and development, which is crucial for understanding crop dynamics, yield and quality formation, and field management. Traditional phenological monitoring relies on field surveys, which are costly and difficult to obtain, failing to meet the spatial requirements of large-scale crop growth monitoring and yield estimation studies for phenological information. Methods suitable for large-scale phenological monitoring are mainly divided into two categories: crop growth model phenological simulation based on meteorological data and phenological monitoring based on remote sensing data characteristics. Crop growth model phenological simulation based on meteorological data relies on crop variety information, meteorological information, and management information, making input data acquisition difficult in spatially heterogeneous regions and parameter determination during phenological monitoring a challenging task. Phenological feature extraction based on remote sensing data can utilize vegetation index VI time series for phenological remote sensing observation, demonstrating significant potential in terms of timeliness and spatial heterogeneity.

[0003] Thresholding, change monitoring, and shape model fitting are the three main methods for remote sensing monitoring of crop phenological periods. Thresholding defines a specific value in the VI time series as a specific phenological period, such as a fixed threshold or a dynamic threshold like the mean, median, or rate over a certain time period. Change monitoring determines specific phenological periods by monitoring characteristic points such as the maximum and minimum values ​​(inflection points) of the VI time series curve. Thresholding and change monitoring methods primarily extract phenological periods based on mathematical definitions, resulting in a limited number of defined phenological periods, mainly focusing on the seedling, heading, and harvest stages. Furthermore, the phenological periods predicted by these methods differ significantly from the actual phenological periods at ground time. Shape model fitting (SMF) uses linear scaling to match the phenological periods throughout the growing season, addressing the issue of limited phenological period definitions. However, the SMF model assumes that the date of a specific crop phenological period is relatively stable on the VI time series curve, assuming all growth stages change synchronously. This not only fails to reflect the growth patterns of winter wheat before and after winter but also neglects the phenological period differences caused by variations in crop growth across heterogeneous regions. Furthermore, using the mean VI time series of a certain region and its corresponding phenological period as a standard to determine the phenological dates of different regions has advantages on a small scale, but the large differences in crop growth in heterogeneous spaces can lead to significant errors. Heterogeneous space, or heterogeneous region, refers to local areas in remote sensing image data of winter wheat that are different in color from the surrounding areas. Summary of the Invention

[0004] This invention addresses the problems of existing winter wheat phenological monitoring methods, such as difficulty in monitoring crop growth over large areas, large monitoring errors in winter wheat in heterogeneous regions, and limited number of phenological stages to be monitored. It provides a method for monitoring winter wheat phenological stages in heterogeneous regions.

[0005] A method for monitoring the phenological stages of winter wheat in heterogeneous regions includes:

[0006] S1. Obtain remote sensing image data of the winter wheat phenological period monitoring area through MODIS satellite, and smooth the remote sensing image data to obtain smoothed remote sensing data;

[0007] S2. Extract band reflectance information from smoothed remote sensing data, and calculate the dual-band enhanced vegetation index EVI2 of the winter wheat phenological monitoring area using the band reflectance information;

[0008] S3. Extract the time series curve of the dual-band enhanced vegetation index EVI2 to obtain the EVI2 time series curve;

[0009] S4. Construct a reference shape fitting model SMF, and obtain the EVI2 time series fitting curve based on the EVI2 time series curve;

[0010] S5. Statistically analyze the measured phenological data of the winter wheat phenological monitoring area, and combine it with the spatial data of the winter wheat phenological monitoring area to obtain the spatial variation law of the phenological monitoring area;

[0011] S6. Construct a geographic weighted shape fitting model GSMF, and optimize the SMF phenological period matching algorithm by combining the spatial variation pattern of the phenological period monitoring area;

[0012] S7. By optimizing the S algorithm through the golden section search and using the Akaike Information Content Criterion (AIC) to select bandwidth parameters, the phenological stages of winter wheat in heterogeneous regions can be extracted.

[0013] In S2, the formula for calculating the dual-band enhanced vegetation index EVI2 is as follows:

[0014]

[0015] In the formula, R NIR R represents the surface reflectance in the near-infrared band. RED This represents the surface reflectance in the red light band.

[0016] In S3, the EVI2 time series curve h(x) is as follows:

[0017]

[0018] In the formula, i and j are the amplitude factors of the EVI2 time series curve, which determine the peak height of the EVI2 time series curve; x is the number of days monitored during the phenological period; f and h are the center positions of the peak of the EVI2 time series curve; a and b are the standard deviations, which determine the distance between the troughs of the EVI2 time series curve.

[0019] In S4, a reference shape fitting model SMF is constructed, and the EVI2 time series curve is averaged to obtain the reference EVI2 time series curve Q(x). Two curve segments are selected from Q(x), and the two curve segments are translated and scaled to obtain the EVI2 time series fitting curve g(x). The calculation process of g(x) is as follows:

[0020] g(x)=yscale1×Q1(xscale1×x+tshift1)+yscale2×Q2(xscale2×x+tshift2);

[0021] In the formula, Q1(x) and Q2(x) are two curves selected from Q(x), yscale1 and yscale2 are the Y-coordinate scaling factors corresponding to the two curves, xscale1 and xscale2 are the X-coordinate scaling factors corresponding to the two curves, and tshift1 and tshift2 are the translation step sizes corresponding to the two curves.

[0022] The Y-axis scaling factor, X-axis scaling factor, and translation step size are all scaling parameters, selected and evaluated using the root mean square error (RMSE). The calculation formula is as follows:

[0023]

[0024] In the formula, f(x) is the predicted quantity of winter wheat in the monitoring area, and m(x) is the actual quantity of winter wheat in the monitoring area.

[0025] In S6, the GSMF model retains the model structure and target curve transformation method of the SMF model, and modifies the phenological period matching algorithm based on the SMF model. It incorporates phenological period differences caused by crop growth variations to estimate the phenological period of winter wheat in the monitoring area. The modified model's phenological period matching algorithm X... est The calculation process is as follows:

[0026] X est =(a 0(ui,vi) +a 1(ui,vi) xscale (ui,vi) )×(X0+a 2(ui,vi) tshift (ui,vi) +ε (ui,vi) );

[0027] In the formula, i = 1, 2, 3, ..., m; m represents the number of observations; a0(ui,vi) a 1(ui,vi) and a 2(ui,vi) This represents the coefficient vector of the phenological period estimation model at position (ui,vi); xscale (ui,vi) tshift represents the scaling factor of the X-coordinate at position (ui,vi); (ui,vi) ε represents the translation step size at position (ui,vi); (ui,vi) X represents the phenological period estimation error at position (ui,vi); X0 is the reference phenological period of the SMF; X est The value represents the estimated phenological period of the target pixel.

[0028] Using the location of a pixel as its centroid, the distances between pixels are calculated to obtain the coefficient vector of the local pixels. as follows:

[0029]

[0030] In the formula, X is the reference phenological period of the GSMF model; X′ is the estimated phenological period of the GSMF model; W (ui,vi) Let represent an n×n diagonal weight matrix, which weights the observations based on their distances from the location (ui, vi) within the phenological observation area; [ ] -1 y is the inverse of the matrix; y is the phenological observation value.

[0031] In S7, the formula for calculating the Akaike Information Content Criterion (AIC) is as follows:

[0032]

[0033] In the formula, AIC value is the Akaike information entropy of the Akaike information content criterion, m represents the number of observations, RSS represents the residual sum of squares, S represents the phenological period observation parameter matrix, and tr(s) is the trace number of matrix S.

[0034] Compared with the prior art, the present invention has the following advantages:

[0035] Based on a thorough consideration of the feasibility and practicality of the technical solution, this invention conducts an in-depth investigation into the phenological information of winter wheat in heterogeneous regions and the ease of obtaining the required data. It proposes a method for monitoring the phenological stages of winter wheat in heterogeneous regions, combining remote sensing information and geographic information of the monitoring area. This method exhibits strong spatial scalability and universality, enabling accurate prediction of key phenological stages of winter wheat on a large regional scale. By modifying the phenological parameter acquisition method in the classic shape model fitting model (SMF), a GSMF model is constructed. This model incorporates geospatial information of latitude, longitude, altitude, and slope of the phenological monitoring area to perform location weighting on the corresponding observations. It calculates phenological differences based on the differences in growing season caused by geographical factors, enabling the extraction of the greening, jointing, booting, heading, flowering, and harvesting stages of different regions based on band reflectance information from remote sensing data. Compared to other remote sensing data, band reflectance information is easier to obtain and can be used to extract phenological information of winter wheat in different regions and growing seasons, exhibiting high accuracy, strong universality, and stability. Meanwhile, this approach can be applied to the monitoring of phenological stages of different remote sensing platforms and different crops (such as corn and rice). In practical applications, the spatiotemporal resolution and collection method of the required remote sensing data can be determined according to specific needs such as the area of ​​the crop monitoring region. Attached Figure Description

[0036] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0037] Figure 1 A flowchart of phenological period monitoring based on a geographically weighted shape fitting model provided in an embodiment of the present invention;

[0038] Figure 2 A 1:1 scatter plot of estimated and measured phenological periods based on the SMF model provided in an embodiment of the present invention;

[0039] Figure 3 A 1:1 scatter plot of estimated and measured phenological periods based on the GSMF model provided in an embodiment of the present invention;

[0040] Figure 4 A 1:1 scatter plot of estimated and measured phenological periods in 2014 based on the GSMF model, provided for embodiments of the present invention;

[0041] Figure 5 A 1:1 scatter plot of estimated and measured phenological periods in 2015 based on the GSMF model, provided for an embodiment of the present invention;

[0042] Figure 6A 1:1 scatter plot of estimated and measured phenological periods in 2016 based on the GSMF model, provided for an embodiment of the present invention;

[0043] Figure 7 A 1:1 scatter plot of estimated and measured phenological periods in 2017 based on the GSMF model, provided for embodiments of the present invention;

[0044] Figure 8 A 1:1 scatter plot of estimated and measured phenological periods in 2018 based on the GSMF model, provided for embodiments of the present invention. Detailed Implementation

[0045] This invention discloses a method for monitoring the phenological stages of winter wheat in heterogeneous regions. Those skilled in the art can refer to the content of this document and appropriately modify the process parameters to achieve the desired result. It should be particularly noted that all similar substitutions and modifications are obvious to those skilled in the art and are considered to be included in this invention. Furthermore, those skilled in the art can clearly modify or appropriately alter and combine the content described herein without departing from the content, spirit, and scope of this invention to implement and apply the technology of this invention.

[0046] In this invention, unless otherwise stated, the scientific and technical terms used herein have the meanings commonly understood by those skilled in the art. To enable those skilled in the art to better understand the technical solutions of this invention, the invention will be further described in detail below with reference to specific embodiments.

[0047] The GSMF model adopts the model structure and target curve transformation method of SMF, and modifies the phenological period matching algorithm based on the SMF model. It combines the geospatial information of latitude, longitude, altitude and slope of the phenological monitoring area to perform position weighting on the corresponding observation values, calculates the differences in crop growth caused by regional differences, and realizes the estimation of the phenological period of winter wheat in heterogeneous regions.

[0048] A method for monitoring the phenological stages of winter wheat in heterogeneous regions includes:

[0049] S1. Obtain remote sensing image data of the winter wheat phenological period monitoring area through MODIS satellite, and smooth the remote sensing image data to obtain smoothed remote sensing data;

[0050] S2. Extract band reflectance information from smoothed remote sensing data, and calculate the dual-band enhanced vegetation index EVI2 of the winter wheat phenological monitoring area using the band reflectance information;

[0051] S3. Extract the time series curve of the dual-band enhanced vegetation index EVI2 to obtain the EVI2 time series curve;

[0052] S4. Construct a reference shape fitting model SMF, and obtain the EVI2 time series fitting curve based on the EVI2 time series curve;

[0053] S5. Statistically analyze the measured phenological data of the winter wheat phenological monitoring area, and combine it with the spatial data of the winter wheat phenological monitoring area to obtain the spatial variation law of the phenological monitoring area;

[0054] S6. Construct a geographic weighted shape fitting model GSMF, and optimize the SMF phenological period matching algorithm by combining the spatial variation pattern of the phenological period monitoring area;

[0055] S7. By optimizing the S algorithm through the golden section search and using the Akaike Information Content Criterion (AIC) to select bandwidth parameters, the phenological stages of winter wheat in heterogeneous regions can be extracted.

[0056] In S2, the formula for calculating the dual-band enhanced vegetation index EVI2 is as follows:

[0057]

[0058] In the formula, R NIR R represents the surface reflectance in the near-infrared band. RED This represents the surface reflectance in the red light band.

[0059] In S3, the EVI2 time series curve h(x) is as follows:

[0060]

[0061] In the formula, i and j are the amplitude factors of the EVI2 time series curve, which determine the peak height of the EVI2 time series curve; x is the number of days monitored during the phenological period; f and h are the center positions of the peak of the EVI2 time series curve; a and b are the standard deviations, which determine the distance between the troughs of the EVI2 time series curve.

[0062] In S4, a reference shape fitting model SMF is constructed, and the EVI2 time series curve is averaged to obtain the reference EVI2 time series curve Q(x). Two curve segments are selected from Q(x), and the two curve segments are translated and scaled to obtain the EVI2 time series fitting curve g(x). The calculation process of g(x) is as follows:

[0063] g(x)=yscale1×Q1(xscale1×x+tshift1)+yscale2×Q2(xscale2×x+tshift2);

[0064] In the formula, Q1(x) and Q2(x) are two curves selected from Q(x), yscale1 and yscale2 are the Y-coordinate scaling factors corresponding to the two curves, xscale1 and xscale2 are the X-coordinate scaling factors corresponding to the two curves, and tshift1 and tshift2 are the translation step sizes corresponding to the two curves.

[0065] The Y-axis scaling factor, X-axis scaling factor, and translation step size are all scaling parameters, selected and evaluated using the root mean square error (RMSE). The calculation formula is as follows:

[0066]

[0067] In the formula, f(x) is the predicted quantity of winter wheat in the monitoring area, and m(x) is the actual quantity of winter wheat in the monitoring area.

[0068] In S6, the GSMF model retains the model structure and target curve transformation method of the SMF model, and modifies the phenological period matching algorithm based on the SMF model. It incorporates phenological period differences caused by crop growth variations to estimate the phenological period of winter wheat in the monitoring area. The modified model's phenological period matching algorithm X... est The calculation process is as follows:

[0069] X est =(a 0(ui,vi) +a 1(ui,vi) xscale (ui,vi) )×(X0+a 2(ui,vi) tshift (ui,vi) +ε (ui,vi) );

[0070] In the formula, i = 1, 2, 3, ..., m; m represents the number of observations; a 0(ui,vi) a 1(ui,vi) and a 2(ui,vi) This represents the coefficient vector of the phenological period estimation model at position (ui,vi); xscale (ui,vi) tshift represents the scaling factor of the X-coordinate at position (ui,vi); (ui,vi) ε represents the translation step size at position (ui,vi); (ui,vi) X represents the phenological period estimation error at position (ui,vi); X0 is the reference phenological period of the SMF; X est The value represents the estimated phenological period of the target pixel.

[0071] Using the location of a pixel as its centroid, the distances between pixels are calculated to obtain the coefficient vector of the local pixels. as follows:

[0072]

[0073] In the formula, X is the reference phenological period of the GSMF model; X′ is the estimated phenological period of the GSMF model; W (ui,vi) Let represent an n×n diagonal weight matrix, which weights the observations based on their distances from the location (ui, vi) within the phenological observation area; [ ] -1 y is the inverse of the matrix; y is the phenological observation value.

[0074] In S7, the formula for calculating the Akaike Information Content Criterion (AIC) is as follows:

[0075]

[0076] In the formula, AIC value is the Akaike information entropy of the Akaike information criterion, m represents the number of observations, RSS represents the sum of squared residuals, S represents the phenological period observation parameter matrix, and tr(s) is the trace of matrix S. By using the Akaike information criterion AIC, the minimum model bandwidth can be selected, reducing the complexity of model construction.

[0077] This invention comprehensively considers factors such as spatial differences in phenological periods and the ease of obtaining the required data. Based on field phenological period survey data, it modifies the phenological period parameter acquisition method in the classic shape model fitting model SMF to extract the greening period, jointing period, booting period, heading period, flowering period, and harvest period of different regions based on the band reflection information in remote sensing data. The invention is applied and verified in the Huang-Huai-Hai region. The RMSE comparison results of the SMF model and the GSMF model at different phenological periods are shown in Table 1.

[0078] Table 1. RMSE tables of SMF and GSMF models at different phenological stages;

[0079]

[0080] The smaller the root mean square error (RMSE), the smaller the difference between the estimated phenological period and the measured phenological period, and the better the model's phenological period estimation effect for winter wheat. As can be seen from Table 1, the RMSE of the GSMF model for each phenological period of winter wheat is smaller than that of the SMF model for each phenological period of winter wheat. Therefore, the GSMF model has a better phenological period estimation effect for winter wheat than the SMF model.

[0081] Figure 1This is a flowchart for phenological period monitoring based on a geographically weighted shape fitting model. Remote sensing image data of the phenological period monitoring area is acquired via MODIS satellite. The remote sensing image data undergoes smoothing processing. Based on the band reflectance information in the remote sensing image data, the crop EVI2 time series curve is calculated. The spatial variation pattern of phenological periods is derived by combining spatial data of latitude, longitude, altitude, and slope of the phenological period monitoring area with measured phenological period data. A geographically weighted shape fitting model is constructed through shape fitting model transformation, estimation algorithm modification, and bandwidth parameter selection to optimize phenological period estimation and obtain phenological period estimation results. The estimation results are validated using independent samples, thus realizing phenological period monitoring based on the geographically weighted shape fitting model.

[0082] Figure 2 This is a 1:1 scatter plot comparing the estimated and measured phenological periods of winter wheat in the Huang-Huai-Hai region based on the SMF model. Figure 2 It can be seen that the estimated phenological period based on the SMF model is generally later than the measured phenological period, and the root mean square error of the estimation results is relatively large, making it impossible to accurately estimate the grain-filling period of winter wheat.

[0083] Figure 3 This is a 1:1 scatter plot comparing the estimated and measured phenological periods of winter wheat in the Huang-Huai-Hai region based on the GSMF model. Figure 3 It can be seen that the phenological period estimation of winter wheat based on the GSMF model has a high degree of overlap with the measured phenological period. Through comparison... Figure 2 and Figure 3 As can be seen from Table 1, the GSMF model has a better estimation effect than the SMF model, and its accuracy and precision are higher for monitoring the phenological stage of winter wheat.

[0084] pass Figures 4 to 8 It can be seen that when monitoring the phenological stages of winter wheat based on the GSMF model, its monitoring effect on the phenological stages of different years is relatively stable, and the model's estimation results for each phenological stage of winter wheat are basically consistent with the actual measured results. The error between the phenological stage estimation results and the actual survey results is small, that is, the GSMF model can accurately estimate multiple phenological stages of wheat in different regions.

[0085] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for monitoring the phenological stages of winter wheat in heterogeneous regions, characterized in that, include: S1. Obtain remote sensing image data of the winter wheat phenological period monitoring area through MODIS satellite, and smooth the remote sensing image data to obtain smoothed remote sensing data; S2. Extract band reflectance information from smoothed remote sensing data, and calculate the dual-band enhanced vegetation index EVI2 of the winter wheat phenological monitoring area using the band reflectance information; S3. Extract the time series curve of the dual-band enhanced vegetation index EVI2 to obtain the EVI2 time series curve; S4. Construct a reference shape fitting model SMF, and obtain the EVI2 time series fitting curve based on the EVI2 time series curve; S5. Statistically analyze the measured phenological data of the winter wheat phenological monitoring area, and combine it with the spatial data of the winter wheat phenological monitoring area to obtain the spatial variation law of the phenological monitoring area; S6. Construct a geographic weighted shape fitting model GSMF, and optimize the SMF phenological period matching algorithm by combining the spatial variation pattern of the phenological period monitoring area; S7. By optimizing the S algorithm through the golden section search and using the Akaike Information Content Criterion (AIC) to select bandwidth parameters, the phenological stages of winter wheat in heterogeneous regions can be extracted. In S4, a reference shape fitting model (SMF) is constructed, and the EVI2 time series curve is averaged to obtain the reference EVI2 time series curve. ,from Two curve segments are selected, and the two curve segments are translated and scaled to obtain the EVI2 time series fitting curve. , The calculation process is as follows: ； In the formula, and From The two curves selected in the text, and The scaling factor for the Y-coordinates of the two curve segments is [missing information]. and The scaling factor for the X-coordinates of the two curve segments is [missing information]. and This represents the translation step size corresponding to the two curve segments; In S6, the GSMF model retains the model structure and target curve transformation method of the SMF model, and modifies the phenological period matching algorithm based on the SMF model. It incorporates phenological period differences caused by crop growth variations to estimate the phenological period of winter wheat in the monitoring area. The modified model's phenological period matching algorithm... The calculation process is as follows: ； In the formula, =1, 2, 3, ... ; Indicates the number of observations; , and Indicates in The coefficient vector of the phenological period estimation model for location; Indicates in The scaling factor for the X-coordinate of the position; Indicates in The translation step size of the position; Indicates in Errors in estimating phenological periods at location; This serves as the reference phenological period for SMF. The value represents the estimated phenological period of the target pixel; Using the location of a pixel as its centroid, the distances between pixels are calculated to obtain the coefficient vector of the local pixels. as follows: ； In the formula, The reference phenological period for the GSMF model; For the phenological period estimated by the GSMF model; express A diagonal weighting matrix, which is based on each observation and location in the phenological observation area. The distances between them are used to weight the corresponding observations based on their location. It is the inverse of the matrix; These are observations from the phenological period.

2. The method for monitoring the phenological stages of winter wheat in heterogeneous regions according to claim 1, characterized in that, In S2, the formula for calculating the dual-band enhanced vegetation index EVI2 is as follows: ； In the formula, The surface reflectance is in the near-infrared band. This represents the surface reflectance in the red light band.

3. The method for monitoring the phenological stages of winter wheat in heterogeneous regions according to claim 2, characterized in that, In S3, the EVI2 time series curve as follows: ； In the formula, and is the amplitude factor of the EVI2 time series curve, which determines the peak height of the EVI2 time series curve; x is the number of days monitored during the phenological period; and This indicates the center position of the peak value of the EVI2 time series curve; and The standard deviation is used to determine the distance between the troughs of the EVI2 time series curve.

4. The method for monitoring the phenological stages of winter wheat in heterogeneous regions according to claim 3, characterized in that, The Y-axis scaling factor, X-axis scaling factor, and translation step size are all scaling parameters, selected and evaluated using the root mean square error (RMSE). The calculation formula is as follows: ； In the formula, To predict the quantity of winter wheat in the monitoring area, To monitor the actual quantity of winter wheat in the region.

5. The method for monitoring the phenological stages of winter wheat in heterogeneous regions according to claim 4, characterized in that, In S7, the Akaike Information Content Criterion The calculation formula is: ； In the formula, The value is the Akaike information entropy, which is the Akaike information content criterion. Indicates the number of observations. Represents the sum of squared residuals. This represents the matrix of observation parameters during the phenological period. For matrix The number of traces.

Images