Method for estimating yield of mature rice by coupling visible light index and growth parameters

Abstract

The present invention discloses a method for estimating the yield of mature rice by coupling visible light index and growth parameters, belonging to the technical field of smart agriculture, which includes collecting unmanned aerial vehicle (UAV) remote sensing data and ground crop data in the test area and performing data processing; estimating the biomass of mature rice based on a single characteristic variable; establishing an estimation model for the biomass of mature rice by integrating the visible light index and growth parameters; and estimating the biomass of mature rice in the test area. By adopting the above method for estimating the yield of mature rice by coupling the visible light index and growth parameters, integrating the visible light index of the rice canopy obtained by UAV remote sensing technology and the rice growth parameters, and combining with the support vector machine regression method, the biomass of rice can be effectively estimated during the mature period of rice, providing a fast and efficient method for estimating the biomass of rice in a relatively large area, and providing a method reference for work such as the growth monitoring and efficient yield estimation of mature rice.

Description

Technical Field

[0001] The present invention relates to the technical field of smart agriculture, and particularly to a method for estimating the yield of mature rice by coupling visible light indices and growth parameters. Background Art

[0002] Accurately estimating the biomass of mature rice is of great significance for ensuring food security and realizing sustainable agricultural development. Traditional measurement of rice biomass often requires destructive sampling, which not only has low efficiency but also is difficult to meet the monitoring needs of large-scale areas. In recent years, due to its high flexibility, high spatio-temporal resolution, and simple operation, unmanned aerial vehicle (UAV) technology has been widely used in crop biomass estimation. The mainstream method is to analyze the relationship between spectral indices in UAV remote sensing images and rice biomass, and use various methods such as linear regression and machine learning to construct rice biomass estimation models, achieving good estimation results. However, there are significant differences in the physiological characteristics and spectral responses of rice at different growth stages. When rice is in the early growth stage, its canopy leaves are sparse and the overlap rate is low, being sensitive to spectral indices. At this time, there is a significant correlation between biomass and spectral indices. However, in the later growth stage, the spectral indices show a saturation phenomenon, resulting in a decrease in the correlation with biomass. Especially in the mature stage, the yellowing and withering of leaves increase significantly, leading to a significant deterioration in the estimation effect of mature rice biomass. The accuracy of estimating mature rice biomass using only the spectral information of images is relatively low.

[0003] To improve the estimation accuracy of crop growth parameters, by introducing texture features and structural features such as crop plant height on the basis of spectral information, it is found that the fusion of multiple features can reduce the influence of spectral index saturation, thereby improving the estimation accuracy of crop biomass. Therefore, through relevant research on crop growth monitoring, the method of multi-feature fusion has achieved good prediction effects in crop yield prediction or biomass estimation in the early growth stage of crops, but the improvement in the estimation effect of crop biomass in the later growth stage is still small. Summary of the Invention

[0004] The object of the present invention is to provide a method for estimating the yield of mature rice by coupling visible light indices and growth parameters. Taking mature rice as the research object, starting from the definition of above-ground crop biomass, a biomass estimation model for mature rice is constructed by combining the visible light indices of rice canopy images with growth parameters such as rice plant height and water content. Among them, the visible light indices mainly provide spectral information on the growth state of rice, and the growth parameters mainly provide direct information on the growth status of rice. Finally, the model is applied to estimate the biomass of mature rice in the test area, and the biomass distribution in the test area is analyzed to further verify the reliability of the model, with a view to providing a new reference for the accurate estimation of mature rice biomass and at the same time providing a more comprehensive scientific basis for rice growth monitoring and yield assessment.

[0005] To achieve the above object, the present invention provides a method for estimating the yield of mature rice by coupling visible light index and growth parameters, including the following steps:

[0006] S1. Collect unmanned aerial vehicle (UAV) remote sensing data and ground crop data in the test area and perform data processing. The UAV remote sensing data includes visible light data and multispectral data; the ground crop data includes rice plant height, fresh weight of rice, and dry weight of rice; the rice plant height is the distance from the bottom to the top of the canopy of rice in the natural growth state, the fresh weight of rice is the weight measured immediately after sampling and bagging, and the dry weight of rice is the weight after drying to a constant weight; the data processing includes the following steps:

[0007] S11. Use Pix4D Mapper software and Yusense Map software to splice and correct the UAV remote sensing data, crop the UAV images according to the position of the sampling area, and take the average reflectance of each sampling area as the spectral reflectance of the sampling point to obtain spectral reflectance data of different bands;

[0008] S12. Construct vegetation index; select visible light index as an indicator for constructing the biomass model of mature rice;

[0009] S13. Calculate the biomass of rice per square meter of land according to the dry weight of rice and the planting density of rice in the sampling area;

[0010] S14. Calculate the water content of rice plants by the ratio of the difference between the fresh weight and dry weight of rice to the fresh weight of rice;

[0011] S2. Estimate the biomass of mature rice based on a single characteristic variable, specifically:

[0012] S21. Use Pearson correlation analysis to evaluate the linear correlation between the visible light index, rice plant height, water content of rice plants, and rice biomass;

[0013] S22. Take the visible light index, rice plant height, and water content of rice plants as independent variables respectively, and use the support vector regression method to construct a rice biomass estimation model, and use the determination coefficient of the training set R 2 _train and the determination coefficient of the test set R 2 _test to represent the accuracy of the training set and the accuracy of the test set respectively;

[0014] S23. Screen the characteristic variables for fusing the visible light index and growth parameters to establish a biomass estimation model of mature rice from the visible light index, rice plant height, and water content of rice plants according to the accuracy of the training set and the accuracy of the test set;

[0015] S3. Establish a biomass estimation model for mature rice by integrating the visible light index and growth parameters, including the following steps:

[0016] S31. According to the coefficient of determination of the test set of the rice biomass estimation model of the visible light index in S2 R 2 _test , classify the visible light index for index optimization, and construct a biomass estimation model for mature rice with different categories of visible light indices as input factors;

[0017] S32. Integrate the rice plant height and the water content of the rice plant to construct a biomass estimation model for mature rice and compare the fitting effects;

[0018] S4. Estimate the biomass of mature rice in the test area, including the following steps:

[0019] S41. Obtain the rice plant height at maturity;

[0020] S42. Obtain the water content of mature rice;

[0021] S43. According to the rice plant height at maturity obtained in S41 and the water content of mature rice obtained in S42, and combining the biomass estimation model for mature rice constructed in S3, estimate the biomass of mature rice in the test area to obtain the biomass distribution of mature rice in the test area.

[0022] Preferably, in S32, the specific process is as follows: the visible light index is successively integrated with the rice plant height, the water content of the rice plant, the rice plant height and the water content of the rice plant, and the three fitting effects are compared.

[0023] Preferably, in S41, the specific process of obtaining the rice plant height at maturity is as follows: process the UAV remote sensing data when the paddy field is bare before rice planting and when the rice is mature to obtain the digital surface model DSM at the corresponding time, which represents the ground height of the paddy field and the overall height of the mature rice respectively; obtain the rice plant height data at maturity through the canopy height model CHM; perform fitting analysis on the plant height data obtained through UAV remote sensing data and the measured plant height.

[0024] Preferably, in S42, the specific process of obtaining the water content of mature rice is as follows: use the UAV to obtain multispectral images and extract the multispectral indices related to the measured water content, combine the reflectance of the red light band, the reflectance of the red edge band and the reflectance of the near-infrared band, and use the random forest regression method to construct a relationship model between the multispectral indices and the water content to estimate the water content of mature rice.

[0025] Therefore, the present invention adopts the above-mentioned method for estimating the yield of mature rice by coupling the visible light index and growth parameters, integrating the visible light index of the rice canopy obtained by the unmanned aerial vehicle remote sensing technology and the rice growth parameters, and combining the support vector machine regression method can effectively estimate the rice biomass at the mature stage of rice, providing a fast and efficient method for estimating the rice biomass in a relatively large area, and can provide a method reference for the work of monitoring the growth of mature rice and efficiently estimating the yield.

[0026] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Brief Description of the Drawings

[0027] Figure 1 is the technical roadmap of the embodiment of the method for estimating the yield of mature rice by coupling the visible light index and growth parameters of the present invention;

[0028] Figure 2 is the correlation analysis of a single characteristic variable and biomass in the embodiment of the method for estimating the yield of mature rice by coupling the visible light index and growth parameters of the present invention;

[0029] Figure 3 is the accuracy of biomass estimation based on a single characteristic variable in the embodiment of the method for estimating the yield of mature rice by coupling the visible light index and growth parameters of the present invention;

[0030] Figure 4 is the accuracy of biomass estimation of the visible light index combined with plant height and water content in the embodiment of the method for estimating the yield of mature rice by coupling the visible light index and growth parameters of the present invention, (a) only using the visible light index; (b) adding the rice plant height; (c) adding the water content of the rice plant; (d) adding the rice plant height and the water content of the rice plant;

[0031] Figure 5 is the distribution map of the plant height of mature rice in the test area in the embodiment of the method for estimating the yield of mature rice by coupling the visible light index and growth parameters of the present invention;

[0032] Figure 6 is the distribution map of the water content of mature rice in the test area in the embodiment of the method for estimating the yield of mature rice by coupling the visible light index and growth parameters of the present invention;

[0033] Figure 7 is the distribution of the biomass of mature rice in the test area in the embodiment of the method for estimating the yield of mature rice by coupling the visible light index and growth parameters of the present invention;

[0034] Figure 8 is the fitting analysis result of the estimated biomass and the measured yield in each area of the test area in the embodiment of the method for estimating the yield of mature rice by coupling the visible light index and growth parameters of the present invention. Detailed Embodiments

[0035] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0036] Unless otherwise defined, the technical terms or scientific terms used in the present invention should have the ordinary meanings understood by those of ordinary skill in the field to which the present invention pertains. The "first", "second" and similar terms used in the present invention do not denote any order, quantity or importance, but are only used to distinguish different components. The terms such as "comprising" or "including" mean that the elements or objects appearing before this word cover the elements or objects listed after this word and their equivalents, without excluding other elements or objects. The terms such as "connected" or "coupled" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The terms such as "upper", "lower", "left", "right" are only used to represent relative positional relationships, and when the absolute position of the object being described changes, the relative positional relationship may also change accordingly.

[0037] Example 1

[0038] The present invention provides a method for estimating the yield of mature rice by coupling visible light index and growth parameters, and its technical route is as Figure 1 shown, including the following steps:

[0039] S1. Collect unmanned aerial vehicle (UAV) remote sensing data and ground crop data in the test area and perform data processing; the UAV remote sensing data includes visible light data and multispectral data; the ground crop data includes rice plant height, fresh weight of rice, and dry weight of rice; in this embodiment, when the UAV collects ground data, the specific settings are as follows: Use a DJI Phantom 4 quadcopter UAV to obtain the original visible light data in the test area, with the flight altitude set at 100 m, the flight speed set at 7 m / s, the forward overlap rate at 80%, the side overlap rate at 70%, and the flight angles at 0° and -45°. Use a DJI Matrice 300 RTK quadcopter UAV, equipped with an MS600 Pro six-channel multispectral sensor to obtain the original multispectral data in the test area, with the flight altitude set at 80 m, the forward overlap rate and the side overlap rate set at 80% and 70% respectively, and the flight angle at 0°.

[0040] The rice plant height refers to the distance from the bottom to the top of the canopy of rice in the natural growth state (unit: cm); the fresh weight of rice is the weight obtained by weighing the rice samples immediately after sampling and bagging (unit: kg); the dry weight of rice is the weight after the rice is dried to a constant weight (unit: kg). In this example, after the rice samples are packaged, they are placed in an oven, blanched at 105 °C for 30 minutes, then dried at 75 °C for about 48 hours to a constant weight, and finally weighed to obtain the dry weight of rice.

[0041] The data processing specifically includes the following steps:

[0042] S11. Use Pix4D Mapper software and Yusense Map software to splice and correct the UAV remote sensing data, crop the UAV images according to the position of the sampling area, and take the average reflectivity of each sampling area as the spectral reflectivity of the sampling point to obtain spectral reflectivity data of different bands.

[0043] S12. Vegetation index construction. The vegetation index is an important parameter to characterize crop growth information through the combination of spectra. According to the estimation requirements of rice biomass, 11 visible light indices are selected in this embodiment as the indicators for constructing the rice mature biomass model. The visible light indices are specifically:

[0044] Red light normalized value (r): R / (R + G + B);

[0045] Green light normalized value (g): G / (R + G + B);

[0046] Blue light normalized value (b): B / (R + G + B);

[0047] Visible light atmospheric impedance vegetation index (VARI): (G - R) / (R + G + B);

[0048] Normalized green - red difference index (GRVI): (G - R) / (G + R);

[0049] Modified green - red vegetation index (MGRVI): (G 2 -R 2 ) / (G 2 +R 2 );

[0050] Blue - green ratio index (BGI): B / G;

[0051] Blue - red ratio index (BRI): B / R;

[0052] Green - red ratio index (GRI): G / R;

[0053] Excess green vegetation index (ExG): ;

[0054] Red - green - blue vegetation index (RGBVI): ;

[0055] Among them, R, G, and B are the three basic color channels in the visible spectrum, and the wavelength ranges are: the red light band R is 620 - 750 nm, the green light band G is 500 - 560 nm, and the blue light band B is 430 - 470 nm.

[0056] S13. Calculate the rice biomass per square meter of land (unit: kg / m 2 ) according to the dry weight of rice and the rice planting density in the sampling area.

[0057] S14. Calculate the water content of rice plants by the ratio of the difference between the fresh weight and the dry weight of rice to its fresh weight.

[0058] S2. Estimate the rice biomass at the maturity stage based on a single characteristic variable, including the following steps:

[0059] S21. Use Pearson correlation analysis to evaluate the linear correlation between the visible light index, rice plant height (Height), rice plant water content (Moisture Content, MC), and rice biomass (Aboveground Biomass, AGB). In this embodiment, the results of the correlation analysis are as Figure 2 . At the maturity stage, the correlation between the visible light index of the rice canopy and the rice biomass is generally low. Among them, the higher correlations are g and RGBVI, and the correlation coefficients are -0.36 and -0.34 respectively; the correlation between some indices and the rice biomass is about 0.1, showing a significant lack of correlation; the correlation coefficients between the rice plant height, rice plant water content, and rice biomass are 0.42 and -0.38 respectively, both higher than the correlation degree between the visible light index and the rice biomass.

[0060] S22. Respectively take the visible light index, rice plant height, and rice plant water content as independent variables, and use the support vector regression method to construct a rice biomass estimation model. Use the determination coefficient R 2 _train of the training set and the determination coefficient R 2 _test to represent the accuracy of the training set and the test set respectively. Support vector regression can well represent the non-linear relationship between variables. Its core idea is to find an optimal hyperplane so that the deviation from the actual value is within an acceptable range. This method can effectively handle high-dimensional data and has excellent generalization ability, but it requires experience and experiments to determine appropriate hyperparameters and kernel functions.

[0061] According to the distribution of the experimental data collected, the support vector regression method is used to estimate the rice biomass. To ensure the comparability between different experiments, the same data partitioning and the same model parameters are used in all experiments in this embodiment: the penalty coefficient C = 6.5, the radial basis kernel function (RBF) is used as the kernel function, and the tolerance deviation value epsilon = 0.5. The visible light index of the rice canopy at the mature stage, the rice plant height, the water content of the rice plant, and the rice biomass data are used as the experimental data set. The experimental data set is divided into a training set and a test set at a ratio of 8.5:1.5, and a rice biomass estimation model at the mature stage is established. The results are as Figure 3 shown:

[0062] (1) The accuracy of the training set for estimating rice biomass using only a single visible light index is generally very low. The coefficient of determination R 2 _train is between 0 and 0.3, and the fitting effect on the training data is poor. Among them, the models constructed based on the green band normalized value g, the red-green-blue vegetation index RGBVI, and the excess green vegetation index ExG R 2 _train are relatively high, which are 0.25, 0.27, and 0.26 respectively.

[0063] (2) The models constructed based on the visible light atmospheric impedance vegetation index VARI, the blue-green ratio index BGI, and the blue-red ratio index BRI R 2 _test are less than 0, indicating that the effects of these three visible light indices on estimating the rice biomass at the mature stage are extremely poor and will be excluded in subsequent experiments; the R 2 _test of other characteristic variables are all greater than 0. Among them, the model based on the red-green-blue vegetation index RGBVI performs the best, R 2 _test and RMSE are 0.55 and 0.45 kg / m 2 .

[0064] (3) Among the growth parameters, the accuracy of the training set for estimating biomass using the rice plant height R 2 _train and the accuracy of the test set for estimating biomass using the water content of the rice plant R 2 _test are both relatively high, and the coefficients of determination are 0.25 and 0.46 respectively.

[0065] S23. Screen the characteristic variables for fusing the visible light index and growth parameters to establish an estimation model for the biomass of mature rice from the visible light index, rice plant height, and water content of rice plants according to the accuracy of the training set and the accuracy of the test set.

[0066] As can be seen from the results of S22, using only a single characteristic variable such as the visible light index or growth parameters may not be able to capture the changes in the biomass of mature rice, and the accuracy of the estimation model for the biomass of mature rice constructed thereby is generally low. Therefore, in this embodiment, 8 visible light indices and 2 growth parameters are selected as the input factors of the model, and multiple types of characteristic variables are fused for modeling analysis.

[0067] S3. Fuse the visible light index and growth parameters to establish an estimation model for the biomass of mature rice, including the following steps:

[0068] S31. According to the coefficient of determination of the test set of the estimation model for the biomass of rice using the visible light index in S2 R 2 _test , classify the visible light indices for index optimization, and construct an estimation model for the biomass of mature rice with different categories of visible light indices as the input factors.

[0069] In this embodiment, the 8 visible light indices selected in S23 are classified into 5 categories for index optimization with the coefficient of determination of the test set of the estimation model for the biomass of rice with a single visible light index R 2 _test as the standard, and the specific classification criteria are shown in Table 1.

[0070] Table 1 Classification of visible light indices

[0071]

[0072] Among them, R 2 _test Based on Figure 3 results.

[0073] S32. Fuse the rice plant height and water content of rice plants to construct an estimation model for the biomass of mature rice and compare the fitting effects. The specific process is as follows: The visible light index is successively fused with the rice plant height, water content of rice plants, and rice plant height and water content of rice plants, and the fitting effects of the three times are compared.

[0074] In this embodiment, the accuracy of the estimation model for the biomass of rice with different categories of visible light indices combined with growth parameters is as Figure 4 shown:

[0075] (1) When constructing an estimation model for the biomass of mature rice with only the visible light index as the model input factor, asFigure 4 As shown in (a) of Figure 4 , the coefficient of determination of the test sets for Type 1 and Type 2 is less than 0, the coefficient of determination of the test set for Type 3 is less than 0.1, and the coefficient of determination of the test sets for Type 4 and Type 5 is around 0.5, indicating that the visible light index preferably has a greater impact on the model accuracy.

[0076] (2) When adding rice plant height on the basis of the visible light index, as Figure 4 shown in (b) of Figure 4 , the coefficient of determination of the test set of the rice biomass estimation model decreases instead, and there are still cases less than 0, and the root mean square error is still large; when only adding the water content of rice plants, as Figure 4 shown in (c) of Figure 4 , the coefficients of determination of the model test sets are significantly improved. The coefficient of determination and the root mean square error of the test set of the model with the highest accuracy are 0.58 and 0.44 kg / m 2 respectively; when adding rice plant height and the water content of rice plants, as Figure 4 shown in (d) of Figure 4 , the fitting effects of the constructed models on the training set are all improved. Among them, the estimation accuracy of the mature rice biomass estimation model of the visible light index category Type 4 reaches the highest, and the coefficient of determination and the root mean square error of the test set both reach the optimal values, which are 0.78 and 0.32 kg / m 2 respectively.

[0077] (3) Optimizing the visible light index and fusing growth parameters are the key links to improve the estimation accuracy of mature rice biomass. Therefore, when the normalized value g of the green band, the red-green-blue vegetation index RGBVI, rice plant height, and the water content of rice plants are used as the input factors of the model, the estimation model of mature rice biomass reaches the optimal estimation effect, indicating that the fusion of growth parameters can make up for the deficiencies of the visible light index and improve the fitting effect of the biomass estimation model on the data.

[0078] S4. Estimate the biomass of mature rice in the test area. To explore the application of the above model in estimating rice biomass in a larger area, by obtaining two visible light indices of the mature rice canopy in the test area, namely the normalized value g of the green light and the red-green-blue vegetation index RGBVI, and two growth parameters, rice plant height and the water content of rice plants, analyze the spatial distribution of the biomass of mature rice in the test area, which specifically includes the following steps:

[0079] S41. Obtain the plant height of rice at the mature stage; process the UAV remote sensing data when the paddy field is bare before rice planting and when the rice is at the mature stage to obtain the digital surface model DSM at the corresponding periods, which represent the ground height of the paddy field and the overall height of the rice at the mature stage respectively; obtain the plant height data of the rice at the mature stage through the canopy height model CHM; perform fitting analysis on the plant height data obtained from the UAV remote sensing data and the measured plant height. In this embodiment, the fitting accuracy between the rice plant height extracted based on the canopy height model and the measured plant height is relatively good, where R 2 is 0.75, RMSE is 0.04 m, Figure 5 and it is the distribution map of the plant height of the rice at the mature stage in the test area.

[0080] S42. Obtain the water content of the rice plants at the mature stage; use the UAV to obtain high-resolution multispectral images and extract the multispectral indices with high correlation with the measured water content, combine the reflectance of the red light band, the reflectance of the red edge band, and the reflectance of the near-infrared band, and adopt the random forest regression method to construct the relationship model between the multispectral indices and the water content to estimate the water content of the rice at the mature stage.

[0081] In this embodiment, a total of 6 multispectral indices such as the difference vegetation index (DVI) with high correlation with the measured water content are extracted from the multispectral images of the test area (see Table 2), and combined with the reflectance of the red light band, the reflectance of the red edge band, and the reflectance of the near-infrared band, and the random forest regression method is used to construct the relationship model between the multispectral indices and the water content. The experimental results show that the estimation accuracy of the water content of the rice at the mature stage using the multispectral data and the random forest regression method is relatively high, the determination coefficient of the model is 0.77, and the deviation degree of the scatter distribution from the 1:1 line is small, and the fitting effect is good. The distribution of the water content of the rice at the mature stage in the test area is obtained as shown in Figure 6 .

[0082] Table 2 Multispectral indices

[0083]

[0084] Among them, NIR represents the near-infrared band with a central wavelength of 840 nm, R’ represents the reflectance of the red light band with a central wavelength of 660 nm, and RE represents the reflectance of the red edge band with a central wavelength of 750 nm.

[0085] S43. According to the plant height of the rice at the mature stage obtained in S41 and the water content of the rice plants at the mature stage obtained in S42, combined with the mature rice biomass estimation model constructed in S3, estimate the mature rice biomass in the test area to obtain the distribution of the mature rice biomass in the test area.

[0086] In this embodiment, finally, the distribution of the mature rice biomass in the test area is as shown in Figures 7 - 8As shown below:

[0087] (1) There is the same magnitude relationship between the estimated value and the measured value of the rice biomass in each region at the maturity stage. The estimated value is generally on the high side, but the difference between them is small, and the absolute value is within 0.15 - 0.39 kg / m 2 . Overall, the model performs well. In terms of the biomass distribution in the test area, the overall level of rice biomass in Region D and Region A is relatively the highest, while the overall level of rice biomass in Region E, B, and C is relatively lower. In terms of the biomass distribution within each region, there are also differences in the rice biomass within the same region.

[0088] (2) The estimated rice biomass in each region of the test area is fitted with the measured yield. The fitting determination coefficient and root mean square error are 0.86 respectively, and the fitting function is y = 0.04x - 0.59. The fitting result shows that there is an obvious linear correlation between the yield and the rice biomass. The situation of the regional yield is consistent with the situation of the rice biomass, that is, the higher the rice biomass value in a region, the relatively higher its yield.

[0089] This result indicates that the model is reliable in the application of estimating the rice biomass at the maturity stage in a relatively large area.

[0090] Therefore, the present invention adopts the above-mentioned method for estimating the yield of rice at the maturity stage by coupling the visible light index and growth parameters, integrating the visible light index of the rice canopy obtained by the unmanned aerial vehicle remote sensing technology and the rice growth parameters, and combining the support vector machine regression method can effectively estimate the rice biomass at the rice maturity stage, providing a fast and efficient method for estimating the rice biomass in a relatively large area, and can provide a method reference for the work of monitoring the growth of rice at the maturity stage and efficiently estimating the yield.

[0091] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that they can still modify or equivalently replace the technical solutions of the present invention, and these modifications or equivalent replacements cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for estimating rice yield at maturity by coupling visible light index with growth parameters, characterized in that: The following steps are involved: S1. Collect and process the UAV remote sensing data and ground crop data of the test area. The UAV remote sensing data includes visible light data and multi-spectral data; the ground crop data includes rice plant height, rice fresh weight and rice dry weight; The rice plant height is the distance from the bottom to the top of the canopy of rice under natural growth conditions; the rice fresh weight is the weight of the rice after sampling and bagging; the rice dry weight is the weight of the rice after drying to a constant weight; Data processing includes the following steps: S11. Use Pix4D Mapper software and Yusense Map software to stitch and correct the UAV remote sensing data, crop the UAV images according to the sampling area location, take the average reflectance of each sampling area as the spectral reflectance of the sampling point, and obtain the spectral reflectance data of different bands; S12, vegetation index construction; select visible light index as an indicator for constructing rice biomass model at maturity; S13. Calculate the rice biomass per square meter of land based on the dry weight of rice and rice planting density in the sampling area; S14, calculating the water content of the rice plant by the ratio of the difference between the fresh weight of the rice and the dry weight of the rice to the fresh weight of the rice; S2. Estimate the rice biomass at maturity based on a single characteristic variable, specifically: S21. Pearson correlation analysis was used to evaluate the linear correlation between visible light index, rice plant height, rice plant water content and rice biomass. S22. Using visible light index, rice plant height, and rice plant water content as independent variables, a rice biomass estimation model was constructed using the support vector regression method. The training set determination coefficient R 2 _train and the test set determination coefficient R 2 _test Represent the accuracy of the training set and the accuracy of the test set respectively; S23, selecting characteristic variables for fusing the visible light index and growth parameters to establish a rice biomass estimation model at maturity from the visible light index, rice plant height, and rice plant water content according to the accuracy of the training set and the accuracy of the test set; S3, integrating the visible light index and growth parameters to establish a rice biomass estimation model at maturity, including the following steps: S31, Test set determination coefficient of the rice biomass estimation model based on the visible light index in S2 R 2 _test , the visible light index was classified for index optimization, and the rice biomass estimation model at maturity was constructed using different categories of visible light index as input factors; S32, integrating rice plant height and rice plant water content to construct a rice biomass estimation model at maturity and compare the fitting results; S4. Estimation of rice biomass at maturity in the experimental area, including the following steps: S41, obtaining the height of rice plants at maturity; S42, obtaining the water content of rice at maturity; S43. Based on the plant height of mature rice obtained in S41 and the water content of mature rice obtained in S42, combined with the mature rice biomass estimation model constructed in S3, the biomass of mature rice in the experimental area is estimated to obtain the distribution of mature rice biomass in the experimental area.

2. The method for estimating rice yield at maturity by coupling visible light index and growth parameters according to claim 1, characterized in that: In S32, the specific process is: the visible light index is sequentially integrated with the rice plant height, the rice plant water content, the rice plant height and the rice plant water content, and the three fitting effects are compared.

3. The method for estimating rice yield at maturity by coupling visible light index and growth parameters according to claim 1, characterized in that: In S41, the specific process of obtaining the plant height of rice at maturity is as follows: the UAV remote sensing data of the rice field before rice planting when it is bare and when the rice is mature are processed to obtain the digital surface model DSM of the corresponding period, which represents the ground height of the rice field and the overall height of the rice at maturity, respectively; the plant height data of rice at maturity is obtained through the canopy height model CHM; the plant height data obtained through the UAV remote sensing data are fitted and analyzed with the measured plant height.

4. The method for estimating rice yield at maturity by coupling visible light index and growth parameters according to claim 1, characterized in that: In S42, the specific process of obtaining the moisture content of rice at maturity is as follows: using drones to obtain multispectral images and extract multispectral indices related to the measured moisture content, combining the red light band reflectivity, red edge band reflectivity and near-infrared band reflectivity, and using the random forest regression method to construct a relationship model between the multispectral index and moisture content, and estimate the moisture content of rice at maturity.

Images