A winter wheat yield estimation method based on multi-source remote sensing data fusion

Abstract

This invention provides a method for estimating winter wheat yield based on multi-source remote sensing data fusion. The method includes selecting MODIS and GF-1WFV data and preprocessing the data; fusing MODIS and GF-1WFV data using the enhanced spatiotemporal adaptive reflectance fusion model ESTARFM, performing consistency checks on the fusion results, and generating time-series data; using a decision tree method to extract winter wheat planting areas based on the temporal variation pattern of winter wheat NDVI; extracting growth monitoring indicators from the multi-source remote sensing fusion data, obtaining meteorological data for the study area using spatial interpolation methods, dividing the study area into stages according to the key growth stages of winter wheat, and statistically calculating the average values ​​of indicators for the key growth stages; constructing a multi-parameter, multi-time-window winter wheat yield estimation model based on the growth indicators and meteorological parameters of the key growth stages of winter wheat; and verifying the yield estimation accuracy. This invention can obtain winter wheat yield estimation results with continuous time and higher spatial resolution.

Description

Technical Field

[0001] This invention relates to the field of field crop growth monitoring, and in particular to a method for estimating winter wheat yield based on multi-source remote sensing data fusion. Background Technology

[0002] Timely, accurate, and wide-ranging acquisition of grain production information is of great significance for guiding agricultural policy formulation, regulating grain prices, and ensuring national food security. Remote sensing technology is an important tool for regional grain production estimation; however, a single remote sensing sensor cannot simultaneously acquire high temporal and spatial resolution images. Therefore, there is an urgent need to fuse remote sensing data from different sensors to achieve complementary advantages among various sensor types.

[0003] Crop yield is the result of the combined effect of multiple factors. The selection of yield estimation indicators and time windows can have a significant impact on the crop yield estimation results. Therefore, it is urgent to determine a set of yield estimation indicators and time window selection methods suitable for fused data, so as to improve the accuracy of yield estimation. Summary of the Invention

[0004] To overcome the limitations of single remote sensing sensors and provide higher spatiotemporal resolution winter wheat yield estimation results, this invention provides a method for winter wheat yield estimation based on multi-source remote sensing data fusion. This invention determines the most suitable yield estimation indicators and time window selection method based on multi-source remote sensing fusion data. This method is applicable to large-scale winter wheat yield estimation and has advantages such as good spatiotemporal continuity, high accuracy, low cost, and ease of use.

[0005] This invention provides a method for estimating winter wheat yield based on multi-source remote sensing data fusion, comprising the following steps:

[0006] 1) Select MODIS data and GF-1 WFV data, and preprocess the data;

[0007] 2) The enhanced spatiotemporal adaptive reflectivity fusion model ESTARFM is used to fuse MODIS and GF-1 WFV data, and the consistency of the fusion results is checked to generate time series data;

[0008] 3) The decision tree method was used, and the winter wheat planting area was extracted based on the temporal variation pattern of winter wheat NDVI;

[0009] 4) Extract growth monitoring indicators from multi-source remote sensing fusion data, obtain meteorological data of the study area using spatial interpolation methods, divide the study area into stages according to the key growth period of winter wheat, and calculate the average values ​​of the indicators for the key growth period.

[0010] 5) Based on the growth indicators and meteorological parameters of winter wheat during key growth stages, a multi-parameter, multi-time-window winter wheat yield estimation model was constructed.

[0011] 6) Verification of yield estimation accuracy.

[0012] In this invention, high spatiotemporal resolution remote sensing data is obtained through a spatiotemporal fusion method, and suitable indicators and time windows for winter wheat yield estimation are determined. This yields winter wheat yield estimation results with continuous temporal information and higher spatial resolution, providing decision support for guiding agricultural production.

[0013] According to an embodiment of the present invention, a method for estimating the yield of winter wheat based on multi-source remote sensing data fusion is provided. In step 1), the preprocessing includes atmospheric correction, geometric correction, mosaicking, and cropping.

[0014] According to an embodiment of the present invention, a method for estimating winter wheat yield based on multi-source remote sensing data fusion is provided. In step 2), the ESTARFM model is used to fuse MODIS and GF-1 WFV data; preferably, the MODIS is the MOD09Q1 product synthesized over 8 days; the time series data consists of fused data and GF-1 WFV data. In this invention, the above-mentioned fusion method can obtain remote sensing data sources with higher spatiotemporal resolution; it enables more accurate and continuous monitoring of winter wheat growth dynamics, thereby improving the accuracy of yield estimation.

[0015] According to an embodiment of the present invention, a method for estimating winter wheat yield based on multi-source remote sensing data fusion is provided. In step 3), the decision tree method is used to extract the winter wheat planting area, and the discrimination criterion is NDVI. March >0.3 and NDVI April NDVI June ;

[0016] Among them, NDVI March This represents the NDVI value in the remote sensing image for early March. April This represents the NDVI value on the remote sensing image in late April. June This represents the NDVI value in early June on the remote sensing image. In this invention, by employing the above classification method and the judgment criteria based on winter wheat phenological changes, the winter wheat planting area can be accurately extracted.

[0017] According to an embodiment of the present invention, a method for estimating the yield of winter wheat based on multi-source remote sensing data fusion is provided. In step 4), the growth monitoring index is the Wide Dynamic Range Vegetation Index (WDRVI), and the calculation formula is as follows:

[0018] (1)

[0019] in, The surface reflectance is in the near-infrared band. This represents the surface reflectance in the red light band.

[0020] According to an embodiment of the present invention, a method for estimating the yield of winter wheat based on multi-source remote sensing data fusion is provided. In step 4), the monitoring time window selects four key growth stages after winter wheat overwintering: the greening stage, the jointing stage, the grain-filling stage, and the milk stage.

[0021] According to an embodiment of the present invention, a method for estimating winter wheat yield based on multi-source remote sensing data fusion is provided. In step 5), the winter wheat yield estimation model adopts a nonlinear random forest (RF) model. Preferably, the WDRVI and meteorological parameters within the four monitoring time windows are simultaneously used as input variables of the model. In this invention, the above-mentioned RF model can effectively handle the nonlinear relationship between yield, growth status, and meteorological factors, and is less prone to overfitting, making the yield estimation model more accurate and comprehensive, thereby improving the yield estimation accuracy.

[0022] According to an embodiment of the present invention, a method for estimating winter wheat yield based on multi-source remote sensing data fusion is provided. In step 5), the winter wheat yield estimation model is as follows:

[0023] (2)

[0024]

[0025] in, To predict yield per unit area, This is a function for calculating yield per unit area. For the input parameters of the i-th reproductive period, is the wide dynamic vegetation index for the i-th growth stage. , Let be the average temperature, maximum temperature, and minimum temperature during the i-th reproductive period, respectively. Let be the precipitation during the i-th reproductive period. Let be the surface temperature during the i-th reproductive period.

[0026] In this invention, the RF model described above can effectively run large datasets and is insensitive to multicollinearity. In the construction of the above model, the growth of winter wheat and external meteorological factors are fully considered in the four key growth stages of winter wheat, which enables the model to better describe the role of various influencing factors in different growth stages of crops on yield formation, thereby making the yield estimation model more accurate and stable.

[0027] According to an embodiment of the present invention, a method for estimating winter wheat yield based on multi-source remote sensing data fusion is provided. In step 6), the yield estimation accuracy verification includes: applying a ten-fold cross-validation method to the established estimation model for cross-validation, and determining the yield based on the coefficient of determination R. 2 The root mean square error (RMSE) and mean absolute error (MAE) are used to assess the accuracy of winter wheat yield estimation models.

[0028] This invention also provides an application of the method for estimating winter wheat yield based on multi-source remote sensing data fusion in agricultural optimization of winter wheat remote sensing yield estimation technology.

[0029] The beneficial effects of this invention are at least as follows:

[0030] (1) Integrate multi-source remote sensing data of the monitoring area to improve the temporal and spatial resolution of the remote sensing data source.

[0031] (2) Taking into account the core factors affecting crop yield: crop growth status and meteorological factors, improve the accuracy of winter wheat yield estimation.

[0032] (3) Taking into full account the key growth periods of crops, construct a yield estimation model based on multiple monitoring time windows to further improve the accuracy of winter wheat yield estimation.

[0033] (4) All data used can be downloaded publicly and free of charge, which greatly reduces costs. The method is simple and easy to implement and can be used for industry needs such as agricultural production management, agricultural futures and agricultural insurance, supporting the widespread application of remote sensing yield estimation of crops. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0035] Figure 1 This is a flowchart illustrating an embodiment of the present invention;

[0036] Figure 2 This is a correlation diagram of the reflectance of GF-1 WFV and real GF-1 WFV in an embodiment of the present invention;

[0037] Figure 3 This is a comparison chart of the production estimation accuracy based on multiple time windows (a) and a single time window (b) in an embodiment of the present invention;

[0038] Figure 4This is a comparison chart of the yield estimation accuracy of fused data (a) and MODIS data (b) in an embodiment of the present invention;

[0039] Figure 5 The following is a spatial distribution map of winter wheat yield per unit area in the study area in 2017 based on different estimation models according to the embodiments of the present invention: (a) fused data and multiple time windows (b) MODIS data and multiple time windows (c) fused data and single time window (d) statistical yield per unit area. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0041] Where specific techniques or conditions are not specified in the embodiments, they shall be performed in accordance with the techniques or conditions described in the literature in this field, or in accordance with the product manual. Instruments and other equipment whose manufacturers are not specified are all conventional products that can be purchased through legitimate channels. Unless otherwise specified, the methods described are conventional methods, and the raw materials described are all obtainable from publicly available commercial sources. Where specific techniques or conditions are not specified in the embodiments, they shall be performed in accordance with the techniques or conditions described in the literature in this field, or in accordance with the product manual.

[0042] In this embodiment of the invention, the yield estimation test data of winter wheat per unit area in 45 counties (districts) of central Hebei Province, including Cangzhou, Hengshui and Shijiazhuang, in 2017 were used.

[0043] Example 1

[0044] Reference Figure 1 This embodiment provides a method for estimating the yield of winter wheat based on multi-source remote sensing data fusion, comprising the following steps:

[0045] (1) Preprocessing of remote sensing data and meteorological data.

[0046] The MOD09Q1 and GF-1 WFV data, which are 250m surface reflectance products synthesized over 8 days from the MODIS data products, covering the period from March to May 2017, were selected. The data were preprocessed, including atmospheric correction, geometric correction, mosaicking, and cropping.

[0047] Meteorological data were obtained from the China Meteorological Science Data Sharing Service Network (http: / / data.cma.cn / ), selecting the average temperature of various meteorological stations within and around the study area. ), highest temperature ( ), minimum temperature ( Precipitation (PRE) and surface temperature (GST) are considered as meteorological change factors.

[0048] (2) Fusion of multi-source remote sensing data.

[0049] The enhanced spatiotemporal adaptive reflectivity fusion model ESTARFM was used to fuse MODIS and GF-1 WFV data, and the consistency of the fusion results was verified. (See...) Figure 2 Time series data consisting of fused data and GF-1 WFV data is generated.

[0050] (3) Extraction from winter wheat planting areas.

[0051] The decision tree method was used to extract winter wheat planting areas, and the discrimination criterion was NDVI. March >0.3 and NDVI April NDVI June .

[0052] Among them, NDVI March This represents the NDVI value in the remote sensing image for early March. April This represents the NDVI value on the remote sensing image in late April. June This represents the NDVI value in the remote sensing image for early June.

[0053] (4) Determination of yield per unit area estimation indicators and time window.

[0054] WDRVI, a growth monitoring index, was extracted from multi-source remote sensing fusion data. The data was smoothed using the SG filtering method, and meteorological data for each district (county) in the study area were generated using the IDW interpolation method of a geographic information system. Based on the division of key growth stages for winter wheat in the study area, the mean values ​​of WDRVI and meteorological parameters were statistically analyzed for four key growth stages: greening stage, jointing stage, grain-filling stage, and milk stage.

[0055] (5) Construction of winter wheat yield estimation model.

[0056] Using growth indicators and meteorological parameters at four key growth stages of winter wheat as input variables and winter wheat yield as output, a multi-parameter, multi-time-window winter wheat yield estimation model was constructed using the Random Forest (RF) algorithm.

[0057]

[0058]

[0059] in, To predict yield per unit area, This is a function for calculating yield per unit area. For the input parameters of the i-th reproductive period, is the wide dynamic vegetation index for the i-th growth stage. , Let be the average temperature, maximum temperature, and minimum temperature during the i-th reproductive period, respectively. Let be the precipitation during the i-th reproductive period. Let be the surface temperature during the i-th reproductive period.

[0060] (6) Verification of the accuracy of yield estimation.

[0061] The established estimation model was cross-validated using the ten-fold cross-validation method, and the coefficient of determination R was used. 2 The root mean square error (RMSE) and mean absolute error (MAE) are used to assess the accuracy of winter wheat yield estimation models. For example... Figure 3 As shown, model R is built based on fused data and multiple time windows. 2 The value is 0.77, the RMSE is less than 600 kg / ha, and the R value of the model built based on fused data and a single time window is [missing information]. 2 With a mean value of only 0.6 and an RMSE higher than 700 kg / ha, it is evident that multiple time windows can more effectively represent the influence of growth vigor and environmental factors at different growth stages of winter wheat on yield formation. For example... Figure 4 As shown, under multiple time windows, the model R built based on MODIS data 2 The accuracy is 0.73, lower than the accuracy of models built based on fused data (R²). 2 =0.77), the increase in RMSE and MSE indicates that the data source has a significant impact on the estimation results of winter wheat yield in the study area, and the improvement of spatial resolution can significantly improve the accuracy of the model.

[0062] (7) Obtain the distribution map of winter wheat yield per unit area in the study area.

[0063] Based on the established process, a distribution map of winter wheat yield per unit area in the study region was finally generated. Here, the estimation results based on different remote sensing data sources (MODIS data) and different time window selection methods (single time window) are used as a comparison to evaluate the yield estimation effect with the actual yield statistics. (See...) Figure 5The winter wheat yield distribution in the study area is uneven, with the eastern region generally yielding higher yields than the central region, and the central region yielding higher yields than the western region. Predictions based on fine spatial resolution data obtained through multi-source remote sensing data fusion are closer to the actual statistical values, with estimation errors within 10% for most counties (districts), particularly showing stronger predictive ability for extremely high-yield areas (greater than 7000 kg / ha). In contrast, estimations based on coarse-resolution MODIS data or single-time-window data show larger estimation errors for each county (district), with more pronounced overestimation or underestimation, making accurate estimation difficult for extremely high-yield or extremely low-yield areas (less than 4000 kg / ha). The yield estimation model based on multi-source remote sensing data fusion and multiple time windows proposed in this invention has better estimation accuracy.

[0064] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for estimating winter wheat yield based on multi-source remote sensing data fusion, characterized in that, Includes the following steps: 1) Select MODIS data and GF-1 WFV data, and preprocess the data; 2) The enhanced spatiotemporal adaptive reflectivity fusion model ESTARFM is used to fuse MODIS and GF-1 WFV data, and the consistency of the fusion results is checked to generate time series data; 3) The decision tree method was used, and the winter wheat planting area was extracted based on the temporal variation pattern of winter wheat NDVI; 4) Extract growth monitoring indicators from multi-source remote sensing fusion data, obtain meteorological data of the study area using spatial interpolation methods, divide the study area into stages according to the key growth period of winter wheat, and calculate the average values ​​of the indicators for the key growth period. 5) Based on the growth indicators and meteorological parameters of winter wheat during key growth stages, a multi-parameter, multi-time-window winter wheat yield estimation model was constructed. 6) Verification of yield estimation accuracy; In step 4), the growth monitoring index is the Wide Dynamic Vegetation Index (WDRVI). In step 4), the monitoring time window selects four key growth stages after winter wheat overwintering: the greening stage, the jointing stage, the grain-filling stage, and the milk stage. In step 5), the winter wheat yield estimation model adopts a nonlinear random forest (RF) model; the WDRVI and meteorological parameters within the four monitoring time windows are simultaneously used as input variables of the model; in step 5), the winter wheat yield estimation model is specifically as follows: (2) in, To predict yield per unit area, This is a function for calculating yield per unit area. For the input parameters of the i-th reproductive period, is the wide dynamic vegetation index for the i-th growth stage. , Let be the average temperature, maximum temperature, and minimum temperature during the i-th reproductive period, respectively. Let be the precipitation during the i-th reproductive period. Let be the surface temperature during the i-th reproductive period.

2. The method for estimating winter wheat yield based on multi-source remote sensing data fusion according to claim 1, characterized in that, In step 1), the preprocessing includes atmospheric correction, geometric correction, mosaicking, and clipping.

3. The method for estimating winter wheat yield based on multi-source remote sensing data fusion according to claim 2, characterized in that, In step 2), the ESTARFM model is used to fuse MODIS and GF-1 WFV data.

4. The method for estimating winter wheat yield based on multi-source remote sensing data fusion according to claim 3, characterized in that, In step 2), MODIS uses the MOD09Q1 product synthesized over 8 days; the time series data consists of fused data and GF-1 WFV data.

5. The method for estimating winter wheat yield based on multi-source remote sensing data fusion according to claim 3, characterized in that, In step 3), the decision tree method is used to extract the winter wheat planting area, and the discrimination criterion is NDVI. March >0.3 and NDVI April NDVI June ; Among them, NDVI March This represents the NDVI value in the remote sensing image for early March. April This represents the NDVI value on the remote sensing image in late April. June This represents the NDVI value in the remote sensing image for early June.

6. The method for estimating winter wheat yield based on multi-source remote sensing data fusion according to claim 1, characterized in that, In step 4), the calculation formula for the growth monitoring indicators is as follows: (1) in, The surface reflectance is in the near-infrared band. This represents the surface reflectance in the red light band.

7. The method for estimating winter wheat yield based on multi-source remote sensing data fusion according to any one of claims 1-6, characterized in that, In step 6), the verification of the yield estimation accuracy includes: applying the ten-fold cross-validation method to the established estimation model for cross-validation, and based on the coefficient of determination R... 2 The root mean square error (RMSE) and mean absolute error (MAE) are used to assess the accuracy of winter wheat yield estimation models.

Images