Poa annua rootstock vegetative propagation growth monitoring method based on clustering processing

Abstract

The invention discloses a method for monitoring vegetative propagation growth of roots and stems of bluegrass based on clustering processing. The method comprises the steps of multi-source monitoring data acquisition, multi-source data preprocessing, growth state monitoring clustering model construction, clustering structure optimization and layered clustering growth state monitoring. The invention relates to the technical field of plant propagation data processing, in particular to a method for monitoring vegetative propagation growth of roots and stems of bluegrass based on clustering processing, which innovatively proposes a bi-clustering mechanism based on a main-sub layered structure, and significantly improves the accuracy, stability and agricultural availability of growth state monitoring. A misclassification point recognition mechanism and a misclassification triggering mechanism are designed, and a cluster adjustment factor is introduced, so that the adaptability and the expression ability of a complex plant growth monitoring behavior mode are enhanced; an optimization algorithm and an optimal parameter combination of a growth state monitoring clustering model are improved by adopting a mirror surface reverse mechanism and a sine and cosine function method, and a clustering structure is optimized, so that the accuracy of a growth state monitoring result is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of plant propagation data processing, and specifically refers to a method for monitoring the asexual propagation growth of Poa pratensis rhizomes based on clustering processing. Background Art

[0002] The method for monitoring the asexual propagation growth of Poa pratensis rhizomes based on clustering processing collects multi-source data on the asexual propagation growth of Poa pratensis rhizomes, and uses a clustering algorithm to group and classify the growth states of Poa pratensis rhizomes, so as to realize the automatic monitoring and dynamic recognition of the growth trend, distribution pattern and population structure during its asexual propagation process.

[0003] However, in the traditional method for monitoring the asexual propagation growth of rhizomes, there is a technical problem that only a single-layer clustering strategy is adopted, which is difficult to simultaneously depict the complex association between the physiological state of plants and external incentives, resulting in insufficient expression ability of the multi-dimensional growth mode of plants and inaccurate classification of the growth monitoring state; there are technical problems of fuzzy clustering boundaries, inaccurate misclassification discrimination, lagging response of the clustering structure and unstable model update in the existing clustering algorithms applicable to plant growth state monitoring, thus leading to inaccurate growth state recognition results; there is a technical problem that the clustering structure parameters in the existing clustering algorithms applicable to growth state monitoring are set unreasonably, resulting in unstable clustering recognition results and insufficient clustering accuracy. Summary of the Invention

[0004] In view of the above situation, to overcome the defects of the prior art, the present invention provides a method for monitoring the asexual reproduction growth of Poa pratensis rhizomes based on clustering processing. Aiming at the technical problems existing in the traditional method for monitoring the asexual reproduction growth of rhizomes, such as only adopting a single-layer clustering strategy, it is difficult to simultaneously depict the complex association between plant physiological states and external incentives, resulting in insufficient expression ability of plant multi-dimensional growth patterns and inaccurate classification of growth monitoring states. This solution innovatively proposes a dual-clustering mechanism based on a master-slave hierarchical structure, and performs two-stage clustering on plant ontology physiological data and environmental operation incentive data respectively, constructing a two-layer label system of main states and sub-states. It can effectively distinguish the subdivision growth response patterns triggered by external changes of plant individuals in similar physiological states, and has higher interpretability of growth monitoring states and label expression accuracy. It realizes the structured hierarchical analysis of multi-source data in the process of plant asexual reproduction, significantly improving the accuracy, stability and agricultural usability of growth state monitoring. Aiming at the technical problems existing in the existing clustering algorithms applicable to plant growth state monitoring, such as fuzzy clustering boundaries, inaccurate misclassification discrimination, lagging response of clustering structures and unstable model updates, resulting in inaccurate growth state recognition results, this solution innovatively introduces a misclassified point recognition mechanism based on cluster radius, constructs a reasonable misclassification judgment boundary, and improves the accuracy and rationality of misclassification recognition. Design a cross-cycle cumulative misclassification trigger mechanism, introduce historical misclassification point statistics and a structure self-adaptive adjustment strategy driven by the Xie-Beni index, and enhance the model's response ability to complex time-varying data structures. Introduce a method for incrementally updating the cluster center driven by a cluster adjustment factor, effectively improving the numerical stability and dynamic adaptability of the model update process. Through the above improvements, the growth state monitoring clustering model constructed by the present invention has strong dynamic self-adaptability, structural adjustment flexibility and state recognition accuracy, can significantly improve the accuracy and temporal stability of growth state classification results, and enhance the adaptability and expression ability to complex plant growth monitoring behavior patterns. Aiming at the technical problems existing in the existing clustering algorithms applicable to growth state monitoring, such as unreasonable setting of clustering structure parameters, resulting in unstable clustering recognition results and insufficient clustering accuracy, this solution uses a mirror reflection mechanism and a sine-cosine function method to improve and optimize the algorithm, globally optimize the key parameters in the clustering model, and obtain the optimal parameter combination, thereby effectively improving the structural expression ability and classification accuracy of the clustering model, and enhancing the accuracy and applicability of growth state monitoring results.

[0005] The technical solution adopted by the present invention is as follows: The method for monitoring the asexual reproduction growth of Poa pratensis rhizomes based on clustering processing provided by the present invention includes the following steps:

[0006] Step S1: Acquisition of multi-source monitoring data;

[0007] Step S2: Preprocessing of multi-source data;

[0008] Step S3: Construct a growth status monitoring clustering model;

[0009] Step S4: Optimize the clustering structure;

[0010] Step S5: Hierarchical clustering for growth status monitoring.

[0011] Furthermore, in step S1, the acquisition of multi-source monitoring data is specifically to collect data information on the growth area of the rhizomes of Poa annua to obtain the original data for asexual reproduction growth monitoring; the original data for asexual reproduction growth monitoring includes historical growth monitoring data and real-time growth monitoring data, and both the historical growth monitoring data and the real-time growth monitoring data include plant physiological data, plant environmental data, and operation behavior data for each monitoring point; the historical growth monitoring data also includes the growth monitoring status.

[0012] Furthermore, in step S2, the preprocessing of multi-source data is specifically to perform data optimization processing on the data to obtain the optimized data for asexual reproduction growth monitoring; it includes the following steps:

[0013] Step S21: Data cleaning processing, specifically to perform missing value processing, outlier processing, and duplicate data processing on the original data;

[0014] Step S22: Data normalization processing, specifically to perform standardization processing on the data using the Min-Max normalization method;

[0015] Step S23: Construction of time-series samples, specifically to slice the continuous monitoring data of each monitoring point according to a set 24-hour time window to construct sample data with consistent time, and use a sliding window mechanism to set the time overlap interval to 4 hours to generate multiple time-series samples with overlapping relationships;

[0016] Step S24: Reference label association, specifically to use the growth monitoring status in the historical growth monitoring data as a reference label for the main clustering operation, which is only used for cluster label selection and does not participate in the clustering data processing process.

[0017] Furthermore, in step S3, the construction of the growth status monitoring clustering model is specifically to establish a clustering model by improving the fuzzy C-means clustering algorithm through a misclassified point recognition mechanism, a misclassification trigger mechanism, and the introduction of a cluster adjustment factor to obtain the growth status monitoring clustering model, including the following steps:

[0018] Step S31: Initialize clustering, specifically to use the initialized number of clustering clusters c and the fuzzy index m to perform the standard fuzzy C-means clustering algorithm on the sample data in the historical growth monitoring data to obtain the set of cluster centers and the membership degree matrix of each data point to each cluster;

[0019] Step S32: Calculate the distances between cluster centers. Specifically, for any two cluster centers in the set of cluster centers and , calculate the distance between them using the Euclidean distance;

[0020] Step S33: Calculate the distances from data points to cluster centers. Specifically, for each data point , calculate the distance from it to each cluster center using the Euclidean distance;

[0021] Step S34: Misclassified point detection. Specifically, by calculating the minimum distance from each data point to all cluster centers and comparing it with the radius of the cluster, determine whether the data point is a misclassified point. If the minimum value of the distances from the data point to all cluster centers is greater than the cluster radius, then the data point is considered a misclassified point; the formula used is as follows:

[0022] ;

[0023] ;

[0024] In the formula, represents the misclassification function, represents the number of clusters in the t-th period, represents the j-th cluster, where j represents the cluster index, represents any data point in the j-th cluster, represents to the Euclidean distance, represents to the Euclidean distance, represents the membership degree of the current period data point to the j-th cluster, represents the radius of the j-th cluster, represents the membership degree value threshold parameter, represents the centroid of the j-th cluster in the current period, represents the i-th sample data point, represents the data in the t-th period;

[0025] Step S35: Adjust the clustering structure. Specifically, if the proportion of the number of misclassified points accumulated up to the current period t in the total amount of data in the current period exceeds the misclassification ratio threshold parameter , then based on the Xie-Beni index as the structure evaluation index, evaluate the structural strength of each cluster number scheme within the range of candidate cluster numbers, select the cluster number with the optimal structure, and re-execute the standard fuzzy C-means clustering algorithm based on the updated number of clusters. Otherwise, update the current cluster centers;

[0026] Step S36: Update the current cluster center, specifically by calculating the temporary cluster center for the current cycle and calculating the cluster adjustment factor , and perform the update of the current cluster center; the formula used is as follows:

[0027] ;

[0028] ;

[0029] ;

[0030] In the formula, represents the membership value of the data point to cluster j in the current cycle, represents the data of the th cycle, represents the center of cluster j in the next cycle, represents the membership value of the data point to cluster j in the previous cycle;

[0031] Step S37: Cluster deletion mechanism, specifically for each clustering cluster existing in the current cycle t, by analyzing its activity changes in consecutive T cycles, determine whether the cluster is in an inactive state, so as to determine whether it is an obsolete cluster. If , then cluster j is considered an obsolete cluster and needs to perform a cluster deletion operation, and the cluster deletion constraint conditions need to be satisfied when performing the cluster deletion operation; among them, represents the fuzzy cardinality of cluster j in the current cycle, that is, the sum of the membership degrees of all data points in the cluster, represents the fuzzy cardinality of cluster j at the time of cycle , and T represents the number of cycles for cluster activity evaluation;

[0032] Step S38: Obtain the final clustering result, specifically if the change amount of all cluster centers is lower than the preset convergence threshold or the current iteration cycle number has reached the preset maximum cycle number , then terminate the clustering iteration and output the final clustering result. The clustering result is specifically to allocate all sample data points to the corresponding clusters according to their final maximum membership degrees, forming multiple cluster sets, and each cluster represents a different growth state under the asexual reproduction of Poa pratensis rhizomes.

[0033] Furthermore, in step S4, the clustering structure optimization specifically includes the following steps:

[0034] Step S41: Construct a search individual position vector, specifically by encoding the parameters of the growth state monitoring clustering model into a search individual position vector; the parameters of the growth state monitoring clustering model include the fuzzy index m, the membership degree threshold and misclassification ratio threshold ;

[0035] Step S42: Design a fitness function, specifically, establish a fitness function through a weighted combination of the average silhouette coefficient and the label jump rate of the current clustering structure;

[0036] Step S43: Obtain the optimal parameter combination of the model, specifically, globally optimize the parameters of the growth state monitoring clustering model through an improved optimization algorithm to obtain the optimal parameter combination of the growth state monitoring clustering model;

[0037] Step S431: Initialize the individuals in the search population. Specifically, randomly initialize the current positions of N search individuals, then optimize the individual positions using the mirror reflection mechanism to obtain the reflected positions of the individuals, and calculate the fitness values F of the individuals in the population through the fitness function i , sort the individuals from the best to the worst according to the fitness values, and select the top N individuals according to the current positions and reflected positions of the individuals to form a new search population. The formula used is as follows:

[0038] ;

[0039] In the formula, represents the reflected position of the i-th individual in the d-th dimension, represents the current position of the i-th individual in the d-th dimension, and represent the minimum and maximum values of the search space in dimension d respectively, represents the scaling factor, which is a constant;

[0040] Step S432: Update the positions of the search individuals. Specifically, update the positions of the search individuals through the sine-cosine function and the non-linear decreasing search. The formula used is as follows:

[0041] ;

[0042] In the formula, represents the position of the i-th individual in the d-th dimension in the -th generation population, represents the position of the i-th individual in the d-th dimension in the -th generation population, represents the position of the optimal individual in the population, represents the number of iterations, represents the maximum number of iterations, represents the search adjustment factor, and , represents the warning value, represents the safety threshold, represents [0, 2 Random numbers within the range, representing random numbers within the range [0, 1], where e represents the base of the natural logarithm;

[0043] Step S433: The search is completed. Specifically, when the fitness value F of the search individual i is higher than the fitness threshold and the maximum number of iterations is reached, the search is terminated and the global optimal position of the search individual is obtained. The global optimal position of the search individual specifically refers to the optimal parameter combination of the growth state monitoring clustering model;

[0044] Step S44: Optimize the clustering parameters. Specifically, adjust the parameters of the growth state monitoring clustering model according to the optimal parameter combination of the growth state monitoring clustering model to obtain the optimal growth state monitoring clustering model.

[0045] Further, in step S5, the hierarchical clustering growth state monitoring specifically identifies multiple types of growth states of bluegrass through the main - sub - clustering structure, forms a two - layer label system, and obtains the growth state monitoring result; it includes the following steps:

[0046] Step S51: The main state recognition stage. Specifically, take the plant physiological data in the real - time growth monitoring data as input data, input it into the optimal growth state monitoring clustering model, perform the first - stage main clustering operation, generate multiple main clustering clusters. In each clustering cluster, according to the associated historical growth monitoring state labels, count the label occurrence frequencies, and select the label with the highest frequency as the main state label of the cluster to obtain the main state recognition result;

[0047] Step S52: The sub - state recognition stage, which is used to analyze the sub - states formed by the external incentives behind the main state and identify the sub - growth response patterns; specifically, it includes the following steps:

[0048] Step S521: Obtain sub - clustering clusters. Specifically, according to the plant physiological data of each main clustering cluster, select its corresponding plant environment data and operation behavior data as the input data for the second - stage clustering, input it into the optimal growth state monitoring clustering model, and perform the second - stage sub - clustering operation to generate multiple sub - clustering clusters;

[0049] Step S522: Output the sub - state recognition result. Specifically, perform statistical processing on the data of each sub - clustering cluster, calculate the data mean, standard deviation, and maximum value to form the feature representation vector of the cluster. Subsequently, construct a set of predefined sub - state labels, each sub - state label consisting of multiple feature parameter interval ranges. Compare the feature representation vector of each sub - clustering cluster with all sub - state labels using cosine similarity, and select the label with the highest similarity as the sub - state label of the cluster to obtain the sub - state recognition result;

[0050] Step S53: Output of the growth status monitoring result, specifically, based on the main status recognition result and the sub-status recognition result, combine them according to the corresponding relationship of the samples within the cluster, and output the double-layer growth status corresponding to each monitoring point to obtain the growth status monitoring result.

[0051] The beneficial effects achieved by the present invention using the above solution are as follows:

[0052] (1) Aiming at the technical problems in the traditional rhizome asexual reproduction growth monitoring method, such as only using a single-layer clustering strategy, it is difficult to simultaneously depict the complex association between plant physiological states and external incentives, resulting in insufficient expression ability of plant multi-dimensional growth patterns and inaccurate classification of growth monitoring states. This solution innovatively proposes a double-clustering mechanism based on the master-slave hierarchical structure, performs two-stage clustering on plant ontological physiological data and environmental operation incentive data respectively, constructs a double-layer label system of main status and sub-status, can effectively distinguish the detailed growth response patterns caused by external changes of plant individuals in similar physiological states, has higher growth monitoring state interpretability and label expression accuracy, realizes the structured hierarchical analysis of multi-source data in the process of plant asexual reproduction, and significantly improves the accuracy, stability and agricultural usability of growth status monitoring.

[0053] (2) Aiming at the technical problems in the existing clustering algorithms applicable to plant growth status monitoring, such as fuzzy clustering boundaries, inaccurate misclassification discrimination, lagging response of clustering structure and unstable model update, resulting in inaccurate growth status recognition results. This solution innovatively introduces a misclassified point recognition mechanism based on cluster radius, constructs a reasonable misclassification judgment boundary, and improves the accuracy and rationality of misclassification recognition; designs a cross-cycle cumulative misclassification trigger mechanism, introduces historical misclassified point statistics and a structure self-adaptive adjustment strategy driven by the Xie-Beni index, and enhances the model's response ability to complex time-varying data structures; introduces a cluster center incremental update method driven by a cluster adjustment factor, effectively improving the numerical stability and dynamic adaptability of the model update process; through the above improvements, the growth status monitoring clustering model constructed by the present invention has strong dynamic self-adaptability, structural adjustment flexibility and state recognition accuracy, can significantly improve the accuracy and temporal stability of the growth status classification result, and enhance the adaptability and expression ability to complex plant growth monitoring behavior patterns.

[0054] (3) Aiming at the technical problems in the existing clustering algorithms applicable to growth status monitoring, such as unreasonable setting of clustering structure parameters, resulting in unstable clustering recognition results and insufficient clustering accuracy. This solution uses a mirror reflection mechanism and a sine-cosine function method to improve and optimize the algorithm, globally optimize the key parameters in the clustering model, and obtain the optimal parameter combination, thereby effectively improving the structural expression ability and classification accuracy of the clustering model, and enhancing the accuracy and applicability of the growth status monitoring result. Description of the Drawings

[0055] Figure 1 It is a schematic flow chart of the method for monitoring the vegetative propagation growth of Poa pratensis rhizomes based on clustering processing provided by the present invention;

[0056] Figure 2 It is a schematic flow chart of step S2;

[0057] Figure 3 It is a schematic flow chart of step S3;

[0058] Figure 4 It is a schematic flow chart of step S4;

[0059] Figure 5 It is a schematic flow chart of step S43;

[0060] Figure 6 It is a schematic flow chart of step S5;

[0061] The accompanying drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention, and do not constitute a limitation to the present invention. Detailed implementation manners

[0062] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying 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 of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0063] In the description of the present invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc. indicating the orientation or positional relationship are based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the system or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation to the present invention.

[0064] Embodiment 1, refer to Figure 1 , the technical solutions adopted by the present invention are as follows: The method for monitoring the vegetative propagation growth of Poa pratensis rhizomes based on clustering processing provided by the present invention includes the following steps:

[0065] Step S1: Acquisition of multi-source monitoring data, specifically, by collecting data information, the original data for monitoring vegetative propagation growth is obtained;

[0066] Step S2: Preprocessing of multi-source data, specifically, performing data cleaning, data normalization, time-series sample construction, and reference label association on the data to obtain optimized data for asexual reproduction growth monitoring;

[0067] Step S3: Constructing a growth state monitoring clustering model for constructing a dynamic and adaptive growth state clustering model to achieve effective classification of multiple types of growth states; specifically, establishing a clustering model by improving the fuzzy C-means clustering algorithm through a misclassification point recognition mechanism, a misclassification trigger mechanism, and introducing a cluster adjustment factor to obtain a growth state monitoring clustering model;

[0068] Step S4: Optimization of the clustering structure, specifically, based on an improved optimization algorithm, constructing a search vector for key model parameters, combining a multi-objective fitness function that fuses the average silhouette coefficient and the label jump rate, completing the initialization of the search population, position update, and global search, finally obtaining an optimal combination of model parameters and using it to update the clustering model structure to obtain an optimal growth state monitoring clustering model;

[0069] Step S5: Hierarchical clustering growth state monitoring for performing structured clustering analysis on the optimized data for asexual reproduction growth monitoring to identify different types of growth states; specifically, by constructing a two-layer clustering structure that combines main state recognition and sub-state recognition, first performing main clustering with plant physiological data as the input to obtain the main state recognition result, then performing sub-clustering based on the environmental and operation data of the main cluster, and completing label assignment through template matching to generate the sub-state recognition result; finally, jointly outputting the main state and sub-state recognition results to obtain the growth state monitoring result.

[0070] Example 2, refer to Figure 1 , this example is based on the above example. In step S1, the acquisition of multi-source monitoring data is specifically to collect data information on the growth area of Poa pratensis rhizomes by deploying a sensor system and a management operation record system to obtain raw data for asexual reproduction growth monitoring; the raw data for asexual reproduction growth monitoring includes historical growth monitoring data and real-time growth monitoring data, and both the historical growth monitoring data and the real-time growth monitoring data include plant physiological data, plant environmental data, and operation behavior data for each monitoring point; the historical growth monitoring data also includes the growth monitoring status;

[0071] The monitoring points are determined by dividing the target planting area into multiple 1m×1m spatial sub-areas and combining the plant distribution density, soil environment heterogeneity, and management operation zoning settings;

[0072] The plant physiological data includes rhizome volume, root resistance value, root zone concentration, root zone concentration, root zone conductivity, and root temperature;

[0073] The plant environmental data includes soil temperature, soil humidity, soil oxygen content, soil pH value, and soil conductivity;

[0074] The operation behavior data includes fertilizer type, fertilizer dosage, irrigation frequency, and irrigation water volume;

[0075] The growth monitoring status includes active growth status, stagnant status, degradation status, abnormal expansion status, and steady state status.

[0076] Example 3, refer to Figure 1 and Figure 2 Based on the above example, the multi-source data preprocessing is used to preprocess the original data of asexual reproduction growth monitoring. Specifically, it performs data cleaning, data normalization, time series sample construction, and reference label association on the data to obtain optimized data for asexual reproduction growth monitoring; it includes the following steps:

[0077] Step S21: Data cleaning, which is used to improve the quality and effectiveness of the data. Specifically, it performs missing value processing, outlier processing, and duplicate data processing on the original data;

[0078] The missing value processing is specifically to fill in the missing values in continuous sampling by interpolating the previous and subsequent time points through the time series linear interpolation method;

[0079] The outlier processing is specifically to automatically identify the upper and lower bounds based on the interquartile range through the box plot method to remove outliers;

[0080] The duplicate data processing is specifically to quickly screen the duplicate data set through the hash check deduplication method;

[0081] Step S22: Data normalization, which is used to standardize the original data with different dimensions and value ranges. Specifically, it performs standardization processing on the data using the Min-Max normalization method;

[0082] Step S23: Time series sample construction, which is used to convert continuous original monitoring data into samples with a unified time scale and structured format. Specifically, it slices the continuous monitoring data of each monitoring point according to a set 24-hour time window to construct sample data with consistent time, and uses a sliding window mechanism to set the time overlap interval to 4 hours to generate multiple time series samples with overlapping relationships;

[0083] Step S24: Reference label association, which is used to provide semantic mapping support for the clustering result. Specifically, it uses the growth monitoring status in the historical growth monitoring data as the reference label for the main clustering operation, which is only used for cluster label selection and does not participate in the clustering data processing process.

[0084] Example 4, refer toFigure 1 and Figure 3 This embodiment is based on the above embodiment. In step S3, the growth state monitoring clustering model is constructed, which is a dynamic clustering model constructed for multi-source monitoring data in the process of rhizome asexual reproduction of Poa annua, and can effectively divide different growth states. Specifically, the clustering model is established by improving the fuzzy C-means clustering algorithm through a misclassification point recognition mechanism, a misclassification trigger mechanism and introducing a cluster adjustment factor, and the growth state monitoring clustering model is obtained, including the following steps:

[0085] Step S31: Initialize clustering. Specifically, use the initialized number of clustering clusters c and the fuzzy index m to execute the standard fuzzy C-means clustering algorithm on the sample data in the historical growth monitoring data, and obtain the set of cluster centers and the membership degree matrix of each data point to each cluster.

[0086] Step S32: Calculate the distance between cluster centers. Specifically, for any two cluster centers and in the set of cluster centers, calculate the distance between them using the Euclidean distance. The formula used is as follows:

[0087] ;

[0088] In the formula, represents the Euclidean distance calculation function, represents the centroid of cluster j in the current cycle, and also represents the center of cluster j; represents the centroid of cluster k in the current cycle, and also represents the center of cluster k; represents the Euclidean norm;

[0089] Step S33: Calculate the distance from the data point to the cluster center. Specifically, for each data point , calculate its distance to each cluster center using the Euclidean distance. The formula used is as follows:

[0090] ;

[0091] In the formula, represents the i-th sample data point, represents the data in the t-th cycle;

[0092] Step S34: Detect misclassified points. Specifically, by calculating the minimum distance from each data point to all cluster centers and comparing it with the radius of the cluster, determine whether the data point is a misclassified point. If the minimum value of the distances from the data point to all cluster centers is greater than the cluster radius, it is considered that the data point is a misclassified point. The formula used is as follows:

[0093] ;

[0094] ;

[0095] Wherein, represents the misclassification function. If represents the sample data point is a misclassified point; otherwise, it represents the sample data point is correctly classified; represents the number of clusters in the t-th cycle, represents the j-th cluster, where j represents the cluster index, represents any data point in the j-th cluster, represents to the Euclidean distance, represents to the Euclidean distance, represents the membership degree of the current cycle data point to the j-th cluster, represents the radius of the j-th cluster, represents the membership degree value threshold parameter, with a range of ;

[0096] Step S35: Adjust the clustering structure. Specifically, if the proportion of the number of misclassified points accumulated up to the current cycle t in the total amount of data in the current cycle exceeds the misclassification ratio threshold parameter , then based on the Xie-Beni index as the structure evaluation index, evaluate the structural strength of each cluster number scheme within the range of candidate cluster numbers, select the cluster number with the optimal structure, and re-execute the standard fuzzy C-means clustering algorithm based on the updated number of clusters; otherwise, update the current cluster center. The formula used is as follows:

[0097] ;

[0098] Wherein, represents the proportion of the number of misclassified points accumulated up to the current cycle t in the total amount of data in the current cycle, represents the cluster number with the optimal structure, represents the set upper limit of the number of clusters, with a value of 5, represents the calculation function of the Xie-Beni index, which is used to evaluate the effectiveness of the clustering result and measure the compactness within the cluster and the separation between clusters, represents selecting the cluster number value that minimizes the evaluation index from the set of candidate cluster numbers;

[0099] Step S36: Update the current cluster center. Specifically, calculate the temporary cluster center of the current cycle and calculate the cluster adjustment factor , and update the current cluster center. The formula used is as follows:

[0100] ;

[0101] ;

[0102] ;

[0103] wherein, represents the membership value of the data point in the current cycle to cluster j, represents the th cycle of data, represents the center of cluster j in the next cycle, represents the membership value of the data point in the previous cycle to cluster j;

[0104] Step S37: Cluster deletion mechanism, specifically for each clustering cluster existing in the current cycle t, by analyzing its activity changes in consecutive T cycles, to determine whether the cluster is in an inactive state, so as to determine whether it is an obsolete cluster. If , then it is considered that cluster j is an obsolete cluster and needs to perform a cluster deletion operation. When performing the cluster deletion operation, it is necessary to meet the cluster deletion constraint conditions. The cluster deletion constraint conditions include a minimum cluster number constraint and a priority deletion constraint. The minimum cluster number constraint is specifically that after deleting a cluster, the remaining number of clusters shall not be less than two; the priority deletion constraint is specifically that if multiple clusters simultaneously meet the deletion conditions, the cluster with the smallest change in fuzzy cardinality shall be deleted first; where represents the fuzzy cardinality of cluster j in the current cycle, that is, the sum of the membership degrees of all data points in the cluster, represents the fuzzy cardinality of cluster j at the time of cycle , and T represents the number of cycles for cluster activity evaluation;

[0105] Step S38: Obtain the final clustering result. Specifically, if the change amount of all cluster centers is lower than a preset convergence threshold or the current iteration cycle number has reached a preset maximum cycle number , then terminate the clustering iteration and output the final clustering result. The clustering result is specifically to allocate all sample data points to the corresponding clusters according to their final maximum membership degrees, forming multiple cluster sets, and each cluster represents a different growth state under the asexual reproduction of Poa pratensis rhizomes.

[0106] By performing the above operations, aiming at the technical problems existing in the existing clustering algorithms applicable to plant growth status monitoring, such as fuzzy clustering boundaries, inaccurate misclassification discrimination, lagging response of clustering structures, and unstable model updates, which lead to inaccurate growth status recognition results, this solution innovatively introduces a misclassified point recognition mechanism based on cluster radius, constructs a reasonable misclassification judgment boundary, and improves the accuracy and rationality of misclassification recognition; designs a cross-cycle cumulative misclassification trigger mechanism, introduces historical misclassification point statistics and a structure self-adaptive adjustment strategy driven by the Xie-Beni index, and enhances the model's response ability to complex time-varying data structures; introduces a cluster center incremental update method driven by a cluster adjustment factor, effectively improving the numerical stability and dynamic adaptability of the model update process; through the above improvements, the growth status monitoring clustering model constructed by the present invention has strong dynamic self-adaptability, flexible structure adjustment, and state recognition accuracy, can significantly improve the accuracy and temporal stability of growth status classification results, and enhance the adaptability and expression ability to complex plant growth monitoring behavior patterns.

[0107] Example Five. Refer to Figure 1 and Figure 4 , based on the above embodiment, in step S4, the clustering structure is optimized by using an improved intelligent optimization algorithm to globally search and self-adaptively adjust the key parameters in the growth status monitoring clustering model, improving the structural performance and time stability of the clustering model in the recognition of complex plant growth behaviors; specifically including the following steps:

[0108] Step S41: Construct a search individual position vector, specifically by encoding the parameters of the growth status monitoring clustering model into a search individual position vector; the parameters of the growth status monitoring clustering model include the fuzzy index m, the membership threshold and the misclassification ratio threshold .

[0109] Step S42: Design a fitness function to construct a multi-objective fitness evaluation function required for clustering structure optimization, comprehensively measuring the performance of the current clustering structure in two dimensions of spatial compactness and time stability. Specifically, a fitness function is established through a weighted combination of the average silhouette coefficient and the label jump rate of the current clustering structure; the formula used is as follows:

[0110] .

[0111] In the formula, represents the position of the i-th search individual, represents the fitness function, represents the average silhouette coefficient calculated from the current clustering result, used to evaluate the geometric structure compactness of the clustering, Indicates the label jump rate of the corresponding clustering result in the time series, which is used to evaluate the stability of clustering in the time series dimension. Indicates the weight factor of the clustering compactness index, and the value range , Indicates the weight factor of the time series stability index. ;

[0112] Step S43: Obtain the optimal parameter combination of the model. Specifically, globally optimize the parameters of the growth state monitoring clustering model through an improved optimization algorithm to obtain the optimal parameter combination of the growth state monitoring clustering model.

[0113] Step S431: Initialize the search population individuals. Specifically, generate the current positions of N search individuals through random initialization, then optimize the individual positions using the mirror reflection mechanism to obtain the reflected positions of the individuals, and calculate the fitness values F of the individuals in the population through the fitness function. i , sort the individuals from excellent to poor according to the fitness values, and select the top N individuals with fitness values according to the current positions and reflected positions of the individuals to form a new search population. The formula used is as follows:

[0114] ;

[0115] In the formula, Represents the reflected position of the i-th individual in the d-th dimension, Represents the current position of the i-th individual in the d-th dimension, and Represent the minimum and maximum values of the search space in dimension d respectively, Represents the scaling factor, which is a constant;

[0116] Step S432: Update the search individual positions. Specifically, update the search individual positions through the sine-cosine function and the non-linear decreasing search. The formula used is as follows:

[0117] ;

[0118] In the formula, Represents the position of the i-th individual in the d-th dimension in the -th generation population, Represents the position of the i-th individual in the d-th dimension in the -th generation population, Represents the position of the optimal individual in the population, Represents the number of iterations, Represents the maximum number of iterations, Represents the search adjustment factor, and , Represents the warning value, Represents the safety threshold. Represents a random number within the range of [0, 2 , and represents a random number within the range of [0, 1], where e represents the base of the natural logarithm;

[0119] Step S433: The search is completed. Specifically, when the fitness value F of the search individual i is higher than the fitness threshold and the maximum number of iterations is reached, the search is terminated and the global optimal position of the search individual is obtained. The global optimal position of the search individual specifically refers to the optimal parameter combination of the growth state monitoring clustering model;

[0120] Step S44: Optimize the clustering parameters. Specifically, adjust the parameters of the growth state monitoring clustering model according to the optimal parameter combination of the growth state monitoring clustering model to obtain the optimal growth state monitoring clustering model.

[0121] By performing the above operations, for the technical problems existing in the existing clustering algorithms applicable to growth state monitoring, such as unreasonable setting of clustering structure parameters, resulting in unstable clustering recognition results and insufficient clustering accuracy, this solution uses a mirror reflection mechanism and a sine-cosine function method to improve and optimize the algorithm, globally optimize the key parameters in the clustering model, and obtain the optimal parameter combination, thereby effectively improving the structure expression ability and classification accuracy of the clustering model, and enhancing the accuracy and applicability of the growth state monitoring results.

[0122] Example Six, refer to Figure 1 and Figure 6 , this example is based on the above example. In step S5, the hierarchical clustering growth state monitoring is used to perform a structured clustering analysis on multi-source monitoring data collected during the rhizome asexual reproduction process of Poa annua, and identify multiple types of growth states presented by plant individuals under different physiological and environmental conditions; specifically, identify multiple types of growth states of Poa annua through the main and sub-clustering structures, form a two-layer label system, and obtain the growth state monitoring result; including the following steps:

[0123] Step S51: The main state recognition stage is used to identify the core growth state presented by the physiological behavior of the Poa annua plant body. Specifically, use the plant physiological data in the real-time growth monitoring data as input data, input it into the optimal growth state monitoring clustering model, perform the first-stage main clustering operation, generate multiple main clustering clusters, and in each clustering cluster, according to the associated historical growth monitoring state labels, count the label occurrence frequency, and select the label with the highest frequency as the main state label of the cluster to obtain the main state recognition result;

[0124] Step S52: The sub-state recognition stage is used to analyze the sub-states formed by the external incentives behind the main state and identify the sub-growth response patterns; specifically, it includes the following steps:

[0125] Step S521: Sub-cluster acquisition, specifically, selecting the corresponding plant environmental data and operational behavior data of each main cluster as input data for the second stage clustering based on the plant physiological data of each main cluster, inputting them into the optimal growth state monitoring clustering model, performing the second stage sub-clustering operation, and generating multiple sub-clusters;

[0126] Step S522: Outputting the sub-state identification results, specifically, performing statistical processing on the data of each sub-cluster, calculating the data mean, standard deviation, and maximum value, forming a feature representation vector for the cluster, and then constructing a set of predefined sub-state labels, each of which consists of multiple feature parameter intervals. The feature representation vector of each sub-cluster is compared with all sub-state labels using cosine similarity, and the label with the highest similarity is selected as the sub-state label for the cluster, thereby obtaining the sub-state identification results. The sub-state labels include a moisture-sensitive sub-label, a stress-sensitive sub-label, a nutrient deficiency sub-label, a management interference sub-label, and an exogenous stability sub-label. The feature parameters include soil moisture, soil temperature, root zone conductivity, CO2 concentration, fertilization frequency, and irrigation dosage. The feature parameter intervals are set based on historical statistical experience and agricultural management knowledge.

[0127] Step S53: Output of the growth state monitoring results is used to form the final multi-level state monitoring results. Specifically, based on the main state identification results and the sub-state identification results, they are combined according to the correspondence between the samples in the cluster, and the double-layer growth state corresponding to each monitoring point is output to obtain the growth state monitoring results.

[0128] By performing the above operations, in order to address the technical problems in traditional rhizome asexual reproduction growth monitoring methods that only use a single-layer clustering strategy, it is difficult to simultaneously characterize the complex correlation between plant physiological states and external inducements, resulting in insufficient expression of plant multi-dimensional growth patterns and inaccurate classification of growth monitoring states, this scheme innovatively proposes a dual clustering mechanism based on a master-sub hierarchical structure, performing two-stage clustering on plant physiological data and environmental operation inducement data respectively, and constructing a two-layer labeling system of main states and sub-states. This can effectively distinguish the subdivided growth response patterns of plant individuals under similar physiological states caused by external changes, and has higher growth monitoring state interpretability and label expression accuracy, realizing structured hierarchical analysis of multi-source data in the process of plant asexual reproduction, and significantly improving the accuracy, stability and agricultural usability of growth state monitoring.

[0129] It should be noted that, in this text, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or elements inherent to such process, method, article or device.

[0130] Although embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention.

[0131] The above describes the present invention and its embodiments, and this description is not restrictive. What is shown in the drawings is only one of the embodiments of the present invention, and the actual structure is not limited thereto. In summary, if those of ordinary skill in the art are inspired by it and, without departing from the purpose of the present invention, design similar structural modes and embodiments to this technical solution without creative efforts, they should all fall within the protection scope of the present invention.

Claims

1. A method for monitoring the asexual reproduction growth of Poa pratensis rhizomes based on clustering processing, characterized in that: The method includes the following steps: Step S1: Acquisition of multi-source monitoring data, specifically, by collecting data information, obtaining the original data for asexual reproduction growth monitoring; Step S2: Preprocessing of multi-source data, specifically, performing data cleaning, data normalization, time-series sample construction, and reference label association on the data to obtain optimized data for asexual reproduction growth monitoring; Step S3: Construction of a growth status monitoring clustering model, specifically, by introducing a misclassification point recognition mechanism, a misclassification trigger mechanism, and a cluster adjustment factor, improving the fuzzy C-means clustering algorithm, constructing a clustering model, and obtaining a growth status monitoring clustering model; including the following steps: initializing clustering, calculating the distance between cluster centers, calculating the distance from data points to cluster centers, misclassification point detection, adjusting the clustering structure, updating the current cluster centers, cluster deletion mechanism, and obtaining the final clustering result; Step S4: Optimization of the clustering structure, specifically, constructing a search vector for the key parameters in the growth status monitoring clustering model, designing a fitness function by integrating the average silhouette coefficient and the label jump rate, improving the optimization algorithm using the mirror image reverse mechanism and the sine-cosine function method, obtaining the optimal parameter combination of the model, and updating the clustering model structure based on this optimal parameter combination to obtain the optimal growth status monitoring clustering model; Step S5: Hierarchical clustering growth status monitoring, specifically, by constructing a two-layer clustering structure combining main status recognition and sub-status recognition, first performing main clustering with plant physiological data as input to obtain the main status recognition result, and then performing sub-clustering based on the environmental and operation data of the main cluster to generate the sub-status recognition result; finally, jointly outputting the main status and sub-status recognition results to obtain the growth status monitoring result.

2. The method for monitoring the growth of Poa pratensis rhizome asexual reproduction based on clustering processing according to claim 1, wherein: In step S3, the construction of the growth status monitoring clustering model specifically includes the following steps: Step S31: Initializing clustering, specifically, using the initialized number of clustering clusters c and the fuzzy index m, performing the standard fuzzy C-means clustering algorithm on the sample data in the historical growth monitoring data to obtain a set of cluster centers and the membership degree matrix of each data point to each cluster; Step S32: Calculate the distances between the cluster centers, specifically, for any two cluster centers in the cluster center set and , calculate the distance between them using the Euclidean distance; Step S33: Calculate the distance from the data point to the cluster center, specifically for each data point , and calculate its distance to each cluster center using the Euclidean distance; Step S34: Misclassified point detection, specifically by calculating the minimum distance from each data point to all cluster centers and comparing it with the cluster radius to determine whether the data point is a misclassified point. If the minimum value of the distances from the data point to all cluster centers is greater than the cluster radius, then the data point is considered a misclassified point. The formula used is as follows: ； ； In the formula, represents the misclassification function, represents the number of clusters in the t-th period, represents the j-th cluster, where j represents the cluster index, represents any data point in the j-th cluster, represents to the Euclidean distance, represents to the Euclidean distance, represents the membership degree of the current period data point to cluster j, represents the radius of the j-th cluster, represents the membership degree value threshold parameter, represents the centroid of cluster j in the current period, represents the i-th sample data point, represents the data in the t-th period; Step S35: Adjust the clustering structure. Specifically, if the proportion of misclassified points accumulated up to the current cycle t in the total amount of data in the current cycle exceeds the misclassification ratio threshold parameter then, based on the Xie-Beni index as the structure evaluation index, evaluate the structural strength of each cluster number scheme within the range of candidate cluster numbers, and select the cluster number with the optimal structure , and re-execute the standard fuzzy C-means clustering algorithm based on the updated number of clusters; otherwise, update the current cluster center. Step S36: Update the current cluster center, specifically by calculating the temporary cluster center for the current cycle and calculating the cluster adjustment factor , and perform the update of the current cluster center; the formula used is as follows: ； ； ； Wherein, represents the membership degree value of the data point in the current period to cluster j, represents the data of the th period, represents the center of cluster j in the next period, represents the membership degree value of the data point in the previous period to cluster j; Step S37: Cluster deletion mechanism. Specifically, for each clustering cluster existing in the current period t, by analyzing its activity changes in consecutive T periods, it is determined whether the cluster is in an inactive state, and thus whether it is an outdated cluster. If , then it is considered that cluster j is an outdated cluster and the cluster deletion operation needs to be performed, and the cluster deletion constraint conditions need to be satisfied when performing the cluster deletion operation; where represents the fuzzy cardinality of cluster j in the current period, that is, the sum of the membership degrees of all data points in the cluster, represents the fuzzy cardinality of cluster j at time , and T represents the number of periods for cluster activity evaluation; Step S38: Obtain the final clustering result. Specifically, if the change amount of all cluster centers is lower than the preset convergence threshold or the current number of iteration cycles has reached the preset maximum number of cycles , then terminate the clustering iteration and output the final clustering result. Specifically, the clustering result is to allocate all sample data points to the corresponding clusters according to the maximum value of their final membership degrees, forming multiple cluster sets, and each cluster represents a different growth state under the asexual reproduction of Poa pratensis rhizomes.

3. The method for monitoring the asexual reproduction growth of Poa pratensis rhizomes based on clustering processing according to claim 1, characterized in that: In step S4, the optimization of the clustering structure specifically includes the following steps: Step S41: Construct a search individual position vector, specifically by encoding the growth state monitoring clustering model parameters into the search individual position vector; the growth state monitoring clustering model parameters include the fuzzy index m, the membership threshold and the misclassification ratio threshold ; Step S42: Designing a fitness function, specifically, establishing a fitness function through the weighted combination of the average silhouette coefficient and the label jump rate of the current clustering structure; Step S43: Obtaining the optimal parameter combination of the model, specifically, globally optimizing the parameters of the growth status monitoring clustering model through the improved optimization algorithm to obtain the optimal parameter combination of the growth status monitoring clustering model; Step S44: Optimization of clustering parameters, specifically, adjusting the parameters of the growth status monitoring clustering model according to the optimal parameter combination of the growth status monitoring clustering model to obtain the optimal growth status monitoring clustering model.

4. The method for monitoring the asexual reproduction growth of Poa pratensis rhizomes based on clustering processing according to claim 1, wherein: Step S43: The obtaining of the optimal parameter combination of the model specifically includes the following steps: Step S431: Initialize the individuals in the search population. Specifically, generate the current positions of N search individuals through random initialization, then optimize the individual positions using the mirror reflection mechanism to obtain the reflected positions of the individuals, and calculate the fitness values F of the search individuals in the population through the fitness function i , sort the individuals from the best to the worst according to the fitness values, and select the top N individuals with the fitness values to form a new search population based on the current positions and reflected positions of the individuals. The formula used is as follows: ； Wherein, represents the reflection position of the i-th individual in the d-th dimension, represents the current position of the i-th individual in the d-th dimension, and respectively represent the minimum and maximum values of the search space in dimension d, represents the scaling factor, which is a constant; Step S432: Updating the search individual position, specifically, updating the search individual position through the sine-cosine function and the non-linear decreasing search, and the formula used is as follows: ； In the formula, represents the position of the d-th dimension of the i-th individual in the th generation population, represents the position of the d-th dimension of the i-th individual in the th generation population, represents the position of the optimal individual in the population, represents the number of iterations, represents the maximum number of iterations, represents the search adjustment factor, and , represents the warning value, represents the safety threshold, represents a random number within the range of [0, 2 , represents a random number within the range of [0, 1], and e represents the base of the natural logarithm; Step S433: Search completed. Specifically, when the fitness value F of the search individual i is higher than the fitness threshold and the maximum number of iterations is reached, the search is terminated and the globally optimal position of the search individual is obtained. The globally optimal position of the search individual specifically refers to the optimal parameter combination of the growth state monitoring clustering model.

5. The method for monitoring the asexual propagation growth of Poa pratensis rhizomes based on clustering processing according to claim 1, characterized in that: In step S5, the monitoring of the hierarchical clustering growth state specifically involves identifying multiple types of growth states of bluegrass through the master-subcluster structure, forming a two-layer label system, and obtaining the growth state monitoring result. It includes the following steps: Step S51: Master state identification stage. Specifically, the plant physiological data in the real-time growth monitoring data is used as the input data and input into the optimal growth state monitoring clustering model to perform the first-stage master clustering operation, generating multiple master clusters. In each cluster, according to the associated historical growth monitoring state labels, the label occurrence frequencies are counted, and the label with the highest frequency is selected as the master state label of the cluster, obtaining the master state identification result. Step S52: Sub-state identification stage, which is used to analyze the subdivision states composed of external incentives behind the master state and identify the subdivision growth response patterns. Specifically, it includes the following steps: Step S521: Obtaining sub-clusters. Specifically, according to the plant physiological data of each master cluster, the corresponding plant environmental data and operation behavior data are selected as the input data for the second-stage clustering and input into the optimal growth state monitoring clustering model to perform the second-stage sub-clustering operation, generating multiple sub-clusters. Step S522: Output of sub-state identification result. Specifically, the data of each sub-cluster is statistically processed to calculate the data mean, standard deviation, and maximum value, forming the feature representation vector of the cluster. Subsequently, a set of predefined sub-state labels are constructed, and each sub-state label consists of multiple characteristic parameter interval ranges. The feature representation vector of each sub-cluster is compared with all sub-state labels in terms of cosine similarity, and the label with the highest similarity is selected as the sub-state label of the cluster, obtaining the sub-state identification result. Step S53: Output of growth state monitoring result. Specifically, based on the master state identification result and the sub-state identification result, they are combined according to the corresponding relationship of the samples within the cluster, and the two-layer growth state corresponding to each monitoring point is output, obtaining the growth state monitoring result.

6. The method for monitoring the asexual reproduction growth of Poa pratensis rhizomes based on clustering processing according to claim 1, characterized in that: In step S1, the acquisition of multi-source monitoring data specifically involves collecting data information on the rhizome growth area of bluegrass to obtain the original data for asexual reproduction growth monitoring. The original data for asexual reproduction growth monitoring includes historical growth monitoring data and real-time growth monitoring data. Both the historical growth monitoring data and the real-time growth monitoring data include the plant physiological data, plant environmental data, and operation behavior data of each monitoring point. The historical growth monitoring data also includes the growth monitoring state.

7. The method for monitoring the asexual propagation growth of Poa pratensis rhizomes based on clustering processing according to claim 1, wherein: In step S2, the preprocessing of multi-source data specifically includes the following steps: Step S21: Data cleaning process. Specifically, missing value processing, outlier processing, and duplicate data processing are performed on the original data. Step S22: Data normalization process. Specifically, the data is standardized using the Min-Max normalization method. Step S23: Construction of time-series samples. Specifically, the continuous monitoring data of each monitoring point is sliced according to a set 24-hour time window to construct sample data with consistent time, and a sliding window mechanism is adopted, setting the time overlap interval to 4 hours, generating multiple time-series samples with overlapping relationships. Step S24: Reference label association, specifically using the growth monitoring status in the historical growth monitoring data as a reference label for the main clustering operation, which is only used for selecting cluster labels and does not participate in the clustering data processing process.

Images