Crop disease and pest short-time forecasting method and device and electronic equipment

Abstract

The invention provides a short-time forecasting method and device for crop diseases and insect pests and electronic equipment, and the method comprises the steps: collecting multi-source heterogeneous data covering a farmland area; preprocessing the collected multi-source heterogeneous data to generate gridding multi-modal fusion data with consistent time and space; performing real-time analysis based on the gridding multi-modal fusion data, identifying and positioning the disease and pest conditions of the current monitoring day through a first preset algorithm model, and generating a gridding disease and pest distribution diagram; and on the basis of the gridded multi-modal fusion data and the pest and disease damage distribution map of multiple historical monitoring days, predicting the pest and disease damage development trend in a future short period through a second preset algorithm model, and generating a future day-by-day gridded pest and disease damage short-time forecast map. According to the invention, unified processing can be carried out on multi-source data of different modals, time-space correlation information among the data is captured, and the problem that accurate short-time forecasting of plant diseases and insect pests is difficult to carry out in the prior art is solved.

Description

Technical Field

[0001] This invention relates to the field of crop disease and pest monitoring technology, and in particular to a method, device and electronic equipment for short-term forecasting of crop diseases and pests. Background Technology

[0002] Crop diseases and pests are a significant factor affecting agricultural production and food security. Traditional pest and disease monitoring mainly relies on manual field surveys, which suffers from low efficiency, limited scope, and strong subjectivity. In recent years, with the development of unmanned aerial vehicle (UAV) remote sensing technology, crop disease and pest monitoring methods based on UAV platforms have gradually been applied. These methods utilize multi-source sensors to acquire morphological, physiological, and temperature information of the crop canopy, thereby enabling the identification and assessment of diseases and pests.

[0003] Existing methods and systems largely rely on single-modal UAV remote sensing data such as images and spectra, lacking the collaborative fusion of multimodal information and making it difficult to comprehensively capture the early characteristics of pests and diseases. Furthermore, the UAV remote sensing data used for modeling and analysis is mostly "point" data, failing to consider the spatial heterogeneity of crop canopy structure and pest and disease transmission characteristics. In addition, existing methods and systems mainly focus on post-event monitoring of crop pests and diseases, lacking the ability to predict short-term development trends and failing to achieve "early warning" and proactive control.

[0004] Chinese invention patent CN117274836A discloses a deep learning-based UAV remote sensing image crop pest and disease monitoring system. It primarily utilizes target detection algorithms to identify pests and diseases from single remote sensing image data collected by the UAV, which cannot meet the needs of large-scale pest and disease monitoring in farmland and lacks short-term pest and disease forecasting. Chinese invention patent CN118883462A discloses a UAV remote sensing technology-based crop pest and disease monitoring method. While it solves the problem of large-scale pest and disease monitoring in farmland, it uses point-source spectral data, resulting in a single data source and ignoring the spatial correlation of crop pest and disease occurrences, leading to reduced accuracy. It also fails to solve the problem of daily-scale short-term pest and disease forecasting.

[0005] There is currently no effective solution to the problem that existing technologies are insufficient for accurate short-term forecasting of pests and diseases. Summary of the Invention

[0006] This invention provides a method, device, and electronic equipment for short-term forecasting of crop diseases and pests, which addresses the shortcomings of existing related technologies in accurately forecasting short-term diseases and pests.

[0007] In a first aspect, the present invention provides a method for short-term forecasting of crop diseases and pests, comprising: Collect multi-source heterogeneous data covering farmland areas; The collected multi-source heterogeneous data is preprocessed to generate spatiotemporally consistent gridded multimodal fusion data; Based on the gridded multimodal fusion data, real-time analysis is performed, and the pest and disease status of the current monitoring day is identified and located through the first preset algorithm model, generating a gridded pest and disease distribution map; Based on the gridded multimodal fusion data from multiple historical monitoring days and the pest and disease distribution map, the development trend of pests and diseases in the near future is predicted by the second preset algorithm model, and a gridded short-term forecast map of pests and diseases is generated for each day in the future.

[0008] According to the present invention, a short-term forecasting method for crop diseases and pests is provided, which collects multi-source heterogeneous data covering farmland areas, including: Visible light images are obtained by acquiring color and morphological information of crop canopies using a visible light camera. Hyperspectral images are obtained by acquiring spectral reflectance information of crop canopies using a hyperspectral camera. Thermal infrared images are obtained by collecting temperature information of the crop canopy using a thermal infrared camera.

[0009] According to a short-term crop disease and pest forecasting method provided by the present invention, the collected multi-source heterogeneous data is preprocessed to generate spatiotemporally consistent gridded multimodal fusion data, including: Geometric and radiometric corrections are performed on the visible light image, the hyperspectral image, and the thermal infrared image. The corrected multi-source heterogeneous data is spatiotemporally registered to ensure that the multi-source heterogeneous data of different modalities maintain consistency in space and time. The multi-source heterogeneous data is processed into a grid to obtain the gridded multimodal fusion data.

[0010] According to a short-term crop disease and pest forecasting method provided by the present invention, the multi-source heterogeneous data is processed into grids to obtain the gridded multimodal fusion data, including: The target farmland is divided into grids based on the crop planting direction, resulting in multiple grid units; The spatiotemporally registered multi-source heterogeneous data is mapped to the corresponding grid cells according to the grid division results, and gridded multimodal fusion data containing visible light images, hyperspectral images and thermal infrared images is generated for each grid cell.

[0011] According to the method for short-term forecasting of crop diseases and pests provided by the present invention, the grid cell has a square shape, and the side length of the square is parallel to the direction of crop planting rows; The grid size of the grid unit is determined based on the characteristics of the crop canopy and the spread of pests and diseases.

[0012] According to the crop disease and pest short-term forecasting method provided by the present invention, the first preset algorithm model is a deep learning model based on multimodal feature fusion, which realizes adaptive fusion of different modal features through attention mechanism; Based on the gridded multimodal fusion data, real-time analysis is performed. A first preset algorithm model is used to identify and locate the pest and disease status for the current monitoring day, generating a gridded pest and disease distribution map, including: The data features of the gridded multimodal fusion data are mapped to a unified dimension through the convolutional layer of the first preset algorithm model; Determine the weighting factors for adaptive fusion of data features from different modalities; Based on the weighting factors, feature fusion is performed on the data features of the gridded multimodal fusion data to obtain a fused feature map; The fused feature map is processed through the convolutional layer and fully connected layer of the first preset algorithm model to obtain a pest distribution map containing the pest category probability distribution and severity level of each grid cell.

[0013] According to the crop disease and pest short-term forecasting method provided by the present invention, the second preset algorithm model is a deep learning model based on multimodal feature spatiotemporal fusion, which realizes the fusion of different modal features through spatiotemporal attention mechanism; Based on the historical gridded multimodal fusion data and the pest and disease distribution map, the development trend of pests and diseases in the near future is predicted through a second preset algorithm model, and a daily gridded short-term forecast map of pests and diseases is generated, including: The gridded multimodal fusion data and the pest and disease distribution map are sorted according to time series. The sorted gridded multimodal fusion data and the pest and disease distribution map are input into the encoder of the second preset algorithm model to extract initial features; Multimodal spatiotemporal fusion features are extracted based on spatiotemporal attention to obtain a spatiotemporal fusion feature map; The spatiotemporal fusion feature map is input into the decoder of the second preset algorithm model for decoding to obtain a gridded short-term forecast map of pests and diseases for each day in the future. The short-term pest and disease forecast map includes a gridded distribution map of pests and diseases in the farmland area for each day of the next 1-5 days. The short-term pest and disease forecast map is used to characterize the pest and disease category and severity level of each grid cell, and to help provide early warning information and prevention and control decision-making suggestions for the pest and disease status of the farmland area.

[0014] According to the present invention, a short-term forecasting method for crop diseases and pests is provided, which extracts multimodal spatiotemporal fusion features based on spatiotemporal attention to obtain a spatiotemporal fusion feature map, including: The gridded multimodal fusion data and the pest distribution map are subjected to spatiotemporal fusion processing of different modal features to obtain a spatiotemporal attention feature map; The spatiotemporal attention feature maps of different modalities are cascaded and fused to obtain the spatiotemporal fused feature map.

[0015] Secondly, the present invention also provides a short-term forecasting device for crop diseases and pests, comprising: The multimodal information acquisition module is used to collect multi-source heterogeneous data covering farmland areas through various sensors mounted on the UAV platform; A multimodal information preprocessing module, connected to the multimodal information acquisition module, is used to preprocess the acquired multi-source heterogeneous data to generate spatiotemporally consistent gridded multimodal fusion data; The pest and disease monitoring module is connected to the multimodal information preprocessing module and is used to perform real-time analysis based on the gridded multimodal fusion data, identify and locate the pest and disease status at the current moment, and generate a gridded pest and disease distribution map. The pest and disease drone short-term forecast module is connected to the pest and disease drone monitoring module. It is used to predict the development trend of pests and diseases in the near future based on the historical gridded multimodal fusion data and the pest and disease distribution map, and generate a gridded pest and disease short-term forecast map for each day in the future. The user interaction module is connected to the multimodal information acquisition module, the pest and disease drone monitoring module, and the pest and disease drone short-term forecast module. It is used to display a gridded short-term forecast map of pests and diseases in farmland crops and to provide users with early warning information and prevention and control decision suggestions.

[0016] Thirdly, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the short-term forecasting method for crop diseases and pests as described in the first aspect above.

[0017] Fourthly, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the short-term forecasting method for crop diseases and pests as described in the first aspect above.

[0018] Fifthly, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the short-term forecasting method for crop diseases and pests as described in the first aspect above.

[0019] Compared with the prior art, the present invention has the following beneficial effects: 1. The short-term crop disease and pest forecasting method provided by this invention, by equipping a multi-source remote sensing imaging sensor, can acquire multimodal information such as visible light images, hyperspectral images, and thermal infrared images of farmland crops, enabling a more comprehensive characterization of crop disease and pest occurrence. Based on this, the remote sensing images are gridded according to crop canopy characteristics and disease and pest propagation characteristics, and spatiotemporal registration is performed to generate three-dimensional data that integrates multimodal data including visible light images, hyperspectral images, and thermal infrared images. This effectively reduces the spatiotemporal heterogeneity of UAV remote sensing multimodal data on crop diseases and pests, and enhances data correlation.

[0020] 2. This invention, by constructing a deep learning model based on multimodal feature fusion, leverages the synergistic effect of multimodal information, enhances the fusion and expression of multimodal features of crop diseases and pests, improves the accuracy of crop disease and pest monitoring, and generates a grid distribution map of disease and pest occurrence to intuitively reflect the occurrence categories and severity levels of crop diseases and pests in different plots of farmland, thus optimizing the monitoring layout.

[0021] 3. Based on the monitoring of pests and diseases by drones, this invention constructs a deep learning model based on the spatiotemporal fusion of multimodal features. By fusing historical multimodal data with pest and disease distribution data, it forecasts the occurrence of crop pests and diseases and generates a daily grid distribution map of short-term forecasts of crop pests and diseases in farmland. This provides users with an intuitive understanding of the types and severity levels of crop pests and diseases in different plots of farmland within the next 1 to 5 days, thus improving timeliness. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0023] Figure 1 This is a flowchart of the short-term forecasting method for crop diseases and pests provided by the present invention; Figure 2 This is a structural block diagram of the crop disease and pest short-term forecasting device provided by the present invention; Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0025] This invention provides a method for short-term forecasting of crop diseases and pests. Figure 1 This is a flowchart of the short-term forecasting method for crop diseases and pests provided by the present invention, such as... Figure 1 As shown, the method includes the following steps: Step S101: Collect multi-source heterogeneous data covering farmland areas; Step S102: Preprocess the collected multi-source heterogeneous data to generate spatiotemporally consistent gridded multimodal fusion data; Step S103: Real-time analysis is performed based on gridded multimodal fusion data. The pest and disease status of the current monitoring day is identified and located through the first preset algorithm model, and a gridded pest and disease distribution map is generated. Step S104: Based on the gridded multimodal fusion data and pest distribution map of multiple historical monitoring days, predict the pest development trend in the short term through the second preset algorithm model, and generate gridded short-term pest forecast maps for each day in the future.

[0026] In this method, firstly, multi-source heterogeneous data covering farmland areas are collected. This multi-source heterogeneous data contains multimodal information about crops in the farmland area, enabling a more comprehensive characterization of crop pest and disease occurrence. Then, the collected multi-source heterogeneous data is preprocessed to generate spatiotemporally consistent gridded multimodal fusion data, effectively reducing the spatiotemporal heterogeneity of crop pest and disease UAV remote sensing multimodal data and enhancing data correlation. Next, real-time analysis is performed based on the gridded multimodal fusion data to identify and locate the current pest and disease status, generating a gridded pest and disease distribution map. Finally, based on historical gridded multimodal fusion data and the pest and disease distribution map, the short-term development trend of pests and diseases is predicted, and daily gridded short-term pest and disease forecast maps are generated. This helps users intuitively understand the future occurrence categories and severity levels of crop pests and diseases, improving the timeliness of crop pest and disease monitoring. Through the above process, multi-source data of different modalities can be processed in a unified manner, capturing the spatiotemporal correlation information between data, providing a richer monitoring perspective, and solving the problem that existing related technologies are difficult to use for accurate short-term forecasting of pests and diseases.

[0027] In some embodiments, step S101 involves collecting multi-source heterogeneous data covering the farmland area, including: acquiring color and morphological information of the crop canopy using a visible light camera to obtain a visible light image; acquiring spectral reflectance information of the crop canopy using a hyperspectral camera to obtain a hyperspectral image; and acquiring temperature information of the crop canopy using a thermal infrared camera to obtain a thermal infrared image.

[0028] In this embodiment, the early characteristics of crop diseases and pests are comprehensively captured through the collaborative work of multiple sensors. Preferably, the acquisition of multi-source heterogeneous data on crop canopies in farmland areas can be based on the different growth cycles of different crops, collecting UAV remote sensing data from different seasons, different weather conditions, and different time periods, so as to carry out monitoring and short-term forecasting of crop diseases and pests throughout the entire growth period.

[0029] In some embodiments, step S102 involves preprocessing the acquired multi-source heterogeneous data to generate spatiotemporally consistent gridded multimodal fusion data. This includes: performing geometric and radiometric corrections on visible light images, hyperspectral images, and thermal infrared images to eliminate distortions and radiometric errors during the imaging process; performing spatiotemporal registration on the corrected multi-source heterogeneous data to ensure consistency between different modalities of multi-source heterogeneous data in space and time; and performing gridded processing on the multi-source heterogeneous data to obtain gridded multimodal fusion data.

[0030] Specifically, the multi-source heterogeneous data is processed into gridded data to obtain gridded multimodal fusion data. This includes: dividing the target farmland into gridded units based on the crop planting direction to obtain multiple grid units; mapping the spatiotemporally registered multi-source heterogeneous data to the corresponding grid units based on the grid division results; and generating gridded multimodal fusion data containing visible light images, hyperspectral images, and thermal infrared images for each grid unit.

[0031] More specifically, the grid unit is square in shape, and the side length of the square is parallel to the direction of the crop planting rows to enhance the rationality of the spatial structure; the grid size is determined according to the characteristics of the crop canopy and the characteristics of pest and disease transmission.

[0032] For example, the formula for determining the optimal mesh size of a mesh cell is as follows:

[0033] in, Indicates the optimal mesh size. This represents the grid size G that minimizes the objective function. This represents the variance of crop canopy characteristics under grid size G. This indicates the risk of missed detection of pests and diseases under grid size G. and Let be the weighting coefficients, and satisfy . α + β =1.

[0034] In this embodiment, the optimal mesh size The value range is from 0.5m×0.5m to 3m×3m, and the aspect ratio of the grid is kept at 1:1. Crop canopy characteristics include, but are not limited to, canopy height, leaf area index, canopy density, etc.

[0035] In some embodiments, step S103, the first preset algorithm model is a deep learning model based on multimodal feature fusion, which achieves adaptive fusion of different modal features through an attention mechanism; real-time analysis is performed based on gridded multimodal fusion data, and the pest and disease status of the current monitoring day is identified and located through the first preset algorithm model to generate a gridded pest and disease distribution map, including: mapping the data features of the gridded multimodal fusion data to a unified dimension through the convolutional layer of the first preset algorithm model; determining the weight factors for the adaptive fusion of data features of different modalities; performing feature fusion on the data features of the gridded multimodal fusion data based on the weight factors to obtain a fused feature map; and processing the fused feature map through the convolutional layer and fully connected layer of the first preset algorithm model to obtain a pest and disease distribution map containing the probability distribution of pest and disease categories and severity levels of each grid unit.

[0036] For example, we first map the features of each modality to a unified dimension using a 1×1 convolution, as shown in the following formula:

[0037]

[0038]

[0039] in, This represents the original feature map of visible light. This represents the original feature map of the hyperspectral image. This represents the original feature map of thermal infrared radiation. , and Let represent the weight matrices for the visible light, hyperspectral, and thermal infrared modes, respectively, and satisfy . , , ,in , This represents a unified feature dimension.

[0040] The weighting factors used for adaptive fusion of features from different modalities are defined as follows:

[0041] in, The weights represent the different modal features. and Represents the weight matrix. and Indicates the bias parameter. and Feature maps representing different modalities and Indicates different modal indices, , and These represent the modal indices for visible light, hyperspectral, and thermal infrared, respectively.

[0042] The formula for multimodal feature fusion is as follows:

[0043] in, This represents the fused feature map.

[0044] The fused feature maps are further processed through convolutional and fully connected layers to finally output the probability distribution and severity level of pest and disease categories for each grid cell, as shown in the following formula:

[0045]

[0046] in, Represents the predicted probability of the category. Represents the classification weight matrix. Indicates the number of pest and disease categories. This represents the feature dimension after flattening. Indicates the flattening operation. Represents the fused feature map. Indicates the category bias term. Indicates the severity level value. Represents the regression weight matrix. This represents the regression bias term. This indicates the highest severity level of pests and diseases. This refers to the Sigmod function.

[0047] In some embodiments, step S104, the second preset algorithm model is a deep learning model based on spatiotemporal fusion of multimodal features, which achieves the fusion of different modal features through a spatiotemporal attention mechanism; based on historical gridded multimodal fusion data and pest distribution maps, the second preset algorithm model predicts the development trend of pests and diseases in the short term and generates a daily gridded short-term forecast map of pests and diseases, including: sorting the gridded multimodal fusion data and pest distribution maps according to time series; inputting the sorted gridded multimodal fusion data and pest distribution maps into the encoder of the second preset algorithm model to extract initial features; extracting multimodal spatiotemporal fusion features based on spatiotemporal attention to obtain a spatiotemporal fusion feature map; inputting the spatiotemporal fusion feature map into the decoder of the second preset algorithm model for decoding to obtain a daily gridded short-term forecast map of pests and diseases. The short-term pest and disease forecast map contains a gridded distribution map of pests and diseases in farmland areas for each day of the next 1-5 days. The short-term pest and disease forecast map is used to characterize the pest and disease categories and severity levels of each grid unit, and to help provide early warning information and prevention and control decision-making suggestions for the pest and disease status of farmland areas.

[0048] For example, the formula for extracting the initial features is as follows:

[0049] in, Indicates the initial features extracted. Indicates encoder, This represents historical multimodal fusion data. A map showing the historical distribution of pests and diseases. Indicates the time step of monitoring. , and These represent the mode indices for visible light, hyperspectral, and thermal infrared, respectively. Based on these, a convolution operation is performed on the initial features, with the specific formula as follows:

[0050] in, These represent query, keyword, and numerical feature maps, respectively. This represents the convolution operation that extracts query, key, and numerical feature maps. This represents the initial features extracted.

[0051] Then, multimodal spatiotemporal fusion features are extracted to obtain a spatiotemporal fusion feature map. This spatiotemporal fusion feature map is then input into the decoder of the second preset algorithm model for decoding to obtain a short-time forecast map. The specific formula is as follows:

[0052] in, This represents a gridded short-term forecast map of pests and diseases in future farmland areas. Indicates the time step of the forecast. Indicates decoder, This represents a spatiotemporal fusion feature map.

[0053] Specifically, based on spatiotemporal attention, multimodal spatiotemporal fusion features are extracted to obtain spatiotemporal fusion feature maps, including: spatiotemporal fusion processing of different modal features on gridded multimodal fusion data and pest distribution maps to obtain spatiotemporal attention feature maps; and cascading fusion of spatiotemporal attention feature maps of different modalities to obtain spatiotemporal fusion feature maps.

[0054] For example, the spatiotemporal attention feature map used for spatiotemporal fusion of different modal features is defined as:

[0055] in, Represents the spatiotemporal attention feature map. , and These represent query, keyword, and numerical feature maps, respectively. and Indexes representing attention in different modalities and dimensions. , and These represent the mode indices for visible light, hyperspectral light, and thermal infrared light, respectively. , and These represent attention indices for time, space, and channel, respectively.

[0056] The formula for spatiotemporal fusion of multimodal features is as follows:

[0057] in, Represents the spatiotemporal fusion feature map. Indicates cascading fusion. This represents a visible light temporal attention feature map. This represents the spatial attention feature map of visible light. This represents the attention feature map of the visible light channel; Represents the hyperspectral temporal attention feature map. Represents the hyperspectral spatial attention feature map. This represents the attention feature map of the hyperspectral channels; This represents the thermal infrared time-attention feature map. This represents the thermal infrared spatial attention feature map. This represents the attention feature map of the thermal infrared channel; , and These represent the mode indices for visible light, hyperspectral light, and thermal infrared light, respectively. , and These represent attention indices for time, space, and channel, respectively. This represents the attention feature map.

[0058] This invention provides a short-term forecasting device for crop diseases and pests. The following is a description of the short-term forecasting device for crop diseases and pests provided by this invention. The short-term forecasting device for crop diseases and pests described below can be referred to in correspondence with the short-term forecasting method for crop diseases and pests described above. Figure 2 This is a structural block diagram of the crop disease and pest short-term forecasting device provided by the present invention, as shown below. Figure 2 As shown, the device includes: The multimodal information acquisition module 201 is used to collect multi-source heterogeneous data covering farmland areas through various sensors mounted on an unmanned aerial vehicle platform; The multimodal information preprocessing module 202 is connected to the multimodal information acquisition module and is used to preprocess the acquired multi-source heterogeneous data to generate spatiotemporally consistent gridded multimodal fusion data. The pest and disease monitoring module 203 is connected to the multimodal information preprocessing module, which is used to perform real-time analysis based on gridded multimodal fusion data, identify and locate the current pest and disease status, and generate a gridded pest and disease distribution map. The pest and disease drone short-term forecast module 204 is connected to the pest and disease drone monitoring module. It is used to predict the development trend of pests and diseases in the short term based on historical gridded multimodal fusion data and pest and disease distribution map, and generate gridded pest and disease short-term forecast maps for each day in the future. The user interaction module 205 is connected to the multimodal information acquisition module, the pest and disease drone monitoring module, and the pest and disease drone short-term forecast module. It is used to display the gridded short-term forecast map of pests and diseases in farmland crops and to provide users with early warning information and prevention and control decision suggestions.

[0059] In operation, this device first collects multi-source heterogeneous data covering the farmland area using the multi-modal information acquisition module 201. This data contains multi-modal information about crops in the farmland area, providing a more comprehensive characterization of crop pest and disease occurrence. Next, the multi-modal information preprocessing module 202 preprocesses the collected multi-source heterogeneous data, generating spatiotemporally consistent gridded multi-modal fusion data. This effectively reduces the spatiotemporal heterogeneity of the crop pest and disease UAV remote sensing multi-modal data and enhances data correlation. The pest and disease UAV monitoring module 203 then performs real-time analysis based on the gridded multi-modal fusion data, identifying and locating the current pest and disease status, and generating a gridded pest and disease distribution map. Finally, the pest and disease UAV short-term forecast module 204 predicts the short-term pest and disease development trend based on historical gridded multi-modal fusion data and the pest and disease distribution map, generating daily gridded short-term pest and disease forecast maps. The user interaction module 205 helps users intuitively understand the types and severity levels of future crop diseases and pests, improving the timeliness of crop disease and pest monitoring. Through the above process, multi-source data of different modalities can be processed in a unified manner, capturing the spatiotemporal correlation information between data, providing a richer monitoring perspective, and solving the problem that existing related technologies are difficult to use for accurate short-term forecasting of diseases and pests.

[0060] Figure 3 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 3 As shown, the electronic device may include: a processor 301, a communication interface 302, a memory 303, and a communication bus 304. The processor 301, communication interface 302, and memory 303 communicate with each other via the communication bus 304. The processor 301 can call logical instructions from the memory 303 to execute a short-term forecasting method for crop diseases and pests. This method includes: Collect multi-source heterogeneous data covering farmland areas; The collected multi-source heterogeneous data is preprocessed to generate spatiotemporally consistent gridded multimodal fusion data; Real-time analysis based on gridded multimodal fusion data is used to identify and locate the current status of pests and diseases, and generate a gridded distribution map of pests and diseases. Based on historical gridded multimodal fusion data and pest distribution maps, the system predicts the development trend of pests and diseases in the near future and generates daily gridded short-term pest and disease forecast maps.

[0061] Furthermore, the logical instructions in the aforementioned memory 303 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a 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 of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0062] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the crop disease and pest short-term forecasting method provided by the above methods, the method including: Collect multi-source heterogeneous data covering farmland areas; The collected multi-source heterogeneous data is preprocessed to generate spatiotemporally consistent gridded multimodal fusion data; Real-time analysis based on gridded multimodal fusion data is used to identify and locate the current status of pests and diseases, and generate a gridded distribution map of pests and diseases. Based on historical gridded multimodal fusion data and pest distribution maps, the system predicts the development trend of pests and diseases in the near future and generates daily gridded short-term pest and disease forecast maps.

[0063] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the crop disease and pest short-term forecasting method provided by the methods described above, the method comprising: Collect multi-source heterogeneous data covering farmland areas; The collected multi-source heterogeneous data is preprocessed to generate spatiotemporally consistent gridded multimodal fusion data; Real-time analysis based on gridded multimodal fusion data is used to identify and locate the current status of pests and diseases, and generate a gridded distribution map of pests and diseases. Based on historical gridded multimodal fusion data and pest distribution maps, the system predicts the development trend of pests and diseases in the near future and generates daily gridded short-term pest and disease forecast maps.

[0064] The device embodiments described above are merely illustrative. 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 modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0065] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0066] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for short-term forecasting of crop diseases and pests, characterized in that, include: Collect multi-source heterogeneous data covering farmland areas; The collected multi-source heterogeneous data is preprocessed to generate spatiotemporally consistent gridded multimodal fusion data; Based on the gridded multimodal fusion data, real-time analysis is performed, and the pest and disease status of the current monitoring day is identified and located through the first preset algorithm model, generating a gridded pest and disease distribution map; Based on the gridded multimodal fusion data from multiple historical monitoring days and the pest and disease distribution map, the development trend of pests and diseases in the near future is predicted by the second preset algorithm model, and a gridded short-term forecast map of pests and diseases is generated for each day in the future.

2. The short-term forecasting method for crop diseases and pests according to claim 1, characterized in that, Collect multi-source heterogeneous data covering farmland areas, including: Visible light images are obtained by acquiring color and morphological information of crop canopies using a visible light camera. Hyperspectral images are obtained by acquiring spectral reflectance information of crop canopies using a hyperspectral camera. Thermal infrared images are obtained by collecting temperature information of the crop canopy using a thermal infrared camera.

3. The short-term forecasting method for crop diseases and pests according to claim 2, characterized in that, The collected multi-source heterogeneous data is preprocessed to generate spatiotemporally consistent gridded multimodal fusion data, including: Geometric and radiometric corrections are performed on the visible light image, the hyperspectral image, and the thermal infrared image. The corrected multi-source heterogeneous data is spatiotemporally registered to ensure that the multi-source heterogeneous data of different modalities maintain consistency in space and time. The multi-source heterogeneous data is processed into a grid to obtain the gridded multimodal fusion data.

4. The short-term forecasting method for crop diseases and pests according to claim 3, characterized in that, The multi-source heterogeneous data is processed into grids to obtain the gridded multimodal fusion data, including: The target farmland is divided into grids based on the crop planting direction, resulting in multiple grid units; The spatiotemporally registered multi-source heterogeneous data is mapped to the corresponding grid cells according to the grid division results, and gridded multimodal fusion data containing visible light images, hyperspectral images and thermal infrared images is generated for each grid cell.

5. The short-term forecasting method for crop diseases and pests according to claim 4, characterized in that, The grid unit has a square shape, and the side length of the square is parallel to the direction of the crop planting ridges. The grid size of the grid unit is determined based on the characteristics of the crop canopy and the spread of pests and diseases.

6. The short-term forecasting method for crop diseases and pests according to claim 1, characterized in that, The first preset algorithm model is a deep learning model based on multimodal feature fusion, which achieves adaptive fusion of different modal features through an attention mechanism; Based on the gridded multimodal fusion data, real-time analysis is performed. A first preset algorithm model is used to identify and locate the pest and disease status for the current monitoring day, generating a gridded pest and disease distribution map, including: The data features of the gridded multimodal fusion data are mapped to a unified dimension through the convolutional layer of the first preset algorithm model; Determine the weighting factors for adaptive fusion of data features from different modalities; Based on the weighting factors, feature fusion is performed on the data features of the gridded multimodal fusion data to obtain a fused feature map; The fused feature map is processed through the convolutional layer and fully connected layer of the first preset algorithm model to obtain a pest distribution map containing the pest category probability distribution and severity level of each grid cell.

7. The short-term forecasting method for crop diseases and pests according to claim 4, characterized in that, The second preset algorithm model is a deep learning model based on spatiotemporal fusion of multimodal features, which achieves the fusion of different modal features through a spatiotemporal attention mechanism; Based on the historical gridded multimodal fusion data and the pest and disease distribution map, the development trend of pests and diseases in the near future is predicted through a second preset algorithm model, and a daily gridded short-term forecast map of pests and diseases is generated, including: The gridded multimodal fusion data and the pest and disease distribution map are sorted according to time series. The sorted gridded multimodal fusion data and the pest and disease distribution map are input into the encoder of the second preset algorithm model to extract initial features; Multimodal spatiotemporal fusion features are extracted based on spatiotemporal attention to obtain a spatiotemporal fusion feature map; The spatiotemporal fusion feature map is input into the decoder of the second preset algorithm model for decoding to obtain a gridded short-term forecast map of pests and diseases for each day in the future. The short-term pest and disease forecast map includes a gridded distribution map of pests and diseases in the farmland area for each day of the next 1-5 days. The short-term pest and disease forecast map is used to characterize the pest and disease category and severity level of each grid cell, and to help provide early warning information and prevention and control decision-making suggestions for the pest and disease status of the farmland area.

8. The short-term forecasting method for crop diseases and pests according to claim 7, characterized in that, Multimodal spatiotemporal fusion features are extracted based on spatiotemporal attention, resulting in a spatiotemporal fusion feature map, including: The gridded multimodal fusion data and the pest distribution map are subjected to spatiotemporal fusion processing of different modal features to obtain a spatiotemporal attention feature map; The spatiotemporal attention feature maps of different modalities are cascaded and fused to obtain the spatiotemporal fused feature map.

9. A short-term forecasting device for crop diseases and pests, characterized in that, include: The multimodal information acquisition module is used to collect multi-source heterogeneous data covering farmland areas through various sensors mounted on the UAV platform; A multimodal information preprocessing module, connected to the multimodal information acquisition module, is used to preprocess the acquired multi-source heterogeneous data to generate spatiotemporally consistent gridded multimodal fusion data; The pest and disease monitoring module is connected to the multimodal information preprocessing module and is used to perform real-time analysis based on the gridded multimodal fusion data, identify and locate the pest and disease status at the current moment, and generate a gridded pest and disease distribution map. The pest and disease drone short-term forecast module is connected to the pest and disease drone monitoring module. It is used to predict the development trend of pests and diseases in the near future based on the historical gridded multimodal fusion data and the pest and disease distribution map, and generate a gridded pest and disease short-term forecast map for each day in the future. The user interaction module is connected to the multimodal information acquisition module, the pest and disease drone monitoring module, and the pest and disease drone short-term forecast module. It is used to display a gridded short-term forecast map of pests and diseases in farmland crops and to provide users with early warning information and prevention and control decision suggestions.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the short-term forecasting method for crop diseases and pests as described in any one of claims 1 to 8.

Images