Bursaphelenchus xylophilus disease propagation path prediction method and device based on artificial intelligence

Abstract

The invention discloses a pine wood nematode disease propagation path prediction method and device based on artificial intelligence. Relates to the technical field of artificial intelligence, and the method comprises the steps: obtaining a plurality of time sequence images in a target range with a target epidemic wood as the center, constructing a node corresponding to each geographic element in the time sequence images, and obtaining a plurality of nodes; obtaining an adjacent matrix corresponding to each time sequence image according to the plurality of nodes and the incidence relation between the nodes; obtaining a weight parameter corresponding to each grid in each time sequence image, and obtaining a target weight matrix corresponding to each time sequence image; and performing propagation path prediction based on the adjacent matrixes corresponding to all the time sequence images and the target weight matrixes corresponding to all the time sequence images through the target strategy network to obtain a prediction result. According to the method and the device, the problem that the propagation path prediction accuracy is relatively low due to the fact that the propagation path of the pine wood nematode disease is predicted by means of manual patrol and fixed quarantine measures in the related technology is solved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and more specifically, to a method and apparatus for predicting the transmission path of pine wilt disease based on artificial intelligence. Background Technology

[0002] Pine wilt disease, a serious biological threat to the health of pine forests worldwide, has always been a crucial issue in forestry conservation. Current technologies largely rely on traditional manual patrols and fixed quarantine measures for monitoring and prediction. While these methods have been effective in the past, their limitations are becoming increasingly apparent in today's rapidly changing forest environment. In recent years, with the expansion of forest cover and the impact of climate change, the spread of pine wilt disease has accelerated in both speed and scope. Against this backdrop, high-precision prediction of the disease's transmission routes has become particularly urgent in order to develop effective control strategies in advance and reduce damage to pine forest resources.

[0003] However, current monitoring and forecasting technologies face several challenges: manual patrols are inefficient and difficult to cover vast forest areas, especially in remote areas and complex terrain conditions; fixed quarantine measures lack flexibility and cannot respond in real time to the dynamic changes of pests and diseases; and data integration is insufficient, as static data from manual sampling is difficult to deeply correlate and analyze with dynamic data from the Internet of Things, remote sensing, etc., which limits the timeliness and accuracy of forecasts.

[0004] There is currently no effective solution to the problem that relying on manual inspections and fixed quarantine measures to predict the transmission path of pine wilt disease in related technologies results in relatively low accuracy in predicting the transmission path. Summary of the Invention

[0005] The main objective of this application is to provide a method and apparatus for predicting the transmission path of pine wilt disease based on artificial intelligence, in order to solve the problem that the accuracy of the prediction of the transmission path of pine wilt disease is relatively low due to the reliance on manual inspection and fixed quarantine measures in related technologies.

[0006] To achieve the above objectives, according to one aspect of this application, an artificial intelligence-based method for predicting the transmission path of pine wilt disease is provided. The method includes: acquiring multiple time-series images within a target area centered on a target infected tree; for each time-series image, constructing nodes corresponding to each geographic feature in the image to obtain multiple nodes, wherein the target infected tree is a pine tree infected with pine wilt disease; obtaining an adjacency matrix corresponding to each time-series image based on the multiple nodes and the relationships between nodes; acquiring weight parameters corresponding to each grid cell in each time-series image to obtain a target weight matrix corresponding to each time-series image; and predicting the transmission path using a target policy network based on the adjacency matrix and the target weight matrix corresponding to all time-series images to obtain a prediction result.

[0007] Furthermore, for each time series image, constructing a node corresponding to each geographic element in the time series image to obtain multiple nodes includes: classifying the geographic elements in each time series image to obtain classification results; labeling the geographic elements in each time series image with preset values ​​based on the classification results to obtain a labeled time series image; and constructing a node corresponding to each geographic element based on the labeled time series image to obtain the multiple nodes.

[0008] Further, obtaining the weight parameters corresponding to each grid cell in each time series image to obtain the target weight matrix corresponding to each time series image includes: obtaining the geographic environment weight parameters corresponding to each grid cell in each time series image to obtain the first weight map corresponding to the natural influence factor; obtaining the human factor weight parameters corresponding to each grid cell in each time series image to obtain the second weight map corresponding to the human influence factor; and performing a weighted calculation on the first weight map and the second weight map to obtain the target weight matrix.

[0009] Furthermore, propagation path prediction is performed by the target policy network based on the adjacency matrix and the target weight matrix corresponding to all temporal images. The prediction result includes: extracting spatial and temporal features based on the adjacency matrix and the target weight matrix to obtain target feature information; processing the target feature information through the target policy network to obtain the probability distribution of each node in the adjacency matrix being infected; and obtaining the prediction result based on the probability distribution of each node being infected.

[0010] Further, spatial and temporal feature extraction is performed based on the adjacency matrix and the target weight matrix to obtain target feature information, including: obtaining first node spatial features based on the adjacency matrix and target weight matrix corresponding to the first temporal image among the plurality of temporal images; calculating multiple second node spatial features by using a graph convolutional network on the first node spatial features, adjacency matrices other than the adjacency matrix corresponding to the first temporal image, and target weight matrices other than the target weight matrix corresponding to the first temporal image; extracting node temporal features by using a long short-term memory network on the multiple second node spatial features; and obtaining target feature information based on the multiple second node spatial features and the node temporal features.

[0011] Further, the target policy network is trained using the following steps: acquiring a set of time-series sample images and acquiring sample feature information corresponding to the set of time-series sample images; processing the sample feature information through an initial policy network to obtain a first predicted probability distribution of infection for each sample node; processing the sample feature information through an initial value network to obtain a current value; processing the sample feature information through a target search algorithm to obtain a second predicted probability distribution of infection for each sample node; and training the initial policy network based on the first predicted probability distribution, the current value, and the second predicted probability distribution to obtain the target policy network.

[0012] Further, processing the sample feature information using a target search algorithm to obtain a second predicted probability distribution of each sample node's infection includes: obtaining the number of times each sample node is visited in the sample neighbor matrix corresponding to the sample feature information; processing the features corresponding to each sample node in the sample feature information using a target value network to obtain the cumulative incentive corresponding to each sample node; and calculating the number of times each sample node is visited, the cumulative incentive corresponding to each sample node, and the first predicted probability distribution to obtain the second predicted probability distribution.

[0013] Furthermore, after processing the sample feature information using a target search algorithm to obtain a second predicted probability distribution of infection for each sample node, the method further includes: updating the sample time series graphs in the time series sample image set according to the second predicted probability distribution of infection for each sample node, to obtain an updated time series sample image set; and training the initial policy network based on the updated time series sample image set to obtain the target policy network.

[0014] To achieve the above objectives, according to another aspect of this application, an artificial intelligence-based device for predicting the transmission path of pine wilt disease is provided. The device includes: a first acquisition unit, configured to acquire multiple time-series images within a target area centered on a target infected tree, and for each time-series image, construct nodes corresponding to each geographic feature in the time-series image to obtain multiple nodes, wherein the target infected tree is a pine tree infected with pine wilt disease; a determination unit, configured to obtain an adjacency matrix corresponding to each time-series image based on the multiple nodes and the association relationships between nodes; a second acquisition unit, configured to acquire weight parameters corresponding to each grid cell in each time-series image to obtain a target weight matrix corresponding to each time-series image; and a prediction unit, configured to predict the transmission path using a target policy network based on the adjacency matrix and the target weight matrix corresponding to all time-series images, to obtain a prediction result.

[0015] Furthermore, the first acquisition unit includes: a classification subunit, used to classify the geographic elements in each time series image to obtain a classification result; a labeling subunit, used to label the geographic elements in each time series image with a preset value based on the classification result to obtain a labeled time series image; and a construction subunit, used to construct a node corresponding to each geographic element based on the labeled time series image to obtain the plurality of nodes.

[0016] Further, the second acquisition unit includes: a first acquisition subunit, used to acquire the geographic environment weight parameters corresponding to each grid in each time series image, to obtain a first weight map corresponding to the natural influence factor; a second acquisition subunit, used to acquire the human factor weight parameters corresponding to each grid in each time series image, to obtain a second weight map corresponding to the human influence factor; and a first calculation subunit, used to perform weighted calculation on the first weight map and the second weight map to obtain the target weight matrix.

[0017] Further, the prediction unit includes: an extraction subunit, used to extract spatial features and temporal features based on the adjacency matrix and the target weight matrix to obtain target feature information; a first processing subunit, used to process the target feature information through a target policy network to obtain the probability distribution of each node in the adjacency matrix being infected; and a determination subunit, used to obtain the prediction result based on the probability distribution of each node being infected.

[0018] Further, the extraction subunit includes: a first determining module, used to obtain a first node spatial feature based on the adjacency matrix and target weight matrix corresponding to the first time-series image among the plurality of time-series images; a calculation module, used to calculate a plurality of second node spatial features by using a graph convolutional network on the first node spatial features, adjacency matrices other than the adjacency matrix corresponding to the first time-series image, and target weight matrices other than the target weight matrix corresponding to the first time-series image; an extraction module, used to extract features from the plurality of second node spatial features by using a long short-term memory network to obtain node temporal features; and a second determining module, used to obtain target feature information based on the plurality of second node spatial features and the node temporal features.

[0019] Further, the target policy network is trained using the following apparatus: a third acquisition unit, used to acquire a set of time-series sample images and acquire sample feature information corresponding to the set of time-series sample images; a first processing unit, used to process the sample feature information through an initial policy network to obtain a first predicted probability distribution of infection for each sample node; a second processing unit, used to process the sample feature information through an initial value network to obtain a current value; a third processing unit, used to process the sample feature information through a target search algorithm to obtain a second predicted probability distribution of infection for each sample node; and a training unit, used to train the initial policy network based on the first predicted probability distribution, the current value, and the second predicted probability distribution to obtain the target policy network.

[0020] Further, the third processing unit includes: a third acquisition subunit, used to acquire the number of times each sample node in the sample neighbor matrix corresponding to the sample feature information is accessed; a second processing subunit, used to process the features corresponding to each sample node in the sample feature information through the target value network to obtain the cumulative incentive corresponding to each sample node; and a second calculation subunit, used to calculate the number of times each sample node is accessed, the cumulative incentive corresponding to each sample node, and the first prediction probability distribution to obtain the second prediction probability distribution.

[0021] Furthermore, the apparatus further includes: an update unit, configured to update the sample time series image in the time series image set according to the second predicted probability distribution of each sample node being infected after processing the sample feature information through a target search algorithm to obtain the second predicted probability distribution of each sample node being infected, thereby obtaining an updated time series image set; and a training unit, configured to train the initial policy network according to the updated time series image set to obtain the target policy network.

[0022] According to another aspect of the present invention, an electronic device is also provided, comprising: a memory storing an executable program; and a processor for running the program, wherein the program executes the artificial intelligence-based method for predicting the transmission path of pine wilt disease during runtime.

[0023] According to another aspect of the present invention, a computer-readable storage medium is also provided, the storage medium storing a program, wherein, when the program is running, the device where the storage medium is located executes any of the above-mentioned artificial intelligence-based method for predicting the transmission path of pine wilt disease.

[0024] In this embodiment, the following steps are adopted: Multiple time-series images are acquired within a target area centered on the target infected tree. For each time-series image, nodes corresponding to each geographic feature in the image are constructed, resulting in multiple nodes. The target infected tree is a pine tree infected with pine wilt disease. Geographical features include: pine trees infected with pine wilt disease, pine trees not infected with pine wilt disease, non-pine trees, rivers, roads, and open spaces. Based on the multiple nodes and the association relationships between nodes, an adjacency matrix corresponding to each time-series image is obtained. The weight parameters corresponding to each grid in each time-series image are obtained, resulting in a target weight matrix corresponding to each time-series image. The propagation path is predicted using a target policy network based on the adjacency matrix and target weight matrix corresponding to all time-series images, and the prediction result is obtained. This solves the technical problem in related technologies where relying on manual inspections and fixed quarantine measures to predict the propagation path of pine wilt disease leads to relatively low accuracy in propagation path prediction.

[0025] In this scheme, high-precision UAV or satellite remote sensing technology is used to acquire images of the target infected trees and their surrounding environment, resulting in time-series images containing multiple time points. Geographical features in each time-series image are identified, including infected pine trees, uninfected pine trees, non-pine trees, rivers, roads, and open spaces. For each identified geographic feature, corresponding nodes are created, such as infected pine trees, healthy pine trees, other tree species, rivers, roads, and open spaces. Based on the spatial relationships and possible interactions between nodes, the association relationships between nodes are constructed and transformed into an adjacency matrix. Based on the geographic environmental characteristics of each grid (such as topography, climate conditions, human activities, etc.), corresponding weight parameters are determined to reflect the influence of that location on the spread of pests and diseases. By combining the weight parameters of each grid in the time-series image, a target weight matrix is ​​generated as a quantitative representation of the impact of environmental factors on the spread of pests and diseases. Finally, the target strategy network predicts the propagation path based on the adjacency matrix and the target weight matrix, and obtains the prediction results. This scheme can achieve accurate prediction of the propagation path of pine wilt disease. Compared with traditional manual inspection and fixed quarantine measures, this scheme overcomes the defects of limited monitoring coverage and poor real-time performance, and can effectively improve the efficiency and accuracy of pest and disease monitoring and early warning. Attached Figure Description

[0026] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0027] Figure 1 A hardware block diagram of a computer terminal for implementing an artificial intelligence-based method for predicting the transmission path of pine wilt disease is shown.

[0028] Figure 2 This is a flowchart of a method for predicting the transmission path of pine wilt disease based on artificial intelligence, according to an embodiment of this application.

[0029] Figure 3 This is a schematic diagram of a method for predicting the transmission path of pine wilt disease based on artificial intelligence, according to an embodiment of this application.

[0030] Figure 4 This is a schematic diagram of an artificial intelligence-based device for predicting the transmission path of pine wilt disease according to an embodiment of this application;

[0031] Figure 5 This is a structural block diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0032] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0033] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0034] It should be noted that the information collected in this application (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data used for analysis, etc.) are information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of this data all comply with relevant laws, regulations, and standards, necessary confidentiality measures have been taken, and they do not violate public order and good morals. Corresponding access points are provided for users to choose to authorize or refuse. For example, interfaces are set up between this system and relevant users or organizations, providing users with corresponding access points to choose to agree to or refuse automated decision-making results; if the user chooses to refuse, the process proceeds to the expert decision-making stage.

[0035] Example 1

[0036] According to an embodiment of this application, an embodiment of a method for predicting the transmission path of pine wilt disease based on artificial intelligence is also provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0037] The method embodiment provided in Embodiment 1 of this application can be executed on a mobile terminal, computer terminal, or similar computing device. Figure 1A hardware block diagram of a computer terminal (or mobile device) for implementing an AI-based method to predict the transmission path of pine wilt disease is shown. Figure 1 As shown, the computer terminal 10 (or mobile device) may include one or more processors 102 (shown as 102a, 102b, ..., 102n in the figure) 102 (processor 102 may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of a BUS bus), a network interface, a power supply, and / or a camera. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, computer terminal 10 may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0038] It should be noted that the aforementioned one or more processors 102 and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be wholly or partially integrated into any other element within the computer terminal 10 (or mobile device). As involved in the embodiments of this application, the data processing circuits serve as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0039] The memory 104 can be used to store software programs and modules for application software, such as the program instructions / data storage device corresponding to the AI-based method for predicting the transmission path of pine wilt disease in this embodiment. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby realizing the aforementioned AI-based method for predicting the transmission path of pine wilt disease. The memory 104 may include high-speed random access memory and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the computer terminal 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0040] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the computer terminal 10. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0041] The display may be, for example, a touchscreen liquid crystal display (LCD) that allows the user to interact with the user interface of the computer terminal 10 (or mobile device).

[0042] Under the aforementioned operating environment, this application provides the following: Figure 2 The method shown is an artificial intelligence-based method for predicting the transmission path of pine wilt disease. Figure 2 This is a flowchart of an AI-based method for predicting the transmission path of pine wilt disease according to Embodiment 1 of this application. The AI-based method for predicting the transmission path of pine wilt disease includes:

[0043] Step S201: Obtain multiple time-series images within the target range centered on the target infected tree, and for each time-series image, construct nodes corresponding to each geographic element in the time-series image to obtain multiple nodes, wherein the target infected tree is a pine tree infected with pine wilt disease.

[0044] Optionally, high-precision image data (the initial data acquisition should identify the morphology, color, and texture of the Masson pine; for cost considerations, the accuracy of subsequent time-series acquisitions can be slightly lower; if the forest stand changes, high-precision data needs to be acquired again) is obtained through drone photography at different time periods or points in time. Coordinate registration is performed on data from different periods in the same area, and sample data is labeled to train a classification target recognition model for pine trees infected with pine wilt disease (divided into early infection stage, wilting stage, etc.), healthy pine trees, and other vegetation based on the full dataset. Then, the infected tree identification model is used to identify the trees and label their geographical coordinates. A sub-map is constructed with the identified infected tree as the center, centered on the spread distance of the pine sawyer beetle (e.g., 200m, 300m, 500m) as the radius (if the radius is 500m, the sub-map is a 1000m × 1000m rectangular area), resulting in the aforementioned multiple time-series images.

[0045] For each time-series image, nodes corresponding to each geographic feature in that image are constructed, resulting in multiple nodes. It should be noted that geographic features include: pine trees infected with pine wilt disease, pine trees not infected with pine wilt disease, non-pine trees, rivers, roads, and open spaces. Pine trees infected with pine wilt disease: serve as sources of transmission and key monitoring nodes. Pine trees not infected with pine wilt disease: potential infection targets, requiring monitoring of their infection risk. Non-pine trees: although not direct transmission paths, their presence may influence the transmission patterns of the pest. Rivers: act as natural buffer zones, potentially limiting the lateral spread of the pest. Roads: areas of intensive human activity, potentially pathways for the leapfrog spread of the pest. Open spaces: areas without vegetation cover, with a lower likelihood of pest transmission. After identifying and classifying these geographic features, each feature is transformed into a node in the graph, resulting in the aforementioned multiple nodes.

[0046] For example, by identifying the coordinates of infected trees, a rectangle with a side length of 1000m is obtained using these coordinates as the center point. The map is then constructed according to geographical features, with infected pine trees as 1, uninfected pine trees as 0, other tree species as -1, rivers as -2, and roads and open spaces as -3. The map is initialized with a 2m grid, and multiple nodes are obtained by marking them.

[0047] Step S202: Based on multiple nodes and the association relationships between nodes, obtain the adjacency matrix corresponding to each time series image.

[0048] Optionally, the adjacency matrix is ​​formed by connecting trees whose distance is less than a threshold (the threshold can be set based on the single flight distance of the longhorn beetle, with an initial threshold of 1 meter that can be optimized later), and not connecting trees that encounter rivers or roads.

[0049] For example, if the distance between node i and node j is less than a set threshold (e.g., 1 meter), and there are no natural or man-made obstacles such as rivers or roads directly blocking their path, then Aij is set to 1, indicating a direct connection. Conversely, if the distance exceeds the threshold, or if there are obstacles, Aij is set to 0, indicating no direct connection. For rivers and roads acting as obstacles, Aij is set to 0 even if the distance between nodes i and j is less than the threshold, reflecting the effectiveness of these obstacles in limiting pest and disease spread. By setting connection thresholds and identifying geographical obstacles, an adjacency matrix reflecting the probability of pest and disease spread is constructed, improving the accuracy of subsequent spread path prediction.

[0050] Step S203: Obtain the weight parameters corresponding to each grid in each time series image to obtain the target weight matrix corresponding to each time series image.

[0051] Optionally, a weight parameter can be assigned to each raster (or node) in each time-series image to reflect the potential impact of that location on the spread of pine wilt disease. A weight matrix is ​​a means of quantifying the influence of geographical environmental factors on the spread process, visually demonstrating which areas are more likely to become bridges for pest and disease transmission, and which areas play an inhibitory role.

[0052] In an optional embodiment, the weighting parameters include, but are not limited to, the weights of natural influencing factors and human factors. Natural environmental factors, such as soil type, climate conditions (temperature, humidity, rainfall), and topographic features (slope, altitude), have a direct or indirect impact on the spread of pests and diseases. For example, a humid climate may favor the activity of pest vector insects, thereby increasing the probability of pest and disease spread; high-slope areas may slow the spread of pests and diseases due to their less accessible environment. The weights of these natural influencing factors can be determined through data analysis and human experience, reflecting their degree of promotion or inhibition of pest and disease spread. Human activities, such as logging, planting, and pesticide spraying, also affect the spread pathways of pests and diseases. For example, logging areas may reduce the risk of spread because they reduce the number of host trees.

[0053] For example, by combining relevant information such as remote sensing image data, meteorological data, terrain data, and human activity data, a comprehensive weight is calculated for each grid location, thereby obtaining the target weight matrix corresponding to each time series image mentioned above.

[0054] Step S204: The propagation path is predicted by the target policy network based on the adjacency matrix and the target weight matrix corresponding to all time series images, and the prediction result is obtained.

[0055] Optionally, the adjacency matrix and the target weight matrix can be combined to form a comprehensive representation of the environmental state. The adjacency matrix reflects the direct connection state between nodes, while the weight matrix quantifies the propagation capability of each node (or grid) under the influence of natural and human factors. The combination of these two can comprehensively describe the potential propagation network and efficiency of pests and diseases at a specific time point. In an optional embodiment, the adjacency matrices and target weight matrices corresponding to all time-series images are converted into feature vectors. These feature vectors are then used as input to a pre-trained target policy network. Finally, the target policy network outputs the probability value of each node being infected with pine wilt disease at a future time point, thus obtaining the prediction result.

[0056] In summary, high-precision UAV or satellite remote sensing technologies were used at different time periods or points in time to acquire images of the target infected trees and their surrounding environment, resulting in time-series images containing multiple time points. Geographical features in each time-series image were identified, including infected pine trees, uninfected pine trees, non-pine trees, rivers, roads, and open spaces. For each identified geographic feature, corresponding nodes were created, such as infected pine trees, healthy pine trees, other tree species, rivers, roads, and open spaces. Based on the spatial relationships and possible interactions between nodes, the association relationships between nodes were constructed and transformed into an adjacency matrix. Based on the geographic environmental characteristics of each grid (such as topography, climate conditions, human activities, etc.), corresponding weight parameters were determined to reflect the influence of that location on the spread of pests and diseases. By combining the weight parameters of each grid in the time-series image, a target weight matrix is ​​generated as a quantitative representation of the impact of environmental factors on the spread of pests and diseases. Finally, the propagation path is predicted based on the adjacency matrix and the target weight matrix to obtain the prediction results. This scheme can achieve accurate prediction of the propagation path of pine wilt disease. Compared with traditional manual inspection and fixed quarantine measures, this scheme overcomes the defects of limited monitoring coverage and poor real-time performance, and can effectively improve the efficiency and accuracy of pest and disease monitoring and early warning.

[0057] Optionally, in the AI-based method for predicting the transmission path of pine wilt disease provided in this application embodiment, for each time series image, constructing nodes corresponding to each geographic element in the time series image to obtain multiple nodes includes: classifying the geographic elements in each time series image to obtain classification results; labeling the geographic elements in each time series image with preset values ​​based on the classification results to obtain labeled time series images; and constructing nodes corresponding to each geographic element based on the labeled time series images to obtain multiple nodes.

[0058] In an optional embodiment, geographic features in each time-series image are classified. These geographic features include, but are not limited to, infected trees (pine trees infected with pine wilt disease), healthy pine trees, non-pine trees, rivers, roads, and open spaces. Based on the classification results, each geographic feature in the time-series image is labeled and assigned a specific numerical value. This step transforms the classification information into a form that the model can directly process. For example, infected trees are labeled as 1, healthy pine trees as 0, non-pine trees as -1, rivers as -2, roads as -3, and open spaces as -4. Based on the labeled time-series images, each geographic feature is transformed into a node in the graph, resulting in the aforementioned multiple nodes.

[0059] The above steps can effectively transform complex geographical information into nodes in a graph structure, thereby enabling accurate prediction of the transmission path of pine wilt disease.

[0060] Optionally, in the AI-based method for predicting the transmission path of pine wilt disease provided in this application embodiment, obtaining the weight parameters corresponding to each grid in each time series image to obtain the target weight matrix corresponding to each time series image includes: obtaining the geographical environment weight parameters corresponding to each grid in each time series image to obtain a first weight map corresponding to natural influencing factors; obtaining the human factor weight parameters corresponding to each grid in each time series image to obtain a second weight map corresponding to human influencing factors; and performing weighted calculation on the first weight map and the second weight map to obtain the target weight matrix.

[0061] In an optional embodiment, natural influencing factors include, but are not limited to, topography, weather, and water system distribution. Human influencing factors include, but are not limited to, human interventions such as logging, pesticide spraying, and isolation. Different values ​​can be set for each natural influencing factor based on human experience and data analysis, and then different weighting coefficients can be set according to specific environmental parameters to obtain a weighted map corresponding to each natural influencing factor. Topography: Initialize the weighting coefficient for topographic influence propagation. Influence factors are composed of factors such as regional DEM elevation model (calculating slope), aspect, and illumination factors, and initialize the topographic influence weighting coefficient a. Water system distribution: Water systems serve as important natural isolation zones for the propagation of the pine sawyer beetle. Rivers with a width greater than 10m (not a limited value, determined based on the maximum single flight distance of the pine sawyer beetle, adjustable) have a resistance coefficient set to 1, or a weighting coefficient of 0. Weather: Analysis of climate data from the past six months, including temperature, humidity, and rainfall analysis, with a weighting coefficient b. Human influencing factors: If the area has undergone treatments such as logging, pesticide spraying, or isolation, the resistance coefficient is set to 1, i.e., the weighting coefficient is set to 0. Roads: Roads are primarily considered a human-influenced factor. The weight of the area within 100m of the road's adjacent edge (not a fixed value, adjustable) is uniformly multiplied by a weighting coefficient c, indicating a strong correlation. It should be noted that the weighting coefficient characterizes the susceptibility to infectious diseases and pests; generally, a higher coefficient indicates greater susceptibility. The resistance coefficient characterizes the degree to which the spread of infectious diseases and pests is hindered; generally, a higher coefficient indicates less susceptibility.

[0062] By calculating the weight map corresponding to each natural influencing factor and then summing them, the first weight map is obtained. The second weight value can be obtained in the same way. Finally, the target weight matrix is ​​obtained by weighting the first and second weight maps.

[0063] In an optional embodiment, the target weight value is calculated using the following formula: ,in, and This is an adjustable parameter.

[0064] By collecting and processing relevant data on natural and anthropogenic influencing factors, a detailed target weight matrix can be constructed. This not only reflects the impact of the environment itself on the spread of pests and diseases, but also takes into account the changes brought about by human activities, thus enabling a more accurate prediction of the transmission path of pine wilt disease.

[0065] Optionally, in the AI-based method for predicting the transmission path of pine wilt disease provided in this application embodiment, the transmission path is predicted by a target policy network based on the adjacency matrix and the target weight matrix corresponding to all time-series images. The prediction results include: extracting spatial features and temporal features based on the adjacency matrix and the target weight matrix to obtain target feature information; processing the target feature information through the target policy network to obtain the probability distribution of each node in the adjacency matrix being infected; and obtaining the prediction results based on the probability distribution of each node being infected.

[0066] In an alternative embodiment, Graph Convolutional Networks (GCNs) or other models suitable for processing graph-structured data can be used to extract spatial features based on adjacency matrices and target weight matrices. The adjacency matrix reveals the connectivity between geographic feature nodes, while the target weight matrix quantifies the propagation capability of each node and the impact of environmental barriers. GCNs can learn the spatial characteristics of each node (such as the propagation potential of its surrounding environment) through information transfer and aggregation between nodes, forming node embedding vectors. Alternatively, temporal models such as Long Short-Term Memory Networks (LSTMs) or Transformers can be used to capture propagation trends and periodic patterns by utilizing the sequence of node state changes in historical time-series images, integrating spatial and temporal features to obtain the aforementioned target feature information.

[0067] The target feature information is input into the target policy network, which is responsible for learning the mapping from the current state to the next action (i.e., predicting the next node that may be infected). The policy network can be designed based on PPO (Proximal Policy Optimization) or other highly adaptive algorithms capable of handling complex decision-making problems.

[0068] After processing the feature information, the target policy network outputs the probability distribution of each node being infected. This is achieved through the output layer of the policy network (usually with a Softmax activation function), ensuring that the sum of the output probabilities is 1, reflecting which node the pest is most likely to spread through in the current state, thus obtaining the prediction results mentioned above.

[0069] By extracting spatial and temporal features and optimizing the policy network, a method for predicting the transmission path of pine wilt disease based on graph structure and environmental weights was developed. This method not only considers the physical characteristics of pest transmission but also incorporates the influence of environmental and anthropogenic factors, making the prediction more accurate and practical.

[0070] Optionally, in the AI-based method for predicting the transmission path of pine wilt disease provided in this application embodiment, spatial feature extraction and temporal feature extraction are performed based on the adjacency matrix and the target weight matrix to obtain target feature information, including: obtaining the spatial features of a first node based on the adjacency matrix and the target weight matrix corresponding to the first temporal image among multiple temporal images; calculating multiple second node spatial features by using a graph convolutional network on the first node spatial features, the adjacency matrix other than the adjacency matrix corresponding to the first temporal image, and the target weight matrix other than the target weight matrix corresponding to the first temporal image; extracting node temporal features by using a long short-term memory network on the multiple second node spatial features; and obtaining target feature information based on the multiple second node spatial features and the node temporal features.

[0071] In an optional embodiment, the image at the earliest moment is selected from multiple temporal images, and the spatial features of the first node are obtained by constructing an adjacency matrix and a target weight matrix based on this image. Starting from the second temporal image, the corresponding adjacency matrix and target weight matrix are processed one by one. The adjacency matrix and target weight matrix corresponding to each temporal image are input into the graph convolutional network (GCN) to calculate the spatial features of each node at that moment, resulting in multiple spatial features of the second node.

[0072] Then, a Long Short-Term Memory (LSTM) network is used to extract temporal features from the spatial features of multiple second-node nodes. LSTM, through its internal memory units and gating mechanisms, analyzes the sequence of spatial features to extract the temporal features of each node. These temporal features reveal the dynamic trends in pest and disease transmission. The spatial features of each node obtained from all time-series images are combined with the temporal features, either through simple concatenation or more complex methods (such as weighted summation or attention mechanisms), to generate the final target feature information.

[0073] In an optional embodiment, a third-generation GCN algorithm using Chebyshev polynomials for approximate fitting is employed, with the inter-layer propagation formula as follows:

[0074]

[0075] in, Let l be the feature matrix of the l-th layer; The ReLU activation function is used; D is the degree matrix, constructed based on the number of edges adjacent to node i; A represents the adjacency matrix, and I is the identity matrix. This is the weight matrix. After stacking two layers of GCN, low-dimensional embedding vectors of the nodes are generated, which serve as inputs for subsequent modules. It is obtained based on the target weight matrix.

[0076] Construct an adjacency matrix A1, A1, A2...At based on time-series data. Input the dynamic adjacency matrix sequence A1, A2...At and the initial node features H(0). Generate the spatiotemporal joint features HST (i.e., the target feature information mentioned above) through ST-GCN:

[0077]

[0078]

[0079]

[0080] Target Policy Network: Input spatiotemporal joint features HST, the structure consists of a fully connected layer (256-128-64-64) + softmax, outputting the action probability distribution Π(a|s), selecting the recognition node as the action space (selecting the next infected node), and then the final prediction result.

[0081] The above steps can effectively extract spatial and temporal features from multiple time-series images, thereby improving the accuracy of predicting the transmission path of pine wilt disease.

[0082] Optionally, in the AI-based method for predicting the transmission path of pine wilt disease provided in this application embodiment, the target policy network is trained using the following steps: acquiring a time-series sample image set and acquiring sample feature information corresponding to the time-series sample image set; processing the sample feature information through an initial policy network to obtain a first predicted probability distribution of infection for each sample node; processing the sample feature information through an initial value network to obtain a current value; processing the sample feature information through a target search algorithm to obtain a second predicted probability distribution of infection for each sample node; and training the initial policy network based on the first predicted probability distribution, the current value, and the second predicted probability distribution to obtain the target policy network.

[0083] In an optional embodiment, time-series images containing information on the spread of pests and diseases over different time periods are included. These time-series images can cover the complete pest and disease transmission cycle, from initial infection to widespread dissemination. Each image contains information on the status and distribution of geographic features, as well as a weighted graph, thus resulting in a set of time-series sample images.

[0084] Then, the initial policy network is used to process the sample feature information to obtain the first predicted probability distribution of each sample node being infected. This step is the foundation for the policy network to learn how to predict the next action (infected node) from the current environmental state. The role of the initial value network is to evaluate the long-term value of a given state, that is, the expected return that can be obtained by following a certain policy from that state. By processing the sample feature information through the value network, the current value (i.e., the estimated expected return of the current state) is obtained, which provides guidance for the optimization of the policy network.

[0085] Secondly, target search algorithms (such as Monte Carlo Tree Search, MCTS) are used to further process the sample feature information to obtain a second predicted probability distribution of infection for each sample node. MCTS can not only explore multiple possible actions, but also evaluate the long-term effects of these actions, providing more valuable data for the policy network.

[0086] By combining the first predicted probability distribution generated by the policy network, the current value given by the value network, and the second predicted probability distribution generated by MCTS, the initial policy network and the initial value network are trained and updated to obtain the final target policy network.

[0087] In an optional embodiment, the loss functions for the policy network and the value network are as follows:

[0088] Policy Networks:

[0089] Value Network:

[0090] in, Let the loss function be the policy network. Let be the loss function of the value network; The ratio between the first and second predicted probability distributions. For the dominant function, This is the trimming parameter, usually set to 0.1 or 0.2. The difference between the current value function prediction and the actual observed short-term returns; It is the state-value function; t is time; It is the value function of the target state.

[0091] By combining the MCTS search algorithm with the value network, the policy network can make decisions based on both immediate and long-term revenue information, thereby improving the accuracy of the prediction model.

[0092] Optionally, in the AI-based method for predicting the transmission path of pine wilt disease provided in this application embodiment, processing sample feature information using a target search algorithm to obtain a second predicted probability distribution of each sample node being infected includes: obtaining the number of times each sample node is visited in the sample neighbor matrix corresponding to the sample feature information; processing the features corresponding to each sample node in the sample feature information using a target value network to obtain the cumulative incentive corresponding to each sample node; and calculating the number of times each sample node is visited, the cumulative incentive corresponding to each sample node, and the first predicted probability distribution to obtain the second predicted probability distribution.

[0093] In an optional embodiment, each node (i.e., a geographic feature, such as infected trees, healthy pine trees, etc.) is considered a node in the MCTS search tree. The root node of the tree can be set as the currently known infected node. The MCTS algorithm expands and searches the nodes of the tree based on policy prior probabilities and value assessments. In each round of the MCTS search, the algorithm selects and visits nodes in the tree according to the probability distribution output by the policy network. The number of visits for each visited node is accumulated. The statistics of the number of visits reflect the extent to which MCTS explores each potential infected node, indirectly indicating its probability as the next step in the propagation path, thus obtaining the number of times each sample node is visited.

[0094] The target value network processes the features corresponding to each sample node in the sample feature information. The task of the value network is to evaluate the long-term benefits or penalties after taking a certain propagation action starting from the current node, that is, to assess the potential risks or benefits of the propagation path. The target value network outputs the immediate incentive value of each sample node, reflecting the immediate benefit of that node as the next step in the propagation path. During the MCTS search process, the algorithm accumulates the immediate incentive of each node to form the cumulative incentive of that node.

[0095] MCTS uses the UCT (Upper Confidence bounds applied to Trees) formula to balance exploration and exploitation to determine the next node to visit. The UCT formula combines the number of times a node is visited, the cumulative incentive, and the prior probability of the policy, and its calculation formula is as follows:

[0096]

[0097] in, This refers to a node in the mesh. The probability of expansion (infection probability) determines the probability that the node will be selected for further expansion during the search process; It is represented as a constant that controls the degree of exploration. This value determines the balance between exploring new options (the model may explore data from the infected tree based on existing data) and utilizing existing information. A larger value will lead to a greater bias towards exploring options that have not been fully evaluated, while a smaller value will tend to choose actions with known higher rewards. Let be the node win rate, where =W / N It's just a node. Cumulative rewards / Number of times visited, W PPO-based value network Computational acquisition (using PPO's value network) It replaces the traditional Rollout, reducing computational overhead. This represents the policy prior probability, which is the first predicted probability distribution mentioned above.

[0098] By using the target search algorithm MCTS to process sample feature information, we can not only obtain the second probability distribution of each node's infection, but also continuously optimize the prediction model through interaction with the policy network and value network, thereby improving the accuracy and stability of prediction.

[0099] Optionally, in the AI-based method for predicting the transmission path of pine wilt disease provided in this application embodiment, after processing the sample feature information through a target search algorithm to obtain a second predicted probability distribution of infection for each sample node, the method further includes: updating the sample time series map in the time series sample image set according to the second predicted probability distribution of infection for each sample node to obtain an updated time series sample image set; and training the initial policy network according to the updated time series sample image set to obtain a target policy network.

[0100] In an optional embodiment, the second predicted probability distribution reflects the likelihood of each sample node being infected under the current policy network. Using this information, the node state of each time-series graph can be updated in the set of time-series sample images. For example, if a node is predicted to have a high probability of being infected, its state in the next time-series sample graph can be marked as "infected". The adjacency matrix and feature matrix of the graph are updated based on the new infection state, and the GCN embedding is recalculated to complete the dynamic update of the graph.

[0101] The search results of MCTS are used as additional training data for PPO to optimize the policy network, so that the policy update of PPO can further understand the search direction of MCTS, thus achieving a closed loop space of policy-search-optimization.

[0102] Through deep search and closed-loop iteration of the policy network in MCTS, the model can more accurately predict the infection probability of each node and the future path of pest and disease transmission. This prediction not only considers the immediate infection risk but also takes into account long-term trends based on historical transmission patterns and environmental factors, significantly improving the accuracy of the prediction.

[0103] In an alternative embodiment, it can be achieved through, as follows: Figure 3 The diagram shown illustrates the prediction of pine wilt disease transmission pathways based on artificial intelligence, including: Step S301: Data acquisition, processing, and training:

[0104] High-precision image data was obtained from drone photography (the initial data acquisition required the identification of the shape, color, and texture characteristics of Masson pine; for cost considerations, the accuracy of subsequent time-series acquisitions could be slightly lower, and high-precision acquisitions were required again when the forest stand changed). Coordinate registration was performed on data from different periods in the same area, and the map was divided into blocks of the same size. Through sample data annotation, the classification target recognition model was trained for pine trees infected with pine wilt disease (divided into early infection stage, withering stage, etc.), healthy pine trees, and other vegetation based on the full dataset.

[0105] Step S302: Identification of Infected Trees and Processing of Influencing Factors: Identification is performed using an infected tree identification model. The identified data is then processed by merging and deduplicating the target bounding box coordinates. Through data analysis and extraction, coordinates within the specified range are extracted for rivers and roads.

[0106] Initialization of Weighted Maps: Natural Influence Factors: Topography: Initialize the weight coefficients for topographic influence propagation. Influence factors are derived from the regional DEM elevation model (calculating slope), aspect, and illumination factors. Initialize the topographic influence weight coefficient 'a'. Roads: Roads are primarily considered anthropogenic influence factors. The weight of the area within 100m of the road's adjacent edge (not a fixed value, adjustable) is uniformly multiplied by coefficient 'c', indicating a strong correlation. Weather: Analysis of climate data from the past six months, including temperature, humidity, and rainfall, with a weight percentage 'b'. Water System Distribution: Water systems serve as important natural buffer zones for the propagation of the pine sawyer beetle. Rivers wider than 10m (not a fixed value, determined based on the maximum single flight distance of the pine sawyer beetle, adjustable) have a resistance coefficient set to 1. Influence Factors of Anthropogenic Treatment Areas: Through data acquisition, if the area has undergone logging, pesticide spraying, or isolation treatments, the resistance coefficient is set to 1.

[0107] Step S303: Graph Construction: Using the identified coordinates of the infected trees as the center point, construct a graph in the original map using rectangles with sides of 1000m, categorized by tree type. Infected pine trees are assigned 1, uninfected pine trees 0, other tree species -1, rivers -2, roads and open spaces -3. Initialize the graph using a 2m grid, marking it to obtain a 500*500 grid data, which serves as the initial adjacency matrix. Calculate and merge the weights of each weighted graph, combining all the weighted graphs corresponding to the above factors into a single comprehensive weighted graph. The value on each grid is the weighted sum of the weights of various factors at the corresponding location.

[0108] Step S304: Training phase: adjacency matrix and feature extraction, PPO strategy and value network design training, and MCTS path search and optimization.

[0109] Step S305: Model Evaluation Phase: Prediction accuracy and F1-Score are used as prediction metrics (verified using historically collected time-series data). The development trajectory of the epidemic is predicted by processing time-series data spanning six months to one year (the shortest period is from the start of insect transmission to the emergence of infected tree characteristics).

[0110] The method for predicting the transmission path of pine wilt disease based on artificial intelligence provided in this application embodiment acquires multiple time-series images within a target area centered on the target infected tree. For each time-series image, nodes corresponding to each geographic element in the image are constructed, resulting in multiple nodes. The target infected tree is a pine tree infected with pine wilt disease, and the geographic elements include: pine trees infected with pine wilt disease, pine trees not infected with pine wilt disease, non-pine trees, rivers, roads, and open spaces. Based on the multiple nodes and the association relationships between nodes, an adjacency matrix corresponding to each time-series image is obtained. The weight parameters corresponding to each grid in each time-series image are obtained, resulting in a target weight matrix corresponding to each time-series image. The transmission path is predicted by a target policy network based on the adjacency matrix and target weight matrix corresponding to all time-series images, and the prediction result is obtained. This solves the technical problem in related technologies where the prediction of the transmission path of pine wilt disease relies on manual inspection and fixed quarantine measures, resulting in relatively low accuracy in the prediction.

[0111] In this scheme, high-precision UAVs or satellite remote sensing technology are used at different time periods or points in time to acquire images of the target infected trees and their surrounding environment, resulting in time-series images containing multiple time points. Geographical features in each time-series image are identified, including infected pine trees, uninfected pine trees, non-pine trees, rivers, roads, and open spaces. For each identified geographic feature, a corresponding node is created, such as infected pine trees, healthy pine trees, other tree species, rivers, roads, and open spaces. Based on the spatial relationships and possible interactions between nodes, the association relationships between nodes are constructed and transformed into an adjacency matrix. Based on the geographic environmental characteristics of each grid (such as topography, climate conditions, human activities, etc.), corresponding weight parameters are determined to reflect the influence of that location on the spread of pests and diseases. By combining the weight parameters of each grid in the time-series image, a target weight matrix is ​​generated as a quantitative representation of the impact of environmental factors on the spread of pests and diseases. Finally, the propagation path is predicted based on the adjacency matrix and the target weight matrix to obtain the prediction results. This scheme can achieve accurate prediction of the propagation path of pine wilt disease. Compared with traditional manual inspection and fixed quarantine measures, this scheme overcomes the defects of limited monitoring coverage and poor real-time performance, and can effectively improve the efficiency and accuracy of pest and disease monitoring and early warning.

[0112] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0113] Example 2

[0114] This application also provides an artificial intelligence-based device for predicting the transmission path of pine wilt disease. It should be noted that this artificial intelligence-based device for predicting the transmission path of pine wilt disease can be used to execute the artificial intelligence-based method for predicting the transmission path of pine wilt disease provided in this application. The following describes the artificial intelligence-based device for predicting the transmission path of pine wilt disease provided in this application.

[0115] According to embodiments of this application, an apparatus is also provided for implementing the above-described artificial intelligence-based method for predicting the transmission path of pine wilt disease, such as... Figure 4 As shown, the device includes: a first acquisition unit 401, a determination unit 402, a second acquisition unit 403, and a prediction unit 404.

[0116] The first acquisition unit 401 is used to acquire multiple time-series images within a target range centered on the target infected tree, and for each time-series image, construct nodes corresponding to each geographic element in the time-series image to obtain multiple nodes, wherein the target infected tree is a pine tree infected with pine wilt disease;

[0117] The determining unit 402 is used to obtain the adjacency matrix corresponding to each time series image based on multiple nodes and the association relationship between nodes;

[0118] The second acquisition unit 403 is used to acquire the weight parameters corresponding to each grid in each time series image, and obtain the target weight matrix corresponding to each time series image;

[0119] The prediction unit 404 is used to predict the propagation path based on the adjacency matrix and the target weight matrix corresponding to all time series images through the target policy network, and obtain the prediction result.

[0120] The artificial intelligence-based pine wilt disease transmission path prediction device provided in this application embodiment acquires multiple time-series images within a target range centered on the target infected tree through a first acquisition unit 401. For each time-series image, it constructs nodes corresponding to each geographical element in the time-series image, resulting in multiple nodes. The target infected tree is a pine tree infected with pine wilt disease, and the geographical elements include: pine trees infected with pine wilt disease, pine trees not infected with pine wilt disease, non-pine trees, rivers, roads, and open spaces. The determination unit 402 obtains the adjacency matrix corresponding to each time-series image based on the multiple nodes and the association relationships between nodes. The second acquisition unit 403 acquires the weight parameters corresponding to each grid in each time-series image, resulting in the target weight matrix corresponding to each time-series image. The prediction unit 404 predicts the transmission path through a target policy network based on the adjacency matrix and the target weight matrix corresponding to all time-series images, and obtains the prediction result. This solves the technical problem in related technologies where relying on manual inspections and fixed quarantine measures to predict the transmission path of pine wilt disease leads to relatively low accuracy in the prediction of the transmission path.

[0121] In this scheme, high-precision UAV or satellite remote sensing technology is used to acquire images of the target infected trees and their surrounding environment, resulting in time-series images containing multiple time points. Geographical features in each time-series image are identified, including infected pine trees, uninfected pine trees, non-pine trees, rivers, roads, and open spaces. For each identified geographic feature, corresponding nodes are created, such as infected pine trees, healthy pine trees, other tree species, rivers, roads, and open spaces. Based on the spatial relationships and possible interactions between nodes, the association relationships between nodes are constructed and transformed into an adjacency matrix. Based on the geographic environmental characteristics of each grid (such as topography, climate conditions, human activities, etc.), corresponding weight parameters are determined to reflect the influence of that location on the spread of pests and diseases. By combining the weight parameters of each grid in the time-series image, a target weight matrix is ​​generated as a quantitative representation of the impact of environmental factors on the spread of pests and diseases. Finally, the target strategy network predicts the propagation path based on the adjacency matrix and the target weight matrix, and obtains the prediction results. This scheme can achieve accurate prediction of the propagation path of pine wilt disease. Compared with traditional manual inspection and fixed quarantine measures, this scheme overcomes the defects of limited monitoring coverage and poor real-time performance, and can effectively improve the efficiency and accuracy of pest and disease monitoring and early warning.

[0122] Optionally, in the AI-based pine wilt disease transmission path prediction device provided in this application embodiment, the first acquisition unit includes: a classification subunit, used to classify the geographic elements in each time series image to obtain classification results; a labeling subunit, used to label the geographic elements in each time series image with preset values ​​according to the classification results to obtain labeled time series images; and a construction subunit, used to construct nodes corresponding to each geographic element according to the labeled time series images to obtain multiple nodes.

[0123] Optionally, in the AI-based pine wilt disease transmission path prediction device provided in this application embodiment, the second acquisition unit includes: a first acquisition subunit, used to acquire the geographic environment weight parameters corresponding to each grid in each time series image to obtain a first weight map corresponding to natural influence factors; a second acquisition subunit, used to acquire the human factor weight parameters corresponding to each grid in each time series image to obtain a second weight map corresponding to human influence factors; and a first calculation subunit, used to perform weighted calculation on the first weight map and the second weight map to obtain a target weight matrix.

[0124] Optionally, in the AI-based pine wilt disease transmission path prediction device provided in this application embodiment, the prediction unit includes: an extraction subunit, used to extract spatial features and temporal features based on the adjacency matrix and the target weight matrix to obtain target feature information; a first processing subunit, used to process the target feature information through a target policy network to obtain the probability distribution of each node in the adjacency matrix being infected; and a determination subunit, used to obtain the prediction result based on the probability distribution of each node being infected.

[0125] Optionally, in the AI-based pine wilt disease transmission path prediction device provided in this application embodiment, the extraction subunit includes: a first determining module, used to obtain first node spatial features based on the adjacency matrix and target weight matrix corresponding to the first time-series image among multiple time-series images; a calculation module, used to calculate multiple second node spatial features by using a graph convolutional network on the first node spatial features, adjacency matrices other than the adjacency matrix corresponding to the first time-series image, and target weight matrices other than the target weight matrix corresponding to the first time-series image; an extraction module, used to extract features from the multiple second node spatial features by using a long short-term memory network to obtain node temporal features; and a second determining module, used to obtain target feature information based on the multiple second node spatial features and node temporal features.

[0126] Optionally, in the AI-based pine wilt disease transmission path prediction device provided in this application embodiment, the target strategy network is trained using the following devices: a third acquisition unit, used to acquire a time-series sample image set and acquire sample feature information corresponding to the time-series sample image set; a first processing unit, used to process the sample feature information through an initial strategy network to obtain a first predicted probability distribution of each sample node being infected; a second processing unit, used to process the sample feature information through an initial value network to obtain a current value; a third processing unit, used to process the sample feature information through a target search algorithm to obtain a second predicted probability distribution of each sample node being infected; and a training unit, used to train the initial strategy network based on the first predicted probability distribution, the current value, and the second predicted probability distribution to obtain the target strategy network.

[0127] Optionally, in the AI-based pine wilt disease transmission path prediction device provided in this application embodiment, the third processing unit includes: a third acquisition subunit, used to acquire the number of times each sample node in the sample neighbor matrix corresponding to the sample feature information is visited; a second processing subunit, used to process the features corresponding to each sample node in the sample feature information through the target value network to obtain the cumulative excitation corresponding to each sample node; and a second calculation subunit, used to calculate the number of times each sample node is visited, the cumulative excitation corresponding to each sample node, and the first prediction probability distribution to obtain the second prediction probability distribution.

[0128] Optionally, in the AI-based pine wilt disease transmission path prediction device provided in this application embodiment, the device further includes: an update unit, used to update the sample time series map in the time series sample image set according to the second predicted probability distribution of each sample node being infected after processing the sample feature information through a target search algorithm; and a training unit, used to train the initial policy network according to the updated time series sample image set to obtain the target policy network.

[0129] It should be noted that the first acquisition unit 401, the determining unit 402, the second acquisition unit 403, and the prediction unit 404 mentioned above correspond to steps S201 to S204 in Embodiment 1. The four units and the corresponding steps implement the same instances and application scenarios, but are not limited to the content disclosed in Embodiment 1. It should be noted that the above modules or units can be hardware or software components stored in memory (e.g., memory 104) and processed by one or more processors (e.g., processors 102a, 102b, ..., 102n). The above modules can also be part of a device and run in the computer terminal 10 provided in Embodiment 1.

[0130] Example 3

[0131] Embodiments of this application may provide an electronic device. Figure 5 This is a structural block diagram of an electronic device according to an embodiment of this application. Figure 5 As shown, the electronic device may include: one or more ( Figure 5 (Only one is shown) processor 502, memory 504, memory controller, and peripheral interface, wherein the peripheral interface is connected to the radio frequency module, audio module and display.

[0132] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the methods and apparatus in the embodiments of this application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby implementing the above-described methods. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0133] The processor can access information and applications stored in memory via a transmission device to perform the following steps: acquire multiple time-series images within a target area centered on the target infected tree; for each time-series image, construct nodes corresponding to each geographic feature in that image to obtain multiple nodes, where the target infected tree is a pine tree infected with pine wilt disease; based on the multiple nodes and the relationships between nodes, obtain the adjacency matrix corresponding to each time-series image; acquire the weight parameters corresponding to each grid cell in each time-series image to obtain the target weight matrix corresponding to each time-series image; and predict the propagation path using a target policy network based on the adjacency matrix and the target weight matrix corresponding to all time-series images to obtain the prediction result.

[0134] The processor can access the information and application programs stored in the memory via the transmission device to perform the following steps: For each time series image, constructing nodes corresponding to each geographic feature in the time series image to obtain multiple nodes includes: classifying the geographic features in each time series image to obtain classification results; labeling the geographic features in each time series image with preset values ​​based on the classification results to obtain labeled time series images; and constructing nodes corresponding to each geographic feature based on the labeled time series images to obtain multiple nodes.

[0135] The processor can access the information and application programs stored in the memory via the transmission device to execute the following steps: obtaining the weight parameters corresponding to each grid cell in each time series image to obtain the target weight matrix corresponding to each time series image, including: obtaining the geographic environment weight parameters corresponding to each grid cell in each time series image to obtain the first weight map corresponding to the natural influence factors; obtaining the human factor weight parameters corresponding to each grid cell in each time series image to obtain the second weight map corresponding to the human influence factors; and performing a weighted calculation on the first weight map and the second weight map to obtain the target weight matrix.

[0136] The processor can access the information and application programs stored in the memory via the transmission device to perform the following steps: predict the propagation path based on the adjacency matrix and the target weight matrix, and obtain the prediction results including: extracting spatial features and temporal features based on the adjacency matrix and the target weight matrix to obtain target feature information; processing the target feature information through the target policy network to obtain the probability distribution of each node in the adjacency matrix being infected; and obtaining the prediction results based on the probability distribution of each node being infected.

[0137] The processor can access information and applications stored in memory via a transmission device to execute the following steps: spatial and temporal feature extraction based on the adjacency matrix and target weight matrix to obtain target feature information, including: obtaining the spatial features of a first node based on the adjacency matrix and target weight matrix corresponding to the first temporal image among multiple temporal images; calculating multiple second node spatial features using a graph convolutional network on the first node spatial features, adjacency matrices other than those corresponding to the first temporal image, and target weight matrices other than those corresponding to the first temporal image; extracting node temporal features from the multiple second node spatial features using a long short-term memory network; and obtaining target feature information based on the multiple second node spatial features and node temporal features.

[0138] The processor can access the information and application programs stored in the memory via the transmission device to execute the following steps: The target policy network is trained using the following steps: acquiring a set of time-series sample images and acquiring the sample feature information corresponding to the set of time-series sample images; processing the sample feature information through the initial policy network to obtain a first predicted probability distribution of infection for each sample node; processing the sample feature information through the initial value network to obtain the current value; processing the sample feature information through the target search algorithm to obtain a second predicted probability distribution of infection for each sample node; and training the initial policy network based on the first predicted probability distribution, the current value, and the second predicted probability distribution to obtain the target policy network.

[0139] The processor can access the information and application programs stored in the memory via the transmission device to execute the following steps: processing the sample feature information using a target search algorithm to obtain a second predicted probability distribution of each sample node's infection, including: obtaining the number of times each sample node is visited in the sample neighbor matrix corresponding to the sample feature information; processing the features corresponding to each sample node in the sample feature information through a target value network to obtain the cumulative incentive corresponding to each sample node; and calculating the number of times each sample node is visited, the cumulative incentive corresponding to each sample node, and the first predicted probability distribution to obtain the second predicted probability distribution.

[0140] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: After processing the sample feature information through the target search algorithm to obtain the second predicted probability distribution of each sample node being infected, the method further includes: updating the sample time series map in the time series sample image set according to the second predicted probability distribution of each sample node being infected, to obtain the updated time series sample image set; training the initial policy network according to the updated time series sample image set to obtain the target policy network.

[0141] Those skilled in the art will understand that Figure 5 The structure shown is for illustrative purposes only. Electronic devices can also be smartphones (such as Android phones, iOS phones, etc.), tablets, PDAs, mobile internet devices (MIDs), PADs, and other terminal devices. Figure 5 This does not limit the structure of the aforementioned electronic device. For example, electronic devices may also include components that are more... Figure 5 The more or fewer components shown (such as network interfaces, display devices, etc.), or having the same Figure 5 The different configurations shown.

[0142] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0143] Example 4

[0144] Embodiments of this application also provide a computer-readable storage medium. Optionally, in this embodiment, the storage medium can be used to store the program code executed by the artificial intelligence-based method for predicting the transmission path of pine wilt disease provided in Embodiment 1.

[0145] Optionally, in this embodiment, the storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals.

[0146] This application also provides a computer program product, which, when executed on a data processing device, is adapted to perform steps of an artificial intelligence-based method for predicting the transmission path of pine wilt disease.

[0147] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0148] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0149] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0150] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0151] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0152] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0153] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for predicting the transmission path of pine wilt disease based on artificial intelligence, characterized in that, The method comprises the following steps: obtaining a plurality of time sequence images in a target range centered on a target tree, and for each time sequence image, constructing a node corresponding to each geographical element in the time sequence image to obtain a plurality of nodes, wherein the target tree is a pine tree infected with pine wood nematode disease; obtaining an adjacency matrix corresponding to each time sequence image according to the plurality of nodes and the association relationship between the nodes; obtaining a target weight matrix corresponding to each time sequence image by obtaining a weight parameter corresponding to each grid in each time sequence image; obtaining a prediction result by propagating path prediction based on the adjacency matrix corresponding to all time sequence images and the target weight matrix corresponding to all time sequence images through a target strategy network.

2. The method of claim 1, wherein, For each time sequence image, constructing a node corresponding to each geographical element in the time sequence image to obtain a plurality of nodes comprises: classifying the geographical elements in each time sequence image to obtain a classification result; labeling each geographical element in each time sequence image with a preset value according to the classification result to obtain a labeled time sequence image; constructing a node corresponding to each geographical element according to the labeled time sequence image to obtain the plurality of nodes.

3. The method of claim 1, wherein, Obtaining a weight parameter corresponding to each grid in each time sequence image to obtain a target weight matrix corresponding to each time sequence image comprises: obtaining a geographical environment weight parameter corresponding to each grid in each time sequence image to obtain a first weight map corresponding to a natural influence factor; obtaining a human factor weight parameter corresponding to each grid in each time sequence image to obtain a second weight map corresponding to a human influence factor; performing weighted calculation on the first weight map and the second weight map to obtain the target weight matrix.

4. The method of claim 1, wherein, Obtaining a prediction result by propagating path prediction based on the adjacency matrix corresponding to all time sequence images and the target weight matrix corresponding to all time sequence images through a target strategy network comprises: performing spatial feature extraction and temporal feature extraction according to the adjacency matrix and the target weight matrix to obtain target feature information; processing the target feature information through the target strategy network to obtain a probability distribution of each node being infected in the adjacency matrix; obtaining the prediction result according to the probability distribution of each node being infected.

5. The method of claim 4, wherein, Performing spatial feature extraction and temporal feature extraction according to the adjacency matrix and the target weight matrix to obtain target feature information comprises: obtaining a first node spatial feature according to the adjacency matrix and the target weight matrix corresponding to the first time sequence image in the plurality of time sequence images; calculating a plurality of second node spatial features through a graph convolution network based on the first node spatial feature, the adjacency matrix excluding the adjacency matrix corresponding to the first time sequence image, and the target weight matrix excluding the target weight matrix corresponding to the first time sequence image; extracting node temporal features through a long short-term memory network based on the plurality of second node spatial features to obtain node temporal features; obtaining target feature information according to the plurality of second node spatial features and the node temporal features.

6. The method of claim 4, wherein, The target strategy network is trained by the following steps: obtaining a time sequence sample image set and obtaining sample feature information corresponding to the time sequence sample image set; The sample feature information is processed by an initial strategy network to obtain a first predicted probability distribution of infection of each sample node; The sample feature information is processed by an initial value network to obtain a current value; The sample feature information is processed by a target search algorithm to obtain a second predicted probability distribution of infection of each sample node; The initial strategy network is trained according to the first predicted probability distribution, the current value and the second predicted probability distribution to obtain the target strategy network.

7. The method of claim 6, wherein, The sample feature information is processed by a target search algorithm to obtain a second predicted probability distribution of infection of each sample node, including: The number of times each sample node is visited in a sample adjacent matrix corresponding to the sample feature information is obtained; The features corresponding to each sample node in the sample feature information are processed by a target value network to obtain the cumulative incentive corresponding to each sample node; The number of times each sample node is visited, the cumulative incentive corresponding to each sample node and the first predicted probability distribution are calculated to obtain the second predicted probability distribution.

8. The method of claim 6, wherein, After the sample feature information is processed by a target search algorithm to obtain a second predicted probability distribution of infection of each sample node, the method further includes: The sample time series graph in the set of time series sample images is updated according to the second predicted probability distribution of infection of each sample node to obtain an updated set of time series sample images; The initial strategy network is trained according to the updated set of time series sample images to obtain the target strategy network. 9.A device for predicting a path of spread of pine wood nematode disease based on artificial intelligence, characterized by, including: The first acquisition unit is configured to acquire a plurality of time series images within a target range centered on a target pine, and for each time series image, construct a node corresponding to each geographic element in the time series image to obtain a plurality of nodes, wherein the target pine is a pine tree infected with pine wood nematode disease; The determination unit is configured to obtain an adjacency matrix corresponding to each time series image according to the plurality of nodes and the association relationship between nodes; The second acquisition unit is configured to obtain a weight parameter corresponding to each grid in each time series image to obtain a target weight matrix corresponding to each time series image; The prediction unit is configured to perform propagation path prediction based on the adjacency matrices corresponding to all time series images and the target weight matrices corresponding to all time series images by a target strategy network to obtain a prediction result.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored executable program, wherein the executable program controls the device where the computer-readable storage medium is located to execute the artificial intelligence-based pine wood nematode disease propagation path prediction method of any one of claims 1 to 8 when the executable program is running.

11. An electronic device, comprising: including: A memory storing an executable program; A processor configured to run the program, wherein the program performs the artificial intelligence-based pine wood nematode disease propagation path prediction method of any one of claims 1 to 8 when the program is running.

Images