Pest identification system and method based on data classification

Abstract

The invention discloses a pest recognition system and method based on data classification, and relates to the technical field of pest recognition. A trained pest classification model is used for recognition and classification, a pest recognition data set is constructed, if the recognized pest damage degree exceeds the expectation, pest outbreak data is collected, and if the recognized pest damage degree exceeds the expectation, the pest outbreak data is sent to a database; performing inducement analysis on insect attack outbreak in the pine forest area by using the trained insect attack outbreak factor identification model, generating an insect attack value according to insect attack outbreak data, and when the insect attack value exceeds an insect attack threshold value, outputting a planting processing scheme by a planting processing scheme library according to correspondence between inducement characteristics and the planting processing scheme; and when the adjustment of the pine forest planting state in the pine forest area does not achieve an expected effect, outputting a corresponding pest treatment scheme by the pest emergency treatment knowledge graph. By constructing a multi-stage alarm mechanism and a multi-stage emergency processing mechanism, when the insect pest breaks out, insect pest processing is realized at different levels, and the influence of the insect pest is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of pest identification, and in particular to a pest identification system and method based on data classification. Background Art

[0002] As a member of the family of the order Coleoptera, the tip beetle is a natural enemy of pine species, especially precious species such as Pinus yunnanensis and Pinus armandii. Their large-scale outbreak can quickly destroy an entire pine forest, causing huge losses of forest resources. In addition, since pine forests play an important role in the ecosystem, such as maintaining water and soil and regulating the climate, the rampant spread of the tip beetle may also have a profound impact on the ecological environment. In order to effectively control the damage caused by the tip beetle, comprehensive prevention and control measures need to be taken, including forestry management, biological control, chemical control, and enhanced quarantine.

[0003] The impact of the tip-cutting borer on Yunnan pine is extremely significant. They can directly feed on pine shoots and hibernate in the trunks in winter. The damage caused by tip and trunk borers causes the growth of Yunnan pine to be stunted, the leaves to turn red and fall off, and eventually the whole tree to die. In addition, the pests may also spread viruses and pathogens, aggravating the occurrence of pests and threatening the forest ecological security and forestry economic production in Yunnan.

[0004] In the Chinese invention patent with application publication number CN113743513A, a pest classification method and device based on multi-feature data is disclosed; the method includes: obtaining a picture to be identified; extracting multiple feature data of the picture; determining the optimal information combination method; combining the feature data with the original picture information according to the optimal information combination method to obtain fusion information; inputting the fusion information into the pest classification model for identification to obtain the pest classification result; the pest classification model is a trained deep convolutional network model. It can effectively identify the pest category, has strong feature expression and classification capabilities, and provides information for subsequent intelligent variable spraying and other pest control; and has good classification accuracy for normal pictures and enlarged images not targeting insects.

[0005] Before preventing and controlling pests, it is necessary to identify and classify various types of pests. The existing identification method usually collects image data of insects through image acquisition devices, and then determines information such as the type and number of insects through image recognition. However, in fact, there may be certain similarities between the morphology of different types of insects. Especially when the current planting status of the pine forest meets the conditions for pest outbreaks, the number of insects with similar species and morphology may increase significantly at the same time, resulting in a greater risk of misidentification and classification of pests. Furthermore, when there may be a simultaneous outbreak of multiple types of pests, if targeted treatment is not carried out in a timely manner, the economic pine forest may also suffer large-scale gnawing, resulting in a large degree of economic losses.

[0006] To this end, the present invention provides a pest identification system and method based on data classification. Summary of the invention

[0007] 1. Technical issues to be resolved

[0008] In view of the deficiencies of the prior art, the present invention provides a pest identification system and method based on data classification, which uses a trained pest classification model for identification and classification, constructs a pest identification data set, and if the identified pest damage exceeds expectations, collects pest outbreak data, uses a trained pest outbreak factor identification model to perform inducement analysis on the pest outbreak in the pine forest area, generates pest values ​​from the pest outbreak data, and when the pest value exceeds the pest threshold, outputs the planting treatment plan from the planting treatment plan library based on the correspondence between the inducement characteristics and the planting treatment plan; when the adjustment of the pine forest planting state in the pine forest area does not achieve the expected effect, the pest emergency treatment knowledge map outputs the corresponding pest treatment plan. By constructing a multi-level alarm mechanism and a multi-level emergency treatment mechanism, when there is already an outbreak of pests, pest treatment is achieved at different levels to reduce the impact of pests, thereby solving the problems in the background technology.

[0009] (II) Technical solution

[0010] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0011] A pest identification method based on data classification includes collecting sound and image data of pests in a pine forest area when the collected pine forest growth status data deviates from a target value beyond expectations;

[0012] Using image and sound features as input, the trained pest classification model is used for identification and classification to build a pest identification data set;

[0013] If the extent of the identified pest damage exceeds expectations, collect pest outbreak data, use the trained pest outbreak factor identification model to analyze the causes of pest outbreaks in the pine forest area, and obtain outbreak cause data;

[0014] Generate pest value from pest outbreak data. When the pest value exceeds the pest threshold, the planting treatment plan library outputs the planting treatment plan based on the correspondence between the cause characteristics and the planting treatment plan.

[0015] When the adjustment of the pine forest planting status in the pine forest area fails to achieve the expected effect, the corresponding pest treatment plan is output from the pest emergency treatment knowledge map based on the characteristics of the pest outbreak.

[0016] Furthermore, the stand density, stand aspect and connectivity of landscape patches in the pine forest area were collected and recorded, and then summarized to generate a data set of pine forest growth status;

[0017] The state coefficient is generated from the pine forest growth state data set. If the state coefficient exceeds the fluctuation threshold, several data collection points are evenly set in the pine forest area, pest images are collected at the data collection points, and the sounds of pest activities are recorded and summarized to generate the pest audio and video data set.

[0018] Furthermore, the sound and image data in the pest audio and video data set are annotated, including the pest species and sound features, and after pre-processing the pest image and sound data, feature extraction is performed on the image and sound data respectively;

[0019] The feature vectors of the image and the sound are concatenated or fused to form a comprehensive feature vector, a multimodal neural network is trained by the comprehensive feature vector, and a trained pest classification model is obtained after testing and verification.

[0020] Furthermore, the collected pest image and sound data are preprocessed, and the corresponding image features and sound features are obtained through feature extraction; the image and sound features are used as input, and the trained pest classification model is used for identification and classification to obtain the corresponding identification and classification results, which are summarized to generate a pest identification data set.

[0021] Furthermore, after obtaining the types and quantities of insect pests in the pine forest area, the types and quantities are used as inputs to perform damage scoring using the trained insect pest damage assessment model to obtain corresponding damage scores;

[0022] The damage scores of several different insect pests are continuously obtained and summarized. If the sum of the damage scores exceeds the preset damage threshold, an inducement analysis instruction is issued to the outside.

[0023] Furthermore, the sensor network in the pine forest area collects environmental data in real time, associates the planting status data, environmental condition data with the pest identification data, and aggregates them to establish a pest outbreak data set;

[0024] After receiving the cause analysis instruction, the data in the pest outbreak data set is used as input, and the trained pest outbreak factor identification model is used to perform cause analysis on the pest outbreak in the pine forest area, identify the outbreak causes that may cause the pest outbreak, and generate a cause data set after aggregation.

[0025] Furthermore, after identifying the pests in the pine forest area, the trained classifier divides the pests into several settlements according to their distribution status, and determines the damage score within each settlement and the location information between each settlement;

[0026] The pest value is constructed based on the hazard score of the pest colony and its location information. If the pest value exceeds the pest threshold, a first-level alarm command is issued to the outside.

[0027] Furthermore, by collecting or formulating several planting treatment plans in advance, a planting treatment plan library is generated after being summarized;

[0028] After receiving the first-level alarm command, the outbreak cause data in the pine forest area is feature extracted to obtain the corresponding cause characteristics; based on the correspondence between the cause characteristics and the planting treatment plan, the planting treatment plan library outputs the corresponding planting treatment plan.

[0029] Further, the planting treatment plan is executed to adjust the planting status in the pine forest area, and after the preset observation period ends, the pest data is continuously collected and identified through the sensor network in the pine forest area;

[0030] If the number of first-level alarm instructions received continuously exceeds expectations, the time node when an alarm instruction is received will be used as the alarm node, and the distribution of alarm nodes and the proportion of each pest value exceeding the pest threshold will be used as the warning ratio.

[0031] Furthermore, the warning value Jtp is constructed based on the distribution of the alarm nodes and the warning ratio. If the obtained warning value Jtp exceeds the warning threshold, a secondary alarm instruction is issued to the outside. The method of constructing the warning value Jtp is as follows:

[0032]

[0033] Where: |t1-t0| is the receiving interval of the first-level alarm command, m is the number of alarm nodes, x i is the warning ratio of the environmental conditions at the i-th alarm node; e can take the value of 2.718; weight coefficient: 0≤α≤1, 0≤β≤1, and α+β=1.

[0034] Furthermore, taking pest outbreak emergency treatment as the target word, a pest emergency treatment knowledge graph is pre-constructed; after receiving the secondary alarm command, feature extraction is performed on each data in the pest outbreak data set to obtain corresponding pest outbreak features; and the corresponding pest treatment plan is output from the pest emergency treatment knowledge graph.

[0035] The pest identification system based on data classification includes: a data collection unit, which collects sound and image data of pests in the pine forest area when the deviation between the collected pine forest growth status data and the target value exceeds the expectation;

[0036] The pest classification and recognition unit takes image and sound features as input, uses the trained pest classification model to perform recognition and classification, and constructs a pest recognition data set;

[0037] The hazard analysis unit collects pest outbreak data if the identified pest damage level exceeds expectations, uses the trained pest outbreak factor identification model to conduct a cause analysis of the pest outbreak in the pine forest area, and obtains outbreak cause data;

[0038] The pest warning unit generates pest values ​​from pest outbreak data. When the pest value exceeds the pest threshold, the planting treatment plan library outputs the planting treatment plan based on the correspondence between the inducement characteristics and the planting treatment plan.

[0039] When the adjustment of the pine forest planting status in the pine forest area fails to achieve the expected effect, the emergency treatment unit outputs the corresponding pest treatment plan based on the pest outbreak characteristics from the pest emergency treatment knowledge map.

[0040] (III) Beneficial effects

[0041] The present invention provides a pest identification system and method based on data classification, which has the following beneficial effects:

[0042] 1. By combining sound features and image features, the accuracy and confidence of pest identification and classification are higher than that of single sound recognition or image recognition and classification.

[0043] 2. Identify the hazard points in the pine forest area and determine the degree of hazard of the current type of pests. This can determine whether the pests need to be treated. Timely treatment of pests can prevent further outbreaks. Determine the causes of pest outbreaks through data analysis and identification. In scenarios where pest outbreaks already exist, targeted treatment of current pests is carried out based on the causes of pest outbreaks, which can delay pest outbreaks and damage.

[0044] 3. The pest value Cop is constructed based on the current pest outbreak status. The current degree of pest outbreak can be quantified and evaluated based on the pest value Cop. If the severity exceeds expectations, this can be used as a starting point and the pest can be dealt with in a timely manner.

[0045] 4. The planting treatment plan library outputs the planting treatment plan, adjusts the current planting status of the pine forest according to the planting treatment plan, conducts preliminary treatment on pests, and reduces the risk of further outbreaks of pests based on the identification and classification of pests in the pine forest area.

[0046] 5. By monitoring and observing the previous treatment measures to establish warning values, a secondary alarm is built based on the warning values ​​to achieve the construction of a multi-level alarm mechanism. If the measures taken previously fail to achieve the expected results, further treatment can be made in a timely manner to avoid the impact of multiple types of pest outbreaks on the growth of the pine forest group.

[0047] 6. Output the corresponding pest treatment plan from the pest emergency treatment knowledge graph to directly treat the pests, such as spraying pesticides, introducing natural enemies of pests, or treating the environmental conditions in the pine forest area; further treat the pests at different levels to avoid further deterioration of the pest outbreak; conduct multi-level judgments on the severity of the pest outbreak, build a multi-level alarm mechanism and a multi-level emergency treatment mechanism, so that when a pest outbreak already exists, pest treatment can be achieved at different levels to reduce the impact of pests. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a schematic diagram of the process flow of the pest identification method based on data classification of the present invention;

[0049] Figure 2 It is a schematic diagram of the structure of the pest identification system based on data classification of the present invention. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] See also Figure 1 The present invention provides a pest identification method based on data classification, comprising:

[0052] Step 1: When the deviation between the collected pine forest growth status data and the target value exceeds expectations, collect the sound and image data of insect pests in the pine forest area;

[0053] The step 1 includes the following contents:

[0054] Step 101, after determining the pine forest area of ​​the pine forest, collect and record the pine forest growth status data such as the stand canopy density Yp, stand slope aspect Yb and landscape patch connectivity Yo of the pine forest area through a collection device carried by the drone, such as an image collection device and a surveying and mapping device, and generate a pine forest growth status data set after aggregation;

[0055] Under dimensionless conditions, the state coefficient Btp is generated from the pine forest growth state data set in the following way:

[0056]

[0057] Weight coefficient: 0≤F1≤1, 0≤F2≤1 and F2+F1=1; the weight coefficient is obtained by referring to the hierarchical analysis method;

[0058] Based on historical data and expectations for pine forest plantation management, fluctuation thresholds are set in advance;

[0059] If the state coefficient Btp exceeds the fluctuation threshold, it means that there may be a certain deviation between the current growth state of the pine forest and the expected value, and the gap between the ideal condition and the expected value is greater than expected. In this case, it may cause an outbreak of pests or pests may already exist, affecting the further growth of the pine forest. At this time, a data collection instruction is issued to the outside.

[0060] When in use, the current growth state of the pine forest area can be judged based on the constructed state coefficient Btp. When the current growth or development state of the pine forest is abnormal, there may be certain insect pest outbreaks in the pine forest area, or insect pests have already occurred. Based on this as a starting point, insect pests can be monitored in advance to ensure the growth state of the pine forest.

[0061] Step 102: after receiving the data collection instruction, several data collection points are evenly set in the pine forest area, and images of pests, such as images of adults and larvae in different environments, are collected at the data collection points using mobile devices, drones or fixed cameras; sounds of pest activities in different environments are recorded using high-sensitivity microphones or acoustic sensors, such as the chirping of locusts and the crawling of aphids, etc., and the sounds are summarized to generate a pest audio and video data set;

[0062] When using, combine the contents in steps 101 and 102:

[0063] By collecting image and sound data of pests in pine forest areas, the classification results are more accurate and reliable when used for pest identification and classification.

[0064] Before preventing and controlling pests, it is necessary to identify and classify various types of pests. The existing identification method usually collects image data of insects through image acquisition devices, and then determines information such as the type and number of insects through image recognition. However, in fact, there may be certain similarities between the morphology of different types of insects. Especially when the current planting status of the pine forest meets the conditions for an insect pest outbreak, the number of insects of similar species and morphology may increase significantly at the same time, resulting in a greater risk of misidentification and classification of pests. Furthermore, when there may be an outbreak of multiple types of pests at the same time, if targeted treatment is not carried out in a timely manner, the economic pine forest may also suffer large-scale gnawing, resulting in a large degree of economic losses.

[0065] Step 2: Using image and sound features as input, use the trained pest classification model to identify and classify, and build a pest identification data set;

[0066] The step 2 includes the following contents:

[0067] Step 201: label the sound and image data in the pest audio and video data set, including the pest species and sound features, such as appearance features, etc.; after pre-processing the pest image and sound data, extract features from the image and sound data, such as Mel frequency cepstrum coefficients and spectrograms, etc.;

[0068] The feature vectors of the image and the sound are concatenated or fused to form a comprehensive feature vector, a multimodal neural network is trained by the comprehensive feature vector, and a trained pest classification model is obtained after testing and verification;

[0069] Step 202: pre-process the collected pest image and sound data, and obtain corresponding image features and sound features through feature extraction;

[0070] Using the features of images and sounds as input, the trained pest classification model is used for identification and classification, and the corresponding identification and classification results, including information such as pest types and confidence levels, are obtained, and then aggregated to generate a pest identification data set;

[0071] When using, combine the contents in steps 201 and 202:

[0072] On the basis of labeling the sound data and image data, a trained pest classification model is obtained. By combining the sound features and image features, the accuracy and confidence of pest identification and classification are higher than that of single sound recognition or image recognition and classification.

[0073] Step 3: If the identified pest damage level exceeds expectations, collect pest outbreak data, use the trained pest outbreak factor identification model to perform cause analysis on the pest outbreak in the pine forest area, and obtain outbreak cause data;

[0074] The step three includes the following contents:

[0075] Step 301: training a neural convolutional network using the labeled sample data to obtain a trained pest damage assessment model;

[0076] After obtaining the types and quantities of insect pests in the pine forest area, the system uses the trained insect pest hazard assessment model as input to perform hazard scoring and obtain the corresponding hazard score. The hazard scores of several different insect pests are continuously obtained and summarized. If the sum of the hazard scores exceeds the preset hazard threshold, an inducement analysis instruction is issued to the outside.

[0077] When in use, after identifying the pest category information in the pine forest area, the hazard points in the pine forest area are identified based on historical data and labeled data, and the degree of hazard of the current type of pests is determined. It can be judged whether the pests need to be treated. Timely treatment of pests can prevent further outbreaks of pests;

[0078] Step 302: Arrange a sensor network in the pine forest area, collect environmental data in real time, such as environmental condition data such as temperature and humidity, associate the planting status data, environmental condition data with the pest identification data, and establish a pest outbreak data set after aggregation;

[0079] Step 303: train a machine learning algorithm with the labeled sample data to obtain a trained pest outbreak factor identification model; after receiving the cause analysis instruction, use the data in the pest outbreak data set as input, use the trained pest outbreak factor identification model to perform cause analysis on the pest outbreak in the pine forest area, identify and obtain the outbreak causes that may cause the pest outbreak, and generate a cause data set after aggregation;

[0080] When using, combine the contents in steps 301 to 303:

[0081] When there is already a trend of pest outbreak in the pine forest area, various data are collected in the pine forest area, and the causes of the pest outbreak are determined through data analysis and identification. In the scenario where a pest outbreak has already occurred, targeted treatment of the current pests is carried out based on the causes of the pest outbreak, which can delay the pest outbreak and damage.

[0082] Step 4: Generate a pest value Cop from the pest outbreak data. When the pest value Cop exceeds the pest threshold, the planting treatment plan library outputs the planting treatment plan based on the correspondence between the inducement characteristics and the planting treatment plan.

[0083] The step 4 includes the following contents:

[0084] Step 401: After identifying the pests in the pine forest area, the pests are divided into several settlements by the trained classifier according to the distribution of the pests, and the damage score in each settlement and the location information between each settlement are determined;

[0085] Under dimensionless conditions, the pest value Cop is constructed based on the damage score of the pest settlement and its location information as follows:

[0086]

[0087] Where: N is the number of pest colonies, X i is the location of the i-th pest settlement, X0 is the center location, k is the parameter controlling distance attenuation, and its value is within [0, 1]. i ) is the weight coefficient of the i-th pest settlement, and its value is within [0, 1]; S(X i ) is the hazard score of the i-th pest colony; weight coefficient: 0≤K1≤1, 0≤K2≤1, and K1+K2=1;

[0088] The pest threshold is set in advance based on historical data and pest management expectations. If the pest value Cop exceeds the pest threshold, it means that the pest situation in the pine forest area may be serious and may be in an outbreak state. At this time, a first-level alarm command is issued to the outside.

[0089] When in use, when pests have already broken out, the pest value Cop is constructed based on the current pest outbreak status. The current pest outbreak degree can be quantified and evaluated based on the pest value Cop. If the severity exceeds expectations, this can be used as a starting point to handle the pests in a timely manner.

[0090] Step 402: by pre-collecting or formulating a number of planting treatment plans, such as irrigation, fertilization or pest control, the planting treatment plan library is generated after being summarized; after receiving the first-level alarm instruction, the outbreak inducement data in the pine forest area is feature extracted to obtain the corresponding inducement feature; according to the correspondence between the inducement feature and the planting treatment plan, the planting treatment plan library outputs the corresponding planting treatment plan;

[0091] When using, combine the contents in steps 401 to 402:

[0092] After the cause of the current pest outbreak has been identified, the planting treatment plan library outputs the planting treatment plan, adjusts the current planting status of the pine forest based on the planting treatment plan, and conducts preliminary treatment on the pests. On the basis of identifying and classifying the pests in the pine forest area, the risk of further pest outbreaks is reduced.

[0093] Step 5: When the adjustment of the pine forest planting status in the pine forest area does not achieve the expected effect, the corresponding pest treatment plan is output from the pest emergency treatment knowledge map according to the characteristics of the pest outbreak;

[0094] The step five includes the following contents:

[0095] Step 501: Execute the planting treatment plan to adjust the planting status in the pine forest area. After the preset observation period ends, the pest data is continuously collected and identified through the sensor network in the pine forest area. If the number of consecutive first-level alarm instructions received exceeds the expectation, the time node of receiving an alarm instruction is used as the alarm node, and the alarm node distribution and the ratio of each pest value exceeding the pest threshold are used as the warning ratio.

[0096] Step 502: Under dimensionless conditions, the warning value Jtp is constructed according to the distribution of alarm nodes and the warning ratio, in the following manner:

[0097]

[0098] Where: |t1-t0| is the receiving interval of the first-level alarm command, m is the number of alarm nodes, x i is the warning ratio of the environmental conditions at the i-th alarm node; e can take the value of 2.718; weight coefficient: 0≤α≤1, 0≤β≤1, and α+β=1;

[0099] Based on historical data and pest management expectations, the warning threshold is pre-set; if the obtained warning value Jtp exceeds the warning threshold, it means that the pest outbreak problem in the pine forest area has not been solved. At this time, a secondary alarm command is issued to the outside;

[0100] When in use, on the basis of having received the first-level alarm instruction and having taken measures, the warning value is established by monitoring and observing the previous treatment measures, and the second-level alarm is established based on the warning value, thereby realizing the establishment of a multi-level alarm mechanism. If the measures taken before fail to achieve the expected results, further treatment can be made in time to avoid the impact of the outbreak of multiple types of pests on the growth of the pine forest group;

[0101] Step 503: Taking emergency treatment of insect pest outbreak as the target word, after deep retrieval and entity relationship construction, a knowledge graph of emergency treatment of insect pest outbreak is pre-constructed;

[0102] After receiving the second-level alarm command, the features of each data in the pest outbreak data set are extracted to obtain the corresponding pest outbreak features; based on the correspondence between the pest outbreak features and the pest treatment plan, the corresponding pest treatment plan is output from the pest emergency treatment knowledge graph;

[0103] When using, combine the contents in steps 501 to 503:

[0104] The pest emergency treatment knowledge graph outputs the corresponding pest treatment plan to directly treat the pests, such as spraying pesticides, introducing natural enemies of pests, or treating the environmental conditions in the pine forest area; by further treating the pests at different levels, the pest outbreak can be avoided from further deterioration;

[0105] Based on the accurate identification of the pest situation in the pine forest area, a multi-level judgment is made on the severity of the pest outbreak, and a multi-level alarm mechanism and a multi-level emergency response mechanism are established. Therefore, when an pest outbreak already exists, pest control can be achieved at different levels to reduce the impact of pests.

[0106] See also Figure 2 The present invention provides a pest identification system based on data classification, comprising:

[0107] The data collection unit collects the sound and image data of insect pests in the pine forest area when the deviation between the collected pine forest growth status data and the target value exceeds the expectation;

[0108] The pest classification and recognition unit takes image and sound features as input, uses the trained pest classification model to perform recognition and classification, and constructs a pest recognition data set;

[0109] The hazard analysis unit collects pest outbreak data if the identified pest damage level exceeds expectations, uses the trained pest outbreak factor identification model to conduct a cause analysis of the pest outbreak in the pine forest area, and obtains outbreak cause data;

[0110] The pest warning unit generates a pest value Cop from the pest outbreak data. When the pest value Cop exceeds the pest threshold, the planting treatment plan library outputs the planting treatment plan based on the correspondence between the inducement characteristics and the planting treatment plan.

[0111] When the adjustment of the pine forest planting status in the pine forest area fails to achieve the expected effect, the emergency treatment unit outputs the corresponding pest treatment plan based on the pest outbreak characteristics from the pest emergency treatment knowledge map.

[0112] The Analytic Hierarchy Process (AHP) is a decision-making method that decomposes decision-related elements into levels such as goals, criteria, and plans, and conducts qualitative and quantitative analysis on this basis. It is particularly suitable for dealing with target systems with hierarchical and staggered evaluation indicators, and when the target value is difficult to describe quantitatively, the AHP is an effective decision-making tool.

[0113] The core of the hierarchical analysis method is to decompose the decision problem into multiple levels to form a hierarchical structure, which usually includes the target level, the criterion level, the sub-criterion level and the solution level. By solving the eigenvector of the judgment matrix, the priority weight of each element of each level to an element of the previous level is obtained, and the final weight of each alternative solution to the total goal is recursively merged by the weighted sum method, so as to find the optimal solution.

[0114] The method of constructing the knowledge graph of pest emergency treatment can refer to the following:

[0115] Collect and organize data. Data sources: Data related to emergency treatment of pests can be collected from multiple channels, including agricultural expert knowledge, literature data, field research data, etc. Data organization: Organize the collected data, including removing duplicate data, correcting erroneous data, and unifying data formats.

[0116] Knowledge modeling, definition of domain knowledge system: Based on the characteristics and needs of emergency treatment of pests, define the domain knowledge system, including the types of pests, symptoms, transmission routes, prevention and control methods, etc. Determine the relationship type: clarify the relationship type between different knowledge points, such as "symptoms-pests", "prevention and control methods-pests", etc.

[0117] Knowledge extraction, semi-structured data extraction: For data that already has a certain structure (such as tables, databases, etc.), it can be extracted through rules. Unstructured data extraction: For unstructured data such as text and images, natural language processing (NLP) and image recognition technology can be used for extraction. For example, NLP technology can be used to automatically obtain knowledge about pests and their control methods from literature and expert data.

[0118] Knowledge fusion, entity recognition and linking: Identify the same entities in different data sources and link them to eliminate redundancy and conflict. Relationship integration: Integrate the relationships between the same entities in different data sources to form a complete knowledge network.

[0119] Knowledge storage, choose a graph database: Choose a graph database suitable for storing knowledge graphs, such as Neo4j. Data import and storage: Import the extracted and fused knowledge into the graph database for storage.

[0120] Knowledge reasoning and updating,Knowledge reasoning: Discover unknown knowledge based on existing knowledge, such as inferring possible pest types based on pest symptoms.,Knowledge updating: With the continuous addition of new data and the continuous development of technology, the knowledge graph needs to be updated and maintained regularly.

[0121] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0122] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0123] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only some logical function divisions. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0124] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, 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 solution of this embodiment.

[0125] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A pest identification method based on data classification, characterized in that: include, When the deviation between the collected pine forest growth status data and the target value exceeds expectations, the sound and image data of insect pests are collected in the pine forest area; the image and sound features are used as input, and the trained insect pest classification model is used for identification and classification to construct an insect pest identification data set; If the degree of damage caused by the identified pests exceeds expectations, pest outbreak data is collected, and the trained pest outbreak factor identification model is used to conduct a cause analysis of the pest outbreak in the pine forest area to obtain outbreak cause data; pest values ​​are generated from the pest outbreak data, and when the pest value exceeds the pest threshold, the planting treatment plan is output from the planting treatment plan library based on the correspondence between the cause characteristics and the planting treatment plan; when the adjustment of the pine forest planting status in the pine forest area does not achieve the expected effect, the pest emergency treatment knowledge map outputs the corresponding pest treatment plan based on the pest outbreak characteristics.

2. The pest identification method based on data classification according to claim 1, characterized in that: Collect and record the stand density, stand aspect and connectivity of landscape patches in the pine forest area, and generate a pine forest growth status data set after aggregation; The state coefficient is generated from the pine forest growth state data set. If the state coefficient exceeds the fluctuation threshold, several data collection points are evenly set in the pine forest area, pest images are collected at the data collection points, and the sounds of pest activities are recorded and summarized to generate the pest audio and video data set.

3. The pest identification method based on data classification according to claim 2, characterized in that: Label the sound and image data in the pest audio and video data set, including the pest species and sound features, and after pre-processing the pest image and sound data, perform feature extraction on the image and sound data respectively; The feature vectors of the image and the sound are concatenated or fused to form a comprehensive feature vector, a multimodal neural network is trained by the comprehensive feature vector, and a trained pest classification model is obtained after testing and verification.

4. The pest identification method based on data classification according to claim 3 is characterized in that: Preprocess the collected pest images and sound data, and obtain corresponding image features and sound features through feature extraction; Using the features of images and sounds as input, the trained pest classification model is used for identification and classification, the corresponding identification and classification results are obtained, and then summarized to generate a pest identification data set.

5. The pest identification method based on data classification according to claim 4, characterized in that: After obtaining the types and quantities of insect pests in the pine forest area, the types and quantities are used as inputs to perform damage scoring using the trained insect pest damage assessment model to obtain the corresponding damage scores; The damage scores of several different insect pests are continuously obtained and summarized. If the sum of the damage scores exceeds the preset damage threshold, an inducement analysis instruction is issued to the outside.

6. The pest identification method based on data classification according to claim 5, characterized in that: The sensor network in the pine forest area collects environmental data in real time, associates the planting status data, environmental condition data with pest identification data, and summarizes them to establish a pest outbreak data set; After receiving the cause analysis instruction, the data in the pest outbreak data set is used as input, and the trained pest outbreak factor identification model is used to perform cause analysis on the pest outbreak in the pine forest area, identify the outbreak causes that may cause the pest outbreak, and generate a cause data set after aggregation.

7. The pest identification method based on data classification according to claim 6, characterized in that: After identifying the pests in the pine forest area, the trained classifier divides the pests into several settlements according to their distribution status, and determines the damage score within each settlement and the location information between each settlement; The pest value is constructed based on the hazard score of the pest colony and its location information. If the pest value exceeds the pest threshold, a first-level alarm command is issued to the outside.

8. The pest identification method based on data classification according to claim 7, characterized in that: By collecting or formulating several planting treatment plans in advance, the planting treatment plan library is generated after being summarized; After receiving the first-level alarm command, the outbreak cause data in the pine forest area is feature extracted to obtain the corresponding cause characteristics; based on the correspondence between the cause characteristics and the planting treatment plan, the planting treatment plan library outputs the corresponding planting treatment plan.

9. The pest identification method based on data classification according to claim 8, characterized in that: Execute the planting treatment plan to adjust the planting status in the pine forest area. After the preset observation period ends, the sensor network in the pine forest area continuously collects and identifies pest data; If the number of first-level alarm instructions received continuously exceeds expectations, the time node when an alarm instruction is received will be used as the alarm node, and the distribution of alarm nodes and the proportion of each pest value exceeding the pest threshold will be used as the warning ratio.

10. The pest identification method based on data classification according to claim 9, characterized in that: The warning value Jtp is constructed based on the distribution of alarm nodes and the warning ratio. If the obtained warning value Jtp exceeds the warning threshold, a secondary alarm instruction is issued to the outside. The method of constructing the warning value Jtp is as follows: Where: |t1-t0| is the receiving interval of the first-level alarm command, m is the number of alarm nodes, x i is the warning ratio of the environmental conditions at the i-th alarm node; e can take the value of 2.718; Weight Coefficient: 0≤α≤1, 0≤β≤1, and α+β=1.

11. The pest identification method based on data classification according to claim 10, characterized in that: With pest outbreak emergency treatment as the target word, a pest emergency treatment knowledge graph is pre-constructed; after receiving the secondary alarm command, feature extraction is performed on the data in the pest outbreak data set to obtain the corresponding pest outbreak features; the corresponding pest treatment plan is output from the pest emergency treatment knowledge graph.

12. A pest identification system based on data classification, characterized in that: include, The data collection unit collects the sound and image data of insect pests in the pine forest area when the deviation between the collected pine forest growth status data and the target value exceeds the expectation; The pest classification and recognition unit takes image and sound features as input, uses the trained pest classification model to perform recognition and classification, and constructs a pest recognition data set; The hazard analysis unit collects pest outbreak data if the identified pest damage level exceeds expectations, uses the trained pest outbreak factor identification model to conduct a cause analysis of the pest outbreak in the pine forest area, and obtains outbreak cause data; The pest warning unit generates pest values ​​from pest outbreak data. When the pest value exceeds the pest threshold, the planting treatment plan library outputs the planting treatment plan based on the correspondence between the inducement characteristics and the planting treatment plan. When the adjustment of the pine forest planting status in the pine forest area fails to achieve the expected effect, the emergency treatment unit outputs the corresponding pest treatment plan based on the pest outbreak characteristics from the pest emergency treatment knowledge map.

Images