A method and system for herbicide screening and recommendation based on field weed monitoring

The method and system enhance weed management by using big data and a herbicide recommendation model to accurately select herbicides based on field weed monitoring, addressing the inefficiencies of traditional methods.

CN118760970BActive Publication Date: 2024-12-03PLANT PROTECTION RES INST OF GUANGDONG ACADEMY OF AGRI SCI
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411223917.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-03
Publication Date
2024-12-03
Estimated Expiration
2044-09-03

AI Technical Summary

Technical Problem

Traditional weed management methods are time-consuming and labor-intensive, and rely on empirical judgments, often failing to accurately select the right herbicide for effective weed control due to varying weed sensitivities and resistances.

Method used

A method and system that utilizes big data retrieval, K-means fuzzy clustering, Pearson correlation coefficients, and a herbicide recommendation model based on a weed control knowledge graph and heterogeneous information network to accurately recommend herbicides based on field weed monitoring.

Benefits of technology

Improves the accuracy and practicality of herbicide selection by leveraging historical weeding instances and environmental characteristics, enhancing the efficiency and sustainability of weed control.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118760970B_ABST
    Figure CN118760970B_ABST
Patent Text Reader

Abstract

The present invention discloses a method and system for screening and recommending herbicides based on field weed monitoring, comprising: obtaining historical weeding instances, using the environmental characteristics and field weed characteristics of each historical weeding instance to classify the instances, and obtaining historical weeding instance classification information; selecting weeding benefit evaluation indicators and weight distribution, constructing weeding benefit evaluation rules to evaluate the weeding benefits of each historical weeding instance, and obtaining weeding benefit evaluation information; constructing a weeding knowledge graph based on the historical weeding instance classification information and the weeding benefit evaluation information, and using the weeding knowledge graph to construct a heterogeneous information network and a herbicide recommendation model; obtaining target field area image information, identifying and analyzing the plants in the target field area and determining the area that needs to be weeded, screening herbicides, and obtaining herbicide recommendation information; formulating herbicide configuration plans based on the herbicide recommendation information and recommending them. The accuracy and practicality of herbicide selection are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of field weeding, and in particular to a method and system for screening and recommending herbicides based on field weed monitoring. Background Art

[0002] In agricultural production, the presence of weeds seriously affects the growth and yield of crops, so timely and effective weed management is crucial. Traditional weed management methods rely on manual inspections and empirical judgments, which is not only time-consuming and labor-intensive, but also often difficult to achieve accurate and efficient weed control. With the development of modern agricultural technology, weed monitoring technology based on machine vision and artificial intelligence has gradually emerged, providing new ideas for weed management.

[0003] However, simply identifying the types and distribution of weeds through weed monitoring technology is not enough to directly guide farmers to take effective weed control measures. Different types of weeds have different sensitivities and resistances to different herbicides, so how to choose the right herbicide is the key to improving weed control effects, reducing chemical residues and protecting the environment. Summary of the invention

[0004] The present invention overcomes the defects of the prior art and provides a method and system for screening and recommending herbicides based on field weed monitoring, the important purpose of which is to improve the accuracy and practicality of herbicide selection.

[0005] To achieve the above-mentioned purpose, the first aspect of the present invention provides a method for screening and recommending herbicides based on field weed monitoring, comprising:

[0006] Based on big data retrieval, historical weeding instances under different field environments are obtained, and the environmental characteristics and field weed characteristics of each historical weeding instance are used to classify each historical weeding instance to obtain historical weeding instance classification information;

[0007] Obtain weed control benefit evaluation elements, select weed control benefit evaluation indicators and assign weights, construct weed control benefit evaluation rules to evaluate the weed control benefits of each historical weed control instance, and obtain weed control benefit evaluation information;

[0008] A weed control knowledge graph is formed according to the historical weed control instance classification information and the weed control benefit evaluation information, a heterogeneous information network is constructed using the weed control knowledge graph, and a herbicide recommendation model is constructed;

[0009] Obtain image information of the target field area, identify and analyze the plants in the target field area and determine the area where weeds need to be cleared, screen herbicides through a herbicide recommendation model, and obtain herbicide recommendation information;

[0010] A herbicide configuration plan is formulated and recommended based on the herbicide recommendation information.

[0011] In this solution, the historical weeding instances under different field environments are obtained based on big data retrieval, and the environmental characteristics and field weed characteristics of each historical weeding instance are used to classify each historical weeding instance to obtain historical weeding instance classification information, which specifically includes:

[0012] Based on big data retrieval, historical weeding instances under different field environments are obtained, and environmental features of each historical weeding instance are extracted, including soil environmental features and field crop features, to obtain environmental feature information of the weeding instance;

[0013] The K-means fuzzy clustering algorithm is introduced to form sample data based on the environmental characteristics of historical weeding instances, preset the number of cluster centers, and calculate the Euclidean distance from each sample data to each cluster center;

[0014] Sort by the Euclidean distance from each sample to each cluster center, and select the target sample with the smallest Euclidean distance from each cluster center based on the sorting results;

[0015] The cluster center corresponding to the smallest Euclidean distance value is used as the cluster center of the target sample data. When all sample data are classified, several clusters are output;

[0016] The field weed characteristics of each historical weeding instance are extracted, including the weed type and weed degree, and the historical weeding instances in the outputted several clusters are subdivided into categories according to the field weed characteristics to obtain the historical weeding instance division information.

[0017] In this scheme, the weeding benefit evaluation elements are obtained, weeding benefit evaluation indicators are selected and weights are allocated, and weeding benefit evaluation rules are constructed to evaluate the weeding benefits of each historical weeding instance to obtain weeding benefit evaluation information, specifically including:

[0018] Based on data retrieval, weed control benefit evaluation factors are obtained, including environmental benefit evaluation factors and weed removal evaluation factors, and the Pearson correlation coefficient between each evaluation factor and weed control benefit is calculated to form a Pearson correlation coefficient matrix;

[0019] Preset a Pearson correlation coefficient threshold, perform threshold processing on the Pearson correlation coefficient matrix, set data less than the Pearson correlation coefficient threshold to zero, obtain the screened weed control benefit evaluation factor according to the Pearson correlation coefficient matrix after the threshold processing, set it as the weed control benefit evaluation index, and obtain the weed control benefit evaluation index information;

[0020] The influence scores of each weed control benefit evaluation index are obtained by expert knowledge method, and the weights of each weed control benefit evaluation index are assigned by intuitive fuzzy analysis method, and an intuitive fuzzy matrix is ​​constructed according to the influence scores.

[0021] The intuitive fuzzy number of each weed control benefit evaluation index is obtained through the intuitive fuzzy matrix, and the intuitive fuzzy entropy of each weed control benefit evaluation index is calculated, and the weight distribution result information is obtained by combining the intuitive fuzzy number and the intuitive fuzzy entropy to distribute the weights of each weed control benefit evaluation index;

[0022] A weeding benefit evaluation rule is constructed according to the weight distribution result information, and historical weeding instance division information is obtained. The weeding benefit evaluation rule is used to evaluate the weeding benefit of each historical weeding instance to obtain weeding benefit evaluation information.

[0023] In this solution, the weeding knowledge graph is formed according to the historical weeding instance classification information and the weeding benefit evaluation information, a heterogeneous information network is constructed using the weeding knowledge graph, and a herbicide recommendation model is constructed, which specifically includes:

[0024] Acquire historical weeding instance classification information and weeding benefit evaluation information, and form entity triples based on the historical weeding instance classification information according to field environment characteristics, field weed characteristics, and weeding scheme characteristics;

[0025] A weed control knowledge graph is constructed according to the entity triples, and a MetaPath random walk is used to perform representation learning on the weed control knowledge graph to obtain entity nodes corresponding to the characteristics of the pastoral environment, and a meta-path is generated according to the connections between the entity nodes;

[0026] The weed control benefit evaluation information is used as an auxiliary feature of the meta-path to form a heterogeneous information network, the heterogeneous information network is represented by a graph, and the cosine similarity between the meta-paths is calculated according to the graph representation;

[0027] Establishing connections between each meta-path based on the calculated cosine similarity, obtaining neighbor nodes corresponding to the farmland weed feature node and the weed control scheme feature node, and constructing an adjacency matrix based on the neighbor nodes;

[0028] The graph representation of the heterogeneous information network is stored in the form of an adjacency matrix, and the graph representation of the heterogeneous information network is learned through a graph neural network, and a herbicide recommendation model is constructed.

[0029] In this solution, the target field area image information is obtained, the plants in the target field area are identified and analyzed, and the area where weeds need to be cleared is determined, and the herbicide recommendation model is used to screen the herbicides to obtain the herbicide recommendation information, which specifically includes:

[0030] Based on the UAV, the target field area is monitored to obtain the target field area image information, and the target field area image information is preprocessed, including grayscale processing, image noise reduction processing and image threshold segmentation processing, to obtain preprocessed image information;

[0031] Constructing a plant recognition model, inputting the preprocessed image information into the plant recognition model to perform recognition analysis on plants in a target field area, and obtaining field plant recognition information;

[0032] Based on the field plant identification information, the plants to be cleaned and the plants that cannot be cleaned in the target field area are marked, and the target field area is divided into a plurality of sub-areas, and the degree and type of weeds in each sub-area are analyzed through the marking results to obtain regional weed analysis information;

[0033] Preset a cleaning area judgment rule, define the area in the target field area that needs to be weeded in combination with the regional weed analysis information, and obtain cleaning area analysis information;

[0034] Acquire soil environment information of a target field area, combine the regional weed analysis information and the cleanup area analysis information and input them into the herbicide recommendation model for herbicide screening, and generate a target node according to the input information;

[0035] Performing first-order neighborhood sampling on the target node to obtain neighbor nodes, and sampling the neighbor nodes to obtain neighbors of the neighbor nodes;

[0036] The attention mechanism is introduced to obtain the attention coefficient, the neighbors of the neighbor node and the corresponding attention coefficient are aggregated with the feature vector of the neighbor node, the feature vector of the neighbor node is updated, and the updated feature vector is averaged;

[0037] The attention coefficient of the neighbor node is obtained by sharing the attention parameter, and the vector is concatenated with the target node. The feature vector of the target node is updated according to the parameter matrix and the activation function. The recommended herbicide is obtained based on the updated feature vector of the target node to obtain the herbicide recommendation information.

[0038] In this solution, the herbicide configuration scheme is formulated and recommended according to the herbicide recommendation information, specifically including:

[0039] Obtain herbicide recommendation information and obtain historical weeding instances corresponding to the recommended herbicides through the constructed heterogeneous information network;

[0040] Extracting the herbicide configuration parameter characteristics and the field weed characteristics when the herbicide is used according to the historical weeding examples corresponding to the recommended herbicide to obtain the first characteristic information;

[0041] Acquire regional weed analysis information, perform difference analysis with the first feature information, calculate the deviation between the actual field weed environment and the field weed environment in the historical weed control instance, and obtain first difference analysis information;

[0042] Sort the historical weeding instances in descending order according to the first difference analysis information, obtain the historical weeding instances with the smallest difference from the actual field weed environment through the sorting results, and extract the corresponding herbicide configuration parameter features to obtain second feature information;

[0043] A herbicide configuration scheme for actual field weeds is formulated in combination with the first feature information, the second feature information and the first difference analysis information, and a herbicide configuration scheme is recommended for the target field area.

[0044] A second aspect of the present invention provides a herbicide screening and recommendation system based on field weed monitoring, the system comprising: a memory, a processor, the memory containing a herbicide screening and recommendation method program based on field weed monitoring, the herbicide screening and recommendation method program based on field weed monitoring being executed by the processor to implement the following steps:

[0045] Based on big data retrieval, historical weeding instances under different field environments are obtained, and the environmental characteristics and field weed characteristics of each historical weeding instance are used to classify each historical weeding instance to obtain historical weeding instance classification information;

[0046] Obtain weed control benefit evaluation elements, select weed control benefit evaluation indicators and assign weights, construct weed control benefit evaluation rules to evaluate the weed control benefits of each historical weed control instance, and obtain weed control benefit evaluation information;

[0047] A weed control knowledge graph is formed according to the historical weed control instance classification information and the weed control benefit evaluation information, a heterogeneous information network is constructed using the weed control knowledge graph, and a herbicide recommendation model is constructed;

[0048] Obtain image information of the target field area, identify and analyze the plants in the target field area and determine the area where weeds need to be cleared, screen herbicides through a herbicide recommendation model, and obtain herbicide recommendation information;

[0049] A herbicide configuration plan is formulated and recommended based on the herbicide recommendation information.

[0050] The present invention discloses a method and system for screening and recommending herbicides based on field weed monitoring, comprising: obtaining historical weeding instances, using the environmental characteristics and field weed characteristics of each historical weeding instance to classify the instances, and obtaining historical weeding instance classification information; selecting weeding benefit evaluation indicators and weight distribution, constructing weeding benefit evaluation rules to evaluate the weeding benefits of each historical weeding instance, and obtaining weeding benefit evaluation information; constructing a weeding knowledge graph based on the historical weeding instance classification information and the weeding benefit evaluation information, and using the weeding knowledge graph to construct a heterogeneous information network and a herbicide recommendation model; obtaining target field area image information, identifying and analyzing the plants in the target field area and determining the area that needs to be weeded, screening herbicides, and obtaining herbicide recommendation information; formulating herbicide configuration plans based on the herbicide recommendation information and recommending them. The accuracy and practicality of herbicide selection are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions in the embodiments or exemplary embodiments of the present invention, the drawings required for use in the embodiments or exemplary descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained according to the drawings without paying creative work.

[0052] Figure 1 A flow chart of a method for screening and recommending herbicides based on field weed monitoring provided by one embodiment of the present invention;

[0053] Figure 2 A recommended flow chart for herbicide screening provided by an embodiment of the present invention;

[0054] Figure 3 A block diagram of a herbicide screening and recommendation system based on field weed monitoring provided by an embodiment of the present invention;

[0055] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0056] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0057] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0058] Figure 1 A flow chart of a method for screening and recommending herbicides based on field weed monitoring provided by one embodiment of the present invention;

[0059] like Figure 1 As shown, the present invention provides a flow chart of a method for screening and recommending herbicides based on field weed monitoring, comprising:

[0060] S102, based on big data retrieval, historical weeding instances under different field environments are obtained, and environmental characteristics and field weed characteristics of each historical weeding instance are used to classify each historical weeding instance to obtain historical weeding instance classification information;

[0061] S104, obtaining weeding benefit evaluation elements, selecting weeding benefit evaluation indicators and assigning weights, constructing weeding benefit evaluation rules, evaluating weeding benefits of each historical weeding instance, and obtaining weeding benefit evaluation information;

[0062] S106, constructing a weeding knowledge graph based on the historical weeding instance classification information and the weeding benefit evaluation information, building a heterogeneous information network using the weeding knowledge graph, and building a herbicide recommendation model;

[0063] S108, obtaining image information of the target field area, identifying and analyzing the plants in the target field area and determining the area where weeds need to be cleared, screening herbicides through a herbicide recommendation model, and obtaining herbicide recommendation information;

[0064] S110, formulating a herbicide configuration plan and making a recommendation based on the herbicide recommendation information.

[0065] It should be noted that the present invention provides a method and system for screening and recommending herbicides based on field weed monitoring, which obtains historical weeding examples under different pastoral environments through big data retrieval, performs initial classification based on the environmental characteristics of each weeding example, and then reclassifies based on the weed characteristics, thereby obtaining historical weeding examples corresponding to various weed characteristics under different environmental categories. Then, the weeding benefit of each historical weeding example is evaluated, a weeding knowledge graph is constructed based on the evaluation results and the classification results, and the weeding knowledge graph is further used to form a heterogeneous information network to train the herbicide recommendation model. Subsequently, the target field area image information obtained in real time is used to identify and analyze the plants in the target field area and determine the area where weeds need to be cleared, and the herbicide is screened through the herbicide recommendation model. Finally, the herbicide configuration scheme is analyzed based on the recommended herbicide, and the appropriate herbicide usage ratio is recommended, etc., which greatly improves the efficiency and effect of field weed control and promotes the sustainable development of agricultural production.

[0066] Further, in a preferred embodiment of the present invention, the historical weeding instances under different field environments are obtained based on big data retrieval, and the environmental characteristics and field weed characteristics of each historical weeding instance are used to classify each historical weeding instance to obtain historical weeding instance classification information, which specifically includes:

[0067] Based on big data retrieval, historical weeding instances under different field environments are obtained, and environmental features of each historical weeding instance are extracted, including soil environmental features and field crop features, to obtain environmental feature information of the weeding instance;

[0068] The K-means fuzzy clustering algorithm is introduced to form sample data based on the environmental characteristics of historical weeding instances, preset the number of cluster centers, and calculate the Euclidean distance from each sample data to each cluster center;

[0069] Sort by the Euclidean distance from each sample to each cluster center, and select the target sample with the smallest Euclidean distance from each cluster center based on the sorting results;

[0070] The cluster center corresponding to the smallest Euclidean distance value is used as the cluster center of the target sample data. When all sample data are classified, several clusters are output;

[0071] The field weed characteristics of each historical weeding instance are extracted, including the weed type and weed degree, and the historical weeding instances in the outputted several clusters are subdivided into categories according to the field weed characteristics to obtain the historical weeding instance division information.

[0072] It should be noted that, first of all, big data retrieval is used to obtain historical weeding examples under different field environments. These examples cover various field conditions, weed conditions and weeding schemes. The environmental characteristics of each historical weeding example are used to perform an initial classification, and the environmental characteristics are used as the major category to obtain historical weeding examples under similar environmental conditions. Then, the categories are subdivided and further classified by weed characteristics, i.e., weed species and weed degree, so as to obtain historical weeding examples corresponding to each weed characteristic under different environmental categories, providing a more scientific basis for subsequent herbicide recommendations and decisions.

[0073] Further, in a preferred embodiment of the present invention, the weeding benefit evaluation elements are obtained, weeding benefit evaluation indicators are selected and weighted, and weeding benefit evaluation rules are constructed to evaluate the weeding benefits of each historical weeding instance to obtain weeding benefit evaluation information, which specifically includes:

[0074] Based on data retrieval, weed control benefit evaluation factors are obtained, including environmental benefit evaluation factors and weed removal evaluation factors, and the Pearson correlation coefficient between each evaluation factor and weed control benefit is calculated to form a Pearson correlation coefficient matrix;

[0075] Preset a Pearson correlation coefficient threshold, perform threshold processing on the Pearson correlation coefficient matrix, set data less than the Pearson correlation coefficient threshold to zero, obtain the screened weed control benefit evaluation factor according to the Pearson correlation coefficient matrix after the threshold processing, set it as the weed control benefit evaluation index, and obtain the weed control benefit evaluation index information;

[0076] The influence scores of each weed control benefit evaluation index are obtained by expert knowledge method, and the weights of each weed control benefit evaluation index are assigned by intuitive fuzzy analysis method, and an intuitive fuzzy matrix is ​​constructed according to the influence scores.

[0077] The intuitive fuzzy number of each weed control benefit evaluation index is obtained through the intuitive fuzzy matrix, and the intuitive fuzzy entropy of each weed control benefit evaluation index is calculated, and the weight distribution result information is obtained by combining the intuitive fuzzy number and the intuitive fuzzy entropy to distribute the weights of each weed control benefit evaluation index;

[0078] A weeding benefit evaluation rule is constructed according to the weight distribution result information, and historical weeding instance division information is obtained. The weeding benefit evaluation rule is used to evaluate the weeding benefit of each historical weeding instance to obtain weeding benefit evaluation information.

[0079] It should be noted that after the historical weeding instances are divided, it is necessary to further evaluate the weeding benefits of each historical weeding instance, and analyze the weed removal effect and environmental impact of the corresponding weeding scheme in each historical weeding instance. By obtaining various weeding benefit evaluation factors, according to the Pearson correlation coefficient between each evaluation factor and the weeding benefit, the evaluation factors with high correlation with the weeding benefit are screened, and then the quality index of the Forbidden City is set. Then, the expert knowledge method and the intuitive fuzzy analysis method are introduced. The influence score of each evaluation index is obtained by the expert knowledge method, which reflects the subjective judgment of the experts on the importance of each index. Then, the intuitive fuzzy analysis method is used to convert these subjective judgments into intuitive fuzzy matrices, and the intuitive fuzzy number and intuitive fuzzy entropy of each evaluation index are further calculated, taking into account the fuzziness and uncertainty of the evaluation index, and quantifying the richness of its information content, providing a more comprehensive and objective basis for weight allocation. Finally, based on the weight distribution results, we constructed a weed control benefit evaluation rule, and conducted a weed control benefit evaluation on historical weed control instances to obtain the corresponding evaluation information, which reflected the actual effects of different weed control methods or techniques and also provided the impact on the environment, providing strong data support for subsequent weed control decisions.

[0080] Furthermore, in a preferred embodiment of the present invention, the weeding knowledge graph is formed according to the historical weeding instance classification information and the weeding benefit evaluation information, a heterogeneous information network is constructed using the weeding knowledge graph, and a herbicide recommendation model is constructed, which specifically includes:

[0081] Acquire historical weeding instance classification information and weeding benefit evaluation information, and form entity triples based on the historical weeding instance classification information according to field environment characteristics, field weed characteristics, and weeding scheme characteristics;

[0082] A weed control knowledge graph is constructed according to the entity triples, and a MetaPath random walk is used to perform representation learning on the weed control knowledge graph to obtain entity nodes corresponding to the characteristics of the pastoral environment, and a meta-path is generated according to the connections between the entity nodes;

[0083] The weed control benefit evaluation information is used as an auxiliary feature of the meta-path to form a heterogeneous information network, the heterogeneous information network is represented by a graph, and the cosine similarity between the meta-paths is calculated according to the graph representation;

[0084] Establishing connections between each meta-path based on the calculated cosine similarity, obtaining neighbor nodes corresponding to the farmland weed feature node and the weed control scheme feature node, and constructing an adjacency matrix based on the neighbor nodes;

[0085] The graph representation of the heterogeneous information network is stored in the form of an adjacency matrix, and the graph representation of the heterogeneous information network is learned through a graph neural network, and a herbicide recommendation model is constructed.

[0086] It should be noted that, firstly, entity triplets are constructed based on the partition information of historical weeding instances. Each triplet consists of three entities: field environment characteristics, field weed characteristics, and weeding scheme characteristics. These entities together describe a specific weeding instance.

[0087] Subsequently, a weed control knowledge graph was constructed using entity triples. Knowledge graph is a method of structured representation of knowledge that can clearly show the relationship between entities. The MetaPath random walk algorithm was used to perform representation learning on the weed control knowledge graph, capturing the potential characteristics of entity nodes (such as pastoral environmental characteristics) and revealing the complex connection patterns between nodes. Metapaths were generated based on the connections between entity nodes. These metapaths are paths connecting different node sequences, reflecting the potential relationships in the knowledge graph. Next, the weed control benefit evaluation information was integrated into this framework as an auxiliary feature of the metapath to form a heterogeneous information network containing rich information. A heterogeneous information network is a special type of graph in which nodes and edges can have different types. The heterogeneous information network is represented by a graph, that is, the nodes and edges in the network are converted into numerical forms that can be used for calculation. The graph representation of the heterogeneous information network is defined as , is a node set, is a set of edge structures. Subsequently, the cosine similarity between each meta-path is calculated to establish the connection between these paths, reflecting the similarity between different weeding instances in terms of environment, weeds and schemes. Based on the cosine similarity, the positions of the field weed feature nodes and the weeding scheme feature nodes in the knowledge graph are further analyzed, and their neighbor nodes are found. These neighbor nodes may represent similar weed species, environmental conditions or effective weeding schemes. This information is used to construct an adjacency matrix, which mathematically represents the connection relationship between nodes in the network. Finally, the graph representation of the heterogeneous information network is stored in the form of an adjacency matrix and is deeply learned and represented by a graph neural network. A graph neural network is a neural network specifically designed to process graph structured data. It can capture the complex relationships between nodes and build a herbicide recommendation model. The model can automatically recommend the most suitable weeding scheme based on the input field environment characteristics and weed characteristics, including the recommended herbicide types and usage methods, thereby helping farmers achieve more efficient and environmentally friendly weeding operations.

[0088] Further, in a preferred embodiment of the present invention, the target field area image information is obtained, the plants in the target field area are identified and analyzed, and the area where weeds need to be cleared is determined, and the herbicide recommendation model is used to screen the herbicide to obtain the herbicide recommendation information, specifically including:

[0089] Based on the UAV, the target field area is monitored to obtain the target field area image information, and the target field area image information is preprocessed, including grayscale processing, image noise reduction processing and image threshold segmentation processing, to obtain preprocessed image information;

[0090] Constructing a plant recognition model, inputting the preprocessed image information into the plant recognition model to perform recognition analysis on plants in a target field area, and obtaining field plant recognition information;

[0091] Based on the field plant identification information, the plants to be cleaned and the plants that cannot be cleaned in the target field area are marked, and the target field area is divided into a plurality of sub-areas, and the degree and type of weeds in each sub-area are analyzed through the marking results to obtain regional weed analysis information;

[0092] Preset a cleaning area judgment rule, define the area in the target field area that needs to be weeded in combination with the regional weed analysis information, and obtain cleaning area analysis information;

[0093] Acquire soil environment information of a target field area, combine the regional weed analysis information and the cleanup area analysis information and input them into the herbicide recommendation model for herbicide screening, and generate a target node according to the input information;

[0094] Performing first-order neighborhood sampling on the target node to obtain neighbor nodes, and sampling the neighbor nodes to obtain neighbors of the neighbor nodes;

[0095] The attention mechanism is introduced to obtain the attention coefficient, the neighbors of the neighbor node and the corresponding attention coefficient are aggregated with the feature vector of the neighbor node, the feature vector of the neighbor node is updated, and the updated feature vector is averaged;

[0096] The attention coefficient of the neighbor node is obtained by sharing the attention parameter, and the vector is concatenated with the target node. The feature vector of the target node is updated according to the parameter matrix and the activation function. The recommended herbicide is obtained based on the updated feature vector of the target node to obtain the herbicide recommendation information.

[0097] It should be noted that, first, the image information of the target field area is captured by the drone, and preprocessed, including grayscale processing, image noise reduction processing and image threshold segmentation processing, to generate preprocessed image information. Next, a plant recognition model is constructed. The preprocessed image information is received as input, the plants in the target field area are identified and analyzed, various plants are distinguished, and field plant identification information is generated. Based on the plant recognition results, the plants in the target field area are marked to clearly distinguish between plants to be cleaned and plants that cannot be cleaned. At the same time, the target field area is divided into several sub-areas, and the degree and type of weeds in each sub-area are analyzed to obtain regional weed analysis information. Subsequently, the cleaning area judgment rules are preset, and the area in the target field area that really needs to be cleaned is defined in combination with the regional weed analysis information. Ensure the pertinence and effectiveness of weeding work and avoid unnecessary waste of resources. After determining the cleaning area, the soil environment information of the target field area is obtained, and input into the herbicide recommendation model together with the regional weed analysis information and the cleaning area analysis information, and appropriate herbicides are recommended, and herbicide recommendation information is generated. In the process of herbicide recommendation, the meta-path of field environment-weed characteristics-weed control scheme is analyzed to recommend herbicides suitable for the current environment. Furthermore, the meta-path information is supplemented according to the corresponding weed control benefit attributes. That is, when the field environment and weed characteristics are determined, there may be multiple herbicide use schemes. The weed control benefits in similar schemes are used to supplement the information of other schemes, and then determine whether the herbicide in the target scheme can be recommended first. Improve the pertinence and benefit of herbicide use to avoid damage to the environment and plants.

[0098] Further, in a preferred embodiment of the present invention, the formulating and recommending a herbicide configuration plan according to the herbicide recommendation information specifically includes:

[0099] Obtain herbicide recommendation information and obtain historical weeding instances corresponding to the recommended herbicides through the constructed heterogeneous information network;

[0100] Extracting the herbicide configuration parameter characteristics and the field weed characteristics when the herbicide is used according to the historical weeding examples corresponding to the recommended herbicide to obtain the first characteristic information;

[0101] Acquire regional weed analysis information, perform difference analysis with the first feature information, calculate the deviation between the actual field weed environment and the field weed environment in the historical weed control instance, and obtain first difference analysis information;

[0102] Sort the historical weeding instances in descending order according to the first difference analysis information, obtain the historical weeding instances with the smallest difference from the actual field weed environment through the sorting results, and extract the corresponding herbicide configuration parameter features to obtain second feature information;

[0103] A herbicide configuration scheme for actual field weeds is formulated in combination with the first feature information, the second feature information and the first difference analysis information, and a herbicide configuration scheme is recommended for the target field area.

[0104] It should be noted that the historical weeding examples of recommended herbicides are further obtained through the constructed heterogeneous information network, and features are extracted from the historical weeding examples, including the herbicide configuration parameter features and the field weed features when the herbicide is used. The herbicide configuration parameter features describe the specific configuration information such as the amount and ratio of the herbicide; the field weed features when the herbicide is used reflect the type, density, growth status, etc. of the field weeds at that time. Further, the regional weed analysis information is obtained, and the difference analysis is performed between it and the field weed features in the first feature information. It aims to calculate the deviation between the actual field weed environment and the field weed environment described in the historical weeding examples, so as to obtain the first difference analysis information, which reveals the degree of difference between the current field environment and the historical records. Based on the first difference analysis information, each historical weeding example is sorted in descending order. According to the similarity between the historical weeding example and the current field weed environment, that is, the smaller the difference, the higher the similarity. The historical weeding example with the smallest difference from the actual field weed environment is obtained through the sorting result. Subsequently, the herbicide configuration parameter features corresponding to the instance are extracted as the second feature information. Finally, the first feature information, the second feature information and the first difference analysis information are integrated to formulate a herbicide configuration plan for actual field weeds, and provide a scientific and reasonable herbicide configuration plan recommendation for the target field area.

[0105] Figure 2 A recommended flow chart for herbicide screening provided by an embodiment of the present invention;

[0106] like Figure 2 As shown, the present invention provides a recommended flow chart for herbicide screening, including:

[0107] S202, acquiring target field area image information, and performing image preprocessing on the target field area image information to obtain preprocessed image information;

[0108] S204, inputting the pre-processed image information into the plant recognition model to perform recognition analysis on the plants in the target field area to obtain field plant recognition information;

[0109] S206, marking the plants to be cleaned and the plants that cannot be cleaned in the target field area based on the field plant identification information, and dividing the target field area into a plurality of sub-areas, analyzing the degree and type of weeds in each sub-area, and obtaining regional weed analysis information;

[0110] S208, defining the area in the target field area where weeds need to be cleared based on the regional weed analysis information, obtaining clearing area analysis information, and performing herbicide screening through a herbicide recommendation model to obtain herbicide recommendation information;

[0111] S210: Formulate a herbicide configuration plan and make a recommendation based on the herbicide recommendation information.

[0112] In addition, the herbicide screening and recommendation method based on field weed monitoring provided by the present invention also includes the following steps:

[0113] Sampling the soil in the target field area after weeding to obtain regional soil sampling information, performing DNA extraction on the regional soil sample information, and performing metagenomic sequencing on the DNA extraction information based on high-throughput metagenomic sequencing technology to obtain sequencing result information;

[0114] Constructing a microbial identification model, inputting the sequencing result information into the microbial identification model, analyzing the soil microbial status after weeding in the target field area, and obtaining microbial identification information;

[0115] Calculating the Shannon index of the soil samples in the target field area based on the microbial identification information, and performing a diversity analysis on the soil in the target area after weeding according to the Shannon index of each soil sample to obtain diversity analysis information;

[0116] Using the regional soil sample information to perform soil environmental testing, detecting soil environmental indicators in the target area after weeding, and combining the diversity analysis information to perform soil capacity assessment to obtain soil capacity assessment information;

[0117] Determine whether the nutritional capacity of the target field area after weeding changes according to the soil capacity assessment information; if changes occur, obtain the soil capacity assessment information of the target field area before weeding, and perform difference analysis with the soil capacity assessment information after weeding;

[0118] Obtain difference analysis result information, extract difference indicators based on the difference analysis result information, determine the cause of soil capacity change based on the extracted difference indicators, and use big data retrieval means to obtain soil fertilizers or microbial inhibitors corresponding to the difference indicators for soil remediation recommendations.

[0119] It should be noted that after weeding the target field area, the herbicides used may destroy the original soil ecological structure while disinfecting weeds, thereby causing a decrease in soil nutritional capacity. Due to different weeding needs, in scenarios where weeding is urgently needed, the amount of herbicides used will increase, which will bring adverse consequences. At this time, the soil after weeding is monitored and analyzed, and its internal microbial structure and soil nutritional indicators are further analyzed to assess whether the soil capacity has changed. When changes occur, the cause of the soil change may be a change in the structure of microorganisms or a change in soil nutritional indicators. Therefore, targeted repairs can be carried out based on the difference in indicators of the target soil before and after weeding to avoid damage to the soil environment due to weeding needs.

[0120] Figure 3 A herbicide screening recommendation system 3 based on field weed monitoring is provided in one embodiment of the present invention. The system comprises: a memory 31 and a processor 32. The memory 31 contains a herbicide screening recommendation method program based on field weed monitoring. When the herbicide screening recommendation method program based on field weed monitoring is executed by the processor 32, the following steps are implemented:

[0121] Based on big data retrieval, historical weeding instances under different field environments are obtained, and the environmental characteristics and field weed characteristics of each historical weeding instance are used to classify each historical weeding instance to obtain historical weeding instance classification information;

[0122] Obtain weed control benefit evaluation elements, select weed control benefit evaluation indicators and assign weights, construct weed control benefit evaluation rules to evaluate the weed control benefits of each historical weed control instance, and obtain weed control benefit evaluation information;

[0123] A weed control knowledge graph is formed according to the historical weed control instance classification information and the weed control benefit evaluation information, a heterogeneous information network is constructed using the weed control knowledge graph, and a herbicide recommendation model is constructed;

[0124] Obtain image information of the target field area, identify and analyze the plants in the target field area and determine the area where weeds need to be cleared, screen herbicides through a herbicide recommendation model, and obtain herbicide recommendation information;

[0125] A herbicide configuration plan is formulated and recommended based on the herbicide recommendation information.

[0126] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0127] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0128] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0129] A person of ordinary skill in the art can understand that: all or part of the steps of implementing the above method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiment; and the aforementioned storage medium includes: a mobile storage device, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

[0130] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

[0131] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for screening and recommending herbicides based on field weed monitoring, characterized in that: include: Based on big data retrieval, historical weeding instances under different field environments are obtained, and the environmental characteristics and field weed characteristics of each historical weeding instance are used to classify each historical weeding instance to obtain historical weeding instance classification information; Obtain weed control benefit evaluation elements, select weed control benefit evaluation indicators and assign weights, construct weed control benefit evaluation rules to evaluate the weed control benefits of each historical weed control instance, and obtain weed control benefit evaluation information; A weed control knowledge graph is constructed based on historical weed control instance classification information and weed control benefit evaluation information, a heterogeneous information network is constructed using the weed control knowledge graph, and a herbicide recommendation model is constructed; Obtain image information of the target field area, identify and analyze the plants in the target field area and determine the area where weeds need to be cleared, screen herbicides through a herbicide recommendation model, and obtain herbicide recommendation information; Formulate and recommend a herbicide configuration plan based on the herbicide recommendation information; The step of constructing a weeding knowledge graph based on the historical weeding instance classification information and the weeding benefit evaluation information, constructing a heterogeneous information network using the weeding knowledge graph, and constructing a herbicide recommendation model specifically includes: Acquire historical weeding instance classification information and weeding benefit evaluation information, and form entity triples of field environment characteristics, field weed characteristics, and weeding scheme characteristics based on the historical weeding instance classification information; A weed control knowledge graph is constructed according to the entity triples, and a MetaPath random walk algorithm is used to perform representation learning on the weed control knowledge graph to obtain entity nodes corresponding to the characteristics of the pastoral environment, and a meta-path is generated according to the connections between the entity nodes; The weed control benefit evaluation information is used as an auxiliary feature of the meta-path to form a heterogeneous information network, the heterogeneous information network is represented by a graph, and the cosine similarity between the meta-paths is calculated according to the graph representation; Establishing connections between each meta-path based on the calculated cosine similarity, obtaining neighbor nodes corresponding to the farmland weed feature node and the weed control scheme feature node, and constructing an adjacency matrix based on the neighbor nodes; The graph representation of the heterogeneous information network is stored in the form of an adjacency matrix, a herbicide recommendation model is constructed based on a graph neural network, and the herbicide recommendation model is deeply learned and trained using the graph representation of the heterogeneous information network to obtain a herbicide recommendation model that meets expectations.

2. The method for screening and recommending herbicides based on field weed monitoring according to claim 1, characterized in that: The historical weeding instances under different field environments are obtained based on big data retrieval, and the environmental characteristics and field weed characteristics of each historical weeding instance are used to classify each historical weeding instance to obtain historical weeding instance classification information, which specifically includes: Based on big data retrieval, historical weeding instances under different field environments are obtained, and environmental features of each historical weeding instance are extracted, including soil environmental features and field crop features, to obtain environmental feature information of the weeding instance; The K-means fuzzy clustering algorithm is introduced to form sample data from the environmental feature information of the weeding instance, the number of cluster centers is preset, and the Euclidean distance from each sample data to each cluster center is calculated; Sort the Euclidean distances from each sample to each cluster center, and select the minimum Euclidean distance value from the target sample data to each cluster center based on the sorting results; The cluster center corresponding to the smallest Euclidean distance value is used as the cluster center of the target sample data. When all sample data are classified, several clusters are output; The field weed characteristics of each historical weeding instance are extracted, including the weed type and weed degree, and the historical weeding instances in the outputted several clusters are subdivided into categories according to the field weed characteristics to obtain the historical weeding instance division information.

3. The herbicide screening and recommendation method based on field weed monitoring according to claim 1, characterized in that: The weeding benefit evaluation elements are obtained, weeding benefit evaluation indicators are selected and weights are allocated, and weeding benefit evaluation rules are constructed to evaluate the weeding benefits of each historical weeding instance, thereby obtaining weeding benefit evaluation information, which specifically includes: Based on data retrieval, weed control benefit evaluation factors are obtained, including environmental benefit evaluation factors and weed removal evaluation factors, and the Pearson correlation coefficient between each evaluation factor and weed control benefit is calculated to form a Pearson correlation coefficient matrix; Preset a Pearson correlation coefficient threshold, perform threshold processing on the Pearson correlation coefficient matrix, set data less than the Pearson correlation coefficient threshold to zero, obtain the screened weed control benefit evaluation factor according to the Pearson correlation coefficient matrix after the threshold processing, set it as the weed control benefit evaluation index, and obtain the weed control benefit evaluation index information; The influence scores of each weed control benefit evaluation index are obtained according to the expert knowledge method, and the intuitive fuzzy matrix is ​​constructed according to the influence scores. The evaluation weights of each weed control benefit evaluation index are set by the intuitive fuzzy analysis method. Specifically, the following are included: The intuitive fuzzy number of each weed control benefit evaluation index is obtained through the intuitive fuzzy matrix, and the intuitive fuzzy entropy of each weed control benefit evaluation index is calculated, and the weight distribution result information is obtained by combining the intuitive fuzzy number and the intuitive fuzzy entropy to distribute the weights of each weed control benefit evaluation index; A weeding benefit evaluation rule is constructed according to the weight distribution result information, and historical weeding instance division information is obtained. The weeding benefit evaluation rule is used to evaluate the weeding benefit of each historical weeding instance to obtain weeding benefit evaluation information.

4. The herbicide screening and recommendation method based on field weed monitoring according to claim 1, characterized in that: The method of obtaining image information of the target field area, identifying and analyzing the plants in the target field area and determining the area where weeds need to be cleared, and screening herbicides through a herbicide recommendation model to obtain herbicide recommendation information specifically includes: Based on the UAV, the target field area is monitored to obtain the target field area image information, and the target field area image information is preprocessed, including grayscale processing, image noise reduction processing and image threshold segmentation processing, to obtain preprocessed image information; Constructing a plant recognition model, inputting the preprocessed image information into the plant recognition model to perform recognition analysis on plants in a target field area, and obtaining field plant recognition information; Based on the field plant identification information, the plants to be cleaned and the plants that cannot be cleaned in the target field area are marked, and the target field area is divided into a plurality of sub-areas, and the degree and type of weeds in each sub-area are analyzed through the marking results to obtain regional weed analysis information; Preset a cleaning area judgment rule, define the area in the target field area that needs to be weeded in combination with the regional weed analysis information, and obtain cleaning area analysis information; Acquire soil environment information of the target field area, input the soil environment information, regional weed analysis information and cleanup area analysis information into the herbicide recommendation model for herbicide screening, and generate a target node according to the input information; Performing first-order neighborhood sampling on the target node to obtain neighbor nodes, and sampling the neighbor nodes to obtain neighbors of the neighbor nodes; The attention mechanism is introduced to obtain the attention coefficient, the neighbors of the neighbor node and the corresponding attention coefficient are aggregated with the feature vector of the neighbor node, the feature vector of the neighbor node is updated, and the updated feature vector is averaged; The attention coefficient of the neighbor node is obtained by sharing the attention parameter, and the vector is concatenated with the target node. The feature vector of the target node is updated according to the parameter matrix and the activation function. The recommended herbicide is obtained based on the updated feature vector of the target node to obtain the herbicide recommendation information.

5. The method for screening and recommending herbicides based on field weed monitoring according to claim 1, characterized in that: The step of formulating a herbicide configuration plan and making a recommendation based on the herbicide recommendation information specifically includes: Obtain herbicide recommendation information and obtain historical weeding instances corresponding to the recommended herbicides through the constructed heterogeneous information network; Extracting the herbicide configuration parameter characteristics and the field weed characteristics when the herbicide is used according to the historical weeding examples corresponding to the recommended herbicide to obtain the first characteristic information; Acquire regional weed analysis information, perform difference analysis with the first feature information, calculate the deviation between the actual field weed environment and the field weed environment in the historical weed control instance, and obtain first difference analysis information; Sort the historical weeding instances in descending order according to the first difference analysis information, obtain the historical weeding instances with the smallest difference from the actual field weed environment through the sorting results, and extract the corresponding herbicide configuration parameter features to obtain second feature information; A herbicide configuration scheme for actual field weeds is formulated in combination with the first feature information, the second feature information and the first difference analysis information, and a herbicide configuration scheme is recommended for the target field area.

6. A herbicide screening and recommendation system based on field weed monitoring, characterized in that: The system includes: a memory and a processor, wherein the memory contains a herbicide screening recommendation method program based on field weed monitoring, and when the herbicide screening recommendation method program based on field weed monitoring is executed by the processor, the following steps are implemented: Based on big data retrieval, historical weeding instances under different field environments are obtained, and the environmental characteristics and field weed characteristics of each historical weeding instance are used to classify each historical weeding instance to obtain historical weeding instance classification information; Obtain weed control benefit evaluation elements, select weed control benefit evaluation indicators and assign weights, construct weed control benefit evaluation rules to evaluate the weed control benefits of each historical weed control instance, and obtain weed control benefit evaluation information; A weed control knowledge graph is constructed based on historical weed control instance classification information and weed control benefit evaluation information, a heterogeneous information network is constructed using the weed control knowledge graph, and a herbicide recommendation model is constructed; Obtain image information of the target field area, identify and analyze the plants in the target field area and determine the area where weeds need to be cleared, screen herbicides through a herbicide recommendation model, and obtain herbicide recommendation information; Formulate and recommend a herbicide configuration plan based on the herbicide recommendation information; The step of constructing a weeding knowledge graph based on the historical weeding instance classification information and the weeding benefit evaluation information, constructing a heterogeneous information network using the weeding knowledge graph, and constructing a herbicide recommendation model specifically includes: Acquire historical weeding instance classification information and weeding benefit evaluation information, and form entity triples of field environment characteristics, field weed characteristics, and weeding scheme characteristics based on the historical weeding instance classification information; A weed control knowledge graph is constructed according to the entity triples, and a MetaPath random walk algorithm is used to perform representation learning on the weed control knowledge graph to obtain entity nodes corresponding to the characteristics of the pastoral environment, and a meta-path is generated according to the connections between the entity nodes; The weed control benefit evaluation information is used as an auxiliary feature of the meta-path to form a heterogeneous information network, the heterogeneous information network is represented by a graph, and the cosine similarity between the meta-paths is calculated according to the graph representation; Establishing connections between each meta-path based on the calculated cosine similarity, obtaining neighbor nodes corresponding to the farmland weed feature node and the weed control scheme feature node, and constructing an adjacency matrix based on the neighbor nodes; The graph representation of the heterogeneous information network is stored in the form of an adjacency matrix, a herbicide recommendation model is constructed based on a graph neural network, and the herbicide recommendation model is deeply learned and trained using the graph representation of the heterogeneous information network to obtain a herbicide recommendation model that meets expectations.