A durian suitable area identification and precision monitoring method and system
Patent Information
- Application Number
- CN202611116716.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-27
- Publication Date
- 2026-10-27
AI Technical Summary
[0005]鉴于此，本发明的目的在于提供一种基于榴莲适生区识别与精准监测方法及系统，以解决或至少部分解决现有技术中所存在的上述问题
本发明提出一种基于榴莲适生区识别与精准监测方法及系统，通过提取榴莲适生区的关键环境因子，并根据关键环境因子实现对榴莲适生区的适宜性等级划分，通过利用经过改进优化的语义分割网络在适宜性等级划分的基础上实现对榴莲实际种植区域的提取，通过对榴莲进行害虫识别，实现对榴莲健康生长的实时监测，并为用户提供针对性的榴莲种植建议。本发明针对榴莲专属场景的定向优化大幅提升了监测精度，为榴莲产业的科学规划与精细化管理提供全链条技术支撑。
Smart Images
Figure CN122990472A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of smart agriculture and remote sensing information technology, and in particular to a method and system for identifying and accurately monitoring suitable durian growing areas. Background Technology
[0002] Durian is a tropical fruit tree with significant economic value and is known as the "King of Fruits." However, the durian industry still faces two major problems: First, suitable planting areas have not been systematically evaluated, and site selection for expansion relies on experience and lacks scientific basis; second, the level of precision in planting management is insufficient, and monitoring of growth status depends on manual inspections, which is inefficient, subjective, and makes it difficult to achieve large-scale precise management.
[0003] In existing technologies, suitable growing area prediction is mainly based on climatic factor analysis. However, this method relies solely on static climate data and fails to utilize real-time ground feature information provided by remote sensing imagery, resulting in discrepancies between the prediction results and the actual plot scale. Durian health monitoring primarily relies on manual inspections, which not only requires a large workforce but is also prone to omissions and errors, leading to losses for users' durian cultivation.
[0004] Furthermore, when existing general remote sensing monitoring solutions are directly applied to durian planting scenarios, due to the unique texture of durian tree crowns and the cloud and fog obstruction in planting areas, there are generally defects such as an extraction error of more than 15% in planting areas and an accuracy rate of less than 60% in identifying early hidden pests, which cannot meet the specific needs of refined management for durian. Summary of the Invention
[0005] Therefore, the purpose of this invention is to provide a method and system for identifying and accurately monitoring suitable durian growing areas, so as to solve or at least partially solve the above-mentioned problems existing in the prior art.
[0006] To achieve the above objectives, the first aspect of the present invention provides a method for identifying and accurately monitoring suitable durian growing areas, the method comprising the following steps: S1. Use satellite remote sensing technology to obtain multi-source satellite remote sensing data of durian suitable growing areas, and preprocess the multi-source satellite remote sensing data; S2. Extract key environmental factors for durian suitable growing areas from preprocessed multi-source satellite remote sensing data; S3. Based on key environmental factors, the suitability of durian growing areas is classified, and a durian planting suitability classification map is generated. S4. Based on the durian planting suitability classification map, extract the actual durian planting area using a semantic segmentation network; S5. Based on the actual durian planting area, use drones to acquire remote sensing image data of durian, and identify pests of durian based on the remote sensing image data. S6. Based on the pest identification results, conduct growth health assessment and early warning for durians; S7. Visualize growth health assessment and early warning information and generate planting suggestions.
[0007] Furthermore, the key environmental factors include surface temperature, vegetation cover, soil moisture index, topographic factors, seasonal temperature variation coefficient, annual accumulated temperature, and spatiotemporal distribution characteristics of precipitation.
[0008] Furthermore, step S3 specifically includes the following steps: S11. Obtain the durian planting environment thresholds corresponding to different soil types and terrains. The durian planting environment thresholds include environmental humidity, soil moisture, soil pH and light conditions. S12. Establish different suitability levels based on the threshold of durian growing environment; S13. Based on multi-source satellite remote sensing data, obtain the soil type and topography of the durian suitable growing area; S14. Dimensionless processing of key environmental factors is performed, and the suitable level of durian growing areas is divided according to the soil type and topography of the durian growing area based on the environmental threshold of durian planting, generating a durian planting suitability classification map.
[0009] Furthermore, the automatic extraction of the actual durian planting area using a semantic segmentation network specifically includes the following steps: S21. Introduce a multi-scale feature feedforward network module into the semantic segmentation network. Use the multi-scale feature feedforward network module to extract multi-scale features of durian based on the durian planting suitability grading map, including the spectral features, texture features and morphological features of the durian canopy. S22. The multi-scale feature feedforward network module is provided with a convolutional position encoding submodule, which is used to fuse the extracted multi-scale features to obtain a unified feature map of durian. S23. An attention mechanism is introduced into the semantic segmentation network to enhance the feature information of the durian unified feature map, resulting in an enhanced durian feature map. S24. Based on the enhanced feature map of durian, the actual planting area of durian is extracted using a semantic segmentation network.
[0010] Furthermore, the enhancement of feature information in the uniform feature map of durian through the attention mechanism specifically includes the following steps: S31. Perform spatial average pooling on the uniform feature map of durian along the horizontal and vertical axes to obtain feature maps along the horizontal and vertical axes. S32. Use convolutional layers to reduce the dimensionality of feature maps in the horizontal and vertical directions, and generate attention weights in the horizontal and vertical directions respectively. S33. Use the attention weights in the horizontal and vertical directions to weight the durian unified feature map to obtain the durian enhanced feature map.
[0011] Furthermore, step S5 specifically includes the following steps: S41. Plan the flight path of the drone according to the actual durian planting area, and use the high-resolution sensor on the drone to obtain remote sensing image data of durian. S42. Perform processing operations on remote sensing image data, including geometric correction and radiometric correction. S43. Based on the processed remote sensing image data, extract the growth status information of durian; S44. Based on growth status information, identify the pest areas in the actual durian planting area, construct a pest identification model, and use the pest identification model to identify the pest type according to the pest area.
[0012] Furthermore, the construction of the pest identification model specifically includes the following steps: S51. Collect and label the pest feature dataset, including pest types and their corresponding pest features, and divide it into training set and validation set. S52. Construct a pest identification model, input the training set into the pest identification model for training, introduce the focus loss function as the loss function of the pest identification model, and adjust the parameters of the pest identification model according to the loss function value. S53. Input the validation set into the trained pest identification model to verify the classification performance of the pest identification model. S54. Input the remote sensing image data containing the pest area into the pest identification model to identify the pest types in the pest area.
[0013] Furthermore, the method of using a pest identification model to identify pest types based on pest areas specifically includes the following steps: S61. Perform dimensionality reduction processing on remote sensing image data containing pest areas and extract the spectral features of pest areas. S62. Extract texture features of pest areas based on remote sensing image data containing pest areas; S63. Combine spectral features and texture features to obtain comprehensive pest features, and input them into the pest identification model. Output the probability of the corresponding pest category based on the comprehensive pest features.
[0014] A second aspect of the present invention provides a system for identifying and accurately monitoring suitable durian growing areas, the system comprising: Data acquisition module: used to acquire multi-source satellite remote sensing data of durian suitable growing areas using satellite remote sensing technology, and to preprocess the multi-source satellite remote sensing data; Environmental factor extraction module: used to extract key environmental factors of durian suitable growing areas from preprocessed multi-source satellite remote sensing data; Suitability Grading Module: Used to construct a durian suitability grading model based on key environmental factors, and generate a durian planting suitability grading map based on the durian suitability grading model; Planting Area Extraction Module: Used to extract the actual planting area of durian based on the durian planting suitability grading map using a semantic segmentation network; Pest identification module: Used to acquire remote sensing image data of durians using drones based on the actual durian planting area, and to identify pests in durians based on the remote sensing image data; Assessment and early warning module: used to assess the growth health of durians and issue early warnings based on pest identification results; Visualization suggestion module: This module is used to visualize growth health assessment and early warning information and generate planting suggestions.
[0015] Compared with the prior art, the beneficial effects of the present invention are: This invention proposes a method and system for identifying and accurately monitoring suitable durian growing areas. By extracting key environmental factors of suitable durian growing areas and classifying these areas according to their suitability, the invention utilizes an improved and optimized semantic segmentation network to extract actual durian planting areas based on this suitability classification. Furthermore, by identifying durian pests, the invention enables real-time monitoring of durian health and provides users with targeted durian planting suggestions. This invention's targeted optimization for the specific durian scenario significantly improves monitoring accuracy, providing full-chain technical support for the scientific planning and refined management of the durian industry. Attached Figure Description
[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only preferred embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0017] Figure 1 This is a schematic diagram of a method for identifying and accurately monitoring suitable durian growing areas, provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of a system for identifying and accurately monitoring suitable durian growing areas, provided as an embodiment of the present invention. Detailed Implementation
[0018] The principles and features of the present invention are described below with reference to the accompanying drawings. The listed embodiments are only used to explain the present invention and are not intended to limit the scope of the present invention.
[0019] Reference Figure 1 This embodiment provides a method for identifying and accurately monitoring suitable durian growing areas, the method comprising the following steps: S1. Use satellite remote sensing technology to obtain multi-source satellite remote sensing data of durian suitable growing areas, and preprocess the multi-source satellite remote sensing data; S2. Extract key environmental factors for durian suitable growing areas from preprocessed multi-source satellite remote sensing data; S3. Based on key environmental factors, the suitability of durian growing areas is classified, and a durian planting suitability classification map is generated. S4. Based on the durian planting suitability classification map, extract the actual durian planting area using a semantic segmentation network; S5. Based on the actual durian planting area, use drones to acquire remote sensing image data of durian, and identify pests of durian based on the remote sensing image data. S6. Based on the pest identification results, conduct growth health assessment and early warning for durians; S7. Visualize growth health assessment and early warning information and generate planting suggestions.
[0020] This embodiment can accurately delineate suitable durian growing areas and further clarify the actual durian planting areas, avoiding resource waste and yield loss caused by blind site selection. Simultaneously, it enables precise monitoring throughout the entire durian growth cycle, allowing for timely identification of pest areas and their types within the actual durian planting area, providing early warnings of risks, and reducing the probability of pest spread. Furthermore, by combining visual displays with targeted planting suggestions, it facilitates differentiated orchard management, achieving efficient control of large-scale orchards, reducing the intensity of manual inspections, improving management efficiency, and stabilizing durian yield and fruit quality.
[0021] In step S1, the preprocessing includes radiometric correction, geometric correction / registration, atmospheric correction, cloud mask denoising, and terrain correction, which are used to eliminate various errors and invalid interferences in multi-source satellite remote sensing data, unify the spatiotemporal scale of multi-source data, accurately extract key environmental factors related to durian growth, and ensure that the modeling and analysis results of durian suitable growing areas are accurate and reliable.
[0022] In step S2, the key environmental factors include surface temperature, vegetation cover, soil moisture index, topographic factors, seasonal temperature variation coefficient, annual accumulated temperature, and spatiotemporal distribution characteristics of precipitation.
[0023] Step S3 specifically includes the following steps: S11. Obtain the environmental thresholds for durian cultivation corresponding to different soil types and terrains. The environmental thresholds for durian cultivation include environmental humidity, soil moisture, soil pH and light conditions. The environmental humidity threshold can be 60%-85%, the soil moisture threshold can be 20%-40%, the soil pH threshold can be pH5.5-pH7.5, and the annual duration threshold for light conditions can be more than 1800 hours. S12. Establish different suitability levels based on the threshold of durian planting environment, including three levels: most suitable, second most suitable, and unsuitable; S13. Based on multi-source satellite remote sensing data, obtain the soil type and topography of the durian suitable growing area; S14. Dimensionless processing of key environmental factors is performed, and the suitable level of durian growing areas is divided according to the soil type and topography of the durian growing area based on the environmental threshold of durian planting, generating a durian planting suitability classification map.
[0024] In step S4, the automatic extraction of the actual durian planting area using a semantic segmentation network specifically includes the following steps: S21. A multi-scale feature feedforward network module is introduced into the semantic segmentation network, and multiple convolutional position encoding sub-modules are set in the multi-scale feature feedforward network module. By adjusting the convolutional kernel size, filling strategy and dilation coefficient, the multi-resolution feature map (i.e. multi-scale feature) of durian is extracted based on the durian planting suitability grading map, including the spectral features, texture features and morphological features of durian crown. S22. Divide the multi-resolution feature map into a first feature map and a second feature map, and process the first feature map and the second feature map through different convolutional position encoding sub-modules to enhance the feature representation capability of the multi-resolution feature map, thereby obtaining a first processed feature map and a second processed feature map. The first processed feature map is represented as follows:
[0025] in, This is the first feature map to be processed. This is the first feature map. For the spatial dimensions of multi-resolution feature maps, This is the second feature map.
[0026] The second processed feature map is represented as follows:
[0027] in, This is the second processing feature map.
[0028] The first processed feature map and the second processed feature map are fused to obtain a unified feature map of durian; S23. An attention mechanism is introduced into the semantic segmentation network to enhance the feature information of the durian unified feature map, resulting in an enhanced durian feature map. S24. Based on the enhanced feature map of durian, the actual planting area of durian is extracted using a semantic segmentation network.
[0029] In this embodiment, by introducing a multi-scale feature feedforward network module into the semantic segmentation network to improve the semantic segmentation network structure, and by further setting a convolutional position encoding submodule, the expressive ability of the semantic segmentation network in analyzing the complex surface features of durian suitable growing areas is enhanced, thereby further improving the extraction accuracy of actual durian planting areas.
[0030] In step S23, the feature information enhancement of the durian unified feature map through the attention mechanism specifically includes the following steps: S31. Perform spatial average pooling on the uniform feature map of durian along the horizontal and vertical axes to obtain feature maps along the horizontal and vertical axes. S32. Dimensionality reduction of the feature maps along the horizontal and vertical axes is achieved using convolutional layers. Attention weights along the horizontal and vertical axes are then generated using two convolutional layers respectively, as shown below:
[0031]
[0032] in, The attention weights are located along the horizontal axis. The attention weights are along the vertical axis. , , These are the weight matrices, and These are feature maps along the horizontal and vertical axes, respectively. Activation function 。
[0033] S33. The durian unified feature map is weighted using attention weights along the horizontal and vertical axes to obtain the durian enhanced feature map, as shown below:
[0034] in, Enhanced feature map of durian, A unified characteristic diagram of durian.
[0035] In this embodiment, by combining the attention mechanism with coordinate information, it is possible not only to focus on the relationship between feature channels, but also to consider the spatial distribution of features, thereby improving the accuracy and robustness of feature representation.
[0036] Step S5 specifically includes the following steps: S41. Plan the flight path of the drone according to the actual durian planting area, and use the high-resolution sensor on the drone to obtain remote sensing image data of durian. S42. Perform processing operations on remote sensing image data, including geometric correction and radiometric correction. S43. Based on the processed remote sensing image data, extract the growth status information of durian; S44. Based on growth status information, a target recognition unit is constructed using a target recognition and localization algorithm. The initial pest region is generated through anchor box association and bounding box regression of the target recognition unit, as shown below:
[0037] in, This is an area where pests are initially detected. This is the radiation correction factor. The normalization coefficient is... For bounding box regression, there is a single-lag variable.
[0038] The initial pest area is corrected to obtain the detection output box of the pest area, as shown below:
[0039]
[0040] in, and The coordinates for detecting pest areas. It is a gray-level co-occurrence matrix. Geometric correction coefficient, This represents the reflectance coefficient in the near-infrared spectral band.
[0041] Based on the detection output box, the pest areas in the actual durian planting area are identified, and a pest identification model is constructed. The pest identification model is then used to identify the pest type based on the pest area.
[0042] In step S54, the construction of the pest identification model specifically includes the following steps: S51. Collect and label the pest feature dataset, including common durian pests such as durian gray spot disease, fruit borer, and root rot disease, as well as their corresponding pest features, and divide them into training set and validation set. S52. Construct a pest identification model, input the training set into the pest identification model for training, and introduce the focus loss function as the loss function of the pest identification model to solve the problem of imbalanced durian pest samples, improve the identification rate of early latent small lesions, and adjust the parameters of the pest identification model according to the loss function value. S53. Input the validation set into the trained pest identification model to verify the classification performance of the pest identification model. S54. Input the remote sensing image data containing the pest area into the pest identification model to identify the pest types in the pest area.
[0043] In step S54, the step of using a pest identification model to identify pest types based on pest areas specifically includes the following steps: S61. Perform dimensionality reduction processing on the remote sensing image data containing pest areas, and extract the spectral features of the pest areas, as shown below:
[0044] in, Spectral characteristics, It is a linear mapping function. For remote sensing image data The grayscale value of each pixel These are the hyperplane bias parameters. The minimum geometric margin from the sample points in the remote sensing image data to the hyperplane. It is a Lagrange multiplier.
[0045] S62. Based on remote sensing image data containing pest areas, extract the texture features of the pest areas, as shown below:
[0046] in, For texture features, This is a local binary coded vector for remote sensing image data. This is the critical number of steps to terminate the iteration. For attribute constraints.
[0047] S63. Combine spectral features and texture features to obtain comprehensive pest features, and input them into the pest identification model. The pest identification model maps the comprehensive pest features to the pest category space through a fully connected layer, thereby outputting the probability of the corresponding pest category, as shown below:
[0048] in, To determine the pooled window size, The batch size of the fully connected layer. Comprehensive characteristics of pests Belongs to the The probability of a specific pest species.
[0049] In this embodiment, a pest identification model is constructed to identify pests in durians, capturing early-stage latent diseases and trace pests in young fruits that are difficult to detect with the human eye, providing early warning of the risk of spread, avoiding durian tree death, fruit rot and drop, and stabilizing orchard yield and fruit quality.
[0050] Step S6 specifically includes: based on the pest identification results, extracting the durian's health indicators, including plant infection index and fruit damage rate, and quantifying these health indicators. According to preset health level standards and combined with the quantified health indicators, the durian's growth and health status is divided into four levels: healthy, slightly abnormal, moderately abnormal, and severely abnormal. Furthermore, based on the health level and pest spread patterns, a risk level (e.g., low, medium, high) is determined. If the risk level reaches medium or high, an early warning mechanism is triggered, sending an alarm signal to the user.
[0051] In step S7, targeted planting suggestions are generated for durian plants with different health levels and different pest types. These suggestions include pest control plans, adjustments to water and fertilizer management, and pruning and maintenance measures, adapting to the differentiated management needs of the orchard. Simultaneously, the distribution of health levels, pest areas, and risk level zoning are visualized using charts and maps, providing a clear overview of the durian orchard's health status and risk distribution.
[0052] Reference Figure 2 This embodiment provides a system for identifying and accurately monitoring suitable durian growing areas, used to implement any of the above-mentioned methods for identifying and accurately monitoring suitable durian growing areas. The system includes: Data acquisition module: used to acquire multi-source satellite remote sensing data of durian suitable growing areas using satellite remote sensing technology, and to preprocess the multi-source satellite remote sensing data; Environmental factor extraction module: used to extract key environmental factors of durian suitable growing areas from preprocessed multi-source satellite remote sensing data; Suitability Grading Module: Used to construct a durian suitability grading model based on key environmental factors, and generate a durian planting suitability grading map based on the durian suitability grading model; Planting Area Extraction Module: Used to extract the actual planting area of durian based on the durian planting suitability grading map using a semantic segmentation network; Pest identification module: Used to acquire remote sensing image data of durians using drones based on the actual durian planting area, and to identify pests in durians based on the remote sensing image data; Assessment and early warning module: used to assess the growth health of durians and issue early warnings based on pest identification results; Visualization Recommendation Module: This module visualizes growth health assessments and early warning information and generates targeted durian planting recommendations tailored to different health levels.
[0053] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for identifying and accurately monitoring suitable durian growing areas, characterized in that, Includes the following steps: S1. Use satellite remote sensing technology to obtain multi-source satellite remote sensing data of durian suitable growing areas, and preprocess the multi-source satellite remote sensing data; S2. Extract key environmental factors for durian suitable growing areas from preprocessed multi-source satellite remote sensing data; S3. Based on key environmental factors, the suitability of durian growing areas is classified, and a durian planting suitability classification map is generated. S4. Based on the durian planting suitability classification map, extract the actual durian planting area using a semantic segmentation network; S5. Based on the actual durian planting area, use drones to acquire remote sensing image data of durian, and identify pests of durian based on the remote sensing image data. S6. Based on the pest identification results, conduct growth health assessment and early warning for durians; S7. Visualize growth health assessment and early warning information and generate planting suggestions.
2. The method for identifying and accurately monitoring suitable durian growing areas according to claim 1, characterized in that, The key environmental factors include surface temperature, vegetation cover, soil moisture index, topographic factors, seasonal temperature variation coefficient, annual accumulated temperature, and spatiotemporal distribution characteristics of precipitation.
3. The method for identifying and accurately monitoring suitable durian growing areas according to claim 1, characterized in that, Step S3 specifically includes the following steps: S11. Obtain the durian planting environment thresholds corresponding to different soil types and terrains. The durian planting environment thresholds include environmental humidity, soil moisture, soil pH and light conditions. S12. Establish different suitability levels based on the threshold of durian growing environment; S13. Based on multi-source satellite remote sensing data, obtain the soil type and topography of the durian suitable growing area; S14. Dimensionless processing of key environmental factors, and in combination with soil type and topography of durian suitable areas, the suitability level of durian suitable areas is divided according to the environmental threshold of durian planting, generating a durian planting suitability classification map.
4. The method for identifying and accurately monitoring suitable durian growing areas according to claim 3, characterized in that, The method of automatically extracting the actual planting area of durian using a semantic segmentation network specifically includes the following steps: S21. Introduce a multi-scale feature feedforward network module into the semantic segmentation network. Use the multi-scale feature feedforward network module to extract multi-scale features of durian based on the durian planting suitability grading map, including the spectral features, texture features and morphological features of the durian canopy. S22. The multi-scale feature feedforward network module is provided with a convolutional position encoding submodule, which is used to fuse the extracted multi-scale features to obtain a unified feature map of durian. S23. An attention mechanism is introduced into the semantic segmentation network to enhance the feature information of the durian unified feature map, resulting in an enhanced durian feature map. S24. Based on the enhanced feature map of durian, the actual planting area of durian is extracted using a semantic segmentation network.
5. The method for identifying and accurately monitoring suitable durian growing areas according to claim 4, characterized in that, The enhancement of feature information in the uniform feature map of durian through an attention mechanism specifically includes the following steps: S31. Perform spatial average pooling on the uniform feature map of durian along the horizontal and vertical axes to obtain feature maps along the horizontal and vertical axes. S32. Use convolutional layers to reduce the dimensionality of feature maps in the horizontal and vertical directions, and generate attention weights in the horizontal and vertical directions respectively. S33. Use the attention weights in the horizontal and vertical directions to weight the durian unified feature map to obtain the durian enhanced feature map.
6. The method for identifying and accurately monitoring suitable durian growing areas according to claim 1, characterized in that, Step S5 specifically includes the following steps: S41. Plan the flight path of the drone according to the actual durian planting area, and use the high-resolution sensor on the drone to obtain remote sensing image data of durian. S42. Perform processing operations on remote sensing image data, including geometric correction and radiometric correction. S43. Based on the processed remote sensing image data, extract the growth status information of durian; S44. Based on growth status information, identify the pest areas in the actual durian planting area, construct a pest identification model, and use the pest identification model to identify the pest type according to the pest area.
7. The method for identifying and accurately monitoring suitable durian growing areas according to claim 6, characterized in that, The construction of the pest identification model specifically includes the following steps: S51. Collect and label the pest feature dataset, including pest types and their corresponding pest features, and divide it into training set and validation set. S52. Construct a pest identification model, input the training set into the pest identification model for training, introduce the focus loss function as the loss function of the pest identification model, and adjust the parameters of the pest identification model according to the loss function value. S53. Input the validation set into the trained pest identification model to verify the classification performance of the pest identification model. S54. Input the remote sensing image data containing the pest area into the pest identification model to identify the pest types in the pest area.
8. The method for identifying and accurately monitoring suitable durian growing areas according to claim 7, characterized in that, The method of using a pest identification model to identify pest types based on pest areas specifically includes the following steps: S61. Perform dimensionality reduction processing on remote sensing image data containing pest areas and extract the spectral features of pest areas. S62. Extract texture features of pest areas based on remote sensing image data containing pest areas; S63. Combine spectral features and texture features to obtain comprehensive pest features, and input them into the pest identification model. Output the probability of the corresponding pest category based on the comprehensive pest features.
9. A system for identifying and accurately monitoring suitable durian growing areas, characterized in that, The system includes: Data acquisition module: used to acquire multi-source satellite remote sensing data of durian suitable growing areas using satellite remote sensing technology, and to preprocess the multi-source satellite remote sensing data; Environmental factor extraction module: used to extract key environmental factors of durian suitable growing areas from preprocessed multi-source satellite remote sensing data; Suitability Grading Module: Used to construct a durian suitability grading model based on key environmental factors, and generate a durian planting suitability grading map based on the durian suitability grading model; Planting Area Extraction Module: Used to extract the actual planting area of durian based on the durian planting suitability grading map using a semantic segmentation network; Pest identification module: Used to acquire remote sensing image data of durians using drones based on the actual durian planting area, and to identify pests in durians based on the remote sensing image data; Assessment and early warning module: used to assess the growth health of durians and issue early warnings based on pest identification results; Visualization suggestion module: This module is used to visualize growth health assessment and early warning information and generate planting suggestions.