Leaf image mode classification method and system for double cropping rice ear fertilization

Abstract

The invention provides a leaf image mode classification method and system for double cropping rice spike fertilization, and the method comprises the steps: collecting a first leaf image parameter array and a second leaf image parameter array at a preset time window of target double cropping rice and at a plurality of positions of a first leaf and a second leaf; carrying out image parameter discrete analysis on the parameter arrays to obtain a first dispersion and a second dispersion, and calculating to obtain a leaf pollution degree by combining a preset dispersion; and carrying out averaging processing and leaf pattern classification on the parameter array, and respectively obtaining two leaf pattern arrays based on the image parameters and the pollution degree, thereby determining a suitable leaf pattern and carrying out double cropping rice fertilization. The technical problem of insufficient fertilization recommendation accuracy on different fields due to insufficient test data volume in the prior art is solved, and the technical effects of realizing accurate regulation and control of the panicle fertilizer and improving the fertilization recommendation accuracy by analyzing pattern classification of double cropping rice leaf image parameters and pollution degrees are achieved.

Description

Technical Field

[0001] The invention relates to the field of image processing, and in particular to a leaf image pattern classification method and system for double-season rice ear fertilization. Background Art

[0002] At present, the commonly used fertilization recommendation methods are mainly based on years of field test data, and the amount of fertilizer for crops is calculated through yield response and agronomic efficiency models. However, due to the problem of insufficient test data, there are obvious differences in the application effects of these methods on different plots, and the accuracy of fertilization recommendations is insufficient, making it difficult to meet the requirements of precision agriculture for fertilization management. Especially for double-season rice, due to the complexity of its growth environment and nutrient requirements, relying solely on calculated fertilization amounts is often difficult to adapt to the actual nutrient conditions of the field, resulting in improper fertilization and affecting yield and quality. Summary of the invention

[0003] The present invention aims to solve the technical problem in the prior art that insufficient test data leads to insufficient accuracy in fertilizer recommendation on different fields, and provides a leaf image pattern classification method and system for double-season rice ear fertilization to solve the problem.

[0004] The technical solution of the present invention to solve the above technical problems is as follows: In a first aspect, the present invention provides a leaf image pattern classification method for double-season rice ear fertilization, comprising: in a preset time window of the target double-season rice, at multiple positions of the first leaf and the second leaf, collecting and obtaining a first leaf image parameter array and a second leaf image parameter array; performing image parameter discrete analysis on the first leaf image parameter array and the second leaf image parameter array to obtain a first discrete degree and a second discrete degree, and calculating and obtaining a first leaf contamination degree and a second leaf contamination degree by combining the preset first discrete degree and the preset second discrete degree; performing averaging processing and leaf pattern classification on the first leaf image parameter array and the second leaf image parameter array to obtain a first leaf pattern array, and performing leaf pattern classification based on the first leaf contamination degree and the second leaf contamination degree to obtain a second leaf pattern array; processing and obtaining leaf patterns based on the first leaf pattern array and the second leaf pattern array to perform double-season rice fertilization.

[0005] In a second aspect, the present invention provides a leaf image pattern classification system for double-season rice ear fertilization, comprising: a leaf acquisition module, configured to acquire a first leaf image parameter array and a second leaf image parameter array at multiple positions on the first leaf and the second leaf within a preset time window of the target double-season rice; a pollution degree analysis module, configured to perform image parameter discrete analysis on the first leaf image parameter array and the second leaf image parameter array to obtain a first dispersion degree and a second dispersion degree, and calculate a first leaf pollution degree and a second leaf pollution degree by combining a preset first dispersion degree and a preset second dispersion degree; a leaf pattern classification module, configured to perform averaging processing and leaf pattern classification on the first leaf image parameter array and the second leaf image parameter array to obtain a first leaf pattern array, and perform leaf pattern classification according to the first leaf pollution degree and the second leaf pollution degree to obtain a second leaf pattern array; and a fertilization decision-making module, configured to process the obtained leaf patterns according to the first leaf pattern array and the second leaf pattern array to perform fertilization on the double-season rice.

[0006] The beneficial effects of the present invention are as follows: Within a preset time window of the target double-season rice, a first leaf image parameter array and a second leaf image parameter array are acquired at multiple positions on the first leaf and the second leaf, so as to respectively acquire the SPAD values (chlorophyll content parameters) at multiple positions on the third leaf from the top (the first leaf) and the fourth leaf from the top (the second leaf) at the critical time point of the double-season rice when the second leaf from the top shows the tip, forming a parameter array, which provides basic data for subsequent analysis. Since there are natural differences in the SPAD values at different positions of the leaf (such as near the leaf vein and at the leaf edge), it is necessary to acquire multiple positions to form an array. Image parameter discrete analysis is performed on the first leaf image parameter array and the second leaf image parameter array to obtain a first dispersion degree and a second dispersion degree, and a first leaf pollution degree and a second leaf pollution degree are calculated by combining a preset first dispersion degree and a preset second dispersion degree; by analyzing the dispersion degree of the SPAD values of the two leaves and comparing it with the preset normal leaf dispersion degree (for example, the difference in SPAD values between the middle and the edge of the leaf is 5-10 units under normal conditions), the pollution degree of the leaf is calculated. When the nitrogen fertilizer of the double-season rice is insufficient, the synthesis of wax and cutin on the leaf surface is reduced, the leaf surface becomes rougher, and it is more likely to adsorb dust, thus affecting the normal distribution of the SPAD value. The nitrogen fertilizer status can be indirectly reflected by the pollution degree.

[0007] The first leaf image parameter array and the second leaf image parameter array are averaged and the leaf pattern classification is performed to obtain the first leaf pattern array. According to the first leaf pollution degree and the second leaf pollution degree, the leaf pattern classification is carried out to obtain the second leaf pattern array. When the ratio of the first leaf image parameter to the second leaf image parameter is greater than 1, it indicates that nitrogen is being transferred to the third top leaf (new leaf), and the fertilization amount needs to be increased; when the ratio is less than 1, it may indicate that the plant has a disease and the fertilization should be reduced or stopped. At the same time, by analyzing the ratio of the first leaf pollution degree to the second leaf pollution degree, when the ratio is large, it indicates that the third top leaf is severely polluted and the nitrogen absorption is low, which may be caused by a disease, and the fertilization should be reduced; when the ratio is small, it indicates that the third top leaf is lightly polluted and the nitrogen absorption is high, and the fertilization needs to be increased. According to the first leaf pattern array and the second leaf pattern array, the leaf pattern is obtained through processing, and the panicle fertilizer application for double-season rice is carried out. By comprehensively combining the results of the two pattern classifications, the mode is taken as the final leaf pattern, and the panicle fertilizer application amount for double-season rice is determined accordingly. According to different leaf patterns, normal fertilization, or increasing the fertilization amount, or reducing the fertilization amount, or even stopping fertilization can be decided, so as to achieve precise regulation of fertilization.

[0008] Through the above technical solution, the present application overcomes the technical problem in the prior art that the accuracy of fertilization recommendation is insufficient on different fields due to insufficient test data volume, and realizes the technical effect of precise regulation of panicle fertilizer by analyzing the pattern classification of double-season rice leaf image parameters and pollution degree, and improving the accuracy of fertilization recommendation. Brief Description of the Drawings

[0009] Figure 1 It is a schematic flow chart of a leaf image pattern classification method for panicle fertilization of double-season rice provided by the present invention; Figure 2 It is a schematic structural diagram of a leaf image pattern classification system for panicle fertilization of double-season rice provided by the present invention.

[0010] In the drawings, the components represented by each reference numeral are as follows: Leaf acquisition module 11, pollution degree analysis module 12, pattern classification module 13, fertilization decision module 14. Detailed Embodiments

[0011] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present invention.

[0012] In the description of the present invention, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of the said features. In the description of the present invention, "a plurality of" means two or more unless otherwise specifically defined.

[0013] In the description of the present invention, the term "for example" is used to mean "serving as an example, illustration, or explanation". Any embodiment described as "for example" in the present invention is not necessarily construed as being more preferred or advantageous than other embodiments. The following description is given to enable any person skilled in the art to implement and use the present invention. In the following description, details are set forth for purposes of explanation. It should be understood that those of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other instances, well-known structures and processes are not elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope that conforms to the principles and features disclosed in the present invention.

[0014] Embodiment 1, as Figure 1 shown, the embodiment of the present invention provides a method for classifying leaf image patterns for double-season rice panicle fertilization, and the method includes: S100: In a preset time window of the target double-season rice, at multiple positions on the first leaf and the second leaf, acquire a first leaf image parameter array and a second leaf image parameter array.

[0015] Specifically, when the target double-season rice grows to a specific stage (i.e., the preset time window), select the first leaf and the second leaf of the double-season rice plant, measure the SPAD values at multiple different positions on the first leaf and the second leaf respectively by a SPAD instrument, and obtain a set of leaf image parameters (SPAD values) respectively, forming a first leaf image parameter array and a second leaf image parameter array. Among them, the SPAD value represents a non-destructive measurement index of chlorophyll content and can reflect the nitrogen nutrition status of plants. The higher the SPAD value of the leaf, the higher the chlorophyll content, indicating that the nitrogen nutrition status of the plant is better. The first leaf image parameter array refers to an ordered data set composed of the SPAD values measured at multiple positions on the first leaf (the third leaf from the top), and can be expressed as, for example, {SPAD 11 , SPAD 12 ,..., SPAD 1n}, where each SPAD value corresponds to a specific position on the leaf; similarly, the second leaf image parameter array refers to an ordered data set composed of the SPAD values measured at multiple positions on the second leaf (the fourth leaf from the top), and can be expressed as, for example, {SPAD 21, SPAD 22 ,..., SPAD 2m}。

[0016] Among them, the first leaf refers to the third leaf from the top of double-cropping rice, and the third leaf from the top is the third leaf counted from the top of the rice plant downwards; the second leaf refers to the fourth leaf from the top of double-cropping rice, and the fourth leaf from the top is the fourth leaf counted from the top of the rice plant downwards; the physiological states of these two leaves can reflect the nitrogen nutrient status of the rice. Multiple positions refer to selecting different sampling points on the same leaf, such as the base, middle, and top of the leaf, or different regions such as near the main vein of the leaf and the leaf edge, so as to obtain comprehensive leaf nutrient status information, avoid errors that may be brought by single-point measurement, and provide more reliable basic data for subsequent leaf pattern classification.

[0017] By collecting the image parameter arrays of the first leaf and the second leaf in a preset time window, it provides basic data for subsequent analysis of the nitrogen nutrient status of double-cropping rice and determination of the panicle fertilizer application strategy.

[0018] S200: Perform image parameter discrete analysis on the first leaf image parameter array and the second leaf image parameter array to obtain a first dispersion degree and a second dispersion degree, and combine a preset first dispersion degree and a preset second dispersion degree to calculate and obtain a first leaf pollution degree and a second leaf pollution degree.

[0019] Specifically, first, perform discrete analysis calculations on the collected first leaf image parameter array and the second leaf image parameter array respectively to obtain a first dispersion degree and a second dispersion degree that reflect the data dispersion degree. This discrete analysis is achieved by calculating the standard deviation of multiple leaf image parameters within the first leaf image parameter array and the second leaf image parameter array. Specifically, by calculating the standard deviation of multiple SPAD values in the first leaf image parameter array and the second leaf image parameter array respectively, the first dispersion degree and the second dispersion degree are obtained. The larger the standard deviation value, the greater the difference in SPAD values at different positions on the leaf and the higher the dispersion degree.

[0020] Then, compare the calculated first dispersion with the preset first dispersion, calculate the magnitude of the deviation of the first dispersion from the preset first dispersion, and obtain the first leaf contamination degree. At the same time, compare the second dispersion with the preset second dispersion, calculate the magnitude of the deviation of the second dispersion from the preset second dispersion, and obtain the second leaf contamination degree. Among them, the preset first dispersion refers to the standard dispersion value that the first leaf (the third leaf from the top) should have under normal growth conditions without being contaminated or interfered by external factors such as dust; the preset second dispersion refers to the standard dispersion value that the second leaf (the fourth leaf from the top) should have under normal growth conditions without being contaminated or interfered by external factors such as dust. The first leaf contamination degree reflects the roughness and dust adsorption situation on the surface of the first leaf due to insufficient nitrogen fertilizer. Under normal circumstances, the difference in SPAD values between the middle part (near the main vein) and the edge of the leaf can reach 5 - 10 units, which is caused by the differences in the structure and physiological state of different parts of the leaf. When the leaf surface becomes rough due to insufficient nitrogen fertilizer and is prone to dust adsorption, this natural difference will change, thus affecting the measurement result of the dispersion. The second leaf contamination degree reflects the roughness and dust adsorption situation on the surface of the second leaf due to insufficient nitrogen fertilizer, and its principle is the same as that of the first leaf contamination degree.

[0021] By obtaining the first leaf contamination degree and the second leaf contamination degree, the nitrogen nutrition status of the leaf can be evaluated, providing a reference basis for subsequent fertilization decisions.

[0022] S300: Perform averaging processing and leaf pattern classification on the first leaf image parameter array and the second leaf image parameter array to obtain a first leaf pattern array. According to the first leaf contamination degree and the second leaf contamination degree, perform leaf pattern classification to obtain a second leaf pattern array.

[0023] Specifically, first, perform averaging calculation processing on the first leaf image parameter array and the second leaf image parameter array respectively to obtain a first leaf image parameter and a second leaf image parameter. The averaging processing refers to calculating the average value of the SPAD values at multiple positions collected to obtain a single parameter that can represent the overall nitrogen status of the entire leaf. For example, the first leaf image parameter refers to the average value of all SPAD values in the first leaf image parameter array; the second leaf image parameter refers to the average value of all SPAD values in the second leaf image parameter array.

[0024] Then, calculate the ratio of the first leaf image parameter to the second leaf image parameter, perform leaf pattern classification, and obtain the first leaf pattern array. The ratio of the first leaf image parameter to the second leaf image parameter reflects the nitrogen distribution relationship between the third top leaf and the fourth top leaf. When this ratio is greater than 1, it indicates that nitrogen is being transferred to the new leaf (the third top leaf), and the plant may need additional fertilization; when this ratio is less than 1, it may indicate that the plant has a disease and needs diagnosis rather than fertilization. At the same time, calculate the ratio of the first leaf contamination degree to the second leaf contamination degree, perform leaf pattern classification, and obtain the second leaf pattern array. The ratio of the first leaf contamination degree to the second leaf contamination degree reflects the relative relationship between the surface roughness and dust adsorption of the third top leaf and the fourth top leaf. When this ratio is larger, it indicates that the contamination degree of the third top leaf is higher, less nitrogen is absorbed, and there may be a disease and no fertilization is required; when this ratio is smaller, it indicates that the contamination degree of the third top leaf is lower, more nitrogen is absorbed, and more nitrogen element supplementation is needed, so fertilization is required.

[0025] By obtaining leaf pattern classification in two different dimensions based on the SPAD value ratio and the contamination degree ratio, the nitrogen nutrition status of double-cropping rice can be comprehensively evaluated from multiple perspectives, providing a comprehensive and accurate reference basis for subsequent panicle fertilizer application decisions.

[0026] S400: According to the first leaf pattern array and the second leaf pattern array, process to obtain a leaf pattern and perform fertilization on the double-cropping rice.

[0027] Specifically, first, select the leaf pattern with the largest proportion of occurrence times from the first leaf pattern array and the second leaf pattern array to obtain a fused leaf pattern. This selection method is similar to a voting mechanism, that is, find the leaf pattern with the highest occurrence frequency from the two pattern arrays respectively as the final basis for leaf pattern decision-making. Selecting the leaf pattern with the most occurrences can reduce the error caused by single judgment and improve the reliability of the decision.

[0028] Then, according to the obtained integrated leaf pattern, determine the application strategy of the panicle fertilizer for double-season rice. For example, for early double-season rice, when the integrated leaf pattern shows that the leaf colors of the third leaf from the top and the fourth leaf from the top are similar (within ±5% of the SPAD value), top-dress the panicle fertilizer according to the original calculated value; when the integrated leaf pattern shows that the leaf color of the third leaf from the top is lighter than that of the fourth leaf from the top (the SPAD value is 5% - 15% lower), apply 50% of the original calculated value as the panicle fertilizer; when the integrated leaf pattern shows that the leaf color of the third leaf from the top is light (the SPAD value is more than 15% lower), then do not apply the panicle fertilizer; when the integrated leaf pattern shows that the leaf color of the third leaf from the top is darker than that of the fourth leaf from the top (the SPAD value is 5% - 15% higher), apply 120% of the original calculated value as the panicle fertilizer; when the integrated leaf pattern shows that the leaf color of the third leaf from the top is dark (the SPAD value is more than 15% higher), apply 140% of the original calculated value as the panicle fertilizer. The fertilization strategy for late double-season rice is similar, but the specific thresholds and fertilization ratios are different. Among them, the original calculated value refers to the recommended application rate of the panicle fertilizer calculated based on parameters such as soil nutrient content, target yield, and planting density.

[0029] Through the fertilization decision-making method based on the leaf pattern, precise fertilization can be carried out according to the actual nitrogen nutrition status of double-season rice, avoiding the blindness that may be brought by traditional empirical fertilization, and also making up for the limitations that may exist in the application of existing fertilization systems in different fields, improving the accuracy and pertinence of the application of the panicle fertilizer for double-season rice, and being conducive to improving fertilizer utilization rate and double-season rice yield.

[0030] Furthermore, in the preset time window of the target double-season rice, at multiple positions on the first leaf and the second leaf, collect the first leaf image parameter array and the second leaf image parameter array, including: S110: At the stage when the second leaf from the top of the target double-season rice shows the tip emerging, select multiple positions on the first leaf and the second leaf of the target double-season rice, where the first leaf is the third leaf from the top and the second leaf is the fourth leaf from the top; S120: At the multiple positions on the first leaf and the second leaf, respectively test and collect the SPAD parameters to obtain the first leaf image parameter array and the second leaf image parameter array.

[0031] In a feasible implementation manner, at the stage when the second leaf from the top of the target double-season rice shows the tip emerging, select multiple positions on the first leaf and the second leaf of the target double-season rice, where the first leaf is the third leaf from the top and the second leaf is the fourth leaf from the top. The stage when the second leaf from the top shows the tip emerging refers to a specific period in the growth process of rice when the second leaf from the top just shows the tip from the leaf sheath. This period is a key stage for the transformation of rice from vegetative growth to reproductive growth. Conducting leaf parameter collection at this time can more accurately reflect the nitrogen nutrition status of rice. At this stage, the third leaf from the top and the fourth leaf from the top have fully expanded, and their physiological characteristics are stable, and they can be used as ideal sampling objects for evaluating the nitrogen status of rice.

[0032] Subsequently, at multiple positions on the first leaf and the second leaf, the SPAD parameters are respectively tested and collected to obtain the first leaf image parameter array and the second leaf image parameter array. Specifically, the SPAD instrument is used to measure at the selected multiple positions, such as different regions like the base, middle, top of the leaf or near the main vein and the leaf margin. One SPAD value is obtained at each position, and these values together constitute the first leaf image parameter array and the second leaf image parameter array. By collecting data at multiple positions, the spatial differences in the nitrogen distribution of the leaf can be more comprehensively understood, providing more reliable basic data for subsequent dispersion analysis and pollution degree calculation.

[0033] Through the above steps, the SPAD value distribution of the third and fourth leaves from the top of double-cropping rice can be obtained within the appropriate time window, providing basic data for subsequent discrete analysis and leaf pattern classification.

[0034] Furthermore, image parameter discrete analysis is performed on the first leaf image parameter array and the second leaf image parameter array to obtain the first dispersion and the second dispersion, including: S210: Perform image parameter discrete analysis calculation on the first leaf image parameter array to obtain the first dispersion. Among them, image parameter discrete analysis is performed by calculating the standard deviation of multiple leaf image parameters within the first leaf image parameter array; S220: Perform image parameter discrete analysis calculation on the second leaf image parameter array to obtain the second dispersion.

[0035] In a preferred embodiment, image parameter discrete analysis is performed by calculating the standard deviation of multiple leaf image parameters within the first leaf image parameter array to obtain the first dispersion. Specifically, the SPAD values at multiple positions on the first leaf (the third leaf from the top) collected are used as a set of data, and the standard deviation of this set of data is calculated to obtain the first dispersion. The standard deviation can effectively reflect the degree of data dispersion. The larger the standard deviation, the greater the difference in SPAD values at different positions of the leaf and the higher the degree of dispersion. Under normal circumstances, there will be a natural difference in SPAD values between the middle and the edge of the leaf. This discrete analysis can capture the degree of this difference. At the same time, image parameter discrete analysis calculation is performed on the second leaf image parameter array to obtain the second dispersion. Specifically, the standard deviation of the SPAD values at multiple positions on the second leaf (the fourth leaf from the top) collected is calculated to quantify the variation degree of SPAD values at different positions of the second leaf, providing a data basis for subsequent analysis of the leaf surface condition.

[0036] Through the above steps, the first dispersion and the second dispersion reflecting the discrete degree of the SPAD value distribution of the first leaf and the second leaf are respectively obtained, which will be used for subsequent calculation of the leaf pollution degree and further evaluation of the nitrogen nutrition status of double-cropping rice.

[0037] Further, in combination with a preset first dispersion degree and a preset second dispersion degree, the first leaf pollution degree and the second leaf pollution degree are calculated as follows: S230: Obtain the first dispersion degree and the second dispersion degree of the same variety of double-cropping rice in a preset time window within a historical time to obtain a sample first dispersion degree set and a sample second dispersion degree set; S240: Calculate and configure the preset first dispersion degree and the preset second dispersion degree according to the sample first dispersion degree set and the sample second dispersion degree set; S250: Calculate the amplitudes by which the first dispersion degree and the second dispersion degree deviate from the preset first dispersion degree and the preset second dispersion degree respectively to obtain the first leaf pollution degree and the second leaf pollution degree.

[0038] Specifically, first, collect the SPAD value dispersion data of the third and fourth leaves from the top of the same variety of double-cropping rice at the tip-exposing stage of the second leaf from the top during multiple growing seasons to form a sample first dispersion degree set and a sample second dispersion degree set. The sample first dispersion degree set and the sample second dispersion degree set can reflect the natural dispersion state of the SPAD values of the double-cropping rice leaves under normal growth conditions without being affected by pollution, providing a reference basis for determining the preset first dispersion degree and the preset second dispersion degree. Then, perform statistical analysis on the sample first dispersion degree set and the sample second dispersion degree set respectively, such as calculating the average value or median, etc., to obtain standard values that can represent the SPAD value dispersion degrees of the third and fourth leaves from the top under normal growth conditions, and use them as the preset first dispersion degree and the preset second dispersion degree respectively, representing the natural dispersion degree of the SPAD values of the leaves without being disturbed by pollution.

[0039] After that, calculate the difference amplitude between the currently measured first dispersion degree and the preset first dispersion degree to obtain the first leaf pollution degree. Specifically, the pollution degree of the first leaf can be quantified by calculating the difference between the first dispersion degree and the preset first dispersion degree, or the ratio of the two. When the first dispersion degree is significantly higher than the preset first dispersion degree, it indicates that the surface of the third leaf from the top may be rough due to insufficient nitrogen fertilizer and is prone to adsorbing dust, resulting in an increase in the difference in SPAD values at different positions, and the first leaf pollution degree is higher. Similarly, calculate the difference amplitude between the currently measured second dispersion degree and the preset second dispersion degree to obtain the second leaf pollution degree. When the second dispersion degree significantly deviates from the preset second dispersion degree, it indicates that the surface state of the fourth leaf from the top is abnormal, and there may be pollution or other influencing factors. By quantifying this deviation degree, the specific value of the second leaf pollution degree can be obtained.

[0040] By obtaining the first leaf pollution degree and the second leaf pollution degree, the nitrogen nutrition status of double-cropping rice can be indirectly evaluated, providing an important basis for subsequent fertilization decisions.

[0041] Further, perform averaging processing and blade pattern classification on the first leaf image parameter array and the second leaf image parameter array to obtain a first blade pattern array. Perform blade pattern classification based on the first leaf contamination degree and the second leaf contamination degree to obtain a second blade pattern array, including: S310: Respectively perform averaging calculation processing on the first leaf image parameter array and the second leaf image parameter array to obtain a first leaf image parameter and a second leaf image parameter; S320: Construct a first blade pattern classifier, where the first blade pattern classifier includes multiple first blade pattern classification branches, and the input feature of each first blade pattern classification branch is the ratio of the first leaf image parameter to the second leaf image parameter, and the output feature is the first blade pattern; S330: Calculate the ratio of the first leaf image parameter to the second leaf image parameter, input it into the first blade pattern classifier for integrated blade pattern classification to obtain a first blade pattern array; S340: Construct a second blade pattern classifier, where the second blade pattern classifier includes multiple second blade pattern classification branches, and the input feature of each second blade pattern classification branch is the ratio of the first leaf contamination degree to the second leaf contamination degree, and the output feature is the second blade pattern; S350: Calculate the ratio of the first leaf contamination degree to the second leaf contamination degree, perform integrated blade pattern classification to obtain a second blade pattern array.

[0042] In a preferred embodiment, respectively perform averaging calculation processing on the first leaf image parameter array and the second leaf image parameter array to obtain a first leaf image parameter and a second leaf image parameter. Specifically, sum the SPAD values at multiple positions of the first leaf and divide by the number of positions to obtain the first leaf image parameter; similarly, sum the SPAD values at multiple positions of the second leaf and divide by the number of positions to obtain the second leaf image parameter. These two parameters respectively represent the overall average chlorophyll content levels of the third-to-last leaf and the fourth-to-last leaf, and can reflect the overall nitrogen nutrient status of the leaves.

[0043] Subsequently, a first leaf pattern classifier is constructed and used for leaf pattern classification. The first leaf pattern classifier includes multiple first leaf pattern classification branches. Each branch uses the ratio of the first leaf image parameter and the second leaf image parameter as the input feature and outputs the corresponding leaf pattern category. By calculating the ratio of the first leaf image parameter to the second leaf image parameter, the nitrogen distribution relationship between the third top leaf and the fourth top leaf can be reflected. When the ratio of the first leaf image parameter to the second leaf image parameter is greater than 1, it indicates that nitrogen is being transferred to the new leaf (the third top leaf), and the plant may need additional fertilization. When the ratio of the first leaf image parameter to the second leaf image parameter is less than 1, it may indicate that the plant has a disease and needs diagnosis rather than fertilization. The calculated ratio is input into the first leaf pattern classifier, and by integrating the results of multiple first leaf pattern classification branches, a first leaf pattern array is obtained.

[0044] Meanwhile, a second leaf pattern classifier is constructed and used for leaf pattern classification in another dimension. The second leaf pattern classifier also includes multiple second leaf pattern classification branches. Each branch uses the ratio of the first leaf contamination degree and the second leaf contamination degree as the input feature and outputs the corresponding leaf pattern category. By calculating the ratio of the first leaf contamination degree to the second leaf contamination degree, the relative relationship of the surface states of the third top leaf and the fourth top leaf can be reflected. When the ratio of the first leaf contamination degree to the second leaf contamination degree is larger, it indicates that the contamination degree of the third top leaf is higher, less nitrogen is absorbed, and there may be a disease, so no fertilization is needed. When the ratio of the first leaf contamination degree to the second leaf contamination degree is smaller, it indicates that the contamination degree of the third top leaf is lower, more nitrogen is absorbed, and more nitrogen needs to be supplemented, so fertilization is needed. The calculated contamination degree ratio is input into the second leaf pattern classifier, and by integrating the results of multiple second leaf pattern classification branches, a second leaf pattern array is obtained.

[0045] Through two different-dimensional leaf pattern classification methods, the nitrogen nutrient status of double-cropping rice can be comprehensively evaluated from the two aspects of SPAD value distribution and leaf surface state, providing a comprehensive and accurate basis for subsequent fertilization decisions.

[0046] Furthermore, constructing the first leaf pattern classifier includes: S321: According to the historical cultivation data of the same-family double-cropping rice, collect a set of sample image parameter ratios, and collect the leaf patterns of the double-cropping rice under each sample image parameter ratio, and label to obtain a set of sample first leaf patterns, where the sample image parameter ratio is the ratio of the first leaf image parameter and the second leaf image parameter of the same-family double-cropping rice; S322: Divide the set of sample image parameter ratios and the set of sample first leaf patterns to obtain multiple first pattern classification training data, where there is an intersection between every two first pattern classification training data; S323: Use machine learning to construct multiple first leaf pattern classification branches; S324: Respectively use the multiple first pattern classification training data to train, validate, and test the multiple first leaf pattern classification branches to obtain a first leaf pattern classifier.

[0047] In a preferred implementation, first, according to the historical cultivation data of the same-family double-season rice, collect a set of sample image parameter ratios and corresponding leaf pattern annotations. Specifically, collect the first leaf image parameters and the second leaf image parameters of the same variety of double-season rice at the tip-exposing stage of the second-to-last leaf during the historical growing season, and calculate the ratio of the first leaf image parameters to the second leaf image parameters to form a set of sample image parameter ratios. At the same time, record the actual leaf growth state corresponding to each sample ratio as the leaf pattern label, so as to obtain a set of sample first leaf patterns. These historical data include the growth performance of double-season rice under different nitrogen distribution states, providing basic data support for training the first leaf pattern classifier.

[0048] Then, divide the set of sample image parameter ratios and the set of sample first leaf patterns to obtain multiple first pattern classification training data. By dividing the set of sample image parameter ratios and the set of sample first leaf patterns into multiple subsets and ensuring that there is a certain intersection between every two subsets, the generalization ability of the model can be enhanced, overfitting can be avoided, and the limited historical data can be fully utilized. Then, use machine learning techniques to construct multiple first leaf pattern classification branches. Among them, different first leaf pattern classification branches can be different types of classification algorithms, such as decision trees, support vector machines, neural networks, etc., or different parameter configurations of the same algorithm. The purpose of constructing multiple first leaf pattern classification branches is to form an ensemble learning model, and by combining the prediction results of multiple basic classifiers, the overall classification performance can be improved.

[0049] After that, respectively use the multiple first pattern classification training data to train, validate, and test the multiple first leaf pattern classification branches. Each classification branch is trained using different training data, then the parameters are adjusted through the validation set, and finally the performance is evaluated on the test set. After this series of processes, a first leaf pattern classifier that can accurately identify different leaf patterns is finally obtained. This first leaf pattern classifier can predict the leaf pattern of the current double-season rice according to the ratio of the first leaf image parameters to the second leaf image parameters, providing a scientific basis for fertilization decisions.

[0050] Similar to the construction of the first leaf pattern classifier, the second leaf pattern classifier also adopts the same construction method. First, based on the historical cultivation data of the same-family double-season rice, collect the ratio of the pollution degree of the first leaf to that of the second leaf to form a sample pollution degree ratio set, and record the actual leaf state corresponding to each ratio as the leaf pattern label to obtain the sample second leaf pattern set. Then, divide the sample data to ensure that there is an intersection between the training data subsets. Next, use machine learning techniques to construct multiple second leaf pattern classification branches, which can use different types of classification algorithms or different parameter configurations of the same algorithm. After that, use the divided training data to train, validate, and test each classification branch, and finally form a second leaf pattern classifier that can predict the leaf pattern based on the ratio of the pollution degree of the first leaf to that of the second leaf.

[0051] In this way, by combining the results of the first leaf pattern classifier and the second leaf pattern classifier, the nitrogen nutrition status of double-season rice can be comprehensively evaluated from two dimensions of the SPAD value and the leaf surface state, providing a reliable basis for the decision-making of panicle fertilizer application.

[0052] Furthermore, according to the first leaf pattern array and the second leaf pattern array, process to obtain a fused leaf pattern for double-season rice fertilization, including: S410: In the first leaf pattern array and the second leaf pattern array, select the leaf pattern with the largest proportion of occurrence times to obtain the fused leaf pattern; S420: Perform double-season rice fertilization according to the fused leaf pattern.

[0053] In a preferred embodiment, first, select the leaf pattern with the largest proportion of occurrence times from the first leaf pattern array and the second leaf pattern array to obtain the fused leaf pattern. Specifically, count the various leaf patterns in the first leaf pattern array and select the leaf pattern with the highest frequency as the judgment result of the first dimension; similarly, count the various leaf patterns in the second leaf pattern array and select the leaf pattern with the highest frequency as the judgment result of the second dimension. Then, fuse the leaf patterns of these two dimensions to form the final fused leaf pattern. This selection method based on the occurrence frequency can reduce the misjudgment risk that may be brought by a single classifier through the integrated judgment of multiple classifiers, improving the accuracy and stability of leaf pattern classification.

[0054] Subsequently, according to the obtained integrated leaf pattern, the application strategy of the panicle fertilizer for double-season rice is determined. Different integrated leaf patterns correspond to different fertilization decisions. For example, when the integrated leaf pattern shows that the leaf colors of the third leaf from the top and the fourth leaf from the top are similar, the panicle fertilizer is top-dressed normally according to the original calculated value; when the integrated leaf pattern shows that the leaf color of the third leaf from the top is lighter than that of the fourth leaf from the top, the application amount of the panicle fertilizer is reduced; when the integrated leaf pattern shows that the leaf color of the third leaf from the top is light, no panicle fertilizer is applied; when the integrated leaf pattern shows that the leaf color of the third leaf from the top is darker than that of the fourth leaf from the top, the application amount of the panicle fertilizer is appropriately increased; when the integrated leaf pattern shows that the leaf color of the third leaf from the top is dark, the application amount of the panicle fertilizer is significantly increased.

[0055] Through the precise fertilization method based on the leaf pattern, the regulation of the panicle fertilizer can be carried out according to the current actual nitrogen nutrition status of double-season rice, avoiding the problems of fertilizer waste or deficiency that may be caused by blind empirical fertilization, and improving the precision of the application of the panicle fertilizer for double-season rice.

[0056] Example 2, as Figure 2 shown, based on the same inventive concept as the leaf image pattern classification method for panicle fertilization of double-season rice provided in Example 1, the embodiment of the present invention further provides a leaf image pattern classification system for panicle fertilization of double-season rice, including: A leaf acquisition module 11, configured to acquire a first leaf image parameter array and a second leaf image parameter array at multiple positions of the first leaf and the second leaf within a preset time window of the target double-season rice; A contamination degree analysis module 12, configured to perform image parameter discrete analysis on the first leaf image parameter array and the second leaf image parameter array to obtain a first discrete degree and a second discrete degree, and calculate and obtain a first leaf contamination degree and a second leaf contamination degree in combination with a preset first discrete degree and a preset second discrete degree; A pattern classification module 13, configured to perform averaging processing and leaf pattern classification on the first leaf image parameter array and the second leaf image parameter array to obtain a first leaf pattern array, and perform leaf pattern classification according to the first leaf contamination degree and the second leaf contamination degree to obtain a second leaf pattern array; A fertilization decision module 14, configured to process and obtain a leaf pattern according to the first leaf pattern array and the second leaf pattern array, and perform fertilization on the double-season rice.

[0057] Further, the leaf acquisition module 11 includes the following execution steps: At the stage when the tip of the second leaf from the top of the target double-season rice emerges, select multiple positions of the first leaf and the second leaf on the target double-season rice, where the first leaf is the third leaf from the top and the second leaf is the fourth leaf from the top; At multiple positions of the first leaf and the second leaf, respectively test and acquire SPAD parameters to obtain a first leaf image parameter array and a second leaf image parameter array.

[0058] Further, the contamination analysis module 12 includes the following execution steps: Perform image parameter discrete analysis and calculation on the first leaf image parameter array to obtain a first dispersion degree, where image parameter discrete analysis is performed by calculating the standard deviation of multiple leaf image parameters within the first leaf image parameter array; Perform image parameter discrete analysis and calculation on the second leaf image parameter array to obtain a second dispersion degree.

[0059] Further, the contamination analysis module 12 further includes the following execution steps: Obtain the first dispersion degree and the second dispersion degree of the same-family double-season rice within a preset time window in historical time to obtain a sample first dispersion degree set and a sample second dispersion degree set; Calculate and configure a preset first dispersion degree and a preset second dispersion degree according to the sample first dispersion degree set and the sample second dispersion degree set; Calculate the amplitudes by which the first dispersion degree and the second dispersion degree deviate from the preset first dispersion degree and the preset second dispersion degree respectively to obtain a first leaf contamination degree and a second leaf contamination degree.

[0060] Further, the pattern classification module 13 includes the following execution steps: Perform averaging calculation processing on the first leaf image parameter array and the second leaf image parameter array respectively to obtain a first leaf image parameter and a second leaf image parameter; Construct a first leaf pattern classifier, where the first leaf pattern classifier includes multiple first leaf pattern classification branches, the input feature of each first leaf pattern classification branch is the ratio of the first leaf image parameter to the second leaf image parameter, and the output feature is the first leaf pattern; Calculate the ratio of the first leaf image parameter to the second leaf image parameter, input it into the first leaf pattern classifier for integrated leaf pattern classification to obtain a first leaf pattern array; Construct a second leaf pattern classifier, where the second leaf pattern classifier includes multiple second leaf pattern classification branches, the input feature of each second leaf pattern classification branch is the ratio of the first leaf contamination degree to the second leaf contamination degree, and the output feature is the second leaf pattern; Calculate the ratio of the first leaf contamination degree to the second leaf contamination degree, perform integrated leaf pattern classification to obtain a second leaf pattern array.

[0061] Further, the pattern classification module 13 further includes the following execution steps: According to the historical cultivation data of homologous double-season rice, collect a set of sample image parameter ratios, and collect the leaf patterns of double-season rice under each sample image parameter ratio, and label to obtain the first set of sample leaf patterns, where the sample image parameter ratio is the ratio of the first leaf image parameter and the second leaf image parameter of homologous double-season rice; Divide the set of sample image parameter ratios and the first set of sample leaf patterns to obtain multiple sets of first-mode classification training data, where there is an intersection between every two sets of first-mode classification training data; Use machine learning to construct multiple first leaf pattern classification branches; Respectively use the multiple sets of first-mode classification training data to train, validate and test the multiple first leaf pattern classification branches to obtain a first leaf pattern classifier.

[0062] Further, the fertilization decision module 14 includes the following execution steps: In the first leaf pattern array and the second leaf pattern array, select the leaf pattern with the largest proportion of occurrence times to obtain a fused leaf pattern; Fertilize the double-season rice according to the fused leaf pattern.

[0063] It should be noted that in the above embodiments, the descriptions of the respective embodiments have their own emphases. For the parts not detailed in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0064] Those skilled in the art should understand that the embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0065] The present invention is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, and the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0066] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means that implement the function specified in one or more processes and / or blocks Figure 1 in the flowchart Figure 1 or blocks or multiple blocks

[0067] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable apparatus provide steps for implementing the function specified in one or more processes and / or blocks Figure 1 in the flowchart Figure 1 or blocks or multiple blocks

[0068] Although the preferred embodiments of the present invention have been described, additional changes and modifications can be made by those skilled in the art once they learn of the basic inventive concept

[0069] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the present invention and its equivalent technologies, the present invention is also intended to include these modifications and variations

Claims

1. A leaf image pattern classification method for double-season rice ear fertilization, characterized in that: The method comprises: In a preset time window of the target double-season rice, at multiple positions of the first leaf and the second leaf, an array of first leaf image parameters and an array of second leaf image parameters are acquired; Performing image parameter discrete analysis on the first leaf image parameter array and the second leaf image parameter array to obtain a first discreteness and a second discreteness, and calculating and obtaining a first leaf contamination degree and a second leaf contamination degree by combining a preset first discreteness and a preset second discreteness; Performing averaging processing and leaf pattern classification on the first leaf image parameter array and the second leaf image parameter array to obtain a first leaf pattern array, and performing leaf pattern classification according to the first leaf contamination degree and the second leaf contamination degree to obtain a second leaf pattern array; The leaf patterns are processed and obtained according to the first leaf pattern array and the second leaf pattern array, and double-season rice fertilization is performed.

2. The leaf image pattern classification method for double-season rice ear fertilization according to claim 1, characterized in that: In a preset time window of the target double-season rice, at multiple positions of the first leaf and the second leaf, the first leaf image parameter array and the second leaf image parameter array are acquired, including: At the tip-out stage of the second leaf of the target double-cropping rice, multiple positions of the first leaf and the second leaf are selected on the target double-cropping rice, wherein the first leaf is the third leaf and the second leaf is the fourth leaf; At multiple positions of the first leaf and the second leaf, SPAD parameters are tested and collected respectively to obtain a first leaf image parameter array and a second leaf image parameter array.

3. The leaf image pattern classification method for double-season rice ear fertilization according to claim 1, characterized in that: Performing image parameter discrete analysis on the first leaf image parameter array and the second leaf image parameter array to obtain a first discreteness and a second discreteness includes: Performing image parameter discrete analysis calculation on the first leaf image parameter array to obtain a first discreteness, wherein the image parameter discrete analysis is performed by calculating a standard deviation of a plurality of leaf image parameters in the first leaf image parameter array; Perform image parameter discrete analysis calculation on the second leaf image parameter array to obtain a second discrete degree.

4. The leaf image pattern classification method for double-season rice ear fertilization according to claim 1, characterized in that: Combining the preset first discreteness and the preset second discreteness, calculating and obtaining the first leaf contamination degree and the second leaf contamination degree, includes: Obtain the first discreteness and the second discreteness of the same family of double-season rice in a preset time window within the historical time, and obtain a sample first discreteness set and a sample second discreteness set; Calculate and configure a preset first discreteness and a preset second discreteness according to the sample first discreteness set and the sample second discreteness set; The amplitudes of the first discreteness and the second discreteness deviating from the preset first discreteness and the preset second discreteness are calculated respectively to obtain the first leaf contamination degree and the second leaf contamination degree.

5. The leaf image pattern classification method for double-season rice ear fertilization according to claim 1, characterized in that: The first leaf image parameter array and the second leaf image parameter array are averaged and leaf pattern classified to obtain a first leaf pattern array, and the leaf pattern is classified according to the first leaf contamination degree and the second leaf contamination degree to obtain a second leaf pattern array, including: Performing averaging calculation processing on the first leaf image parameter array and the second leaf image parameter array respectively to obtain first leaf image parameters and second leaf image parameters; Constructing a first leaf pattern classifier, wherein the first leaf pattern classifier comprises a plurality of first leaf pattern classification branches, an input feature of each first leaf pattern classification branch is a ratio of a first leaf image parameter to a second leaf image parameter, and an output feature is a first leaf pattern; Calculating a ratio of the first leaf image parameter to the second leaf image parameter, inputting the ratio into the first leaf pattern classifier for integrated leaf pattern classification, and obtaining a first leaf pattern array; Constructing a second leaf pattern classifier, wherein the second leaf pattern classifier comprises a plurality of second leaf pattern classification branches, an input feature of each second leaf pattern classification branch is a ratio of the first leaf contamination degree to the second leaf contamination degree, and an output feature is a second leaf pattern; The ratio of the first leaf contamination degree to the second leaf contamination degree is calculated, and integrated leaf pattern classification is performed to obtain a second leaf pattern array.

6. The leaf image pattern classification method for double-season rice ear fertilization according to claim 1, characterized in that: Construct the first leaf pattern classifier, including: According to the historical cultivation data of the same family double-season rice, a set of sample image parameter ratios is collected, and the leaf pattern of the double-season rice under each sample image parameter ratio is collected, and the sample first leaf pattern set is obtained by annotation, wherein the sample image parameter ratio is the ratio of the first leaf image parameter to the second leaf image parameter of the same family double-season rice; Dividing the sample image parameter ratio set and the sample first leaf pattern set to obtain a plurality of first pattern classification training data, wherein there is an intersection between every two first pattern classification training data; Using machine learning, multiple first leaf pattern classification branches are constructed; The plurality of first pattern classification training data are respectively used to train, verify and test the plurality of first leaf pattern classification branches to obtain a first leaf pattern classifier.

7. The leaf image pattern classification method for double-season rice ear fertilization according to claim 1, characterized in that: According to the first leaf pattern array and the second leaf pattern array, a fused leaf pattern is obtained by processing to perform double-season rice fertilization, comprising: In the first blade pattern array and the second blade pattern array, selecting the blade pattern with the largest number of occurrences to obtain a fused blade pattern; According to the fused leaf pattern, double-season rice fertilization is carried out.

8. A leaf image pattern classification system for double-season rice ear fertilization, characterized in that: A system for implementing the leaf image pattern classification method for double-season rice ear fertilization as claimed in any one of claims 1 to 7, comprising: A leaf acquisition module is used to acquire a first leaf image parameter array and a second leaf image parameter array at multiple positions of the first leaf and the second leaf in a preset time window of the target double-season rice; A pollution degree analysis module is used to perform image parameter discrete analysis on the first leaf image parameter array and the second leaf image parameter array to obtain a first discrete degree and a second discrete degree, and calculate the first leaf pollution degree and the second leaf pollution degree by combining the preset first discrete degree and the preset second discrete degree; a leaf pattern classification module, configured to perform averaging processing and leaf pattern classification on the first leaf image parameter array and the second leaf image parameter array to obtain a first leaf pattern array, and perform leaf pattern classification according to the first leaf contamination degree and the second leaf contamination degree to obtain a second leaf pattern array; The fertilization decision module is used to process and obtain the leaf pattern according to the first leaf pattern array and the second leaf pattern array, and perform double-season rice fertilization.

Images