Soybean yield remote sensing estimation method suitable for black soil area slope cropland

Abstract

The invention relates to the field of soybean yield estimation methods, and discloses a soybean yield remote sensing estimation method suitable for a black soil area slope cropland, which comprises the following steps of: 1, acquiring remote sensing images, DEM data and ridge direction data in a bare soil period and a whole growth period of a yield estimation area; 2, extracting a soil degradation index according to the bare soil period image, extracting a plurality of time sequence vegetation indexes according to the whole growth period image, extracting a plurality of topographic factors according to DEM data, and constructing a slope ridge angle indication ridge culture mode in combination with DEM and ridge direction data; 3, processing the extracted various vegetation indexes and environmental factors; 4, an RF modeling method is used, the double-time-phase NDVI is selected to be combined with DEM, SRA and SDI, a soybean yield estimation model is constructed, and the soybean yield of the black soil area slope cropland is estimated.The method adopts the double-time-phase NDVI to estimate the yield, a new SRA index is constructed to indicate the ridge culture mode, SDI is introduced to represent the soil degradation degree, and higher soybean yield estimation precision is achieved.

Description

Technical Field

[0001] This invention relates to the field of soybean yield estimation methods, specifically a remote sensing estimation method for soybean yield on sloping farmland in black soil regions. Background Technology

[0002] Soybeans are an important oilseed crop. With the improvement of living standards, the supply-demand gap for soybeans is gradually widening, making timely and accurate estimation of soybean yield crucial. Crop yield estimation models are mainly divided into empirical statistical models, semi-empirical models, and coupled models. Among them, machine learning algorithms combined with various spectral vegetation indices can improve the potential for crop yield estimation. Since crop yield is affected by factors such as climate, topography, soil, and agricultural management practices, incorporating auxiliary information can effectively improve the reliability of the model. Soil erosion is a major problem faced by sloping farmland in the black soil region, leading to reduced soybean yields. Traditional ridge cultivation is the main farming method on sloping farmland in the black soil region, and farming methods affect organic matter content, thus affecting crop yield. Currently, yield estimation models mostly rely on single vegetation indices or single growth period data; research on using time series data to screen for multiple growth periods for soybean yield estimation is still relatively lacking. Furthermore, existing research is still incomplete in combining environmental factors to improve the accuracy of remote sensing estimation of soybean yield on sloping farmland in the black soil region. In particular, the impact of soil erosion and ridge cultivation on yield has not been fully quantified. Existing yield estimation models rarely consider the influence of soil degradation and ridge cultivation on soybean yield on sloping farmland. Therefore, it is unclear whether the above factors can improve the accuracy of the yield estimation model. To this end, this application proposes a remote sensing estimation method for soybean yield on sloping farmland in the black soil region. Summary of the Invention

[0003] The purpose of this invention is to provide a remote sensing estimation method for soybean yield on sloping farmland in black soil regions, in order to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a remote sensing estimation method for soybean yield on sloping farmland in black soil regions, comprising the following steps:

[0005] Step 1: Obtain remote sensing images and DEM data of the bare soil period and the entire growth period in the estimated yield area, as well as ridge orientation data;

[0006] Step 2: Extract soil degradation index from bare soil images, extract various temporal vegetation indices from full-growth period images, extract various topographic factors from DEM data, and construct ridge planting method by combining DEM and ridge direction data;

[0007] Step 3: Process the extracted vegetation indices and environmental factors, including projection, resampling, and masking by plot;

[0008] Step 4: Using the RF modeling method, select two-phase NDVI combined with three factor types: DEM, SRA, and SDI, to construct a soybean yield estimation model and estimate the soybean yield on sloping farmland in the black soil region.

[0009] Preferably, step two involves extracting the soil degradation index from the bare soil period images, specifically as follows:

[0010]

[0011] Wherein, NIR stands for near-infrared band, R for red band, and B for blue band; in Sentinel-2, these correspond to bands 8, 4, and 2, respectively, and SMMI. 50pn and SI 50pn s1 and s2 are the normalized SMMI and SI; s1 and s2 are the corresponding weights, and both s1 and s2 are 0.5.

[0012] Preferably, in step two, based on the DEM data, various terrain factors are extracted using Saga 9.5.1, including aspect, slope, plan curvature, profile curvature, general curvature, concavity / convexity, terrain humidity index, terrain roughness index, and terrain location index.

[0013] Preferably, step two, which combines DEM and ridge direction data to construct the ridge angle indicator ridge planting method, specifically involves:

[0014] Angle = Aspect - Ridge - 90° (4)

[0015]

[0016] Where Angle is the angle between the slope aspect and the ridge aspect, Aspect is the slope aspect, and Ridge is the ridge aspect. The closer the extracted SRA is to 1, the closer the ridge aspect and slope aspect are to a perpendicular trend, i.e., a transverse ridge. The closer the absolute value is to 0, the closer the trend is to a parallel trend, i.e., a downhill ridge. A negative value is taken when the slope is on the shaded side, and a positive value is taken when the slope is on the sunny side. A value of 0 represents a plane.

[0017] Preferably, the dual-phase NDVI is NDVIa and NDVIb, which represent the NDVI of the two growth periods with the highest correlation to the measured yield, specifically June 24 and July 4.

[0018] Preferably, the formula for estimating the output in step four is:

[0019]

[0020] Where y is the predicted soybean yield, n is the number of decision trees, and f t(x) is the prediction function of the t-th decision tree, and x is the input feature vector, specifically NDVIa, NDVIb, DEM, SRA, and SDI.

[0021] Compared with the prior art, the beneficial effects of this invention are as follows:

[0022] This invention uses dual-phase NDVI for yield estimation and constructs a new SRA index to indicate ridge cultivation. It also introduces SDI to represent the degree of soil degradation, thereby improving the accuracy of the estimation model and achieving high yield estimation accuracy for soybeans. Attached Figure Description

[0023] Figure 1 This is a flowchart of the method of the present invention;

[0024] Figure 2 This is a schematic diagram of the SRA extracted in this invention;

[0025] Figure 3 This is a comparison chart of the production estimation accuracy of the present invention;

[0026] Figure 4 Scatter plot for verifying the accuracy of the optimal production estimation model of this invention;

[0027] Figure 5 This is a flowchart illustrating the multi-model construction process of an embodiment of the present invention. Detailed Implementation

[0028] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0029] Example

[0030] Please see Figures 1-4 The diagram illustrates a remote sensing method for estimating soybean yield on sloping farmland in black soil regions, comprising the following steps:

[0031] Step 1: Obtain remote sensing images and DEM data of the bare soil period and the entire growth period in the estimated yield area, as well as ridge orientation data;

[0032] Step 2: Extract soil degradation index from bare soil images, extract various temporal vegetation indices from full-growth period images, extract various topographic factors from DEM data, and construct ridge planting method by combining DEM and ridge direction data;

[0033] Step 3: Process the extracted vegetation indices and environmental factors, including projection, resampling, and masking by plot;

[0034] Step 4: Using the RF modeling method, select two-phase NDVI combined with three factor types: DEM, SRA, and SDI, to construct a soybean yield estimation model and estimate the soybean yield on sloping farmland in the black soil region.

[0035] In this embodiment, the soil degradation index is extracted based on the bare soil period image, specifically as follows:

[0036]

[0037]

[0038] SDI = s1 × SMMI 50pn +s2×SI 50pn (3)

[0039] Wherein, NIR stands for near-infrared band, R for red band, and B for blue band; in Sentinel-2, these correspond to bands 8, 4, and 2, respectively, and SMMI. 50pn and SI 50pn s1 and s2 are the normalized SMMI and SI; s1 and s2 are the corresponding weights, and both s1 and s2 are 0.5.

[0040] Furthermore, based on the DEM data, various topographic factors were extracted using Saga 9.5.1, including aspect, slope, plan curvature, profile curvature, general curvature, concavity / convexity, topographic moisture index, topographic roughness index, and topographic location index. Combining the DEM and ridge aspect data, a slope-ridge angle indicator for ridge planting was constructed as follows:

[0041] Angle = Aspect - Ridge - 90° (4)

[0042]

[0043] Where Angle is the angle between the slope direction and the ridge direction, and Ridge is the ridge direction. The closer the extracted SRA is to 1, the closer the ridge direction and slope direction are to the vertical trend, i.e., transverse ridge; the closer the absolute value is to 0, the closer the trend is to the parallel trend, i.e. downhill ridge. It takes a negative value when it is on the shady slope and a positive value when it is on the sunny slope. A value of 0 represents a plane.

[0044] Multiple temporal vegetation indices were extracted from images throughout the entire growth period, as shown in Table 1:

[0045] Table 1

[0046]

[0047]

[0048] NIR stands for near-infrared band, R for red band, G for green band, and B for blue band; in Sentinel-2, these correspond to bands 8, 4, 3, and 2, respectively.

[0049] Furthermore, to verify the model accuracy of the estimation method provided by this invention, three methods—MLR, SVM, and XGBoost—were selected and modeled with different factor combinations to verify the accuracy of the soybean estimation results of the model provided by this invention. For specific steps, please refer to [link / reference needed]. Figure 5 .

[0050] The Heshan Farm, managed by the Jiufeng Branch of Beidahuang State Farms Group, was selected as the yield estimation area for the experiment. The vector range of Heshan Farm was uploaded to the GEE platform, and Sentinel-2 remote sensing images of the bare soil period and the entire growth period in 2024 were selected. The time of the growth period images is shown in Table 2. The spatial resolution is 10m, which is suitable for crop yield estimation within the farm area. The required images were exported to Google Drive and saved to obtain DEM data and ridge data.

[0051] Table 2

[0052]

[0053] Use ArcGIS 10.8 image analysis and raster calculator tools to extract various vegetation indices and environmental factors.

[0054] Soil erosion is a major factor leading to soil degradation and reduced crop productivity in the Northeast Black Soil Region. The Soil Degradation Index (SDI) can quantitatively reflect the degree of soil quality degradation. After determining the weights of each index using fuzzy mathematics, it is widely used in soil assessment. Using bare soil images from Heshan Farm on May 2nd, and combining the Soil Moisture Monitoring Index (SMMI) and Soil Salinity Index (SI), a Soil Degradation Index (SDI) was constructed.

[0055]

[0056] SDI = s1 × SMMI 50pn +s2×SI 50pn (3)

[0057] Wherein, NIR stands for near-infrared band, R for red band, and B for blue band; in Sentinel-2, these correspond to bands 8, 4, and 2, respectively, and SMMI. 50pn and SI 50pn s1 and s2 are the normalized SMMI and SI; s1 and s2 are the corresponding weights, where s1 and s2 are both set to 0.5.

[0058] Soybean growth has a clear periodicity, and vegetation indices can reflect the growth and development of plants. During the crop growth cycle, vegetation indices will change significantly. In order to improve the accuracy of yield estimation and advance the yield estimation time, it is crucial to select appropriate remote sensing time phases. Therefore, time-series vegetation indices are extracted from images of the entire growth period.

[0059] Topographic features of farmland are the most important factor influencing crop yield variability. Traditional ridge cultivation is the main farming method on sloping farmland in the Northeast Black Soil Region. Cultivation patterns affect soil organic matter content, thus impacting crop yield. Therefore, multiple topographic factors were extracted based on the DEM (Digital Elevation Model), and combined with ridge aspect data to construct a slope-ridge angle (SRA) indicator to show ridge cultivation patterns.

[0060] Angle = Aspect - Ridge - 90° (4)

[0061]

[0062] Where Angle represents the angle between the slope aspect and the ridge direction, and Ridge represents the ridge direction. The closer the extracted SRA value is to 1, the closer the ridge direction and slope aspect are to being perpendicular, i.e., a transverse ridge; the closer the absolute value is to 0, the closer the ridge direction is to being parallel, i.e., a downslope ridge. Slope aspect analysis is then overlaid, with negative values ​​for shady slopes and positive values ​​for sunny slopes. A value of 0 represents a planar surface. The extraction results are as follows: Figure 2 As shown.

[0063] In ArcGIS 10.8, the extracted indices are processed through steps such as projection, resampling, and masking by plot to ensure that the data have the same projection, spatial resolution, and range. Based on the latitude and longitude of the estimated production area, the projection is selected as "WGS_1984_UTM_Zone_51N", and the data is resampled to 10m to match the spatial resolution of Sentinel-2. Then, the data is masked according to the soybean planting plots.

[0064] Based on 68 measured soybean yield data collected from September 21 to 24, 2024, Pearson correlation analysis was used to screen for vegetation indices and environmental covariates that were significantly correlated with the measured yield. The screening results are shown in Table 3. A total of five factors were involved in the model construction. Using ArcGIS 10.8, the five selected yield estimation factors were extracted to vector points using the multi-value extraction to point tool and exported as a CSV table.

[0065] Table 3

[0066]

[0067] Among them, NDVIa and NDVIb represent the NDVI of the two growth periods with the highest correlation to the measured yield, namely June 24 and July 4, respectively; DEM is the elevation of the estimated yield area; SRA is the extracted slope angle index; and SDI is the extracted soil degradation index.

[0068] This invention divides the training and validation sets into an 8:2 ratio, with 56 sampling points in the training set and 12 sampling points in the validation set. Four methods—MLR, SVM, XGBoost, and RF—are selected and modeled using different factor combinations. The factor of determination (R) is used. 2 The accuracy of the production estimation model is evaluated using the root mean square error (RMSE), as shown in the following formula:

[0069]

[0070]

[0071] Where n is the number of samples, and Yi is the measured value of the i-th sampling point. Ei represents the average value of the actual yield measurement, and Ei is the predicted value of the i-th sampling point. To predict the average yield, typically R 2 The range is 0-1. The closer the value is to 1, the closer the predicted value is to the actual measured value. RMSE is used to test the degree of difference between the actual measured value and the predicted value. The smaller the value, the smaller the difference between the two. The accuracy of different models is shown in Table 4.

[0072] Table 4

[0073]

[0074]

[0075]

[0076]

[0077]

[0078] The results in Table 4 clearly show that the estimation model constructed using the RF method, which combines dual-temporal NDVI with three factors—DEM, SRA, and SDI—achieves the highest accuracy. 2 The value was 0.87, and the RMSE was 0.56 t / hm. 2 Accuracy verification scatter plot as follows Figure 3 As shown, this invention uses this method to estimate soybean yield on sloping farmland in black soil regions. The yield estimation formula is as follows:

[0079]

[0080] Where y is the predicted soybean yield, n is the number of decision trees, and f t (x) is the prediction function of the t-th decision tree, and x is the input feature vector, specifically NDVIa, NDVIb, DEM, SRA, and SDI.

[0081] The present invention uses RF for modeling because the relationship between crop yield and vegetation index and environmental variables is indirect or nonlinear. Therefore, machine learning models have certain advantages in yield estimation, while RF models are good at handling data with complex nonlinear relationships and have certain effectiveness and high accuracy in rigorous crop yield estimation.

[0082] Compared with existing yield estimation models, this invention uses dual-temporal NDVI for yield estimation and constructs a new SRA index to indicate ridge cultivation. It also introduces SDI to represent the degree of soil degradation. The study found that the model accuracy improved after introducing the soil degradation index and slope-ridge angle. After introducing SRA, R... 2 The maximum increase was 0.49, and the maximum decrease in RMSE was 0.5 t / hm. 2 After introducing SDI, R 2 The maximum increase was 0.48, and the maximum decrease in RMSE was 0.48 t / hm. 2 This demonstrates that incorporating SRA and SDI into the soybean yield estimation model for sloping farmland in the black soil region is effective and has high estimation accuracy.

[0083] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0084] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A remote sensing method for estimating soybean yield on sloping farmland in black soil regions, characterized in that, The steps include: Step 1: Obtain remote sensing images and DEM data of the bare soil period and the entire growth period in the estimated yield area, as well as ridge orientation data; Step 2: Extract soil degradation index from bare soil images, extract various temporal vegetation indices from full-growth period images, extract various topographic factors from DEM data, and construct ridge planting method by combining DEM and ridge direction data; Step 3: Process the extracted vegetation indices and environmental factors, including projection, resampling, and masking by plot; Step 4: Using the RF modeling method, select two-phase NDVI combined with three factor types: DEM, SRA, and SDI, to construct a soybean yield estimation model and estimate the soybean yield on sloping farmland in the black soil region.

2. The remote sensing estimation method for soybean yield on sloping farmland in black soil regions according to claim 1, characterized in that: Step two, extracting the soil degradation index based on the bare soil period images, specifically involves: SDI=s1×SMMI s0pn +s2×SI s0pn (3) Wherein, NIR stands for near-infrared band, R for red band, and B for blue band; in Sentinel-2, these correspond to bands 8, 4, and 2, respectively, and SMMI. 50pn and SI 50pn s1 and s2 are the normalized SMMI and SI; s1 and s2 are the corresponding weights, and both s1 and s2 are 0.

5.

3. The remote sensing estimation method for soybean yield on sloping farmland in black soil regions according to claim 2, characterized in that: Step two involves using Saga 9.5.1 to extract various topographic factors based on the DEM data, including aspect, slope, plan curvature, profile curvature, general curvature, concavity / convexity, topographic humidity index, topographic roughness index, and topographic location index.

4. The remote sensing estimation method for soybean yield on sloping farmland in black soil regions according to claim 3, characterized in that: Step two, which combines DEM and ridge direction data to construct the ridge angle indicator ridge cultivation method, specifically involves: Angle = Aspect - Ridge - 90° (4) Where Angle is the angle between the slope aspect and the ridge aspect, Aspect is the slope aspect, and Ridge is the ridge aspect. The closer the extracted SRA is to 1, the closer the ridge aspect and slope aspect are to a perpendicular trend, i.e., a transverse ridge. The closer the absolute value is to 0, the closer the trend is to a parallel trend, i.e., a downhill ridge. A negative value is taken when the slope is on the shaded side, and a positive value is taken when the slope is on the sunny side. A value of 0 represents a plane.

5. The remote sensing estimation method for soybean yield on sloping farmland in black soil regions according to claim 4, characterized in that: The dual-phase NDVI refers to NDVIa and NDVIb, which represent the NDVI of the two growth periods with the highest correlation to measured yield, specifically June 24 and July 4.

6. The remote sensing estimation method for soybean yield on sloping farmland in black soil regions according to claim 5, characterized in that: The formula for estimating the output in step four is as follows: Where y is the predicted soybean yield, n is the number of decision trees, and f t (x) is the prediction function of the t-th decision tree, and x is the input feature vector, specifically NDVIa, NDVIb, DEM, SRA, and SDI.

Images