Termite control decision-making method, device and equipment based on search enhancement generation

Abstract

The application discloses a termite control decision-making method, device and equipment based on retrieval augmented generation. The retrieval augmented generation technology is deeply integrated with a large language model to construct an intelligent decision-making system, multi-modal termite monitoring data and historical control cases are collected, and a special knowledge base is built. The RAG technology is used to retrieve and match the optimal scheme of similar working conditions, and a large language model is used for reasoning to generate a targeted control decision. The termite monitoring data and the optimized scheme are accurately matched, and the control efficiency and ecological safety can be improved.

Description

Technical Field

[0001] This application belongs to the field of termite control technology, and in particular relates to a termite control decision-making method, apparatus and equipment based on search-enhanced generation. Background Technology

[0002] In water conservancy projects, structures such as earth-rock dams and earthen embankments are susceptible to termite infestation. Termite nests form penetrating seepage channels, directly threatening the safety of the project. Current termite control relies on manual inspections and single-sensor monitoring, which suffers from low efficiency, high missed detection rates (deeply buried nests are difficult to locate), and fragmented data. Traditional control decisions depend on human experience and fail to effectively integrate historical termite infestation cases with ecological constraints, resulting in low accuracy and overuse of pesticides (polluting water / soil). Summary of the Invention

[0003] This application provides a termite control decision-making method, apparatus, and equipment based on retrieval enhancement generation.

[0004] In a first aspect, embodiments of this application provide a termite control decision-making method based on retrieval enhancement, including: Obtain multimodal data sets of termites The multimodal dataset includes vibration feature vectors, environmental numerical vectors, visual trace quantization vectors, and risk prediction vectors, wherein... A data vector representing the nth type; A historical dataset was built based on multiple historical ant infestation data points from different historical time periods. ,in Indicates the first A subset of historical data at the moment of writing. Indicates the first Pen Moment The data types in each subset of historical data are the same as those in the multimodal data set. Based on the historical dataset and its corresponding historical prevention and control effect data, determine multiple sets of optimized prevention and control decisions at different historical time points. ,in Indicates the first The optimized prevention and control decision subset at the moment of writing. Indicates the first The first moment of writing The number of data points for each optimized prevention and control decision subset is determined based on the operating condition information. Based on the correspondence between time points, a bijective relationship is established between the historical dataset and the optimized prevention and control decision set. This yields a set of monitoring data and prevention / control decisions. The text embedding model is used to perform clustering and matching retrieval on the multimodal dataset, historical dataset, and monitoring data-prevention decision set to infer the retrieval results; The search results, current engineering structure parameters, and ecologically sensitive area constraints are input into a pre-trained large language model to infer and generate a precise prevention and control decision that matches the current working conditions. The precise prevention and control decision includes at least ant nest targeted treatment parameters.

[0005] Optionally, acquiring the multimodal data set of termites includes: Real-time collection of acoustic vibration signals, carbon dioxide concentration, soil / air temperature and relative humidity via intelligent monitoring terminals yields vibration data and environmental data. The images to be inspected are obtained through drone inspections, and then the images are input into a pre-trained deep learning model for termite trace recognition to obtain termite activity trace data. Based on the meteorological forecast data within the preset period, the specific time window for the breeding and swarming period is predicted, and the risk area for the spread of termite populations is simulated by combining the termite nest location information, thus obtaining risk prediction data; The vibration data, environmental data, termite activity trace data, and risk prediction data are time-stamp aligned and spatial coordinate registered to obtain vibration feature vectors, environmental numerical vectors, visual trace quantization vectors, and risk prediction vectors, which constitute the termite multimodal data set at the current moment.

[0006] Optionally, the historical ant infestation data includes historical prevention and control plans corresponding to each historical time point; The step of determining a set of optimized prevention and control decisions for multiple historical time points based on the historical dataset and its corresponding historical prevention and control effect data includes: For historical datasets with multiple historical time points and their corresponding historical prevention and control effect data, the following steps are performed in parallel: Historical prevention and control measures are quantitatively scored based on historical prevention and control effectiveness data from multiple dimensions and their corresponding preset weight values. Historical prevention and control measures with scores greater than a first preset threshold are marked and used as a candidate prevention and control decision set. For any two candidate prevention and control decisions in the candidate prevention and control decision set, the similarity is calculated using a cosine similarity algorithm based on parameter vectors. If the similarity is greater than a second preset threshold, the candidate prevention and control decision with the lower score is removed from the candidate prevention and control decision set to obtain the optimized prevention and control decision subset corresponding to that historical time point. Multiple subsets of optimization and prevention decisions corresponding to different historical time points are combined into an optimization and prevention decision set.

[0007] Optionally, the step of using a text embedding model to perform clustering and matching retrieval on the multimodal dataset, historical dataset, and monitoring data-prevention decision set to infer retrieval results includes: Multimodal data sets and historical datasets Simultaneously, the text embedding model is used for cluster analysis to obtain a set of cluster categories. ,in Represents a multimodal data set Clustering categories, Represents historical datasets Clustering categories for each subset of historical data; Establish a multimodal dataset Historical datasets and cluster category set bijective relation ; Sure Corresponding cluster categories ; Filtering and Collection of similar historical data cluster categories: ; Sure Corresponding historical data subset ; Sure The corresponding optimized prevention and control decision set yielded the following search results: .

[0008] Optionally, the targeted treatment parameters for the ant nest include at least: the three-dimensional spatial coordinates of the ant nest center, the number of holes, the drilling angle, the maximum grouting pressure, and the reagent ratio.

[0009] Optionally, the current engineering structural parameters include at least: the dam material properties, the distribution of the seepage prevention structure, and the maximum allowable grouting pressure.

[0010] Secondly, embodiments of this application provide a termite control decision-making device based on retrieval enhancement generation, comprising: The acquisition module is used to acquire multimodal data sets of termites. The multimodal dataset includes vibration feature vectors, environmental numerical vectors, visual trace quantization vectors, and risk prediction vectors, wherein... A data vector representing the nth type; The first module is used to build a historical dataset based on multiple historical ant infestation data from different historical time points. ,in Indicates the first A subset of historical data at the moment of writing. Indicates the first Pen Moment The data types in each subset of historical data are the same as those in the multimodal data set. The determination module is used to determine a set of optimized prevention and control decisions for multiple historical time points based on the historical dataset and its corresponding historical prevention and control effect data. ,in Indicates the first The optimized prevention and control decision subset at the moment of writing. Indicates the first The first moment of writing The number of data points for each optimized prevention and control decision subset is determined based on the operating condition information. The second module is used to establish a bijective relationship between the historical dataset and the optimized prevention and control decision set based on the correspondence between time points. This yields a set of monitoring data and prevention / control decisions. The retrieval module is used to perform clustering and matching retrieval on the multimodal dataset, historical dataset, and monitoring data-prevention decision set using a text embedding model, and infer the retrieval results. The decision module is used to input the retrieval results, current engineering structure parameters, and ecologically sensitive area constraints into a pre-trained large language model, and to infer and generate a target-precision prevention and control decision that matches the current working conditions. The target-precision prevention and control decision includes at least ant nest targeted treatment parameters.

[0011] Thirdly, embodiments of this application provide an electronic device, which includes: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, it implements the steps of the termite control decision-making method based on retrieval enhancement generation as described in any embodiment of the first aspect.

[0012] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer program instructions, which, when executed by a processor, implement the steps of the termite control decision-making method based on retrieval enhancement as described in any embodiment of the first aspect.

[0013] Fifthly, embodiments of this application provide a computer program product, which is stored in a storage medium and executed by at least one processor to implement the steps of the termite control decision-making method based on retrieval enhancement provided in the first aspect of embodiments of this application.

[0014] The termite control decision-making method, apparatus, and equipment based on retrieval-enhanced generation (RAG) technology in this application deeply integrates retrieval-enhanced generation (RAG) technology with a large language model to construct an intelligent decision-making system. It collects multimodal termite monitoring data and historical control cases to build a dedicated knowledge base. Through RAG technology, it retrieves and matches the optimal solution for similar working conditions, and then generates targeted control decisions through large language model reasoning. This achieves accurate matching between termite monitoring data and optimized solutions, improving control efficiency and ecological security. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating a termite control decision-making method based on retrieval enhancement provided in an embodiment of this application; Figure 2 This is a schematic diagram of the structure of a termite control decision-making device based on retrieval enhancement provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0017] Figure label: The termite control decision-making device 200 based on enhanced retrieval comprises an acquisition module 201, a first establishment module 202, a determination module 203, a second establishment module 204, a retrieval module 205, and a decision-making module 206. Electronic device 300, processor 301, memory 302, communication interface 303, bus 310. Detailed Implementation

[0018] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.

[0019] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0020] It should be noted that the acquisition, storage, use, and processing of data in this application embodiment all comply with the relevant provisions of national laws and regulations.

[0021] In water conservancy projects, structures such as earth-rock dams and earthen embankments are susceptible to termite infestation. Termite nests form penetrating seepage channels, directly threatening the safety of the project. Current termite control relies on manual inspections and single-sensor monitoring, which suffers from low efficiency, high missed detection rates (deeply buried nests are difficult to locate), and fragmented data. Traditional control decisions depend on human experience and fail to effectively integrate historical termite infestation cases with ecological constraints, resulting in low accuracy and overuse of pesticides (polluting water / soil).

[0022] While existing technologies incorporate monitoring methods such as AI vision and ground-penetrating radar, they lack a coherent knowledge base linking "monitoring data" and "prevention and control decisions," and also lack the ability to accurately match specialized domain knowledge. Retrieval-enhanced generation (RAG) and large language models have enabled data-driven decision-making in areas such as water supply scheduling, but they are not adapted to termite control scenarios—they cannot integrate multimodal monitoring data (vibration / gas / visual) and historical prevention and control effect data for collaborative reasoning, resulting in delayed decision-making and poor ecological compatibility.

[0023] Although current termite control technologies for water conservancy projects are trending towards intelligence, precision, ecology, and collaboration, and are upgrading from single monitoring to multimodal fusion and AI recognition, and from indiscriminate extermination to precise positioning and targeted application, there are still gaps such as insufficient intelligence in the decision-making process, weak cross-scenario versatility, and a lack of closed-loop feedback on effects.

[0024] To address the problems in the related technologies, embodiments of this application provide a termite control decision-making method, apparatus, and equipment based on retrieval enhancement generation.

[0025] The termite control decision-making method based on retrieval enhancement generation provided in this application will be described in detail below with reference to the accompanying drawings, through specific embodiments and application scenarios.

[0026] Figure 1 A flowchart illustrating the termite control decision-making method based on retrieval enhancement generation, according to an embodiment of this application, is shown. Figure 1 As shown, the termite control decision-making method based on retrieval enhancement can specifically include the following steps: S101. Obtain the multimodal data set of termites. The multimodal dataset includes vibration feature vectors, environmental numerical vectors, visual trace quantization vectors, and risk prediction vectors, wherein... A data vector representing the nth type; S102. Establish a historical dataset based on multiple historical ant infestation data from different historical time points. ,in Indicates the first A subset of historical data at the moment of writing. Indicates the first Pen Moment The data types in each subset of historical data are the same as those in the multimodal data set. S103. Based on the historical dataset and its corresponding historical prevention and control effect data, determine multiple sets of optimized prevention and control decisions at different historical time points. ,in Indicates the first The optimized prevention and control decision subset at the moment of writing. Indicates the first The first moment of writing The number of data points for each optimized prevention and control decision subset is determined based on the operating condition information. S104. Based on the correspondence between time points, establish a bijective relationship between the historical dataset and the optimized prevention and control decision set. This yields a set of monitoring data and prevention / control decisions. S105. Using a text embedding model, clustering and matching retrieval are performed on the multimodal dataset, historical dataset, and monitoring data-prevention and control decision set to infer the retrieval results; S106. Input the search results, current engineering structure parameters, and ecologically sensitive area constraints into a pre-trained large language model, and infer to generate a target-precision prevention and control decision that matches the current working conditions. The target-precision prevention and control decision includes at least ant nest targeted treatment parameters.

[0027] Therefore, by deeply integrating Retrieval Enhancement Generation (RAG) technology with large language models, an intelligent decision-making system is constructed. Multimodal termite monitoring data and historical control cases are collected to build a dedicated knowledge base. The optimal solution for similar working conditions is retrieved and matched through RAG technology, and targeted control decisions are generated through reasoning by large language models. This achieves accurate matching between termite monitoring data and optimized solutions, which can improve control efficiency and ecological security.

[0028] The specific implementation methods for each of the above steps are described below.

[0029] In some embodiments, in S101, multi-source sensor data, termite activity trace data, and risk prediction data are acquired, and these heterogeneous data are time-stamp aligned and spatial coordinate registered to construct a termite multimodal data set at the current time t.

[0030] In practice, intelligent monitoring terminals deployed on dams and canals collect monitoring data in real time and transmit the data back via NB-IoT or LoRaWAN networks. Piezoelectric vibration sensors collect acoustic vibration signals in the frequency range of 0.1Hz-500Hz. After bandpass filtering, specific time-frequency characteristics of termites gnawing on wood or soil are extracted, including but not limited to pulse duration and energy spectral density. Simultaneously, the carbon dioxide concentration (ppm), soil / air temperature (°C), and relative humidity (%RH) around the monitoring points are also collected.

[0031] In practice, high-definition images from drone inspections and video streams from ground inspection robots are used to obtain images to be inspected. These images are then input into a pre-trained deep learning model for termite trace recognition to obtain information on termite trails, swarming holes, and mud cover traces, including but not limited to the length and width of termite trails, the number and size of swarming holes, and the percentage of mud cover area.

[0032] In practice, based on meteorological forecast data within a preset period (such as temperature, humidity, and rainfall probability in the next 7-15 days), the specific time window for the breeding and swarming period is predicted, and the potential spread risk area of ​​the termite population is simulated by combining the ant nest location information.

[0033] In practice, a unified global coordinate system is established with the water conservancy project's benchmark control point as the origin. The X-axis is defined along the dam axis, the Y-axis is perpendicular to the dam axis, and the Z-axis is defined as the elevation / depth direction. The monitoring area is divided into a standard three-dimensional voxel grid. For fixed vibration and environmental sensors, their installation coordinates are measured and mapped to the corresponding grid nodes in the unified coordinate system. For termite activity trace data, the device pose parameters at the time of collection are obtained, and the image pixel coordinates are back-projected onto the three-dimensional spatial surface. The data is then corrected according to the type of termite trace to obtain the estimated three-dimensional coordinates of the termite nest. Based on a unified timestamp, the continuous trajectory of the mobile device is smoothed. Multi-source heterogeneous data with a spatial distance of less than a preset threshold within the same time window are aggregated into the same grid cell to generate a multimodal data set with spatial confidence labels. Therefore, this multimodal dataset includes vibration feature vectors, environmental numerical vectors, visual trace quantification vectors, and risk prediction vectors obtained after aligning heterogeneous data with timestamps and registering spatial coordinates. These vectors represent the activity characteristics of the same ant colony. The data vector representing the nth type.

[0034] In some embodiments, in S102, ant infestation case records at different historical time points are extracted from the water conservancy project management database, and the original data is cleaned to remove records with missing key fields or incorrect labeling. For the processed historical ant infestation data, each historical record is mapped to a subset of historical data with the same data format as the multimodal dataset, thereby forming a historical dataset. ,in Indicates the first A subset of historical data at the moment of writing. Indicates the first Pen Moment Historical data of various types. That is, .

[0035] In some embodiments, in S103, for each historical moment, the corresponding historical prevention and control schemes and prevention and control effect data are retrieved, including but not limited to ant nest removal rate, ecological impact score on water / soil, and changes in engineering structure stability; the scoring weights corresponding to the prevention and control effect data of each dimension are set, and the historical prevention and control schemes are quantitatively scored based on the prevention and control effect data of multiple dimensions and their corresponding weight values; historical prevention and control schemes with scores greater than a first preset threshold are marked and used as candidate prevention and control decision sets.

[0036] In another embodiment, based on the operating condition information at various historical moments, multiple prevention and control strategies are combined using parameter vectors, including but not limited to targeted pesticide application and the application of anti-termite materials. The operating condition information and prevention and control strategies are input into a termite ecological simulation model, the simulation model is run, and the prevention and control effect score of each strategy under the corresponding operating conditions is predicted. Strategies with scores greater than a first preset threshold are selected as candidate prevention and control decision sets. The first preset threshold is determined based on the goal of minimizing termite nest removal and ecological impact.

[0037] Furthermore, for any two candidate prevention and control strategies in the candidate prevention and control decision set, the similarity is calculated using a cosine similarity algorithm based on the parameter vector; if the similarity is greater than a second preset threshold, the candidate prevention and control strategy with the lower score is removed from the candidate prevention and control decision set, resulting in the final optimized prevention and control decision subset at the m-th time step. ,in Indicates the first The first moment of writing Optimize prevention and control decisions.

[0038] Thus, we can obtain multiple sets of optimized prevention and control decisions at different historical points in time. .Right now, It should be noted that the number of subsets at each moment may be an empty set or a set with an indefinite number of terms, depending on the distribution of termites, engineering structure, and ecological constraints.

[0039] In some embodiments, in S104, monitoring data—prevention and control decisions—are generated by optimizing the prevention and control decision set based on the corresponding time points using historical datasets. ,in For monitoring data—prevention and control decision set, Representing the Historical data subset of the pen moment , Representing the Optimization and prevention decision subset at the pen moment Therefore, the historical dataset and the optimized prevention and control decision set form a bijective relationship. .

[0040] In some embodiments, in S105, the multimodal data set and historical datasets Simultaneously, the text embedding model is used for cluster analysis. Finally, the clustering results can be compared to determine the category to which each subset belongs, i.e., the cluster category set. ,in Represents a multimodal data set Clustering categories, Represents historical datasets The clustering categories of each subset within the set of clustering categories. These clustering categories may consist of two or more categories; therefore, the categories within each subset of the clustering category set may share some similarities or differ from each other. (Multimodal dataset) Historical datasets and cluster category set It belongs to a bijective relationship. .

[0041] Furthermore, by clustering category sets The scheduling decisions that belong to the same cluster category are identified and used as the retrieval results. The calculation process is as follows: By bijective relation Get current data Corresponding clustering categories: ; Filtering and Collection of similar historical data cluster categories: ; By bijective relation inverse mapping , obtain Corresponding historical dataset: ; By bijective relation , obtain The corresponding optimized prevention and control decision set, i.e., the search results: .

[0042] This is understandable, because of the multimodal data set. and cluster category set It belongs to a bijective relation, so it can be used to find multimodal datasets. Clustering categories .pass and subcollections Intersection finds the same cluster categories Because of historical datasets and cluster category set Since it belongs to a bijective relation, we can find the same cluster categories. Historical termite data sets corresponding to the same cluster category Because the historical termite data set and the optimized prevention and control decision set have a bijective relationship, the monitoring data—prevention and control decision set—can be analyzed. It can find the set of optimized prevention and control decisions with the same cluster category as the search results. .

[0043] Optionally, the text embedding model may employ an unsupervised clustering model, including but not limited to the K-means model, self-organizing map model, nearest neighbor propagation algorithm model, hierarchical clustering model, mean shift algorithm model, DBSCAN model, and binary K-means model.

[0044] Optionally, a multimodal semantic mapping mechanism can be used to uniformly map real-time monitoring data, historical operating condition data, and decision knowledge base to a high-dimensional semantic vector space. Through vector similarity calculation and density clustering analysis, data matching and decision inference can be achieved.

[0045] Furthermore, in some embodiments, in S106, the retrieval results and previous stage auxiliary decisions are simultaneously input into a large language model so that precise prevention and control decisions can be inferred through analysis using the large language model. The previous stage auxiliary decisions refer to termite control-related information previously released before the current time point. For example, if it is predicted that a dam area will be in the termite swarming period within the next 7 days, and the termite nest is currently located at a depth of 2 meters in the dam, a 0.05% concentration imidacloprid slow-release agent needs to be used for targeted irrigation at a rate of 2 L / m²; simultaneously, anti-termite composite material is laid on the dam body in this area, covering a 5-meter radius around the termite nest; and termite natural enemies such as *Triplophysa guani* are released at a density of 50 wasps / m² to control the spread of termites during the swarming period.

[0046] In practical implementation, the search results are used as an empirical reference, while the dam material properties, anti-seepage structure distribution, and maximum allowable grouting pressure of the current project are used as structural hard constraints. The ecological protection zone level, prohibited chemical list, and groundwater protection threshold of the work area are used as ecological hard constraints. These are combined to form a structured prompt. A large language model, finely tuned with professional termite control knowledge, is called to compare the differences in multimodal characteristics between the current working conditions and historical cases. Process parameters in historical solutions that do not meet the current project's structural safety threshold or violate ecological red lines are automatically eliminated, and the three-dimensional spatial coordinates of the termite nest are deduced. Under the premise of satisfying all constraints, the optimal number of holes, drilling angle, grouting pressure, and agent ratio are recalculated. Thus, the model outputs a targeted and precise prevention and control decision containing parameters for targeted treatment of the termite nest.

[0047] Optionally, the parameters for targeted treatment of ant nests should include at least: spatial positioning parameters, pesticide application parameters, and process control parameters. Spatial positioning parameters include the three-dimensional coordinates of the ant nest center, the radius of effect, and the drilling layout (hole location, hole depth, and inclination angle); pesticide application parameters include the type of pesticide adapted to ecological constraints, the precise concentration, and the injection volume per hole and the total injection volume; process control parameters include the maximum grouting pressure, grouting rate, and segmented construction strategy based on structural safety calculations.

[0048] Optionally, the model output can be a standardized JSON format decision instruction, along with a trace of the reasoning basis, for direct execution by the construction terminal or manual review.

[0049] Therefore, this application addresses the decision-making gap by using multimodal monitoring data and historical cases as the foundation of the RAG knowledge base, and AI recognition results and ant nest location data as real-time input to a large language model. Through retrieval, matching, and reasoning, it fills the gaps in automated professional decision-making, forming a complete closed loop and achieving a systematic technological upgrade. This realizes a closed-loop process encompassing monitoring, retrieval, decision-making, execution, and feedback, solving the problem of professional decision-making bias in general models, adapting to multiple scenarios such as reservoirs and dikes, improving the accuracy of prevention and control and ecological safety, and promoting the upgrade of termite control from single-point technology to systematic intelligent decision-making.

[0050] As can be seen, the termite control decision-making method based on retrieval enhancement in this application has the following beneficial technical effects: Firstly, it addresses the pain points of existing technology decision-making relying on human experience and lacking intelligence. By using RAG technology to retrieve and match historical best cases, and combining large language model reasoning to generate professional decisions, it avoids the decision bias and "illusion" problem of general models, greatly improving decision accuracy and response speed, and can quickly output adapted solutions without human intervention. Secondly, by integrating multimodal monitoring data with ant nest location results, personalized targeted prevention and control strategies can be generated. Compared with traditional blind extermination, this reduces the waste of environmentally friendly agents and soil and water pollution, while also reducing secondary damage to water conservancy structures, thus balancing prevention and control effectiveness with ecological safety.

[0051] Third, it breaks through the limitations of existing patents in terms of weak cross-scenario adaptability, and can be flexibly adapted to various types of water conservancy projects such as reservoirs, dikes, and canals. It eliminates the need for separate design schemes for a single scenario, reduces operation and maintenance costs, and adapts to the large-scale prevention and control needs of large-scale water conservancy facilities.

[0052] Fourth, construct a closed-loop system of "monitoring-decision-execution-feedback" to continuously update the knowledge base and model reasoning logic through prevention and control effect data, realize dynamic iterative optimization of decision-making strategies, improve the stability and efficiency of termite prevention and control in the long term, and form a sustainable intelligent prevention and control system.

[0053] It should be noted that the above description describes some embodiments of this application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in a different order than that shown in the above embodiments and still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0054] Based on the same technical concept, and corresponding to any of the above embodiments, this application also provides a termite control decision-making device 200 based on retrieval enhancement generation.

[0055] like Figure 2 As shown, the termite control decision-making device 200 based on retrieval enhancement may include: Module 201 is used to acquire a multimodal data set of termites. The multimodal dataset includes vibration feature vectors, environmental numerical vectors, visual trace quantization vectors, and risk prediction vectors, wherein... A data vector representing the nth type; The first module 202 is used to establish a historical dataset based on multiple historical ant infestation data from different historical time points. ,in Indicates the first A subset of historical data at the moment of writing. Indicates the first Pen Moment The data types in each subset of historical data are the same as those in the multimodal data set. The determining module 203 is used to determine a set of optimized prevention and control decisions for multiple historical time points based on the historical dataset and its corresponding historical prevention and control effect data. ,in Indicates the first The optimized prevention and control decision subset at the moment of writing. Indicates the first The first moment of writing The number of data points for each optimized prevention and control decision subset is determined based on the operating condition information. The second module 204 is used to establish a bijective relationship between the historical dataset and the optimized prevention and control decision set based on the correspondence between time points. This yields a set of monitoring data and prevention / control decisions. The retrieval module 205 is used to perform clustering and matching retrieval on the multimodal dataset, historical dataset, and monitoring data-prevention and control decision set using a text embedding model, and infer the retrieval results. The decision module 206 is used to input the retrieval results, current engineering structure parameters, and ecologically sensitive area constraints into a pre-trained large language model, and to infer and generate a target-precision prevention and control decision that matches the current working conditions. The target-precision prevention and control decision includes at least ant nest targeted treatment parameters.

[0056] It should be noted that, for ease of description, the above devices are described in terms of function, divided into various modules. Of course, in implementing this application, the functions of each module can be implemented in one or more software and / or hardware.

[0057] The apparatus described above is used to implement the termite control decision-making method based on retrieval enhancement in any of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0058] Based on the same technical concept, corresponding to any of the above embodiments, this application also provides an electronic device.

[0059] Figure 3 A schematic diagram of a more specific electronic device hardware structure provided in this embodiment is shown.

[0060] The electronic device 300 may include a processor 301 and a memory 302 storing computer program instructions.

[0061] Specifically, the processor 301 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0062] Memory 302 may include mass storage for data or instructions. For example, and not limitingly, memory 302 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 302 may include removable or non-removable (or fixed) media. Where appropriate, memory 302 may be internal or external to the integrated gateway disaster recovery device. In a particular embodiment, memory 302 is non-volatile solid-state memory.

[0063] In certain embodiments, the memory may include read-only memory (ROM), random access memory (RAM), disk storage media devices, optical storage media devices, flash memory devices, and electrical, optical, or other physical / tangible memory storage devices. Thus, generally, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method according to one aspect of this application.

[0064] The processor 301 reads and executes computer program instructions stored in the memory 302 to implement any of the termite control decision-making methods based on retrieval enhancement in the above embodiments.

[0065] In some examples, the electronic device 300 may also include a communication interface 303 and a bus 310. For example, Figure 3 As shown, the processor 301, memory 302, and communication interface 303 are connected through bus 310 and complete communication with each other.

[0066] The communication interface 303 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0067] Bus 310 includes hardware, software, or both, that couples components of an online data traffic metering device together. For example, and not as a limitation, bus 310 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 310 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, any suitable bus or interconnect is contemplated herein.

[0068] For example, the electronic device 300 can be a mobile phone, tablet computer, laptop computer, handheld computer, in-vehicle electronic device, ultra-mobile personal computer (UMPC), netbook, or personal digital assistant (PDA), etc.

[0069] Based on the same technical concept, corresponding to any of the methods in the above embodiments, this application also provides a non-transitory computer-readable storage medium. This computer-readable storage medium stores computer program instructions; when executed by a processor, these computer program instructions implement any of the termite control decision-making methods based on retrieval enhancement generation in the above embodiments. Examples of computer-readable storage media include non-transitory computer-readable storage media such as portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, etc.

[0070] Based on the same technical concept, corresponding to any of the above embodiments, this application also provides a computer program product, which includes computer program instructions. In some embodiments, the computer program instructions can be executed by one or more processors of a computer to cause the computer and / or the processor to perform the termite control decision-making method based on retrieval enhancement. Corresponding to the execution entity for each step in each embodiment of the termite control decision-making method based on retrieval enhancement, the processor executing the corresponding step may belong to the corresponding execution entity.

[0071] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0072] The functional blocks shown in the above-described structural diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0073] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0074] The aspects of this application have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by dedicated hardware performing the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0075] The above description is merely a specific implementation of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.

Claims

1. A termite control decision-making method based on retrieval enhancement generation, characterized in that, include: Obtain multimodal data sets of termites The multimodal dataset includes vibration feature vectors, environmental numerical vectors, visual trace quantization vectors, and risk prediction vectors, wherein... A data vector representing the nth type; A historical dataset was built based on multiple historical ant infestation data points from different historical time periods. ,in Indicates the first A subset of historical data at the moment of writing. Indicates the first Pen Moment The data types in each subset of historical data are the same as those in the multimodal data set. Based on the historical dataset and its corresponding historical prevention and control effect data, determine multiple sets of optimized prevention and control decisions at different historical time points. ,in Indicates the first The optimized prevention and control decision subset at the moment of writing. Indicates the first The first moment of writing The number of data points for each optimized prevention and control decision subset is determined based on the operating condition information. Based on the correspondence between time points, a bijective relationship is established between the historical dataset and the optimized prevention and control decision set. This yields a set of monitoring data and prevention / control decisions. The text embedding model is used to perform clustering and matching retrieval on the multimodal dataset, historical dataset, and monitoring data-prevention decision set to infer the retrieval results; The search results, current engineering structure parameters, and ecologically sensitive area constraints are input into a pre-trained large language model to infer and generate a precise prevention and control decision that matches the current working conditions. The precise prevention and control decision includes at least ant nest targeted treatment parameters.

2. The method according to claim 1, characterized in that, The acquisition of the multimodal data set of termites includes: Real-time collection of acoustic vibration signals, carbon dioxide concentration, soil / air temperature and relative humidity via intelligent monitoring terminals yields vibration data and environmental data. The images to be inspected are obtained through drone inspections, and then the images are input into a pre-trained deep learning model for termite trace recognition to obtain termite activity trace data. Based on the meteorological forecast data within the preset period, the specific time window for the breeding and swarming period is predicted, and the risk area for the spread of termite populations is simulated by combining the termite nest location information, thus obtaining risk prediction data; The vibration data, environmental data, termite activity trace data, and risk prediction data are time-stamp aligned and spatial coordinate registered to obtain vibration feature vectors, environmental numerical vectors, visual trace quantization vectors, and risk prediction vectors, which constitute the termite multimodal data set at the current moment.

3. The method according to claim 1, characterized in that, The historical ant infestation data includes historical prevention and control plans corresponding to each historical time point. The step of determining a set of optimized prevention and control decisions for multiple historical time points based on the historical dataset and its corresponding historical prevention and control effect data includes: For historical datasets with multiple historical time points and their corresponding historical prevention and control effect data, the following steps are performed in parallel: Historical prevention and control measures are quantitatively scored based on historical prevention and control effectiveness data from multiple dimensions and their corresponding preset weight values. Historical prevention and control measures with scores greater than a first preset threshold are marked and used as a candidate prevention and control decision set. For any two candidate prevention and control decisions in the candidate prevention and control decision set, the similarity is calculated using a cosine similarity algorithm based on parameter vectors. If the similarity is greater than a second preset threshold, the candidate prevention and control decision with the lower score is removed from the candidate prevention and control decision set to obtain the optimized prevention and control decision subset corresponding to that historical time point. Multiple subsets of optimization and prevention decisions corresponding to different historical time points are combined into an optimization and prevention decision set.

4. The method according to claim 1, characterized in that, The text embedding model is used to perform clustering and matching retrieval on the multimodal dataset, historical dataset, and monitoring data-prevention decision set to infer the retrieval results, including: Multimodal data sets and historical datasets Simultaneously, the text embedding model is used for cluster analysis to obtain a set of cluster categories. ,in Represents a multimodal data set Clustering categories, Represents historical datasets Clustering categories for each subset of historical data; Establish a multimodal dataset Historical datasets and cluster category set bijective relation ; Sure Corresponding cluster categories ; Filtering and Collection of similar historical data cluster categories: ; Sure Corresponding historical data subset ; Sure The corresponding optimized prevention and control decision set yielded the following search results: .

5. The method according to claim 4, characterized in that, The parameters for targeted treatment of ant nests include at least: the three-dimensional spatial coordinates of the ant nest center, the number of holes, the drilling angle, the maximum grouting pressure, and the reagent ratio.

6. The method according to claim 4, characterized in that, The current engineering structural parameters include at least: the dam material properties, the distribution of the seepage prevention structure, and the maximum allowable grouting pressure.

7. A termite control decision-making device based on retrieval enhancement generation, characterized in that, include: The acquisition module is used to acquire multimodal data sets of termites. The multimodal dataset includes vibration feature vectors, environmental numerical vectors, visual trace quantization vectors, and risk prediction vectors, wherein... A data vector representing the nth type; The first module is used to build a historical dataset based on multiple historical ant infestation data from different historical time points. ,in Indicates the first A subset of historical data at the moment of writing. Indicates the first Pen Moment The data types in each subset of historical data are the same as those in the multimodal data set. The determination module is used to determine a set of optimized prevention and control decisions for multiple historical time points based on the historical dataset and its corresponding historical prevention and control effect data. ,in Indicates the first The optimized prevention and control decision subset at the moment of writing. Indicates the first The first moment of writing The number of data points for each optimized prevention and control decision subset is determined based on the operating condition information. The second module is used to establish a bijective relationship between the historical dataset and the optimized prevention and control decision set based on the correspondence between time points. This yields a set of monitoring data and prevention / control decisions. The retrieval module is used to perform clustering and matching retrieval on the multimodal dataset, historical dataset, and monitoring data-prevention decision set using a text embedding model, and infer the retrieval results. The decision module is used to input the retrieval results, current engineering structure parameters, and ecologically sensitive area constraints into a pre-trained large language model, and to infer and generate a target-precision prevention and control decision that matches the current working conditions. The target-precision prevention and control decision includes at least ant nest targeted treatment parameters.

8. An electronic device, characterized in that, The device includes: a processor and a memory storing computer program instructions; when the processor invokes the computer program instructions, it implements the termite control decision-making method based on retrieval enhancement as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions, which, when invoked by a processor, implement the termite control decision-making method based on retrieval enhancement as described in any one of claims 1-6.

10. A computer program product, characterized in that, When the instructions in the computer program product are executed by the processor of the electronic device, the electronic device performs the termite control decision-making method based on retrieval enhancement as described in any one of claims 1-6.

Images