Sorghum canopy nitrogen concentration prediction method and device, medium and product

Abstract

The invention discloses a sorghum canopy nitrogen concentration prediction method and device, a medium and a product, and relates to the technical field of nitrogen concentration prediction, and the sorghum canopy nitrogen concentration prediction method comprises the steps: obtaining a target multispectral image; the target multispectral image is a multispectral image comprising the canopy of the sorghum to be predicted in the current period; based on the target multispectral image, determining a background-removed multispectral orthographic image of the sorghum to be predicted; determining a plurality of optimal remote sensing variables and a plurality of optimal texture features of the to-be-predicted sorghum based on the background-removed multispectral orthographic image of the to-be-predicted sorghum; and inputting the plurality of optimal remote sensing variables and the plurality of optimal texture features of the to-be-predicted sorghum into the nitrogen concentration prediction model to obtain a predicted value of the canopy nitrogen concentration of the to-be-predicted sorghum in the current period. The method can be used for predicting the nitrogen concentration of the sorghum canopy in real time.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of nitrogen concentration prediction, in particular to a sorghum canopy nitrogen concentration prediction method and device, medium and product. BACKGROUND

[0002] Sorghum is widely planted in the world as an important food and feed crop due to its drought tolerance and strong adaptability. Nitrogen fertilizer is a key factor for improving crop yield and quality, and its reasonable use is an important link to achieve efficient and sustainable agricultural development. However, unreasonable nitrogen management not only reduces nitrogen use efficiency, but also can lead to soil degradation and environmental pollution. Therefore, the precise management and non-destructive diagnosis of canopy nitrogen concentration are of great significance for improving nitrogen use efficiency, reducing environmental pollution and achieving sustainable agricultural development. Traditional canopy nitrogen concentration prediction methods mostly rely on sampling and analysis of sorghum and soil, which can accurately reflect the nitrogen nutrition status of crops, but the process is complicated, time-consuming and can cause damage to sorghum, and cannot meet the needs of large-area, rapid and real-time monitoring. SUMMARY

[0003] The purpose of the present application is to provide a sorghum canopy nitrogen concentration prediction method, device, medium and product to solve the problem of real-time prediction of sorghum canopy nitrogen concentration.

[0004] To achieve the above-mentioned purpose, the present application provides the following solutions:

[0005] In a first aspect, the present application provides a sorghum canopy nitrogen concentration prediction method, comprising:

[0006] obtaining a target multispectral image; the target multispectral image is a multispectral image including the canopy of the sorghum to be predicted at the current stage, and the current stage is the jointing stage or the heading stage;

[0007] determining a multispectral orthographic image of the sorghum to be predicted after removing the background based on the target multispectral image;

[0008] determining a plurality of optimal remote sensing variables and a plurality of optimal texture features of the sorghum to be predicted based on the multispectral orthographic image of the sorghum to be predicted after removing the background; the plurality of optimal remote sensing variables are obtained by screening a plurality of initial remote sensing variables, and the plurality of optimal texture features are obtained by screening a plurality of initial texture features;

[0009] inputting the plurality of optimal remote sensing variables and the plurality of optimal texture features of the sorghum to be predicted into a nitrogen concentration prediction model to obtain a predicted value of the canopy nitrogen concentration of the sorghum to be predicted at the current stage; the nitrogen concentration prediction model is obtained by screening a trained initial network, and the initial network includes a random forest network, a support vector machine, a partial least squares regression method and a back propagation neural network.

[0010] In an embodiment, the target multi-spectral image is collected by a UAV.

[0011] In an embodiment, based on the target multi-spectral image, a multi-spectral orthographic image of the to-be-predicted sorghum after removing a background is determined, including:

[0012] A radiation correction parameter of a diffuse reflection gray cloth when the UAV collects the target multi-spectral image is obtained;

[0013] A two-dimensional multi-spectral reconstruction is performed on the target multi-spectral image by using the radiation correction parameter, to obtain a multi-spectral orthographic image of the to-be-predicted sorghum;

[0014] A mask processing is performed on the multi-spectral orthographic image of the to-be-predicted sorghum, to obtain a multi-spectral orthographic image of the to-be-predicted sorghum after removing the background.

[0015] In an embodiment, based on the multi-spectral orthographic image of the to-be-predicted sorghum after removing the background, a plurality of optimal remote sensing variables and a plurality of optimal texture features of the to-be-predicted sorghum are determined, including:

[0016] A canopy spectral value of a plurality of wave bands of the to-be-predicted sorghum is determined according to the multi-spectral orthographic image of the to-be-predicted sorghum after removing the background;

[0017] A plurality of optimal remote sensing variables of the to-be-predicted sorghum are determined according to the canopy spectral value of the plurality of wave bands of the to-be-predicted sorghum;

[0018] A plurality of optimal texture features of the to-be-predicted sorghum are determined according to the multi-spectral orthographic image of the to-be-predicted sorghum after removing the background.

[0019] In an embodiment, the plurality of wave bands include a green wave band, a red wave band, a red edge wave band, and a near-infrared wave band.

[0020] In an embodiment, when the current period is the jointing period, the plurality of optimal remote sensing variables include the canopy spectral value of the near-infrared wave band and the triangular vegetation index;

[0021] When the current period is the heading period, the plurality of optimal remote sensing variables include a green ratio vegetation index, a green normalized difference vegetation index, a green edge chlorophyll index, a red edge chlorophyll index, a normalized difference red edge index, and a nitrogen reflection index;

[0022] When the current period is the jointing period, the plurality of optimal texture features include average gray scale values of the green wave band, the red wave band, the red edge wave band, and the near-infrared wave band of each region of interest in the multi-spectral orthographic image of the to-be-predicted sorghum after removing the background;

[0023] The plurality of optimal texture features include: average gray values of green light bands, red light bands, red edge bands and near-infrared bands of each region of interest in the multispectral orthographic image of the sorghum to be predicted after background removal.

[0024] In an embodiment, the determination process of the nitrogen concentration prediction model comprises:

[0025] An initial data set is obtained, and the initial data set comprises: a plurality of optimal remote sensing variables, a plurality of optimal texture features and measured values of canopy nitrogen concentrations of a plurality of sample sorghums;

[0026] The initial data set is divided into a training set and a test set according to a preset ratio;

[0027] The initial network is trained using the training set to obtain a plurality of trained initial networks;

[0028] The determination coefficients of each trained initial network are determined using the test set;

[0029] The trained initial network with the largest determination coefficient is determined as the nitrogen concentration prediction model.

[0030] In a second aspect, the present application provides a computer device, comprising: a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor executes the computer program to implement the sorghum canopy nitrogen concentration prediction method described above.

[0031] In a third aspect, the present application provides a computer readable storage medium having a computer program stored thereon, and the computer program is executed by a processor to implement the sorghum canopy nitrogen concentration prediction method described above.

[0032] In a fourth aspect, the present application provides a computer program product comprising a computer program, and the computer program is executed by a processor to implement the sorghum canopy nitrogen concentration prediction method described above.

[0033] According to the specific embodiments provided by the present application, the following technical effects are disclosed:

[0034] The application discloses a sorghum canopy nitrogen concentration prediction method, device, medium and product. First, a target multispectral image is acquired; the target multispectral image is a multispectral image including a canopy of a sorghum to be predicted at a current stage, and the current stage is the jointing stage or the heading stage. Then, based on the target multispectral image, a multispectral orthographic image of the sorghum to be predicted after background removal is determined. Subsequently, based on the multispectral orthographic image of the sorghum to be predicted after background removal, a plurality of optimal remote sensing variables and a plurality of optimal texture features of the sorghum to be predicted are determined. The plurality of optimal remote sensing variables are obtained by screening a plurality of initial remote sensing variables, and the plurality of optimal texture features are obtained by screening a plurality of initial texture features. Finally, the plurality of optimal remote sensing variables and the plurality of optimal texture features of the sorghum to be predicted are input into a nitrogen concentration prediction model to obtain a prediction value of the canopy nitrogen concentration of the sorghum to be predicted at the current stage. The nitrogen concentration prediction model is obtained by screening a trained initial network, and the initial network includes a random forest network, a support vector machine, a partial least squares regression method and a back propagation neural network. The optimal remote sensing variables obtained by screening the initial remote sensing variables and the optimal texture features obtained by screening the plurality of initial texture features are input into the nitrogen concentration prediction model obtained by screening the trained initial network, so that the prediction value of the canopy nitrogen concentration of the sorghum to be predicted is automatically determined. Compared with the traditional sampling analysis of the sorghum and the soil for nitrogen concentration prediction, the sorghum canopy nitrogen concentration can be predicted in real time. BRIEF DESCRIPTION OF DRAWINGS

[0035] In order to more clearly illustrate the technical solutions in the embodiments of the application or the related art, the following will briefly introduce the drawings needed in the embodiments. Obviously, the drawings in the following description only some embodiments of the application, and for those skilled in the art, other drawings can be obtained without creative labor based on these drawings.

[0036] Figure 1 The application environment diagram of the sorghum canopy nitrogen concentration prediction method in an embodiment of the application;

[0037] Figure 2 The sorghum canopy nitrogen concentration prediction method flowchart provided by an embodiment of the application;

[0038] Figure 3 The RFE importance ranking result diagram of the remote sensing variables of the sorghum at the jointing stage;

[0039] Figure 4 The decision tree importance ranking result diagram of the remote sensing variables of the sorghum at the jointing stage;

[0040] Figure 5 The RFE importance ranking result diagram of the remote sensing variables of the sorghum at the heading stage;

[0041] Figure 6 Figure for ranking the importance of remote sensing variables of sorghum at the heading stage by decision tree;

[0042] Figure 7 Figure for ranking the importance of texture features of sorghum at the jointing stage by RFE;

[0043] Figure 8 Figure for ranking the importance of texture features of sorghum at the jointing stage by decision tree;

[0044] Figure 9 Figure for ranking the importance of texture features of sorghum at the heading stage by RFE;

[0045] Figure 10 Figure for ranking the importance of texture features of sorghum at the heading stage by decision tree;

[0046] Figure 11 Figure for the relationship between the measured value and the predicted value;

[0047] Figure 12 Figure for the structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0048] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0049] The purpose of the present application is to provide a sorghum canopy nitrogen concentration prediction method, device, medium and product, aiming to predict the sorghum canopy nitrogen concentration in real time.

[0050] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0051] The sorghum canopy nitrogen concentration prediction method provided in the embodiments of the present application can be applied to, for example Figure 1The application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data required by the server 104 to process. The data storage system can be set up separately, or integrated on the server 104, or placed on the cloud or other servers. The terminal 102 can send the target multispectral image to the server 104, and the server 104 receives the target multispectral image. For the target multispectral image, the server 104 determines the multispectral orthographic image of the to-be-predicted sorghum after removing the background based on the target multispectral image; based on the multispectral orthographic image of the to-be-predicted sorghum after removing the background, determine a plurality of optimal remote sensing variables and a plurality of optimal texture features of the to-be-predicted sorghum; input the plurality of optimal remote sensing variables and the plurality of optimal texture features of the to-be-predicted sorghum into the nitrogen concentration prediction model to obtain the predicted value of the canopy nitrogen concentration of the to-be-predicted sorghum in the current period. The server 104 can feed back the obtained predicted value of the canopy nitrogen concentration of the to-be-predicted sorghum in the current period to the terminal 102. In addition, in some embodiments, the sorghum canopy nitrogen concentration prediction method can also be realized by the server 104 or the terminal 102 alone, such as directly predicting the sorghum canopy nitrogen concentration for the target multispectral image by the terminal 102, or obtaining the target multispectral image from the data storage system by the server 104 and predicting the sorghum canopy nitrogen concentration for the target multispectral image.

[0052] In an exemplary embodiment, as shown, a sorghum canopy nitrogen concentration prediction method is provided, comprising: Figure 2

[0053] Step 1: Obtain a target multispectral image; the target multispectral image is a multispectral image including the canopy of the to-be-predicted sorghum in the current period, and the current period is the jointing stage or the heading stage.

[0054] Specifically, the sorghum grows rapidly in the jointing stage, the leaves become longer, and the stems become thicker. The above-ground internodes elongate rapidly, the base of the stem near the ground surface becomes round and can be touched, the growth rate and nutrient absorption rate of the sorghum reach the maximum value, and the growth point is above the ground surface. During this period, the height of the sorghum grows rapidly, and the pistil inflorescence is formed, generally 35-75 days after emergence. The heading stage is the period when the ear (inflorescence) of sorghum grows out of the flag leaf (top leaf) and begins to flower and pollinate, usually 60-90 days after sowing.

[0055] As an optional implementation, the target multispectral image is collected by a drone.

[0056] Specifically, the drone is DJI drone Mavic3Multispectral.

[0057] Step 2: Determine the multispectral orthographic image of the to-be-predicted sorghum after removing the background based on the target multispectral image. ​

[0058] As an optional implementation, step 2 comprises:

[0059] Step 21: Obtain the radiation correction parameter of the diffuse reflection gray cloth when the unmanned aerial vehicle collects the target multi-spectral image.

[0060] Step 22: Perform two-dimensional multi-spectral reconstruction on the target multi-spectral image by using the radiation correction parameter, to obtain a multi-spectral orthographic image of the to-be-predicted sorghum.

[0061] Specifically, the pixel value of a pixel point of the multi-spectral orthographic image corresponds to the reflectivity of each wave band.

[0062] Step 23: Perform mask processing on the multi-spectral orthographic image of the to-be-predicted sorghum, to obtain a multi-spectral orthographic image of the to-be-predicted sorghum after removing the background.

[0063] Specifically, the mask processing on the multi-spectral orthographic image of the to-be-predicted sorghum can remove the background part of the multi-spectral orthographic image.

[0064] Step 3: Determine a plurality of optimal remote sensing variables and a plurality of optimal texture features of the to-be-predicted sorghum based on the multi-spectral orthographic image of the to-be-predicted sorghum after removing the background.

[0065] The plurality of optimal remote sensing variables are obtained by screening a plurality of initial remote sensing variables, and the plurality of optimal texture features are obtained by screening a plurality of initial texture features.

[0066] Specifically, when the current period is the jointing period, the plurality of optimal remote sensing variables of the to-be-predicted sorghum in step 3 are a plurality of optimal remote sensing variables of the to-be-predicted sorghum in the jointing period; and when the current period is the heading period, the plurality of optimal remote sensing variables of the to-be-predicted sorghum in step 3 are a plurality of optimal remote sensing variables of the to-be-predicted sorghum in the heading period.

[0067] As an optional implementation, step 3 comprises:

[0068] Step 31: Determine the canopy spectral value of each wave band of the to-be-predicted sorghum according to the multi-spectral orthographic image of the to-be-predicted sorghum after removing the background.

[0069] Specifically, the mean value of the reflectivity of any wave band at all pixel points in the multi-spectral orthographic image after removing the background is the canopy spectral value of the wave band.

[0070] As an optional implementation, the plurality of wave bands comprise a green wave band, a red wave band, a red edge wave band, and a near-infrared wave band.

[0071] Specifically, the wavelength of the green wave band is 560 nm, the wavelength of the red wave band is 650 nm, the wavelength of the red edge wave band is 730 nm, and the wavelength of the near-infrared wave band is 860 nm.

[0072] Step 32: determining a plurality of optimal remote sensing variables of the to-be-predicted sorghum according to the canopy spectral values of a plurality of wave bands of the to-be-predicted sorghum.

[0073] Specifically, the screening process of the plurality of optimal remote sensing variables of the current period includes:

[0074] Step 321: obtaining the canopy spectral values of a plurality of wave bands of the sample sorghum in the current period, and determining a plurality of initial remote sensing variables of the current period according to the canopy spectral values of the plurality of wave bands of the sample sorghum in the current period. The initial remote sensing variables are shown in Table 1.

[0075] Table 1: initial remote sensing variable table

[0076]

[0077] wherein SQRT represents square root.

[0078] S12: using the Recursive Feature Elimination (RFE) based on cross-validation to perform importance sorting and feature screening on the 18 initial remote sensing variables of the current period, and taking the initial remote sensing variables with importance values greater than 0.05 as the optimal remote sensing variables of the current period screened by RFE. The importance sorting results of the initial remote sensing variables of the jointing stage and the heading stage screened by RFE are shown in Figure 3 and Figure 5 respectively. Figure 3 and Figure 5 The vertical axis in the figures represents the importance value. The optimal remote sensing variables of the jointing stage screened by RFE include: NIR, RDVI, TVI, DVI; and the optimal remote sensing variables of the heading stage screened by RFE include: NRI, GRVI, GNDVI, CIgreen, RECI, NDRE.

[0079] S13: using the decision tree to perform importance sorting and feature screening on the 18 initial remote sensing variables of the current period, and taking the initial remote sensing variables with importance values greater than 0.05 as the optimal remote sensing variables of the current period screened by the decision tree. The importance sorting results of the initial remote sensing variables of the jointing stage and the heading stage screened by the decision tree are shown in Figure 4 and Figure 6 respectively. Figure 4 and Figure 6 The vertical axis in the figures represents the importance value. The optimal remote sensing variables of the jointing stage screened by the decision tree include: NIR, TVI, CIgreen, RVI; and the optimal remote sensing variables of the heading stage screened by the decision tree include: NRI, NDRE, CIgreen, GNDVI, GRVI, SRCI, RECI.

[0080] S14: determining the intersection of the optimal remote sensing variables of the jointing stage screened by the RFE and the optimal remote sensing variables of the jointing stage screened by the decision tree as the multiple optimal remote sensing variables of the jointing stage; determining the intersection of the optimal remote sensing variables of the heading stage screened by the RFE and the optimal remote sensing variables of the heading stage screened by the decision tree as the multiple optimal remote sensing variables of the heading stage.

[0081] Step 33: determining the multiple optimal texture features of the to-be-predicted sorghum according to the background-removed multispectral orthographic image of the to-be-predicted sorghum.

[0082] Specifically, the screening process of the multiple optimal texture features of the current stage includes:

[0083] S21: obtaining the background-removed multispectral orthographic image of the sample sorghum in the current stage, and determining the multiple initial texture features of the current stage according to the background-removed multispectral orthographic image of the sample sorghum in the current stage. The initial texture features are shown in Table 2.

[0084] Table 2: initial texture features table

[0085]

[0086] Wherein, X_mean represents the average gray value of a certain band of a certain region of interest in the background-removed multispectral orthographic image (X is G, which is the green band; X is R, which is the red band; X is Rededge, which is the red edge band; X is NIR, which is the near-infrared band), X_var represents the variance of a certain band of a certain region of interest in the background-removed multispectral orthographic image; X_hom represents the homogeneity of a certain band of a certain region of interest in the background-removed multispectral orthographic image; X_Con represents the contrast of a certain band of a certain region of interest in the background-removed multispectral orthographic image; X_dis represents the difference of a certain band of a certain region of interest in the background-removed multispectral orthographic image; X_ent represents the entropy of a certain band of a certain region of interest in the background-removed multispectral orthographic image; X_sm represents the second moment of a certain band of a certain region of interest in the background-removed multispectral orthographic image; X_cor represents the correlation of a certain band of a certain region of interest in the background-removed multispectral orthographic image.

[0087] S22: performing importance sorting and feature screening on the 32 initial texture features of the current stage by using the RFE, and taking the initial texture features with the importance value greater than 0.03 as the optimal texture features of the current stage screened by the RFE. The importance sorting results of the initial texture features of the jointing stage and the heading stage screened by the RFE are shown in Figure 7 and Figure 9 Figure 7 and Figure 9 ​The middle longitudinal axis represents the importance value. The optimal texture features of the jointing stage after RFE screening include Rededge_mean, G_mean, R_mean, and NIR_mean. The optimal texture features of the heading stage after RFE screening include Rededge_mean, G_mean, R_mean, and NIR_mean.

[0088] S23: The importance of the 32 initial texture features of the current stage is sorted by using the decision tree, and the initial texture features with the importance value greater than 0.03 are taken as the optimal texture features of the current stage after decision tree screening. The importance sorting results of the initial texture features of the jointing stage and the heading stage by the decision tree are shown in FIGS. Figure 8 Figure 10 Figure 8 Figure 10 The middle longitudinal axis represents the importance value. The optimal texture features of the jointing stage after RFE screening include Rededge_mean, G_mean, R_mean, and NIR_mean. The optimal texture features of the heading stage after RFE screening include Rededge_mean, G_mean, R_mean, and NIR_mean.

[0089] S24: The intersection of the optimal texture features of the jointing stage after RFE screening and the optimal texture features of the jointing stage after decision tree screening is determined as the multiple optimal texture features of the jointing stage. The intersection of the optimal texture features of the heading stage after RFE screening and the optimal texture features of the heading stage after decision tree screening is determined as the multiple optimal texture features of the heading stage.

[0090] As an optional implementation, when the current stage is the jointing stage, the multiple optimal remote sensing variables include the canopy spectral value of the near-infrared wave band and the triangular vegetation index.

[0091] When the current stage is the heading stage, the multiple optimal remote sensing variables include the green wave band ratio vegetation index, the green normalized difference vegetation index, the green edge chlorophyll index, the red edge chlorophyll index, the normalized red edge difference index, and the nitrogen reflection index.

[0092] When the current stage is the jointing stage, the multiple optimal texture features include the average gray value of the green wave band, the average gray value of the red wave band, the average gray value of the red edge wave band, and the average gray value of the near-infrared wave band of each region of interest in the multispectral orthographic image of the to-be-predicted sorghum after removing the background.

[0093] When the current stage is the heading stage, the multiple optimal texture features include the average gray value of the green wave band, the average gray value of the red wave band, the average gray value of the red edge wave band, and the average gray value of the near-infrared wave band of each region of interest in the multispectral orthographic image of the to-be-predicted sorghum after removing the background.​​​

[0094] Step 4: input the plurality of optimal remote sensing variables and the plurality of optimal texture features of the to-be-predicted sorghum into the nitrogen concentration prediction model to obtain a prediction value of the canopy nitrogen concentration of the to-be-predicted sorghum at the current stage.

[0095] The nitrogen concentration prediction model is obtained by screening an initial network, and the initial network includes a random forest network (RF), a support vector machine (SVM), a partial least squares regression (PLSR), and a backpropagation neural network (BPNN).

[0096] As an optional implementation, in step 4, the determination process of the nitrogen concentration prediction model includes:

[0097] Step 41: obtaining an initial data set, the initial data set including a plurality of optimal remote sensing variables, a plurality of optimal texture features, and a measured value of the canopy nitrogen concentration of a plurality of sample sorghums.

[0098] Specifically, after sampling at the jointing stage, the leaf blades are dried to determine the dry weight, and at the heading stage, the leaf blades and ears of the sorghum are dried to determine the dry weight of each. The canopy nitrogen concentration at the jointing stage is the nitrogen concentration of the leaf blades, and the key steps for determining the nitrogen concentration are as follows: (a) after the leaf blade sample is dried, the leaf veins are removed, and only the leaf pulp is crushed and sieved (the sieve is not less than 50 mesh); (b) whether the Dumas combustion method or the Kjeldahl method is used, each sample is set to have more than 3 repetitions, and preferably 3-4 repetitions. Among them, a maximum of 40 groups of samples are measured, and a standard sample is set for reference, and preferably 20-30 groups of samples are set for reference. The actual value of the canopy nitrogen concentration at the heading stage = (leaf blade total nitrogen concentration x leaf blade dry weight + ear total nitrogen concentration x ear dry weight) / (leaf blade dry weight + ear dry weight), and the key steps for determining the nitrogen concentration are as follows: (a) after the sorghum sample is collected, the leaf blades and ears are separated, the leaf blades are crushed after the leaf veins are removed, and the ears are crushed and sieved through a 50-mesh sieve; (b) whether the Dumas combustion method or the Kjeldahl method is used, each sample is set to have more than 3 repetitions, and preferably 3-4 repetitions. Among them, a maximum of 40 groups of samples are measured, and a standard sample is set for reference, and preferably 20-30 groups of samples are set for reference.

[0099] Step 42: dividing the initial data set according to a preset ratio to obtain a training set and a test set.

[0100] Step 43: training the initial network using the training set to obtain a plurality of trained initial networks.

[0101] Step 44: Determine the determination coefficient of each trained initial network respectively by using the test set.

[0102] Step 45: Determine the trained initial network with the largest determination coefficient as the nitrogen concentration prediction model.

[0103] Specifically, when the determination coefficient R 2 is determined, the root mean squared error (RMSE) of each trained initial network can also be determined, R 2 The higher the determination coefficient is, the lower the RMSE is, which proves that the model accuracy is higher. The determination coefficient and the root mean squared error of each trained initial network are shown in Table 3.

[0104] Table 3 Determination coefficient and root mean squared error table

[0105]

[0106] It can be seen that the nitrogen concentration monitoring accuracy of sorghum at the heading stage is higher than that at the jointing stage, the R 2 of each model at the jointing stage of sorghum ranges from 0.75 to 0.84, the R 2 of RF is the highest, and the corresponding RMSE is also the lowest, which is 1.37 g·kg -1 , the R 2 of each method modeling at the heading stage of sorghum ranges from 0.83 to 0.89, the R 2 of the BPNN method is the highest, and the corresponding RMSE is also the lowest, which is 0.85 g·kg -1 , indicating that the present application can be suitable for non-destructive determination of the canopy nitrogen concentration of general sorghum at the jointing stage and the heading stage.

[0107] The present application also verifies the method of the present application by using the verification set, and the relationship curve between the measured value and the predicted value of the canopy nitrogen concentration of the verification set is shown in Figure 11 .

[0108] In an exemplary embodiment, a computer device is provided, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor executing the computer program to implement the sorghum canopy nitrogen concentration prediction method.

[0109] In an exemplary embodiment, a computer readable storage medium is provided, which stores a computer program, the computer program being executed by a processor to implement the sorghum canopy nitrogen concentration prediction method.

[0110] In an exemplary embodiment, a computer program product is provided, comprising a computer program, the computer program being executed by a processor to implement the sorghum canopy nitrogen concentration prediction method.

[0111] In an exemplary embodiment, a computer device, which can be a server or a terminal, has an internal structure as shown in Figure 12 The computer device includes a processor, a memory, an input / output interface (I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is configured to exchange information between the processor and external devices. The communication interface of the computer device is configured to communicate with external terminals through a network connection. The computer program is executed by the processor to implement a method for predicting nitrogen concentration in a sorghum canopy.

[0112] Those skilled in the art can understand that Figure 12 The structure shown in the above is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0113] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiments of each method. Any reference to memory, databases or other media used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0114] The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a blockchain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., without being limited thereto.

[0115] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant regulations.

[0116] The technical features of the above embodiments can be combined arbitrarily. In order to make the description simple, not all possible combinations of the technical features in the above embodiments are described, but as long as the combinations of the technical features do not exist contradictory, they should be considered as the scope of the present application.

[0117] The principles and implementations of the present application are described in detail with specific examples in this paper, and the above examples are only used to help understand the method of the present application and its core idea; at the same time, for those skilled in the art, according to the idea of the present application, the specific implementation and application range will be changed. In summary, the content of the specification should not be understood as a limitation of the present application.

Claims

1. A method for predicting the nitrogen concentration in the canopy of sorghum, characterized by, The method comprises the following steps: Obtaining a target multispectral image; the target multispectral image is a multispectral image including a canopy of the sorghum to be predicted at a current stage; the current stage is the jointing stage or the heading stage; Based on the target multispectral image, a multispectral orthographic image of the sorghum to be predicted after removing the background is determined; Based on the multispectral orthographic image of the sorghum to be predicted after removing the background, a plurality of optimal remote sensing variables and a plurality of optimal texture features of the sorghum to be predicted are determined; the plurality of optimal remote sensing variables are obtained by screening a plurality of initial remote sensing variables, and the plurality of optimal texture features are obtained by screening a plurality of initial texture features; The plurality of optimal remote sensing variables and the plurality of optimal texture features of the sorghum to be predicted are input into a nitrogen concentration prediction model to obtain a predicted value of the nitrogen concentration of the canopy of the sorghum to be predicted at the current stage; the nitrogen concentration prediction model is obtained by screening a trained initial network; the initial network includes a random forest network, a support vector machine, a partial least squares regression method and a back propagation neural network.

2. The sorghum canopy nitrogen concentration prediction method of claim 1, wherein, The target multispectral image is collected by using a UAV. 3.The method of claim 2, wherein, Based on the target multispectral image, a multispectral orthographic image of the sorghum to be predicted after removing the background is determined, which comprises the following steps: Obtaining a radiation correction parameter of a diffuse reflection gray cloth when the UAV collects the target multispectral image; Using the radiation correction parameter, a two-dimensional multispectral reconstruction is performed on the target multispectral image to obtain a multispectral orthographic image of the sorghum to be predicted; Performing mask processing on the multispectral orthographic image of the sorghum to be predicted to obtain a multispectral orthographic image of the sorghum to be predicted after removing the background. 4.The method of claim 1, wherein, Based on the multispectral orthographic image of the sorghum to be predicted after removing the background, a plurality of optimal remote sensing variables and a plurality of optimal texture features of the sorghum to be predicted are determined, which comprises the following steps: According to the multispectral orthographic image of the sorghum to be predicted after removing the background, canopy spectral values of a plurality of wave bands of the sorghum to be predicted are determined; According to the canopy spectral values of the plurality of wave bands of the sorghum to be predicted, a plurality of optimal remote sensing variables of the sorghum to be predicted are determined; According to the multispectral orthographic image of the sorghum to be predicted after removing the background, a plurality of optimal texture features of the sorghum to be predicted are determined. 5.The method of claim 4, wherein, The plurality of wave bands include a green wave band, a red wave band, a red edge wave band and a near-infrared wave band. 6.The method of claim 2, wherein, When the current stage is the jointing stage, the plurality of optimal remote sensing variables include the canopy spectral value of the near-infrared wave band and the triangular vegetation index; When the current stage is the heading stage, the plurality of optimal remote sensing variables include a green wave band ratio vegetation index, a green normalized difference vegetation index, a green edge chlorophyll index, a red edge chlorophyll index, a normalized red edge difference index and a nitrogen reflection index; When the current stage is the jointing stage, the plurality of optimal texture features include the average gray values of the green wave band, the average gray values of the red wave band, the average gray values of the red edge wave band and the average gray values of the near-infrared wave band of each region of interest in the multispectral orthographic image of the sorghum to be predicted after removing the background; When the current stage is the heading stage, the plurality of optimal texture features include the average gray values of the green wave band, the average gray values of the red wave band, the average gray values of the red edge wave band and the average gray values of the near-infrared wave band of each region of interest in the multispectral orthographic image of the sorghum to be predicted after removing the background. 7.The method of claim 6, wherein, The determination process of the nitrogen concentration prediction model comprises: obtaining an initial data set, wherein the initial data set comprises a plurality of optimal remote sensing variables, a plurality of optimal texture features and measured values of the crown layer nitrogen concentration of a plurality of samples of sorghum; dividing the initial data set according to a preset ratio to obtain a training set and a test set; training the initial network using the training set to obtain a plurality of trained initial networks; determining the determination coefficients of each trained initial network using the test set; determining the trained initial network with the largest determination coefficient as the nitrogen concentration prediction model.

8. A computer apparatus comprising: A memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement the sorghum crown layer nitrogen concentration prediction method in any one of claims 1-7.

9. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the sorghum crown layer nitrogen concentration prediction method in any one of claims 1-7.

10. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the sorghum crown layer nitrogen concentration prediction method in any one of claims 1-7.

Images