Agricultural decision determination method and device, electronic equipment, product and storage medium

Abstract

The invention provides an agricultural decision determination method and device, electronic equipment, a product and a storage medium, and relates to the technical field of intelligent decision. The method comprises the following steps: determining multi-modal agricultural data and an agricultural task target; based on the multi-modal agricultural data and the agricultural task target, constructing a candidate multi-hop reasoning path set in a preset agricultural knowledge graph; determining the score of each candidate multi-hop reasoning path in the candidate multi-hop reasoning path set based on real-time environment data; determining the candidate multi-hop reasoning path with the maximum score corresponding to each candidate multi-hop reasoning path as an agricultural decision path, and determining an agricultural decision based on the agricultural decision path; the agricultural decision is used for solving the agricultural task goal. According to the invention, the defects of low reasoning efficiency and influence on accuracy and timeliness of decision making caused by difficulty in dynamically adjusting the reasoning path according to real-time environment change and task requirements in the prior art can be solved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent decision-making technology, and in particular to an agricultural decision-making method, device, electronic equipment, product and storage medium. Background Art

[0002] As global agricultural production continues to increase its requirements for efficiency and sustainability, the application of smart agricultural technology has gradually become an important means to improve agricultural productivity. In this field, the widespread application of technologies such as sensors, drones, image recognition and data analysis has enabled agricultural production to be managed accurately. Existing technologies usually provide decision support by building fixed rule engines, which can quickly achieve decision support, especially when there is clear knowledge and experience.

[0003] However, the complexity and variability of agricultural scenarios make traditional methods that rely on fixed rules incapable of dealing with practical problems. It is difficult to dynamically adjust the reasoning path according to real-time environmental changes and task requirements, resulting in low reasoning efficiency and affecting the accuracy and timeliness of agricultural decision-making. Summary of the invention

[0004] The present invention provides an agricultural decision-making method, device, electronic device, product and storage medium to solve the defect that the prior art is difficult to dynamically adjust the reasoning path according to real-time environmental changes and task requirements, resulting in low reasoning efficiency and affecting the accuracy and timeliness of decision-making.

[0005] The present invention provides an agricultural decision-making method, comprising: Identify multimodal agricultural data and agricultural mission objectives; Based on the multimodal agricultural data and the agricultural task objectives, constructing a set of candidate multi-hop reasoning paths in a preset agricultural knowledge graph; Determining a score of each candidate multi-hop reasoning path in the set of candidate multi-hop reasoning paths based on the real-time environment data; The candidate multi-hop reasoning path with the largest score corresponding to each of the candidate multi-hop reasoning paths is determined as the agricultural decision path, and an agricultural decision is determined based on the agricultural decision path; the agricultural decision is used to solve the agricultural task goal.

[0006] According to an agricultural decision determination method provided by the present invention, the candidate multi-hop reasoning path set is constructed based on the following method: Based on the multimodal agricultural data and the agricultural task objectives, determining a path starting point and a path end point in a preset agricultural knowledge graph; Based on the path starting point and the path ending point, determining a plurality of jump nodes located between the path starting point and the path ending point in the preset agricultural knowledge graph; Constructing a candidate multi-hop reasoning path set based on a plurality of jump nodes between the path starting point and the path end point; Each candidate multi-hop reasoning path in the candidate multi-hop reasoning path set includes multiple jump nodes.

[0007] According to an agricultural decision determination method provided by the present invention, the score of any candidate multi-hop reasoning path in the candidate multi-hop reasoning path set is determined based on the following method: Determine the matching probability and weight of each jump node in the candidate multi-hop reasoning path based on the real-time environment data; Determine the score corresponding to each jump node based on the matching probability and weight of each jump node; The score of the candidate multi-hop reasoning path is determined based on the score corresponding to each jump node in the candidate multi-hop reasoning path.

[0008] According to an agricultural decision determination method provided by the present invention, the preset agricultural knowledge graph is determined based on historical agricultural data, and based on the multimodal agricultural data and the agricultural task objectives, a candidate multi-hop reasoning path set is constructed in the preset agricultural knowledge graph, including: When the multimodal agricultural data is logically inconsistent with the historical agricultural data in the preset agricultural knowledge graph, the preset agricultural knowledge graph is updated based on the multimodal agricultural data to obtain an updated preset agricultural knowledge graph; Based on the multimodal agricultural data and the agricultural task objectives, a set of candidate multi-hop reasoning paths is constructed in the updated preset agricultural knowledge graph.

[0009] According to an agricultural decision determination method provided by the present invention, the multimodal agricultural data is determined based on the following method: Acquire agricultural text data, agricultural image data, agricultural video data, and agricultural sensor data; Based on the weight of the agricultural text data, the weight of the agricultural image data, the weight of the agricultural video data and the weight of the agricultural sensor data, the agricultural text data, the agricultural image data, the agricultural video data and the agricultural sensor data are weightedly fused to determine multimodal agricultural data.

[0010] According to an agricultural decision determination method provided by the present invention, any of the jump nodes is composed of a triplet, and the triplet includes a first entity, a second entity, and a relationship between the first entity and the second entity.

[0011] The present invention also provides an agricultural decision-making device, comprising: A data determination module, which is used to determine multimodal agricultural data and agricultural task objectives; A construction module, which is used to construct a set of candidate multi-hop reasoning paths in a preset agricultural knowledge graph based on the multimodal agricultural data and the agricultural task objectives; A score determination module, which is used to determine the score of each candidate multi-hop reasoning path in the candidate multi-hop reasoning path set based on real-time environment data; An agricultural decision determination module is used to determine the candidate multi-hop reasoning path with the largest score corresponding to each of the candidate multi-hop reasoning paths as the agricultural decision path, and determine the agricultural decision based on the agricultural decision path; the agricultural decision is used to solve the agricultural task objectives.

[0012] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the agricultural decision-making method described above is implemented.

[0013] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the agricultural decision-making methods described above.

[0014] The present invention also provides a computer program product, comprising a computer program, which, when executed by a processor, implements any of the agricultural decision-making methods described above.

[0015] The agricultural decision determination method, device, electronic device, product and storage medium provided by the present invention include determining multimodal agricultural data and agricultural task objectives; constructing a set of candidate multi-hop reasoning paths in a preset agricultural knowledge graph based on the multimodal agricultural data and the agricultural task objectives; determining the score of each candidate multi-hop reasoning path in the candidate multi-hop reasoning path set based on real-time environmental data; determining the candidate multi-hop reasoning path with the largest score corresponding to each of the candidate multi-hop reasoning paths as the agricultural decision path, and determining the agricultural decision based on the agricultural decision path; the agricultural decision is used to solve the agricultural task objectives; the present invention ensures a comprehensive understanding of the current agricultural scenario by determining multimodal agricultural data and agricultural task objectives; determining the score of each candidate multi-hop reasoning path in the candidate multi-hop reasoning path set through real-time environmental data, and determining the candidate multi-hop reasoning path with the highest score as the agricultural decision path, ensuring that the decision process is based on the latest and most relevant information, thereby reducing decision errors caused by delayed or inaccurate information, and improving the accuracy and timeliness of agricultural decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 It is a flow chart of the agricultural decision-making determination method provided by the present invention.

[0018] Figure 2 It is a structural schematic diagram of the agricultural decision-making determination device provided by the present invention.

[0019] Figure 3 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are 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.

[0021] In the process of agricultural production, with the continuous advancement of technology, more and more agricultural practitioners have begun to rely on intelligent decision support systems to improve production efficiency and sustainability. However, after in-depth practice and observation, the applicant of this patent found that most of the existing agricultural decision support relies on fixed rule engines or single data sources, and cannot flexibly respond to the complexity and variability in the agricultural production environment. Specifically, when faced with multimodal agricultural data, such as sensor data, image information, and climate data, these often cannot achieve effective adaptive data fusion and dynamic reasoning path adjustment, resulting in low reasoning efficiency and insufficient decision accuracy.

[0022] In further research, the applicant noted that although the existing technology has attempted to solve agricultural decision-making problems in some aspects by introducing rule engines and models based on specific data types, such as quickly responding to pest outbreaks or providing specific crop management recommendations, these methods still have obvious shortcomings. First, fixed rule engines are rigid and difficult to adjust dynamically when dealing with real-time changing agricultural environments; second, the application of single data source models limits the comprehensiveness and accuracy of the decision-making process, making it often impossible for farmers to obtain effective support in complex decision-making scenarios.

[0023] Therefore, the applicant believes that although the existing technology has been effective in certain specific scenarios, it still has significant limitations in overall decision support capabilities. This understanding prompted the applicant to develop a new agricultural decision determination method to fill these technical gaps and provide more intelligent and efficient agricultural decision support.

[0024] In view of the above problems, the present invention proposes the following embodiments.

[0025] Figure 1 It is a flow chart of the agricultural decision-making method provided by the present invention, such as Figure 1 As shown, the method includes the following: Step 110, determining multimodal agricultural data and agricultural task objectives.

[0026] Here, multimodal agricultural data refers to a data set consisting of agricultural-related data from different sources and types, including but not limited to agricultural text data, agricultural image data, agricultural video data, and agricultural sensor data. Agricultural task goals refer to specific goals or tasks that need to be achieved in the agricultural production process. As the core of decision support, agricultural task goals can guide the reasoning direction and decision-making process, which can directly affect the subsequent reasoning path selection and data processing. Clear agricultural task goals can ensure that the decision-making process revolves around user needs.

[0027] The specific method for determining multimodal agricultural data can be referred to the following embodiments, which will not be elaborated here.

[0028] Step 120, based on the multimodal agricultural data and the agricultural task objectives, construct a set of candidate multi-hop reasoning paths in a preset agricultural knowledge graph.

[0029] Here, the preset agricultural knowledge graph is a structured knowledge representation method that contains multiple entities in the agricultural field, such as crop types, climate conditions, soil moisture, etc., as well as relationships between entities, such as causal relationships, association relationships, etc. The construction of the preset agricultural knowledge graph helps to understand and reason about complex information in agricultural scenarios. A multi-hop reasoning path refers to gradually approaching the agricultural task goal by crossing multiple related entities and relationships during the reasoning process. The candidate multi-hop reasoning path set includes multiple candidate multi-hop reasoning paths.

[0030] For example, in agricultural production scenarios, decision-making often depends on multiple factors, which are closely connected through the relationships in the knowledge graph. For example, assuming that the agricultural task goal is to determine the "best time to apply pesticides", and the multimodal agricultural data includes "crop type", the reasoning process is as follows: The first jump: derive "current pests and diseases" from "crop type". At this time, multimodal agricultural data, such as sensor monitoring and image recognition, can be used to detect the health of crops and identify existing pests and diseases; the second jump: derive "optimal prevention and control plan" from "current pests and diseases". In this process, the recommended prevention and control strategy will be generated based on the types of pests and diseases and their impacts, combined with the knowledge and historical data of agricultural experts in the preset agricultural knowledge graph. The third jump: derive "optimal time for pesticide application" from "optimal prevention and control plan". In this step, external environmental factors such as weather forecasts and crop growth stages can be determined from multimodal agricultural data to ensure the best pesticide application effect. In this reasoning process, the optimal path is dynamically selected through the reinforcement learning algorithm, which means that the reasoning is not fixed along a certain path, but the reasoning path is flexibly adjusted according to the context of the current problem and the changes in the actual environment. In this way, it can quickly adapt to the complex and changing agricultural environment and achieve accurate and efficient decision support.

[0031] Step 130 : determining the score of each candidate multi-hop reasoning path in the candidate multi-hop reasoning path set based on the real-time environment data.

[0032] Here, real-time environmental data refers to the instant data obtained through sensors, weather stations and other monitoring methods in agricultural production. These data include but are not limited to temperature, humidity, soil conditions, crop health status, etc., and can reflect changes in the agricultural environment in real time.

[0033] Here, the path score is an indicator that quantifies the importance, adaptability, and effectiveness of each candidate reasoning path. The higher the score, the better the applicability and effectiveness of the candidate multi-hop reasoning path under the current environmental conditions. Since the agricultural environment changes rapidly, the calculation of the score needs to be dynamically adjusted based on real-time data. This flexibility ensures that the optimal path can be selected based on the latest information when making decisions.

[0034] Step 140, determining the candidate multi-hop reasoning path with the largest score corresponding to each of the candidate multi-hop reasoning paths as the agricultural decision path, and determining an agricultural decision based on the agricultural decision path; the agricultural decision is used to solve the agricultural task goal.

[0035] Here, agricultural decisions are specific action plans or guidelines based on agricultural decision pathways, aimed at solving specific agricultural task goals, such as crop management, irrigation needs assessment, or pest and disease control.

[0036] Here, the agricultural decision path refers to the reasoning chain that is finally determined by combining multimodal agricultural data and real-time environmental data in the multi-hop reasoning process.

[0037] For example, after calculating the corresponding scores for each candidate multi-hop reasoning path, the path with the largest score is identified. This path represents the reasoning chain that best meets the agricultural task objectives under the current real-time environmental conditions. Once the best agricultural decision path is selected, specific agricultural decisions will be made based on the reasoning information in the path. For example, if the path involves "determining crop irrigation needs", irrigation or adjustment of irrigation amount may be recommended based on the reasoning results of the path.

[0038] The agricultural decision determination method provided by the embodiment of the present invention ensures a comprehensive understanding of the current agricultural scenario by determining multimodal agricultural data and agricultural task objectives; determines the score of each candidate multi-hop reasoning path in the candidate multi-hop reasoning path set through real-time environmental data, and determines the candidate multi-hop reasoning path with the highest score as the agricultural decision path, ensuring that the decision-making process is based on the latest and most relevant information, thereby reducing decision-making errors caused by delayed or inaccurate information, and improving the accuracy and timeliness of agricultural decisions.

[0039] Based on any of the above embodiments, in this method, the candidate multi-hop reasoning path set is constructed based on the following method: Based on the multimodal agricultural data and the agricultural task objectives, determining a path starting point and a path end point in a preset agricultural knowledge graph; Based on the path starting point and the path ending point, determining a plurality of jump nodes located between the path starting point and the path ending point in the preset agricultural knowledge graph; Constructing a candidate multi-hop reasoning path set based on a plurality of jump nodes between the path starting point and the path end point; Each candidate multi-hop reasoning path in the candidate multi-hop reasoning path set includes multiple jump nodes.

[0040] Here, the path start point refers to the starting point of reasoning, which is usually an entity or concept related to the agricultural task goal, such as "soil moisture" or "crop type". The path end point refers to the end point of reasoning, which is related to the final agricultural decision. Jump nodes refer to key entities or concepts between the path start point and the end point during the reasoning process. They provide necessary contextual information for the reasoning process and help form more comprehensive decisions.

[0041] Exemplarily, the starting point and end point of reasoning are first identified based on multimodal agricultural data, such as sensor data, climate data, crop data, etc., and agricultural task objectives, such as "determining whether crops need irrigation". For example, if the agricultural task goal is to evaluate irrigation needs, the starting point of the path may be "soil moisture" and the end point may be "irrigation plan". After determining the starting point and end point of the path, the relevant jump nodes connecting these starting points and end points will be found in the preset agricultural knowledge graph. Jump nodes refer to key entities or concepts between the starting point and the end point in the reasoning process, which can help provide additional information required in the reasoning process. For example, from "soil moisture" to "climate conditions", and then to the node of "crop growth stage". Based on the determined path starting point, end point and jump node, multiple candidate multi-hop reasoning paths will be constructed. These paths show different reasoning links from the starting point through the jump node to the end point. Each candidate path will contain one or more jump nodes.

[0042] For example, the candidate multi-hop reasoning path can be determined in the preset agricultural knowledge graph based on the following formula: in, is the total probability of the candidate multi-hop reasoning path, T is the number of jump nodes contained in the candidate multi-hop reasoning path, It is The probability of jumping, is the discount factor, which indicates the contribution of each jump node to the final path selection. This formula calculates the probability of the complete path from the starting state to the target state, and the overall selection probability of the candidate multi-hop reasoning path is obtained by summing.

[0043] The agricultural decision determination method provided by the embodiment of the present invention first determines the starting point and end point of the path based on multimodal agricultural data and agricultural task objectives. Then, multiple jump nodes located between the starting point and the end point are identified to form a series of reasoning paths connecting the starting point, the jump node and the end point. Through this process, it is possible to integrate various information, enhance the flexibility and accuracy of reasoning, and provide more comprehensive support for decision-making. This not only helps to improve the scientificity and pertinence of decision-making, but also provides diversified solutions for agricultural management, thereby optimizing agricultural production efficiency and sustainability.

[0044] Based on any of the above embodiments, in the method, the score of any candidate multi-hop reasoning path in the candidate multi-hop reasoning path set is determined based on the following method: Determine the matching probability and weight of each jump node in the candidate multi-hop reasoning path based on the real-time environment data; Determine the score corresponding to each jump node based on the matching probability and weight of each jump node; The score of the candidate multi-hop reasoning path is determined based on the score corresponding to each jump node in the candidate multi-hop reasoning path.

[0045] It should be noted that the matching probability of each jump node is calculated based on real-time environmental data, such as current soil moisture, climate conditions, etc. This matching probability represents the adaptability of the jump node in the current scenario. For example, in an environment with low humidity, the matching probability of the soil moisture node will be higher. The weight refers to the relative importance of each jump node in the entire reasoning process. Some jump nodes are crucial to decision-making, so they are assigned higher weights. For example, the "severity of pests and diseases" jump node may have a higher weight in crop management tasks because it directly affects the decision-making of prevention and control plans.

[0046] Here, the node score refers to the score calculated based on the matching probability and weight of each jump node. This score reflects the contribution of the node in the current environment. The higher the score, the more significant the role of the node in the reasoning path. The score of the candidate multi-hop reasoning path refers to the total score of the candidate reasoning path obtained by summarizing the scores of all jump nodes in the candidate multi-hop reasoning path. The higher the score of the path, the higher the applicability of the path in the current environment, and the higher the priority. The score of the candidate reasoning path can be calculated based on the following formula: in, is the total path score, is the weight coefficient of each hopping node, is the matching probability of each jump node, based on the current state and the target state of the next hop node For example, the current real-time environmental data indicates the current state The air humidity and the sudden humidity are low, then assuming that the target state of the next jump node is is "wet soil", then the current state and the target state of the next hop node The matching probability is lower.

[0047] Through this process, the optimal path can be dynamically selected in each hop to ensure the most accurate and effective inference results.

[0048] Exemplarily, the weight coefficient may be updated based on the following formula: in, represents the updated weight, represents the weight before updating, is the learning rate, is a loss function based on the current state.

[0049] The agricultural decision determination method provided by the embodiment of the present invention calculates the score of each jump node based on the matching probability and weight of the node, and finally obtains the final score of the entire reasoning path by aggregating the scores of these nodes. This process ensures that the path that is most likely to lead to the best decision can be selected, thereby improving the accuracy and timeliness of agricultural decisions. Through this dynamic and adaptive reasoning method, not only can accurate decision support based on the current environment be provided, but also more optimized and reasonable reasoning choices can be made in complex agricultural situations.

[0050] Based on any of the above embodiments, in the method, the preset agricultural knowledge graph is determined based on historical agricultural data, and the candidate multi-hop reasoning path set is constructed in the preset agricultural knowledge graph based on the multimodal agricultural data and the agricultural task goal, including: When the multimodal agricultural data is logically inconsistent with the historical agricultural data in the preset agricultural knowledge graph, the preset agricultural knowledge graph is updated based on the multimodal agricultural data to obtain an updated preset agricultural knowledge graph; Based on the multimodal agricultural data and the agricultural task objectives, a set of candidate multi-hop reasoning paths is constructed in the updated preset agricultural knowledge graph.

[0051] Here, historical agricultural data refers to agricultural data collected in the past, which is obtained through long-term agricultural research and experience accumulation and is used to build the initial agricultural knowledge graph. The preset agricultural knowledge graph refers to a graph pre-built based on historical agricultural data to support agricultural reasoning and decision-making. This graph usually contains a large amount of agricultural domain knowledge and can provide a basic framework for reasoning.

[0052] In one embodiment, by combining multimodal agricultural data and historical agricultural data, the preset agricultural knowledge graph is dynamically updated to construct a more accurate and reliable set of multi-hop reasoning paths. Specifically, when the existing multimodal agricultural data is inconsistent with the knowledge constructed based on historical data in the preset agricultural knowledge graph, the preset knowledge graph will be updated based on these multimodal data. The updated graph will contain new information, which may include new agricultural scenarios, environmental changes, or new task requirements, so that the reasoning path is more in line with the current agricultural production situation.

[0053] For example, in the dynamic update process of the knowledge graph, it is necessary to deal with the docking and conflict between new and old knowledge. In agricultural scenarios, new multimodal agricultural data, such as agricultural sensor data, climate data, etc., may be inconsistent with historical data. At this time, a series of steps are needed to ensure the accuracy and real-time nature of the knowledge graph.

[0054] Specifically, new knowledge is extracted from multimodal agricultural data and needs to be integrated with the existing knowledge graph, which is first completed through entity alignment and relationship mapping. For example, if the new data indicates that the suitable growing season for a certain crop has changed, the consistency of the new information with the existing knowledge graph will be ensured by identifying the new growing season (entity) and associating it with the crop data (such as suitable season) in the existing knowledge graph.

[0055] Optionally, when new knowledge conflicts with existing knowledge, a selection problem is faced. For example, historical data may indicate that a crop needs irrigation in a particular season, but real-time soil moisture sensor data shows that the soil in the area is already moist enough and does not need irrigation. At this time, the conflict is resolved through the reinforcement learning model. The confidence of each data point is calculated based on multiple factors (such as historical data, real-time data, and the reliability of the data source), and the knowledge with higher confidence is selected as the basis for the final decision. Specifically: Triplet groups that frequently appear in historical data (such as crop irrigation needs) usually have higher confidence. Real-time sensor data in multimodal agricultural data directly reflects the current actual situation and may have higher confidence. The confidence is adjusted according to different data sources and situations through the reinforcement learning model to ensure that the knowledge that best fits the current situation is retained.

[0056] It should be noted that during the update of the preset agricultural knowledge graph, the reasoning path also needs to be adjusted. The reasoning path may change due to the addition of new knowledge. For example, if new sensor data indicates that the soil moisture is sufficient in a certain season, the reasoning path from "crop-season-irrigation" to "crop-soil moisture-irrigation" may be adjusted. The reasoning path will be automatically adjusted to ensure that the reasoning process is based on the latest and most accurate knowledge.

[0057] Exemplarily, the process of updating the preset agricultural knowledge graph can be determined based on the following formula: in, is the updated knowledge graph, It is the old knowledge graph. Represents a section that is updated based on new data, such as real-time sensor data.

[0058] The agricultural decision-making determination method provided by the embodiment of the present invention can flexibly adjust the decision path according to the latest multimodal agricultural data and environmental changes, avoid the use of outdated information, thereby optimizing the accuracy and timeliness of decisions, improving the level of intelligence in agricultural management, and ultimately providing more scientific and efficient decision support for agricultural production.

[0059] Based on any of the above embodiments, in this method, the multimodal agricultural data is determined based on the following method: Acquire agricultural text data, agricultural image data, agricultural video data, and agricultural sensor data; Based on the weight of the agricultural text data, the weight of the agricultural image data, the weight of the agricultural video data and the weight of the agricultural sensor data, the agricultural text data, the agricultural image data, the agricultural video data and the agricultural sensor data are weightedly fused to determine multimodal agricultural data.

[0060] Specifically, the weight of each data type can be pre-set, and various types of agricultural data are fused according to the actual importance (i.e., "weight") of each data type to obtain a comprehensive multimodal agricultural data. This means that the importance of each data type in the final decision will be determined based on the actual situation. For example, some agricultural tasks may rely more on agricultural sensor data, while in other tasks, agricultural image data may be more critical. Weighted fusion enables the system to dynamically adjust the influence of various types of data on decision-making.

[0061] Exemplarily, multi-modal data can be obtained by weighted fusion of multiple data based on the following formula: ; in, Represents agricultural multimodal data, Represents agricultural text data, represents agricultural image data, Represents agricultural video data, Represents agricultural sensor data; represents the weight of agricultural text data, Represents the weight of agricultural image data, represents the weight of agricultural video data, represents the weight of agricultural sensor data; The agricultural decision-making determination method provided by the embodiment of the present invention allows the system to adjust the weights of various types of data according to the specific needs of agricultural tasks through weighted fusion, so that the system can respond flexibly to different situations and improve the adaptability of decision-making. By reasonably configuring the weights of various types of data, the system can rely on the most critical and reliable data sources in a complex and changeable agricultural environment to ensure the accuracy and practicality of decision-making.

[0062] Based on any of the above embodiments, in the method, any of the jump nodes is composed of a triplet, and the triplet includes a first entity, a second entity, and a relationship between the first entity and the second entity.

[0063] For example, in agricultural scenarios, multi-hop reasoning is needed for decision support. The reasoning of each jump node is based on the triples in the knowledge graph. One jump node corresponds to one triple, and each triple has its specific matching probability and weight, which determines whether the path is suitable for the current environment and task objectives. Triples are the basic building blocks in the knowledge graph. Each triple consists of the first entity-relationship-second entity, describing the specific relationship between the two entities. For example: "Crop X-infection-pest and disease Y" means that there is an infection relationship between crop X and pest and disease Y. A path may have multiple jump nodes, that is, multiple triples; for example: the first jump: jump from "crop X" to "pest and disease Y" (that is, the triple "crop X-infection-pest and disease Y"); the second jump: jump from "pest and disease Y" to "control plan Z" (that is, the triple "pest and disease Y-response-control plan Z"). For example, if the agricultural task goal needs to determine whether a crop is suitable for growth in the current season, it may choose the following three-hop path: The first jump: "Crop-best growing period-season", for example: "Corn-growing period-spring"; The second step: "Season-climate characteristics-temperature", for example: "Spring-mild-suitable for growth"; The third jump: "Temperature-Crop Adaptability-Growth", for example: "Mild-Suitable-Corn"; Through the jumps of these triples, the target judgment result is finally obtained. During the reasoning process, the matching probability and weight of each triple are evaluated based on the real-time environmental data and the current agricultural task objectives. The most appropriate path is dynamically selected for reasoning based on these evaluation results. When performing reasoning, the applicability of each jump node (that is, each triple) in the current task must be evaluated, and this applicability is reflected by the matching probability. The matching probability is calculated based on the current environment (such as season, temperature, humidity, etc.) and historical data, indicating whether the triple is likely to be established in the current scenario. The matching probability of each triple and the importance of the task determine its weight. Triples with high weights will be given priority in the reasoning process to ensure that the reasoning path is highly relevant to the task goal.

[0064] The agricultural decision determination method provided by the embodiment of the present invention can store agricultural knowledge in a clear and structured manner in the form of triples, which is convenient for subsequent reasoning and decision-making. The triple structure allows new knowledge to be easily added to the knowledge graph while maintaining its relevance to existing knowledge, and has good scalability. By crossing multiple jump nodes (i.e., multiple triples), a more complex reasoning process can be achieved to solve complex agricultural management problems, such as crop planting, pest control and other tasks.

[0065] The agricultural decision determination device provided by the present invention is described below. The agricultural decision determination device described below and the agricultural decision determination method described above can be referenced to each other.

[0066] FIG2 is a schematic diagram of the structure of the agricultural decision-making device provided by the present invention. As shown in FIG2 , the agricultural decision-making device includes: A data determination module 210, which is used to determine multimodal agricultural data and agricultural task objectives; A construction module 220, which is used to construct a set of candidate multi-hop reasoning paths in a preset agricultural knowledge graph based on the multimodal agricultural data and the agricultural task objectives; A score determination module 230, which is used to determine the score of each candidate multi-hop reasoning path in the candidate multi-hop reasoning path set based on the real-time environment data; The agricultural decision determination module 240 is used to determine the candidate multi-hop reasoning path with the largest score corresponding to each of the candidate multi-hop reasoning paths as the agricultural decision path, and determine the agricultural decision based on the agricultural decision path; the agricultural decision is used to solve the agricultural task objectives.

[0067] Figure 3 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 3 As shown, the electronic device may include: a processor 310, a communication interface 320, a memory 330 and a communication bus 340, wherein the processor 310, the communication interface 320 and the memory 330 communicate with each other through the communication bus 340. The processor 310 may call the logic instructions in the memory 330 to execute the agricultural decision determination method, which includes: determining multimodal agricultural data and agricultural task objectives; constructing a candidate multi-hop reasoning path set in a preset agricultural knowledge graph based on the multimodal agricultural data and the agricultural task objectives; determining the score of each candidate multi-hop reasoning path in the candidate multi-hop reasoning path set based on real-time environmental data; determining the candidate multi-hop reasoning path with the largest score corresponding to each of the candidate multi-hop reasoning paths as an agricultural decision path, and determining an agricultural decision based on the agricultural decision path; the agricultural decision is used to solve the agricultural task objective.

[0068] In addition, the logic instructions in the above-mentioned memory 330 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0069] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the agricultural decision determination method provided by the above-mentioned methods, which includes: determining multimodal agricultural data and agricultural task objectives; based on the multimodal agricultural data and the agricultural task objectives, constructing a set of candidate multi-hop reasoning paths in a preset agricultural knowledge graph; based on real-time environmental data, determining the score of each candidate multi-hop reasoning path in the candidate multi-hop reasoning path set; determining the candidate multi-hop reasoning path with the largest score corresponding to each of the candidate multi-hop reasoning paths as an agricultural decision path, and determining an agricultural decision based on the agricultural decision path; the agricultural decision is used to solve the agricultural task objectives.

[0070] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the agricultural decision determination method provided by the above-mentioned methods, the method comprising: determining multimodal agricultural data and agricultural task objectives; constructing a set of candidate multi-hop reasoning paths in a preset agricultural knowledge graph based on the multimodal agricultural data and the agricultural task objectives; determining the score of each candidate multi-hop reasoning path in the set of candidate multi-hop reasoning paths based on real-time environmental data; determining the candidate multi-hop reasoning path with the largest score corresponding to each of the candidate multi-hop reasoning paths as an agricultural decision path, and determining an agricultural decision based on the agricultural decision path; the agricultural decision is used to solve the agricultural task objectives.

[0071] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0072] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0073] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for determining agricultural decisions, characterized in that: include: Identify multimodal agricultural data and agricultural mission objectives; Based on the multimodal agricultural data and the agricultural task objectives, constructing a set of candidate multi-hop reasoning paths in a preset agricultural knowledge graph; Determining a score of each candidate multi-hop reasoning path in the set of candidate multi-hop reasoning paths based on the real-time environment data; The candidate multi-hop reasoning path with the largest score corresponding to each of the candidate multi-hop reasoning paths is determined as the agricultural decision path, and an agricultural decision is determined based on the agricultural decision path; the agricultural decision is used to solve the agricultural task goal.

2. The agricultural decision-making method according to claim 1, characterized in that: The candidate multi-hop reasoning path set is constructed based on the following method: Based on the multimodal agricultural data and the agricultural task objectives, determining a path starting point and a path end point in a preset agricultural knowledge graph; Based on the path starting point and the path ending point, determining a plurality of jump nodes located between the path starting point and the path ending point in the preset agricultural knowledge graph; Constructing a candidate multi-hop reasoning path set based on a plurality of jump nodes between the path starting point and the path end point; Each candidate multi-hop reasoning path in the candidate multi-hop reasoning path set includes multiple jump nodes.

3. The agricultural decision-making method according to claim 2, characterized in that: The score of any candidate multi-hop reasoning path in the candidate multi-hop reasoning path set is determined based on the following method: Determine the matching probability and weight of each jump node in the candidate multi-hop reasoning path based on the real-time environment data; Determine the score corresponding to each jump node based on the matching probability and weight of each jump node; The score of the candidate multi-hop reasoning path is determined based on the score corresponding to each jump node in the candidate multi-hop reasoning path.

4. The agricultural decision-making method according to claim 1, characterized in that: The preset agricultural knowledge graph is determined based on historical agricultural data, and based on the multimodal agricultural data and the agricultural task objectives, a candidate multi-hop reasoning path set is constructed in the preset agricultural knowledge graph, including: When the multimodal agricultural data is logically inconsistent with the historical agricultural data in the preset agricultural knowledge graph, the preset agricultural knowledge graph is updated based on the multimodal agricultural data to obtain an updated preset agricultural knowledge graph; Based on the multimodal agricultural data and the agricultural task objectives, a set of candidate multi-hop reasoning paths is constructed in the updated preset agricultural knowledge graph.

5. The agricultural decision-making method according to claim 1, characterized in that: The multimodal agricultural data is determined based on the following method: Acquire agricultural text data, agricultural image data, agricultural video data, and agricultural sensor data; Based on the weight of the agricultural text data, the weight of the agricultural image data, the weight of the agricultural video data and the weight of the agricultural sensor data, the agricultural text data, the agricultural image data, the agricultural video data and the agricultural sensor data are weightedly fused to determine multimodal agricultural data.

6. The agricultural decision-making method according to claim 2, characterized in that: Any of the jump nodes is composed of a triplet, and the triplet includes a first entity, a second entity, and a relationship between the first entity and the second entity.

7. An agricultural decision-making device, characterized in that: include: A data determination module, which is used to determine multimodal agricultural data and agricultural task objectives; A construction module, which is used to construct a set of candidate multi-hop reasoning paths in a preset agricultural knowledge graph based on the multimodal agricultural data and the agricultural task objectives; A score determination module, which is used to determine the score of each candidate multi-hop reasoning path in the candidate multi-hop reasoning path set based on real-time environment data; An agricultural decision determination module is used to determine the candidate multi-hop reasoning path with the largest score corresponding to each of the candidate multi-hop reasoning paths as the agricultural decision path, and determine the agricultural decision based on the agricultural decision path; the agricultural decision is used to solve the agricultural task objectives.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the agricultural decision-making method as described in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the agricultural decision-making method as described in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the agricultural decision-making method as described in any one of claims 1 to 6 is implemented.

Images