Intelligent response type forage grass stress resistance and salt tolerance regulation and control system

Abstract

The invention discloses an intelligent response type forage grass stress resistance and salt tolerance regulation and control system, and the system comprises a machine vision module which is used for carrying out the saline-alkali stress recognition of a leaf image of awnless brome planted in saline-alkali soil through a trained deep learning model, and obtaining the saline-alkali stress level of the awnless brome; the metabonomics data module is used for performing metabolite analysis on nutritional organs of awnless brome by utilizing a chromatography-mass spectrometry technology and a random forest algorithm to obtain a key metabolite set of the awnless brome in the saline-alkali soil; the growth trend prediction module is used for analyzing the key metabolite set and the soil environment data set of the saline-alkali soil by using the constructed multivariable regression model to obtain a growth state prediction result of the awnless brome in the saline-alkali soil; and the soil regulation and control module is used for regulating a water and fertilizer management strategy of the saline-alkali soil according to the growth state prediction result and the saline-alkali stress level of the awnless brome. The stress resistance of the awnless brome in the saline-alkali soil can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent planting and salt resistance regulation of Bromus inermis Leyss. in saline-alkali soil, and relates to, but is not limited to, an intelligent response type pasture resistance and salt tolerance regulation system. BACKGROUND

[0002] Bromus inermis Leyss. is a kind of pasture with strong adaptability, which has certain potential in the planting of pasture in saline-alkali soil. However, due to the inhibitory effect of salt and alkaline components in the soil of saline-alkali soil on the growth of pasture, the yield and survival rate of Bromus inermis Leyss. are reduced. The traditional management method of pasture in saline-alkali soil relies on manual experience regulation, lacks real-time feedback mechanism and precise resource management, and often cannot effectively cope with adversity stress. Therefore, a more intelligent system is needed to realize the precise regulation of Bromus inermis Leyss. and improve its resistance in saline-alkali soil. SUMMARY

[0003] Based on the above problems, the present application provides an intelligent response type pasture resistance and salt tolerance regulation system, which aims to realize the accurate identification of the growth state of Bromus inermis Leyss. and the salt-alkali stress level in saline-alkali soil by fusing machine vision, metabolomics and growth prediction model, so as to dynamically regulate the water and fertilizer management strategy of saline-alkali soil by means of the identified growth state prediction result and salt-alkali stress level, and further improve the resistance of Bromus inermis Leyss. in saline-alkali adversity.

[0004] The technical scheme of the present application embodiment is as follows: An intelligent response type pasture resistance and salt tolerance regulation system, the system comprising: a machine vision module, configured to use a trained deep learning model to perform salt-alkali stress identification on leaf images of Bromus inermis Leyss. planted in saline-alkali soil, and obtain a salt-alkali stress level of Bromus inermis Leyss.; a metabolomics data module, configured to use chromatography-mass spectrometry technology and random forest algorithm to analyze metabolites of the vegetative organs of Bromus inermis Leyss., and obtain a key metabolite set of Bromus inermis Leyss. in saline-alkali soil; a growth trend prediction module, configured to use a constructed multivariate regression model to analyze the key metabolite set and a soil environment data set of saline-alkali soil, and obtain a growth state prediction result of Bromus inermis Leyss. in saline-alkali soil; a soil regulation module, configured to adjust the water and fertilizer management strategy of saline-alkali soil according to the growth state prediction result and the salt-alkali stress level of Bromus inermis Leyss.

[0005] In some embodiments, the system further comprises: The intelligent decision module is configured to collect soil state data of the saline-alkali land after the water and fertilizer management adjustment and growth response data of the brome, and optimize the water and fertilizer management strategy based on the growth response data and the soil state data of the brome by using a reinforcement learning algorithm to obtain an optimized water and fertilizer management strategy. The intelligent feedback module is configured to regulate the soil of the saline-alkali land based on the optimized water and fertilizer management strategy.

[0006] In some embodiments, the system further comprises a first model construction module, wherein the first model construction module comprises: The sample image acquisition unit is configured to acquire a sample image set, the sample image set comprising a plurality of leaf images of the brome and salt damage degree annotation information corresponding to each leaf image. The model improvement unit is configured to improve the YOLOv7 model by using an edge attention operator to generate an improved YOLOv7 model. The model training unit is configured to iteratively train the improved YOLOv7 model by using the sample image set until a trained deep learning model is obtained, wherein the trained deep learning model satisfies a loss of an output result.

[0007] In some embodiments, the model improvement unit is specifically configured to embed the edge attention operator after one or more basic convolution modules in a backbone network of the YOLOv7 model to generate the improved YOLOv7 model.

[0008] In some embodiments, the metabolomics data module comprises: The metabolite data acquisition unit is configured to perform metabolite analysis on the vegetative organs of the brome planted in the saline-alkali land and normal conditions by using a chromatography-mass spectrometry technique to obtain raw metabolite spectrum data of the brome. The metabolite screening unit is configured to screen high-variation metabolites with a variation coefficient greater than a preset variation threshold from the raw metabolite spectrum data, and screen a set of key salt-resistant metabolites related to salt-alkali stress response from the high-variation metabolites by using a random forest algorithm.

[0009] In some embodiments, the system further comprises: The soil detection module is configured to acquire a soil environment data set composed of the conductivity, salt concentration, temperature, humidity, and pH value of the saline-alkali land by using a soil sensor deployed in the saline-alkali land.

[0010] In some embodiments, the system further comprises a second model construction module, wherein the second model construction module comprises: The modeling data acquisition unit is configured to acquire historical soil data of the saline-alkali land and historical key metabolites of the brome to obtain a modeling data set. The variable screening unit is configured to screen, from the modeling data set, variables having a correlation higher than a preset correlation threshold with the growth condition of the A. sterilis by a principal component analysis method, to obtain a plurality of input feature variables. The model establishing unit is configured to establish a multiple linear regression model by taking the plurality of input feature variables as independent variables and taking the growth condition index of the A. sterilis as a dependent variable. The model fitting unit is configured to train and fit the multiple linear regression model by using the modeling data set, to obtain a constructed multiple variable regression model.

[0011] In some embodiments, the soil regulation module is specifically configured to execute a first water and fertilizer management strategy when the salinity stress level is severe and the growth state prediction result is growth recession, the first water and fertilizer management strategy including: increasing the application of phosphorus and potassium fertilizer and organic fertilizer to improve the soil, and performing quantitative irrigation by using micro-sprinkling irrigation; execute a second water and fertilizer management strategy when the salinity stress level is moderate and the growth state prediction result is growth stagnation, the second water and fertilizer management strategy including: increasing the application of calcium sulfate or sulfur powder to reduce the pH value of the soil, and simultaneously increasing the irrigation water amount to perform salt compression; execute a third water and fertilizer management strategy when the salinity stress level is slight and the growth state prediction result is growth increase, the third water and fertilizer management strategy including: balanced fertilization according to the normal fertilizer requirement, and performing water-saving and soil-moisture conservation by using drip irrigation.

[0012] The technical scheme provided by the embodiments of the present application has at least the following beneficial effects: The intelligent response type pasture stress tolerance and salt resistance regulation system disclosed by the embodiments of the present application comprises: a machine vision module, which is used for identifying salt and alkali stress of leaf images of Bromus inermis planted in a saline-alkali land by using a trained deep learning model, and obtaining a salt and alkali stress level of the Bromus inermis, that is, the machine vision module is used to convert the morphological diagnosis relying on artificial experience in traditional technology into a relatively objective identification signal, i.e., the salt and alkali stress level, and the deep learning model can capture early leaf subtle discoloration, texture change and other features that cannot be detected by the human eye, so that the accuracy of the identified salt and alkali stress level can be further improved; a metabolomics data module, which is used for analyzing metabolites of the vegetative organs of the Bromus inermis by using a chromatography-mass spectrometry combined technology and a random forest algorithm, and obtaining a key metabolite set of the Bromus inermis in the saline-alkali land, that is, the chromatography-mass spectrometry combined technology and the random forest algorithm are used to convert the response of the Bromus inermis to stress to a molecular function level, and the key metabolite set of the internal molecular mechanism of the stress response can be accurately revealed from the physiological level; a growth trend prediction module, which is used for analyzing the key metabolite set and a soil environment data set of the saline-alkali land by using a constructed multivariate regression model, and obtaining a growth state prediction result of the Bromus inermis in the saline-alkali land, that is, the multivariate regression model is used to quantitatively predict the growth trajectory (growth, stagnation or decline) of the Bromus inermis in a future period of time; and a soil regulation module, which is used for adjusting a water and fertilizer management strategy of the saline-alkali land according to the growth state prediction result and the salt and alkali stress level of the Bromus inermis, wherein the “stress level” (present situation severity) and “growth prediction” (future trend) are comprehensively considered to form a multi-dimensional decision matrix, so that the strategy matching is more fine and scientific. That is to say, through the cooperative work of the four core modules, a “perception-diagnosis-prediction-decision-execution” integrated precision agriculture closed loop is constructed, that is, through the fusion of machine vision, metabolomics and growth prediction models, the growth state and the salt and alkali stress level of the Bromus inermis in the saline-alkali land are accurately identified, so that the water and fertilizer management strategy of the saline-alkali land is dynamically regulated by means of the identified growth state prediction result and the salt and alkali stress level, and then the stress resistance of the Bromus inermis in the saline-alkali adversity is improved.

[0013] It should be understood that the foregoing general description and the following detailed description are only exemplary and explanatory, but not limiting the technical solutions provided by the embodiments of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0014] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings. Figure 1A first intelligent response type pasture grass stress resistance and salt tolerance regulation system provided in the embodiments of the present application is shown in the following composition diagram: Figure 2 A second intelligent response type pasture grass stress resistance and salt tolerance regulation system provided in the embodiments of the present application is shown in the following composition diagram: Figure 3 A third intelligent response type pasture grass stress resistance and salt tolerance regulation system provided in the embodiments of the present application is shown in the following composition diagram: Figure 4 A fourth intelligent response type pasture grass stress resistance and salt tolerance regulation system provided in the embodiments of the present application is shown in the following composition diagram: Figure 5 A fifth intelligent response type pasture grass stress resistance and salt tolerance regulation system provided in the embodiments of the present application is shown in the following composition diagram. DETAILED DESCRIPTION

[0015] To make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the following will be combined with the drawings in the embodiments of the present application to make a clear and complete description of the technical solutions in the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application, but not all the embodiments. The following embodiments are used to illustrate the present application, but not to limit the scope 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 scope of the present application.

[0016] In the following description, “some embodiments” are related to a subset of all possible embodiments, but it can be understood that “some embodiments” can be the same subset or different subset of all possible embodiments, and can be combined with each other without conflict.

[0017] It should be noted that the terms “first\second\third” involved in the embodiments of the present application are only to distinguish similar objects, and do not represent the specific order of the objects. It can be understood that “first\second\third” can be interchanged with specific order or sequence as allowed, so that the embodiments of the present application described here can be implemented in an order other than that illustrated or described here.

[0018] Those skilled in the art can understand that, unless otherwise defined, all terms (including technical terms and scientific terms) used herein have the same meaning as that generally understood by those of ordinary skill in the art to which the embodiments of the present application belong. It should also be understood that terms such as those defined in general dictionaries should be understood as having meanings consistent with those in the prior art, and should not be interpreted with idealized or overly formal meanings unless specifically defined as such.

[0019] Embodiment one Referring toFigure 1 Fig. 1 shows a schematic diagram of a first intelligent response type pasture grass stress tolerance and salt resistance regulation system 100 provided by an embodiment of the present application, wherein the intelligent response type pasture grass stress tolerance and salt resistance regulation system 100 comprises: a machine vision module 110, configured to use a trained deep learning model to perform salt-alkali stress identification on leaf images of Bromus inermis planted in a saline-alkali land, and obtain a salt-alkali stress level of the Bromus inermis.

[0020] a metabolomics data module 120, configured to use a chromatography-mass spectrometry technology and a random forest algorithm to perform metabolite analysis on vegetative organs of the Bromus inermis, and obtain a key metabolite set of the Bromus inermis in the saline-alkali land.

[0021] a growth trend prediction module 130, configured to use a constructed multivariate regression model to analyze the key metabolite set and a soil environment data set of the saline-alkali land, and obtain a growth state prediction result of the Bromus inermis in the saline-alkali land.

[0022] a soil regulation module 140, configured to adjust a water and fertilizer management strategy of the saline-alkali land according to the growth state prediction result and the salt-alkali stress level of the Bromus inermis.

[0023] In some embodiments of the present application, the intelligent response type pasture grass stress tolerance and salt resistance regulation system 100 can further comprise a data acquisition module (not shown in the figure), which can use a high-resolution camera or a drone device to regularly take leaf images of the Bromus inermis planted in the saline-alkali land, so as to obtain the leaf images of the Bromus inermis planted in the saline-alkali land. The leaf images can present information such as leaf withering, leaf yellowing, and leaf lesion. Figure 1 In some embodiments of the present application, the intelligent response type pasture grass stress tolerance and salt resistance regulation system 100 can further comprise an intelligent decision module and an intelligent feedback module (not shown in the figure):

[0024] Figure 1 the intelligent decision module is configured to acquire soil state data of the saline-alkali land after the water and fertilizer management adjustment and growth response data of the Bromus inermis, and use a reinforcement learning algorithm to optimize the water and fertilizer management strategy according to the growth response data and the soil state data of the Bromus inermis, so as to obtain an optimized water and fertilizer management strategy. the intelligent feedback module is configured to regulate the soil of the saline-alkali land according to the optimized water and fertilizer management strategy.

[0025]

[0026] ​​In some embodiments of the present application, an intelligent decision module is adopted. Firstly, real-time state feedback of the soil ecosystem of the saline-alkali land (such as dynamic changes of salt concentration, pH value, nutrient content, etc.) and the physiological and phenotypic response indexes of the brome (such as biomass growth, leaf area index, metabolite spectrum change, etc.) are acquired. Then, a reinforcement learning algorithm is adopted to formalize the water and fertilizer management process into a sequential decision problem. In this framework, the soil state data collected above is defined as the environment state, different water and fertilizer management schemes (i.e. the water and fertilizer management strategy of the saline-alkali land) are defined as the action of the agent, and the growth response data of the brome is quantified as the reward signal. Through continuous interaction with the environment (i.e. the saline-alkali farmland system), the reinforcement learning algorithm can autonomously learn and discover the water and fertilizer management strategy that can obtain the best long-term and cumulative growth reward in a complex and dynamic saline-alkali stress environment, so as to dynamically generate an optimized strategy that surpasses the initial experience, is more accurate and more efficient, i.e. the optimized water and fertilizer management strategy.

[0027] Correspondingly, an intelligent feedback module is adopted to directly convert the optimized water and fertilizer management strategy from a theoretical decision into specific farmland operation instructions and accurately implement them in the target saline-alkali land. In this way, a complete closed loop from "perception-decision-learning" to "execution" is completed, and active, directional and adaptive regulation of the soil environment of the saline-alkali land is realized.

[0028] In some embodiments of the present application, the intelligent decision module can further analyze the change trend of the collected soil state data and the growth response data of the brome, and correspondingly determine a new regulation strategy, as shown in Table 1 below. Among them, the data combination mode composed of the change trend of the electrical conductivity (EC) in the soil state data, the change trend of the metabolites (proline, betaine), lesion, etc. in the growth response data corresponds to the adjustment of the new water and fertilizer management strategy.

[0029] Table 1 Water and fertilizer management strategy corresponding to different data combination modes In Table 1, "↑" means increase, and "↓" means decrease.

[0030] In this way, by collecting the environmental data (soil state data) and crop feedback data (growth response data) after water and fertilizer management, and using a reinforcement learning algorithm for closed-loop optimization, a leap from static strategy execution to dynamic strategy evolution is realized. It can adapt to the complex and changeable saline-alkali land environment, and actively discover the optimal water and fertilizer formula and irrigation scheme that can maximize the growth benefit of the brome under a specific soil state, so as to realize accurate and efficient improvement of the saline-alkali soil and adaptive and accurate support for the growth of the brome, and finally significantly improve the productivity and resource utilization efficiency of the saline-alkali agricultural ecosystem.

[0031] In some embodiments of the present application, the machine vision module 110 can employ the trained deep learning model to perform salt-alkali stress identification on multiple leaf images of the B. israelica planted in saline-alkali land, obtain the salt-alkali stress level corresponding to each leaf image in the multiple leaf images, and integrate (e.g., take the average value, etc.) the salt-alkali stress levels corresponding to the multiple leaf images to obtain the salt-alkali stress level of the B. israelica planted in the saline-alkali land.

[0032] In some embodiments of the present application, the salt-alkali stress level can be represented using heavy, moderate, and light, etc., or can be represented using I, II, and III, and the present application does not make any limitation thereto.

[0033] In some embodiments of the present application, regarding the construction of the trained deep learning model, in the intelligent response type pasture stress tolerance and salt regulation system 100, the first model construction module 150 shown in the figure can be employed to implement, wherein the first model construction module 150 can include: Figure 2 The sample image acquisition unit 1501 is configured to acquire a sample image set, and the sample image set includes multiple leaf images of the B. israelica and salt damage degree annotation information corresponding to each leaf image.

[0034] The model improvement unit 1502 is configured to employ an edge attention operator to improve the YOLOv7 model to generate an improved YOLOv7 model.

[0035] The model training unit 1503 is configured to employ the sample image set to iteratively train the improved YOLOv7 model until the trained deep learning model with a loss of output result satisfying a convergence condition is obtained.

[0036] In some embodiments of the present application, the sample image set can include a data set of 10,000 leaf images of the B. israelica; wherein 6,000 are normal leaf images, and 4,000 are leaf images of different salt damage degrees. Here, the LabelImg tool can be used to label the lesion and yellowing area in the 4,000 leaf images of different salt damage degrees.

[0037] It should be noted that the salt damage degree annotation information in the multiple leaf images of the B. israelica and the salt damage degree annotation information corresponding to each leaf image included in the sample image set can be severe, moderate, mild, and none, and the present application does not make any specific limitation thereto.

[0038] In some embodiments of the present application, the model improvement unit 1502 is specifically configured to embed the edge attention operator after one or more basic convolution modules in the backbone network of the YOLOv7 model to generate the improved YOLOv7 model. ​

[0039] In some embodiments of the present application, the YOLOv7 model also continues the sight framework of the YOLO series in the prior art: backbone network → neck network → head network. The present application can be embedded in an edge attention operator after each or any or multiple basic convolution modules of the backbone network of the YOLOv7 model, such as the CBS module (which specifically includes: Conv + BatchNorm + SiLU activation function).

[0040] Here, the edge attention operator is a differentiable calculation process that calculates the edge information of the input feature map and converts it into spatial attention weights, thereby re-weighting the original features. The core goal is to enable the feature extraction process of the network to have edge perception ability, thereby paying more attention to object boundaries and improving the modeling of geometric structures.

[0041] In some embodiments of the present application, the edge attention operator can be an image gradient-based edge attention operator, such as the Sobel operator, the Scharr operator, etc., or a learnable edge attention operator, such as an operator composed of multiple convolution layers, etc., which is not limited by the present application.

[0042] Here, because the salt injury symptoms in the leaf image of the salt-tolerant bird cherrygrass planted in saline-alkali soil are usually concentrated in the edge area of the leaf, the present application embeds an edge attention operator in the YOLOv7 model to enhance the salt injury feature extraction ability of the edge area of the leaf. In this way, by introducing an edge attention operator with an edge attention mechanism into the YOLOv7 model, the improved YOLOv7 model generated can focus on the salt injury reaction area of the leaf edge, thereby reducing the interference of the background and non-diseased tissues, and further improving the recognition accuracy of the salt injury features.

[0043] In some embodiments of the present application, the model training unit 1503 is configured to use the sample image set as a training set to iteratively train the improved YOLOv7 model until a trained deep learning model is obtained whose output result loss satisfies the convergence condition; wherein during the training of the improved YOLOv7 model, the Adam optimizer can be used, the learning rate is initially set to 0.001, and the training is performed for 300 epochs.

[0044] In this way, through the joint cooperation of the sample image acquisition unit 1501, the model improvement unit 1502, and the model training unit 1503 inside the first model construction module 150, the perception and extraction ability of the trained deep learning model for the salt damage features of the leaf edge of the no awn brome can be effectively improved, so as to realize more accurate identification and grading of the salt damage degree, and the generated trained deep learning model has higher detection accuracy and robustness in application, providing a reliable technical tool for subsequent growth monitoring and accurate evaluation of no awn brome planted in saline-alkali soil.

[0045] In some embodiments of the present application, the metabolomics data module 120 is configured to use chromatography-mass spectrometry and random forest algorithm to perform metabolite analysis on the vegetative organs, such as roots and / or leaves, of no awn brome planted in saline-alkali soil, to obtain a set of key metabolites of no awn brome in saline-alkali soil.

[0046] It should be noted that the vegetative organs, such as roots and / or leaves, of no awn brome planted in saline-alkali soil can be obtained by relevant technical personnel using sterile gloves and sealed bags to collect leaf and / or root samples of no awn brome planted in saline-alkali soil. Here, after obtaining the samples of the vegetative organs, the samples need to be immediately placed in a portable refrigerator (the temperature can be set to 4°C) and delivered to the laboratory within 2 hours. After receiving the samples, the laboratory first washes and liquid nitrogen freezes the samples, then grinds them into powder, uses a methanol-water solution (the corresponding volume ratio can be 8:2) to perform ultrasonic extraction, and after centrifugation (the specific operation data can be 12000 rpm for 10 minutes), the supernatant is obtained to obtain samples for chromatography-mass spectrometry analysis.

[0047] In some embodiments of the present application, the chromatography-mass spectrometry can be gas chromatography-mass spectrometry (GC-MS) or liquid chromatography-mass spectrometry (LC-MS).

[0048] In some embodiments of the present application, as shown in Figure 3 The metabolomics data module 120 includes: A metabolite data acquisition unit 1201 is configured to use chromatography-mass spectrometry to perform metabolite analysis on the vegetative organs of no awn brome planted in saline-alkali soil and normal conditions, to obtain the original metabolite spectrum data of no awn brome; The metabolite screening unit 1202 is used to screen highly variable metabolites with a coefficient of variation greater than a preset variation threshold from the original metabolite profile data, and to use a random forest algorithm to screen a set of key salt-resistant metabolites related to salt-alkali stress response from the highly variable metabolites.

[0049] In some embodiments of this application, the metabolic data acquisition unit 1201 can analyze metabolite components using GC-MS or LC-MS to monitor salt damage reactions in awnless bromegrass, particularly changes in antioxidants, amino acids, sugars, etc. Here, the raw metabolite profile data of awnless bromegrass can be represented by an identification list, such as including: proline, malic acid, betaine, etc., along with their respective retention times and mass-to-charge ratios.

[0050] In some embodiments of this application, the metabolite screening unit 1202 can use the Principal Component Analysis (PCA) method and the Python Scikit-learn library to screen 300 metabolites with a coefficient of variation greater than 20% from the original metabolite profile data, and further screen a set of key salt-tolerant metabolites composed of 10 key salt-tolerant metabolites such as proline and betaine through a random forest regression model.

[0051] Thus, firstly, global metabolite data of awnless bromeliads grown in saline-alkali land are obtained using chromatography-mass spectrometry (GC-MS). Then, based on the coefficient of variation, a threshold is set to screen out highly variable metabolites with poor intragroup repeatability and high potential susceptibility to environmental disturbances. Finally, the high-precision variable screening capability of the random forest algorithm is used to identify the key metabolites that contribute the most to the classification of saline-alkali stress from the highly variable metabolites. In other words, the metabolite profile information of awnless bromeliads is obtained using GC-MS or LC-MS, and the concentration of metabolites is correlated with the plant growth status under saline-alkali stress using the random forest algorithm, thereby constructing a high-confidence set of salt-tolerant metabolite markers, namely the key salt-tolerant metabolite set.

[0052] In some embodiments of this application, the intelligent responsive forage stress and salt tolerance regulation system 100 may further include: a soil detection module 160, such as... Figure 4 As shown, where: The soil detection module 160 is used to acquire a soil environmental dataset consisting of electrical conductivity, salt concentration, temperature, humidity and pH value of the saline-alkali land by using soil sensors deployed in the saline-alkali land.

[0053] In some embodiments of the present application, the soil detection module 160 monitors the pH value, EC, salt concentration, temperature and other parameters of the saline-alkali soil in real time through the deployed soil sensors, to ensure accurate feedback of the soil environment data. Here, after obtaining the initial soil environment data, the median filtering algorithm is used, with a window size of 5, to remove abnormal data points, to obtain a soil environment data set with higher accuracy.

[0054] Correspondingly, the growth trend prediction module 130 first receives the soil environment data set delivered by the soil detection module 160 and the key metabolite set delivered by the metabolomics data module 120; then, using the constructed multivariate regression model, the key metabolite set and the soil environment data set are analyzed to obtain the growth state prediction result of Bromus inermis in saline-alkali land.

[0055] In some embodiments of the present application, each key metabolite in the key metabolite set M and each soil environment data in the soil environment data set E can be substituted into the constructed multivariate regression model shown in formula (1) below, to correspondingly obtain the growth state prediction result of Bromus inermis in saline-alkali land : ; formula (1); wherein n is the number of key metabolites included in the key metabolite set, and m is the number of soil environment data included in the soil environment data set; is the i-th key metabolite in the key metabolite set; is the j-th soil environment data in the soil environment data set; and is a regression coefficient, which is determined in the construction stage (training stage) of the constructed multivariate regression model, is an error term.

[0056] In some embodiments of the present application, the constructed multivariate regression model can be used to predict the growth trend of crops (Bromus inermis) under different saline-alkali stress conditions.

[0057] Here, the growth state prediction result of Bromus inermis in saline-alkali land can be represented using: growth recession, growth stagnation and growth increase.

[0058] In this way, by jointly inputting the key salt-resistant metabolite set and the soil environment data into the constructed multivariate regression model, quantitative prediction of the growth state of Bromus inermis in saline-alkali land is achieved. The fusion of intrinsic physiological response of plants and external environmental stress factors not only significantly improves the accuracy and reliability of growth prediction, but also reveals the deep regulatory relationship among metabolites, environment and phenotype, providing data-driven decision basis for crop growth monitoring and control strategy formulation in precision agriculture in saline-alkali land.

[0059] In some embodiments of the present application, the constructed multivariate regression model can be implemented using the second model construction module 170 in the intelligent response type pasture grass stress resistance and salt tolerance regulation system 100, as shown in the figure, wherein the second model construction module 170 comprises: Figure 5 The modeling data acquisition unit 1701 is configured to acquire historical soil data of saline-alkali land and historical key metabolites of B. israelica, to obtain a modeling data set.

[0060] In some embodiments of the present application, the soil physicochemical property data of saline-alkali land (such as pH value, conductivity, sodium ion concentration, etc.) characterizing the soil environment stress factor, and the B. israelica key endogenous metabolite spectrum data (such as proline, betaine, etc. osmotic adjustment substance content) characterizing the internal physiological response state of B. israelica are used as the multidimensional modeling data set for subsequent machine learning, to lay a data foundation for revealing the environment-plant interaction relationship.

[0061] The variable screening unit 1702 is configured to screen, from the modeling data set, variables with a correlation higher than a preset correlation threshold with the growth condition of B. israelica by a principal component analysis method, to obtain a plurality of input feature variables.

[0062] In some embodiments of the present application, the variable screening unit 1702 converts the original high-dimensional and possibly collinear soil and metabolite variables in the modeling data set to a set of new independent comprehensive variables (principal components) by using PCA. By analyzing the correlation of these principal components with the growth condition indicators (such as biomass, plant height) of B. israelica, the contribution of each original variable to the explanation of plant growth variation is inversely traced and quantified. Finally, according to a preset statistical significance threshold, the key environmental and physiological feature variables with a strong driving relationship with the growth condition are accurately screened, so as to realize the optimization and dimensionality reduction of the feature space.

[0063] The model establishment unit 1703 is configured to establish a multivariate linear regression model with the plurality of input feature variables as independent variables and the growth condition indicators of B. israelica as dependent variables.

[0064] The model fitting unit 1704 is configured to train and fit the multivariate linear regression model by using the modeling data set, to obtain the constructed multivariate regression model.

[0065] ​In some embodiments of the present application, the model establishing unit 1703 and the model fitting unit 1704 work together to define the mathematical relationship between the environmental factors and the phenotype of the brome by using the screened key soil and metabolite characteristics as independent variables (predictors) and the core growth indicators of the brome as dependent variables (response variables) in the form of multiple linear regression, that is, to establish a multiple linear regression model. The constructed modeling data set is used to statistically estimate and fit the parameters (including the weight coefficients of each variable and the intercept term) in the multiple linear regression model, and finally a parameter-determined multivariate regression model with actual prediction ability is generated, which can be used to infer the growth status of new samples.

[0066] In this way, by integrating and screening the saline-alkali environment data and plant metabolomics data, the key factors driving the growth of brome are effectively extracted, and on this basis, a quantitative multiple linear regression model is established. The constructed multivariate regression model not only can accurately predict the growth status of brome under specific saline-alkali stress conditions, but also the parameters of the constructed multivariate regression model have clear biological interpretation significance, which can clearly reveal the influence degree of different soil properties and the internal metabolic process of brome on the final growth performance, thereby providing data-driven decision basis and theoretical insight for precise improvement of saline-alkali land and cultivation management of stress-resistant crops.

[0067] In some embodiments of the present application, the soil regulation module 140 is configured to adjust the water and fertilizer management strategy, i.e., the amount of fertilizer and irrigation, of the saline-alkali land according to the growth status prediction result and the saline-alkali stress level of the brome.

[0068] In some embodiments of the present application, the soil regulation module 140 can be specifically configured to perform the following operations: When the saline-alkali stress level is severe and the growth status prediction result is growth recession, a first water and fertilizer management strategy is executed, which includes increasing the application of phosphorus and potassium fertilizer and organic fertilizer to improve the soil, and using micro-sprinkling irrigation for fixed amount irrigation.

[0069] When the saline-alkali stress level is moderate and the growth status prediction result is growth stagnation, a second water and fertilizer management strategy is executed, which includes increasing the application of calcium sulfate or sulfur powder to reduce the soil pH value, and simultaneously increasing the irrigation water amount to press salt.

[0070] When the saline-alkali stress level is mild and the growth status prediction result is growth increase, a third water and fertilizer management strategy is executed, which includes balanced fertilization according to the normal fertilizer requirement, and using drip irrigation for water saving and soil conservation.

[0071] In some embodiments, the saline-alkali stress level and the growth state prediction result can be converted into numerical values respectively, such as: the value corresponding to the severe saline-alkali stress level is 0.5; the value corresponding to the moderate saline-alkali stress level is 0.3; the value corresponding to the mild saline-alkali stress level is 0.2. At the same time, the value corresponding to the growth recession growth state prediction result is 0.5; the value corresponding to the growth stagnation growth state prediction result is 0.3; the value corresponding to the growth increase growth state prediction result is 0.2. The value corresponding to the saline-alkali stress level and the value corresponding to the growth state prediction result are added to obtain a comprehensive value, and the corresponding control response can be implemented according to the comprehensive value, and the comprehensive value is different, the corresponding warning level of the saline-alkali soil is also different; as shown in Table 2: Table 2 Different comprehensive values correspond to different control responses In some embodiments of the present application, the soil control module 140 can also obtain the current growth stage of the B. inermis, and determine the corresponding saline-alkali condition in the set parameter library according to the growth stage, so as to control the corresponding water and fertilizer management strategy by means of the determined saline-alkali condition, the growth state prediction result and the saline-alkali stress level of the B. inermis. Here, the set parameter library is an empirical value parameter library for storing multi-dimensional control parameters (tolerance, water and fertilizer demand, etc.) of different grass varieties (including: B. inermis) under different growth stages and saline-alkali conditions. That is, the control strategy can be automatically adjusted according to the variety characteristics and the growth stage to adapt to different types of saline-alkali soil. In this way, according to the real-time data query parameter library, the irrigation amount, the fertilizer amount, etc. are automatically adjusted through a specific association relationship, which can realize the precise control of water and fertilizer management of the saline-alkali soil.

[0072] In this way, by linking the saline-alkali stress level and the growth state prediction result of the B. inermis, an intelligent water and fertilizer management scheme with precision and situational awareness is given. It can dynamically match and automatically trigger the optimal customized strategy for different combinations of stress and growth situations such as severe stress-growth recession, moderate stress-growth stagnation, and mild stress-growth increase, thereby realizing the transformation from passive response to active intervention, from uniform management to differentiated strategy, effectively improving the utilization efficiency of water and fertilizer resources, and precisely promoting the healthy growth of B. inermis and the sustainable improvement of the soil environment under saline-alkali adversity.

[0073] The intelligent response type pasture stress tolerance and salt tolerance regulation system disclosed by the embodiments of the present application is described as follows: first, a machine vision module is used to identify the salt stress of the leaf images of Bromus inermis planted in saline-alkali land by using a trained deep learning model, so as to obtain the salt stress level of Bromus inermis; second, a metabolomics data module is used to analyze the metabolites of the vegetative organs of Bromus inermis by using a chromatography-mass spectrometry technology and a random forest algorithm, so as to obtain a key metabolite set of Bromus inermis in the saline-alkali land; third, a growth trend prediction module is used to analyze the key metabolite set and a soil environment data set of the saline-alkali land by using a constructed multivariate regression model, so as to obtain a growth state prediction result of Bromus inermis in the saline-alkali land; and finally, a soil regulation module is used to adjust the water and fertilizer management strategy of the saline-alkali land according to the growth state prediction result and the salt stress level of Bromus inermis. In this way, through the collaborative work of the four core modules, a "perception-diagnosis-prediction-decision-execution" integrated precision agriculture closed loop is constructed, that is, through the fusion of machine vision, metabolomics and growth prediction model, the growth state and the salt stress level of Bromus inermis in the saline-alkali land are accurately identified, so as to dynamically regulate the water and fertilizer management strategy of the saline-alkali land by means of the identified growth state prediction result and the salt stress level, and thus the stress tolerance of Bromus inermis in the saline-alkali environment is improved.

[0074] As a person skilled in the art, it should be known that although the prior art provides various ways to improve the salt tolerance and growth environment of crops, there are certain limitations in chemical, physical or biological methods. For example, chemical methods may cause environmental pollution and dependence problems, physical methods are complex to operate and have limited effect, and biological methods face technical barriers, high cost and poor environmental adaptability. Therefore, the prior art has not fundamentally solved the problem of efficient and sustainable regulation of salt-tolerant crops.

[0075] Therefore, the embodiments of the present application provide an intelligent response type pasture stress tolerance and salt tolerance regulation system, that is, an intelligent response type Bromus inermis stress tolerance and salt tolerance regulation system based on machine vision and metabolomics data. By constructing a dynamic evaluation model of salt stress (a constructed multivariate regression model and a trained deep learning model), the growth state, salt stress response and soil environment of Bromus inermis can be monitored and analyzed in real time, the release of nutrients such as nitrogen, phosphorus and potassium in the saline-alkali land can be accurately adjusted on demand, and a regulation scheme suitable for different soil types can be provided, so as to improve the yield of Bromus inermis planted in the saline-alkali land, ensure high yield, high survival rate and high-quality growth of Bromus inermis. In the present scheme, the following innovations are achieved: 1. Multi-module collaborative architecture: A collaborative system architecture composed of machine vision module, metabolomics data module, growth trend prediction module, and soil regulation module. Each module is connected to the data bus through a specific communication method, realizing the full-process functions of data collection, processing, analysis, and regulation for the planting of Bromus inermis in saline-alkali land, which is different from the traditional system of single module independent operation.

[0076] 2. Multi-source data fusion analysis: The salt stress level of Bromus inermis output by the machine vision image data, the key metabolite set of Bromus inermis in saline-alkali land output by the metabolomics data module, and the soil environment data set of saline-alkali land are fused and analyzed. Specifically, through deep learning analysis of images (such as a deep learning model constructed by an improved YOLOv7 model to detect leaf salt damage symptoms), PCA and random forest regression screening of key salt-resistant metabolites, and filtering processing of soil data, the comprehensive evaluation of Bromus inermis salt damage condition is realized, breaking through the limitations of traditional single data type analysis.

[0077] 3. Dynamic evaluation model construction: Based on the key metabolite set output by the metabolomics data module and the soil environment data set closely related to plant growth characteristics, a multi-variable regression model is established by means of multiple linear regression model, which is a dynamic evaluation model of Bromus inermis in saline-alkali environment. The regression algorithm is used to correlate metabolite concentration, soil environment data, and plant growth state under salt stress, and the plant growth trend is predicted by the multi-variable regression model, providing data support for precise regulation.

[0078] In addition, the embodiments of the present application can further realize real-time feedback and multi-level early warning, that is, a mechanism for real-time feedback and salt damage three-level early warning is realized by combining multi-source data. When the salt stress level output by the machine vision module and the growth state prediction result predicted by the growth trend prediction module based on the key metabolite set and the soil environment data meet the preset threshold at the same time, the early warning is triggered, the system automatically queries the salt-resistant regulation parameter library and generates the corresponding regulation instruction (adjusting irrigation, fertilization, leaching, etc.). Here, the visual data, metabolomics analysis results, and soil monitoring data can be combined to respond in real time and automatically adjust the growth environment of Bromus inermis. For example, when the soil salt content is too high, the irrigation or fertilization amount is automatically adjusted to reduce the accumulation of salt and promote the healthy growth of Bromus inermis.

[0079] In summary, the intelligent response type pasture stress tolerance and salt resistance regulation system provided by the embodiments of the present application can provide an efficient and intelligent solution for the planting of pasture (such as Bromus inermis) in saline-alkali land, which can significantly improve the growth and yield of Bromus inermis and other pastures in saline-alkali soil, while reducing environmental pollution and resource waste, thereby not only improving agricultural production efficiency, but also providing technical support for promoting ecological agriculture and sustainable agriculture development.

[0080] That is, the intelligent response type pasture stress tolerance and salt tolerance regulation system provided by the embodiments of the present application has the following improvements compared with the prior art. 1. Improved salt and alkali resistance: through deep learning algorithm, the growth state of the brome is automatically identified and analyzed, it is judged whether it is subjected to salt and alkali stress, and the change trend of plant growth can be captured in real time, abnormal phenomena can be timely warned, and through analysis of metabolites in the brome, the influence of salt and alkali stress on the physiological state of the brome is revealed, and a dynamic evaluation model is established to predict the influence of salt damage on the brome. In this way, through accurate salt damage evaluation and intelligent regulation, the stress resistance of the brome in the saline-alkali soil is improved, and the growth, survival rate and yield of the brome are improved.

[0081] 2. Real-time monitoring and accurate regulation: it can respond to environmental changes in real time, automatically adjust relevant parameters according to the growth state of the brome and the change of the saline-alkali soil, that is, dynamically adjust the supply of fertilizer and water, which not only can optimize the growth environment of the brome in the saline-alkali soil to improve the growth efficiency of the brome in the saline-alkali soil, but also can avoid resource waste.

[0082] 3. Multi-dimensional data fusion: through the fusion of machine vision, metabolomics and soil monitoring data, the growth state of the brome is comprehensively evaluated, and the scientificity and accuracy of regulation are improved.

[0083] 4. Strong adaptability: it can automatically adjust the management strategy according to different types of saline-alkali soil and pasture varieties, and has strong adaptability and wide application scenarios.

[0084] The present application can effectively reduce pollution to soil and environment, avoid problems such as soil acidification and ecological imbalance caused by traditional chemical modifiers, and provide a more green and environmentally friendly crop improvement technology. At the same time, the present application can provide a simple, long-lasting and effective crop improvement method, which can simplify the operation while improving the stress resistance of crops, especially the adaptability in harsh environments. And for the problems of complex implementation, high cost and limited application range of the biological method (such as genetic engineering, gene sequencing selection breeding and strain growth promotion) in the prior art for improving the salt resistance and growth environment of crops, the present application can provide a solution that simplifies the biological improvement process and has wide adaptability, that is, it can effectively improve the performance of crops in various environments while reducing technical barriers and implementation costs.

[0085] It is to be understood that the terminology "one embodiment" or "an embodiment" used throughout this specification means that a particular feature, structure or characteristic described is included in at least one embodiment of the application. Therefore, appearances of the phrases "in one embodiment" or "in an embodiment" in various places throughout this specification are not necessarily referring to the same embodiment. Furthermore, the particular features, structures or characteristics can be implemented in any suitably-arranged

[0086] It has to be understood that the terms "including", "containing", or any other similar term are intended to be recited in a non-exclusive manner, such that a process, a method, an article or an apparatus that comprises a list of elements does not necessarily comprise those only elements but can also comprise other elements not expressly listed or inherent to such a process, method, article or apparatus. An element specified by the phrase "comprising a" does not exclude the presence of additional identical elements in the process, method, article or apparatus that includes that element.

[0087] In several embodiments provided in the present application, it should be understood that the disclosed system can be implemented in other manners. The above described system embodiments are merely schematic. For example, the division of the units is only a logical function division, and there can be another division manner in actual implementation. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed coupling, or direct coupling or communication connection between the components can be indirect coupling or communication connection through some interfaces, devices or units, and can be electrical, mechanical or other forms.

[0088] The units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units; they can be located in one place, or distributed on a plurality of network units; and some or all of the units can be selected according to actual needs to achieve the purposes of the embodiments of the present application.

[0089] In addition, each functional unit in the embodiments of the present application can be integrated into one processing unit, or each unit can be a separate physical unit, or two or more units can be integrated into one unit; and the integrated unit can be implemented in the form of hardware or hardware plus software function unit.

[0090] Alternatively, the above-mentioned integrated units of the present application, if implemented in the form of software functional modules and sold or used as independent products, can also be stored in a computer-readable storage medium. Based on such understanding, the technical solutions of the embodiments of the present application can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable the equipment automatic test line to perform all or part of the methods described in the embodiments of the present application. The foregoing storage medium includes mobile storage devices, ROM, magnetic discs or optical discs, and various media that can store program codes.

[0091] The systems disclosed in the several system embodiments provided by the present application can be combined arbitrarily without conflict to obtain new system embodiments.

[0092] The features disclosed in the several system embodiments provided by the present application can be combined arbitrarily without conflict to obtain new system embodiments.

[0093] The above is only an implementation of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A smart responsive forage stress and salt tolerance regulation system, characterized in that, The system includes: The machine vision module is used to identify salt and alkali stress in leaf images of awnless bromeliads planted in saline-alkali land using a trained deep learning model, and to obtain the salt and alkali stress level of the awnless bromeliads. The metabolomics data module is used to analyze the metabolites of the vegetative organs of awnless brome using chromatography-mass spectrometry and random forest algorithm, and to obtain the key metabolite set of awnless brome in saline-alkali land. The growth trend prediction module is used to analyze the key metabolite set and soil environment dataset of saline-alkali land using the constructed multivariate regression model to obtain the prediction results of the growth status of awnless brome in saline-alkali land. The soil regulation module is used to adjust the water and fertilizer management strategy for saline-alkali land based on the predicted growth status and the salt and alkali stress level of awnless bromeliad.

2. The system according to claim 1, characterized in that, The system also includes: The intelligent decision-making module is used to collect soil state data of saline-alkali land after water and fertilizer management adjustments, as well as growth response data of awnless bromegrass. Based on the growth response data of awnless bromegrass and soil state data, the module uses reinforcement learning algorithm to optimize the water and fertilizer management strategy, resulting in an optimized water and fertilizer management strategy. The intelligent feedback module is used to regulate the soil in saline-alkali land based on the optimized water and fertilizer management strategy.

3. The system according to claim 1 or 2, characterized in that, The system further includes: a first model construction module, wherein the first model construction module includes: The sample image acquisition unit is used to acquire a sample image set, which includes: multiple leaf images of awnless bromeliads and salt damage labeling information corresponding to each leaf image; The model improvement unit is used to improve the YOLOv7 model by employing the edge attention operator, thereby generating an improved YOLOv7 model. The model training unit is used to iteratively train the improved YOLOv7 model using a set of sample images until a trained deep learning model is obtained whose output loss satisfies the convergence condition.

4. The system according to claim 3, characterized in that, The model improvement unit is specifically used to embed the edge attention operator into one or more basic convolutional modules in the backbone network of the YOLOv7 model to generate an improved YOLOv7 model.

5. The system according to claim 1, characterized in that, The metabolomics data module includes: The metabolic data acquisition unit is used to perform metabolite analysis on the vegetative organs of awnless bromegrass grown under saline-alkali land and normal conditions using chromatography-mass spectrometry (GC-MS) technology, and to obtain raw metabolite profile data of awnless bromegrass. The metabolite screening unit is used to screen highly variable metabolites with a coefficient of variation greater than a preset variation threshold from the original metabolite profile data, and to use the random forest algorithm to screen a set of key salt-resistant metabolites related to salt-alkali stress response from the highly variable metabolites.

6. The system according to claim 1, characterized in that, The system also includes: The soil testing module is used to acquire a soil environmental dataset consisting of electrical conductivity, salt concentration, temperature, humidity, and pH value of the saline-alkali land using soil sensors deployed in the land.

7. The system according to claim 1, characterized in that, The system further includes: a second model construction module, wherein the second model construction module includes: The modeling data acquisition unit is used to acquire historical soil data of saline-alkali land and historical key metabolites of awnless bromegrass to obtain a modeling dataset. The variable selection unit is used to select variables from the modeling dataset that have a higher correlation with the growth status of awnless bromeliad than a preset correlation threshold using the principal component analysis method, thereby obtaining multiple input feature variables. The model building unit is used to establish a multiple linear regression model with multiple input feature variables as independent variables and the growth status index of awnless bromeliad as the dependent variable. The model fitting unit is used to train and fit the multiple linear regression model using the modeling dataset to obtain the constructed multivariate regression model.

8. The system according to claim 1, characterized in that, The soil regulation module is specifically used to implement the following water and fertilizer management strategies when the salt and alkali stress level is severe and the growth status prediction result is growth decline: First water and fertilizer management strategy includes: increasing the application of phosphorus and potassium fertilizers and organic fertilizers to improve the soil, and using micro-sprinkler irrigation for fixed-rate irrigation; when the salt and alkali stress level is moderate and the growth status prediction result is growth stagnation, second water and fertilizer management strategy is implemented, second water and fertilizer management strategy includes: increasing the application of calcium sulfate or sulfur powder to lower the soil pH value, and simultaneously increasing the irrigation water volume to suppress salt; when the salt and alkali stress level is mild and the growth status prediction result is growth growth, third water and fertilizer management strategy is implemented, third water and fertilizer management strategy includes: applying fertilizer evenly according to normal fertilizer requirements, and using drip irrigation to conserve water and moisture.

Images