Paddy field drought intelligent diagnosis system and method

Abstract

The invention relates to the technical field of agricultural drought prevention and drought resistance, in particular to a paddy field drought intelligent diagnosis system and method. Acquiring time-space acquisition data and rice growth periods corresponding to the inspection points; respectively inputting the paddy field image, the space-time acquisition data and the rice growth period corresponding to each patrol point into a patrol point drought grade diagnosis model for identification to obtain a drought grade corresponding to each patrol point; based on the drought level corresponding to each patrol point, determining the corresponding area proportion of each drought level in the patrol area; and finally, calculating the drought level corresponding to the patrol area based on the corresponding area proportion of each drought level in the patrol area. According to the method, the rice growth period and the time-space acquisition data are comprehensively considered, so that the drought grade corresponding to each patrol point and the drought grade corresponding to the patrol area can be accurately identified.

Description

Technical Field

[0001] The present invention relates to the technical field of agricultural drought prevention and drought resistance, and particularly to an intelligent diagnosis system and method for drought conditions in paddy fields. Background Art

[0002] As a typical water-demanding crop, rice has significant differences in water requirements during different growth stages. These complex water requirement characteristics make the drought diagnosis in paddy fields more complex than that in general farmlands, and differential evaluation needs to be carried out in combination with specific growth stages.

[0003] Currently, the following problems mainly exist in the drought diagnosis and identification of paddy fields in China: There are problems such as inaccurate reporting, untimely reporting, and strong subjectivity in the traditional manual reporting method; the image quality of remote sensing monitoring technology is severely affected under the cloudy and rainy climate conditions in the south, so the accuracy of drought condition identification for different growth stages of rice is not high enough; in addition, the joint intelligent analysis and application of multi-factor massive data such as rice images, time and space are not sufficient; most of the existing methods adopt a unified drought judgment standard, ignoring the water requirement differences in different growth stages of rice. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent diagnosis system and method for drought conditions in paddy fields to improve the recognition accuracy.

[0005] To achieve the above purpose, the present invention provides an intelligent diagnosis method for drought conditions in paddy fields, and the method includes: Navigate to each inspection point within the inspection area to obtain the paddy field images corresponding to each inspection point; Obtain the spatio-temporal acquisition data and rice growth stage corresponding to each inspection point; the spatio-temporal acquisition data includes position information, image acquisition time, and meteorological data; Input the paddy field images, the spatio-temporal acquisition data, and the rice growth stage corresponding to each inspection point into the inspection point drought level diagnosis model for identification to obtain the drought level corresponding to each inspection point; Determine the area ratio corresponding to each drought level in the inspection area based on the drought level corresponding to each inspection point; Calculate the drought level corresponding to the inspection area based on the area ratio corresponding to each drought level in the inspection area.

[0006] Optionally, the method further includes: Taking the paddy field images, spatio-temporal acquisition data, and rice growth stage corresponding to each historical inspection point as the data input set, and taking the drought level corresponding to each historical inspection point as the data output set, training with the multi-modal agricultural scene perception network MASPN to obtain the inspection point drought level diagnosis model.

[0007] Optionally, using the paddy field images, spatio-temporal acquisition data, and rice growth stages corresponding to each historical inspection point as the data input set, and the drought levels corresponding to each historical inspection point as the data output set, training is performed using the Multimodal Agricultural Scene Perception Network (MASPN) to obtain a drought level diagnosis model for the inspection points, which specifically includes: Performing feature extraction on the paddy field images corresponding to each historical inspection point based on the ConvNeXt-V2 deep convolutional neural network architecture to obtain image visual feature vectors; Performing feature extraction on the spatio-temporal acquisition data corresponding to each historical inspection point to obtain spatio-temporal feature vectors; Performing feature extraction based on the water requirement characteristics of the rice growth stages corresponding to each historical inspection point to obtain growth stage feature vectors; Using an adaptive cross-modal attention mechanism, training is performed based on the image visual feature vectors, the spatio-temporal feature vectors, the growth stage feature vectors, and the drought levels corresponding to each historical inspection point, and the total loss value is calculated; Determining whether the total loss value meets the set requirements; if it meets the set requirements, a drought level diagnosis model for the inspection points is constructed based on the weights corresponding to the three feature vectors; if it does not meet the set requirements, the weights corresponding to the three feature vectors are adjusted through the backpropagation algorithm.

[0008] Optionally, the method further includes: Performing spatial clustering analysis on the drought data corresponding to each inspection point in the inspection area to generate a spatial distribution heat map and hot spots corresponding to the inspection area; the drought data includes longitude and latitude, the drought levels corresponding to each inspection point, and the encoding of the growth stage.

[0009] Optionally, the method further includes: Using a method that combines inverse distance weighted interpolation and Kriging interpolation to perform spatial interpolation on the drought levels corresponding to each inspection point to generate a continuous drought distribution surface.

[0010] The present invention also discloses a rice paddy drought intelligent diagnosis system, which includes: A mobile terminal with a drought inspection system, a server, and a service end; the mobile terminal is connected to the service end through the server; The mobile terminal includes: a camera and a GPS positioning module; the camera is used to obtain the paddy field images corresponding to each inspection point; the GPS positioning module is used to position each inspection point to obtain location information; The drought inspection system includes: a data acquisition module and an intelligent recognition module; The data acquisition module is used to obtain the paddy field images, spatio-temporal acquisition data, and rice growth stages corresponding to each inspection point; the spatio-temporal acquisition data includes location information, image acquisition time, and meteorological data; The intelligent recognition module is used to input the paddy field images, the spatio-temporal acquisition data, and the rice growth stages corresponding to each inspection point into the inspection point drought level diagnosis model for recognition, so as to obtain the drought levels corresponding to each inspection point; The server includes: an intelligent analysis module; The intelligent analysis module is used to receive the drought levels corresponding to each inspection point sent by the mobile terminal; and determine the area ratio corresponding to each drought level in the inspection area based on the drought levels corresponding to each inspection point; calculate the drought level corresponding to the inspection area based on the area ratio corresponding to each drought level in the inspection area; The server side includes: a data display module; The data display module is used to display the drought levels corresponding to each inspection point and / or the drought level corresponding to the inspection area sent by the server.

[0011] Optionally, the drought inspection system further includes: A user management module, which is used to integrate functions such as inspection personnel type selection, personal information management, and inspection task management; A map navigation module, which is used to implement paddy field inspection position positioning, inspection point marking, inspection range query, inspection map zooming, and inspection path planning.

[0012] Optionally, the server further includes: A data processing module, which is used to receive the paddy field images uploaded by the mobile terminal, perform format standardization and quality inspection processing, and use them as the data input set for subsequent optimization of the inspection point drought level diagnosis model; A storage module, which is used to adopt a distributed storage architecture and use MongoDB to store the processed data.

[0013] Optionally, the intelligent analysis module is further used to perform spatial clustering analysis on the drought situation data corresponding to each inspection point in the inspection area, generate a spatial distribution heat map and hot spots corresponding to the inspection area, and send them to the server side for display; the drought situation data includes longitude and latitude, the drought levels corresponding to each inspection point, and the codes of the growth stages.

[0014] According to the specific embodiments provided by the present invention, the following technical effects are disclosed by the present invention: The present invention discloses an intelligent diagnosis system and method for drought conditions in paddy fields. First, paddy field images corresponding to each inspection point are acquired; spatio-temporal acquisition data and the rice growth stage corresponding to each inspection point are obtained. Secondly, the paddy field images, the spatio-temporal acquisition data, and the rice growth stage corresponding to each inspection point are respectively input into the drought level diagnosis model of the inspection point for identification to obtain the drought level corresponding to each inspection point. Then, based on the drought levels corresponding to each inspection point, the area ratios corresponding to each drought level in the inspection area are determined. Finally, the drought level corresponding to the inspection area is calculated based on the area ratios corresponding to each drought level in the inspection area. The present invention comprehensively considers the rice growth stage and spatio-temporal acquisition data, and thus can accurately identify the drought level corresponding to each inspection point and the drought level corresponding to the inspection area. The present invention can provide accurate, timely, and objective decision-making support for drought prevention and control, and significantly improve the intelligent level of drought monitoring and drought resistance work in paddy fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.

[0016] Figure 1 Structural diagram of the intelligent diagnosis system for drought conditions in paddy fields according to the embodiment of the present invention; Figure 2 Specific flowchart for training the drought level diagnosis model of the inspection point by the multi-modal agricultural scenario perception network MASPN according to the embodiment of the present invention; Figure 3 Flowchart of the intelligent diagnosis method for drought conditions in paddy fields according to the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0018] The purpose of the present invention is to provide an intelligent diagnosis system and method for drought conditions in paddy fields to improve the recognition accuracy.

[0019] To make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below with reference to the drawings and specific embodiments.

[0020] Embodiment 1 Figure 1 This is the structural diagram of the intelligent drought diagnosis system for paddy fields in the embodiments of the present invention. As Figure 1 shown, the present invention discloses an intelligent drought diagnosis system for paddy fields. The system includes: a mobile terminal 1, a server 2, and a server end 3; the mobile terminal 1 is connected to the server end 3 through the server 2. The mobile terminal 1 of the present invention includes a drought inspection system, a camera, a GPS positioning module, and a network module; the drought inspection system is set on the mobile terminal 1 in the form of an APP; the camera is used to obtain paddy field images corresponding to each inspection point; the camera supports autofocus, image stabilization, and light compensation functions; the GPS positioning module is used to position each inspection point to obtain position information; the network module is used to realize data transmission between the mobile terminal 1 and the server 2. The paddy field images mentioned in the present invention are JPEG format images with a resolution of 1024×768 pixels. A firewall is set between the network module and the server 2. The network module is constructed through a 4G / 5G mobile network or a WiFi network and has a function of resuming data transmission after network interruption. When the network is interrupted, it can automatically reconnect and retransmit data. The server end 3 is connected to the server 2 through a dedicated network or a VPN to ensure the security and stability of data transmission.

[0021] The server end 3 of the present invention is used to query and statistically analyze the drought distribution. The server end 3 adopts a B / S architecture, supports multi-user concurrent access, and has a complete permission management and data security protection mechanism.

[0022] The mobile terminal 1 of the present invention can be a smart phone, a tablet computer, or a dedicated device with GPS positioning function and camera shooting function. Its camera can only be called and turned on within the specified paddy field inspection area. The legality of the inspection position is verified through GPS positioning. When it exceeds the specified range, the camera function is automatically disabled, and the drought inspection system can be called. The drought inspection system adopts a front-end and back-end separation architecture. The front end communicates with the back end (i.e., the mobile terminal 1) through a 4G / 5G wireless network and realizes two-way data interaction with the back end (i.e., the mobile terminal 1).

[0023] The drought inspection system provides two versions, Android and iOS, with exactly the same functions. After the inspector downloads the "Drought Inspection System" and installs it on the mobile terminal 1 according to the operating system type of the mobile terminal 1, he can complete registration and login and then use it.

[0024] The drought inspection system of the present invention includes: a data acquisition module, an intelligent recognition module, a user management module, and a map navigation module. Each module will be discussed in detail below: The data acquisition module is used to acquire paddy field images, spatio-temporal acquisition data, and rice growth stages corresponding to each inspection point; the spatio-temporal acquisition data includes location information, image acquisition time, and meteorological data, and the meteorological data includes temperature, humidity, and rainfall; the location information includes longitude and latitude; the meteorological data can be obtained by accessing relevant meteorological network data; the above location information and image acquisition time can be directly obtained through the mobile terminal 1. The rice growth stage disclosed in the present invention is one of the tillering stage, jointing stage, heading stage, filling stage, and maturity stage.

[0025] The user management module integrates functions such as the selection of inspection personnel types, personal information management, and inspection task management. The drought inspection system provides two types of inspection personnel, namely township agricultural technicians and reservoir inspectors, and different inspection areas are allocated according to the identities of different inspectors, and there is no conflict or overlap between the inspection areas.

[0026] When the user first enters the drought inspection system, a loading page will be displayed, and then it will enter the inspection personnel type selection interface. The inspector selects the corresponding township / reservoir for inspection according to their identity. When a certain inspection point is selected, the system automatically locks the area, and other inspectors cannot select it repeatedly; to release the inspection point, the administrator of the server 3 needs to process the database.

[0027] The personal information management function includes the addition and modification of personal information, where the mobile phone number and name are required fields, and the position and unit are optional fields. The inspector can modify personal information by clicking on the personal center in the upper right corner of the home page.

[0028] The inspection task management provides functions such as inspection assistance, modification of inspection types, and inspection process management. Status prompts of "Start Inspection" or "Continue Inspection" are displayed below the home page interface, and the two are switched dynamically. After clicking Start Inspection, it will change to the Continue Inspection state.

[0029] The map navigation module integrates functions such as paddy field inspection location positioning, inspection point marking, inspection range query, inspection map zooming, and inspection path planning. The map navigation module has an offline map function and supports the switching display of satellite maps and electronic maps to ensure normal use in remote areas with poor network signals.

[0030] The inspector views the real-time map distribution through the map navigation module and performs GPS precise positioning centered on himself. The gestures of spreading and merging two fingers can achieve the zooming in and out of the map, and multi-level zooming is supported. After the inspector determines the inspection area range, the inspection points can be marked and saved on the inspection map. The system records the inspector's trajectory data in real time and transmits it to the server 2, and further ensures the authenticity and effectiveness of the inspection through trajectory verification.

[0031] Precise annotation is carried out according to the four-level drought standard before training: Normal water layer (Level 0): There is an obvious water layer on the field surface, with a depth of 2 - 5 cm.

[0032] Soil exposure (Level 1): The water layer on the field surface disappears, and the soil is exposed but without cracks.

[0033] Absence of water layer (Level 2): The field surface is dry, and the soil shrinks with fine cracks appearing.

[0034] Cracks on the field surface (Level 3): The field surface is severely cracked, and the crack width > 1 cm.

[0035] The intelligent recognition module is used to input the paddy field images, spatio-temporal acquisition data, and rice growth stages corresponding to each inspection point into the drought level diagnosis model of the inspection point for recognition, and obtain the drought level corresponding to each inspection point; the intelligent recognition module is also used to use the paddy field images, spatio-temporal acquisition data, and rice growth stages corresponding to each historical inspection point as the data input set, and the drought levels corresponding to each historical inspection point as the data output set, and train using the multi-modal agricultural scene perception network MASPN to obtain the drought level diagnosis model of the inspection point; the drought levels corresponding to each historical inspection point mentioned in the present invention are normal water layer, soil exposure, absence of water layer, and cracks on the field surface. The MASPN mentioned in the present invention adopts three feature extraction branches and one cross-modal fusion. The three feature extraction branches are the image feature branch, the spatio-temporal feature branch, and the growth stage feature branch. The image feature branch is actually the branch corresponding to the image features, the spatio-temporal feature branch is actually the branch corresponding to the spatio-temporal features, and the growth stage feature branch is actually the branch corresponding to the growth stage features.

[0036] As Figure 2 shown, the present invention discloses using the paddy field images, spatio-temporal acquisition data, and rice growth stages corresponding to each historical inspection point as the data input set, and the drought levels corresponding to each historical inspection point as the data output set, and training using the multi-modal agricultural scene perception network MASPN to obtain the drought level diagnosis model of the inspection point, specifically including: Image feature branch: Based on the ConvNeXt-V2 deep convolutional neural network architecture, extract features from the paddy field images corresponding to each historical inspection point to obtain image visual feature vectors, specifically including: Perform size standardization and pixel value normalization on the paddy field images corresponding to each historical inspection point through the data processing layer to obtain the preprocessed paddy field images; the preprocessed paddy field images are RGB images of 1024×768×3.

[0037] Extract features from the preprocessed paddy field images stage by stage through 4 stages of the ConvNeXt-V2 deep convolutional neural network to obtain spatial feature maps; each stage contains multiple ConvNeXt blocks.

[0038] The spatial feature map is processed successively through the global average pooling layer and the feature mapping layer to obtain a 128-dimensional image visual feature vector.

[0039] Compared with traditional convolutional neural networks, the ConvNeXt-V2 of the present invention has stronger feature expression ability and better generalization performance, and is particularly suitable for agricultural scenario image analysis. The network adopts layer normalization (LayerNorm) and Gaussian error linear unit (GELU) activation functions, and has better training stability.

[0040] Spatio-temporal feature branch: Feature extraction is performed based on the spatio-temporal acquisition data corresponding to each historical inspection point to obtain a spatio-temporal feature vector, specifically including: The position information corresponding to each historical inspection point is standardized through the data processing layer, the image acquisition time corresponding to each historical inspection point is converted, and the meteorological data corresponding to each historical inspection point is normalized.

[0041] The preprocessed data is input into a two-layer fully connected neural network for feature encoding to obtain a 128-dimensional spatio-temporal feature vector; among them, the first layer: the input dimension is 7 (longitude, latitude, timestamp, temperature, humidity, rainfall, acquisition date), the output dimension is 256, and the rectified linear unit (ReLU) is used for activation; the second layer: the input dimension is 256, the output dimension is 128, and the rectified linear unit (ReLU) is used for activation.

[0042] These spatio-temporal features provide important environmental context information for drought judgment.

[0043] Growth stage feature branch: Feature extraction is performed based on the water demand characteristics of the rice growth stage corresponding to each historical inspection point to obtain a growth stage feature vector, specifically including: The rice growth stage is encoded into a digital identifier based on the digital encoding layer.

[0044] The digital identifier is mapped into a 64-dimensional continuous vector using the embedding layer.

[0045] The 64-dimensional continuous vector is extended to a 128-dimensional growth stage feature vector using the fully connected layer.

[0046] Cross-modal fusion training: An adaptive cross-modal attention mechanism is adopted to train based on the image visual feature vector, the spatio-temporal feature vector, the growth stage feature vector, and the drought level corresponding to each historical inspection point, and the total loss value is calculated.

[0047] Judge whether the total loss value meets the set requirements; if it meets the set requirements, a drought level diagnosis model for the inspection point is constructed based on the weights corresponding to the three feature vectors; if it does not meet the set requirements, the weights corresponding to the three feature vectors are adjusted through the backpropagation algorithm.

[0048] The adaptive cross-modal attention mechanism disclosed in the present invention can dynamically adjust the weights of the three feature extraction branches according to the water demand characteristics at different growth stages, so as to achieve more accurate drought recognition.

[0049] The weights corresponding to the three feature vectors are specifically:

[0050] Among them, 、 and respectively represent the attention weights of the image feature branch, the spatio-temporal feature branch, and the growth stage feature branch, 、 and respectively represent the query weight matrices, 、 and respectively represent the feature vectors of the three branches, and respectively represent the key weight matrices, 、 and respectively represent the learnable bias terms, and softmax is the normalized exponential function.

[0051] The specific formula for calculating the total loss value is: L = L main + λ 1 ×L stage + λ 2 ×L climate + λ 3 ×L contrastive Among them, L represents the total loss value; L main is the loss of the main task, that is, the loss of drought level recognition for four classifications; L stage is the loss of the auxiliary task, that is, the loss of growth stage recognition; L climate is the loss of the auxiliary task, that is, the loss of regional climate type recognition; L contrastive is the contrastive learning loss, which is used to enhance the feature discrimination between different drought levels, so that the model can better distinguish different degrees of drought; λ 1 , λ 2 , λ3 They are all adjustable hyperparameters used to balance the weights of different task losses.

[0052] The adaptive cross-modal attention mechanism makes it pay more attention to image features and spatio-temporal features during the sensitive water demand period of rice (such as the heading stage), because the water condition has a significant impact on rice growth at this time; while in the mature stage with relatively less water demand, the sensitivity to minor drought changes will be appropriately reduced. Through this dynamic weighted fusion, MASPN can automatically identify four drought levels in paddy fields, namely field surface cracks, water layer loss, soil exposure, and normal water layer, in the photos taken of paddy fields according to the water demand characteristics of different growth stages of rice.

[0053] Server 2 serves as the core data processing center, supporting load balancing and high-availability deployment; in addition, Server 2 includes: a data processing module, a storage module, an intelligent analysis module, and an information publishing module. The specific functions of each module are as follows: The data processing module is used to receive the paddy field images uploaded by Mobile Terminal 1, perform format standardization and quality detection processing, and use them as the data input set for optimizing the drought level diagnosis model of each inspection point; the image data is uniformly converted to the JPEG format, and the resolution is standardized to 1024×768 pixels. Quality detection includes blur detection, illumination evaluation, occlusion detection, etc. Unqualified images will be marked and re-acquisition will be required.

[0054] The storage module is used to adopt a distributed storage architecture and use the MongoDB database management system to store the processed data, support massive data storage and fast retrieval, and support automatic backup and disaster recovery.

[0055] The intelligent analysis module is used to receive the drought levels corresponding to each inspection point sent by Mobile Terminal 1; and determine the area proportion corresponding to each drought level in the inspection area based on the drought levels corresponding to each inspection point; calculate the drought level corresponding to the inspection area based on the area proportion corresponding to each drought level in the inspection area, specifically including: Calculate the drought degree index of the inspection area based on the area proportion corresponding to each drought level in the inspection area. The specific formula is: DI = Σ(W i × P i × S i ), where DI is the drought degree index of the inspection area, i ranges from 1 to 4, W i is the drought level weight coefficient (W1 = 0 for normal water layer, W2 = 1 for soil exposure, W3 = 2 for water layer loss, W4 = 3 for field surface cracks), P i is the area proportion of the i-th drought level in the area, and S iIt is the growth period adjustment coefficient (S = 1.0 in the tillering stage, S = 1.2 in the jointing stage, S = 1.5 in the heading stage, S = 1.3 in the filling stage, and S = 0.8 in the maturity stage).

[0056] The drought level is divided according to the drought degree index DI value of the inspection area: DI ≤ 0.5 is mild drought, 0.5 < DI ≤ 1.5 is moderate drought, 1.5 < DI ≤ 2.5 is severe drought, and DI > 2.5 is extreme drought.

[0057] The drought degree index of the inspection area of the present invention adopts a weighted calculation method, comprehensively considering the drought level weight coefficient, the area ratio of each level and the rice growth period adjustment coefficient for quantitative evaluation to obtain the drought degree index of the inspection area, and uses the regional drought degree index to conduct real-time diagnosis and evaluation of the current regional drought degree. The current regional drought degree is one of the four drought levels: mild, moderate, severe, and extreme.

[0058] The intelligent analysis module is also used to perform spatial clustering analysis on the drought situation data corresponding to each inspection point in the inspection area, generate a spatial distribution heat map and hot spots corresponding to the inspection area, and send them to the server 3 for display; the drought situation data includes longitude and latitude, the drought level corresponding to each inspection point, and the coding of the growth period; the number of clusters is adaptively determined according to the size of the area. Specifically, machine learning methods such as the K-means clustering algorithm and statistical analysis are used to comprehensively analyze the uploaded drought situation data, classify and count the drought level distribution of each area and each growth period according to different rice growth periods, automatically calculate the regional drought degree index, and generate an evaluation of the current drought degree of the regional paddy fields and a spatial distribution heat map.

[0059] The intelligent analysis module is also used to perform spatial interpolation on the drought level corresponding to each inspection point by combining inverse distance weight interpolation and Kriging interpolation to generate a continuous drought situation distribution surface, and send it to the server 3 for display. The interpolation accuracy is optimized through cross-validation, and the mean absolute error is controlled within 0.15.

[0060] The information release module is used to send at least one of the drought level corresponding to the inspection point, the drought level corresponding to the inspection area, the drought situation statistical report, the spatial distribution heat map and hot spots corresponding to the inspection area, and the continuous drought situation distribution surface of the area to multiple servers 3 for display; the spatial distribution heat map corresponding to the inspection area uses different colors to represent different drought levels, supports multi-level zooming and layer overlay; the drought situation statistical report includes statistical information such as the area and proportion of each drought level.

[0061] The server 3 of the present invention includes: an inspection management module, a data display module, and an information query module. The specific functions of each module are as follows: The data display module is used to display the drought levels corresponding to each inspection point sent by the server 2 and / or the drought levels corresponding to the inspection areas, and is also used to display drought situation statistical reports, spatial distribution heat maps and hot spots corresponding to the inspection areas, and the drought situation distribution surfaces of continuous areas.

[0062] The inspection management module specifically includes: (1) Inspection task management: Administrators can create, assign, and modify inspection tasks, and set inspection areas, inspection frequencies, responsible persons, etc. It supports task division for water systems (river basins, irrigation areas, reservoirs). The system automatically generates an inspection schedule and pushes it to the corresponding inspectors through the APP.

[0063] (2) Inspection progress monitoring: Real-time monitoring of the inspection progress in each area, displaying completed, in-progress, and overdue uncompleted inspection tasks. Intuitively display the inspection coverage through the map, and un-covered areas are marked with different colors. Support for inspection track playback to verify the authenticity of inspections.

[0064] (3) Inspection result review: Quality review of the drought situation data uploaded by inspectors, including image quality inspection, GPS position verification, logical consistency check, etc. Automatic review rules and manual review processes can be set to ensure data quality.

[0065] (4) Permission management: Establish a hierarchical permission management system, and allocate different operation permissions according to user roles (system administrator, regional administrator, ordinary user). Support data access permission control to ensure data security.

[0066] The information query module specifically includes: (1) Multi-dimensional query: Support combined queries according to multiple dimensions such as time, area, drought level, rice growth period, etc. The time dimension supports different granularities such as day, week, month, quarter, year, etc.; the area dimension supports different spatial ranges such as point, line, and surface; query conditions and sorting rules can be customized.

[0067] (2) Statistical analysis: Provide drought situation area statistics, grade distribution statistics, time series statistics, spatial distribution statistics, etc. Support year-on-year and month-on-month analysis, and various statistical charts (bar charts, pie charts, line charts, scatter plots, etc.) can be generated.

[0068] (3) Data export: Support exporting query results to multiple formats such as Excel, PDF, CSV, etc., to meet the data usage needs of different users, and the exported fields and format templates can be customized.

[0069] The present invention constructs a multi-modal agricultural scenario perception network MASPN, which integrates image features, spatio-temporal environmental information, and rice growth stage features, and realizes dynamic weight adjustment through an adaptive cross-modal attention mechanism. Compared with traditional single-image recognition methods, the recognition accuracy is increased by 10-15%.

[0070] In view of the differences in water demand characteristics of rice at different growth stages, the present invention establishes a special growth stage feature encoding branch and the corresponding adjustment coefficient for this branch, realizes differential drought assessment, and effectively solves the technical defect that the prior art ignores the growth stage differences.

[0071] The present invention constructs a complete intelligent system from data collection to decision support. Through machine learning algorithms such as K-means clustering and statistical analysis, it can automatically generate a regional drought degree index and a spatial distribution heat map, providing quantitative support for drought prevention and control decisions.

[0072] Embodiment 2 As Figure 3 shown, the present invention also discloses an intelligent diagnosis method for drought in paddy fields, and the method includes: Step S1: Navigate to each inspection point within the inspection area to obtain the paddy field images corresponding to each inspection point.

[0073] Step S2: Obtain the spatio-temporal acquisition data and rice growth stage corresponding to each inspection point; the spatio-temporal acquisition data includes position information, image acquisition time, and meteorological data.

[0074] Step S3: Input the paddy field images, the spatio-temporal acquisition data, and the rice growth stage corresponding to each inspection point into the inspection point drought level diagnosis model for recognition respectively, to obtain the drought level corresponding to each inspection point.

[0075] Step S4: Determine the area ratio corresponding to each drought level in the inspection area based on the drought level corresponding to each inspection point.

[0076] Step S5: Calculate the drought level corresponding to the inspection area based on the area ratio corresponding to each drought level in the inspection area.

[0077] As an optional implementation manner, the present invention uses the paddy field images, spatio-temporal acquisition data, and rice growth stage corresponding to each historical inspection point as the data input set, and the drought level corresponding to each historical inspection point as the data output set, and trains using the multi-modal agricultural scenario perception network MASPN to obtain the inspection point drought level diagnosis model, specifically including: Extract features from the paddy field images corresponding to each historical inspection point based on the ConvNeXt-V2 deep convolutional neural network architecture to obtain image visual feature vectors.

[0078] Extract features based on the spatio-temporal acquisition data corresponding to each historical inspection point to obtain spatio-temporal feature vectors.

[0079] Extract features based on the water requirement characteristics of the rice growth stage corresponding to each historical inspection point to obtain growth stage feature vectors.

[0080] Adopt an adaptive cross-modal attention mechanism to train and calculate the total loss value based on the image visual feature vector, the spatio-temporal feature vector, the growth stage feature vector, and the drought level corresponding to each historical inspection point.

[0081] Judge whether the total loss value meets the set requirements; if it meets the set requirements, construct a drought level diagnosis model for the inspection points based on the weights corresponding to the three feature vectors; if it does not meet the set requirements, adjust the weights corresponding to the three feature vectors through the backpropagation algorithm.

[0082] As an optional implementation manner, the method of the present invention further includes: Conduct spatial clustering analysis on the drought data corresponding to each inspection point in the inspection area to generate a spatial distribution heat map and hot spots corresponding to the inspection area; the drought data includes longitude and latitude, the drought level corresponding to each inspection point, and the encoding of the growth stage.

[0083] As an optional implementation manner, the method of the present invention further includes: Adopt a method combining inverse distance weighted interpolation and Kriging interpolation to perform spatial interpolation on the drought level corresponding to each inspection point to generate a continuous drought distribution surface.

[0084] For the same parts as in Embodiment 1, specifically refer to Embodiment 1 and will not be repeated here.

[0085] Embodiment 3 The present invention also discloses a drought inspection method, including the following steps: S1. User login and navigation: The inspection personnel hold a mobile terminal, log in through the user management module of the drought inspection system, select the inspection type, and use the GPS positioning function of the map navigation module to navigate to each inspection point in the inspection area.

[0086] S2. On-site data collection: Start the camera in the specified inspection area, take paddy field images of typical scenes reflecting the drought degree of the paddy field, and at the same time record the specific rice growth stage and spatio-temporal acquisition data of the current rice in the drought inspection system.

[0087] S3. Intelligent recognition and confirmation: The intelligent recognition module automatically starts the multi-modal agricultural scene perception network MASPN, analyzes the paddy field images in combination with the water requirement characteristics of the rice growth stage and spatio-temporal acquisition data, gives the recognition result and confidence level among the four drought levels, and the inspector uploads it to the server after confirmation or correction.

[0088] S4. Intelligent analysis and diagnosis: After the server receives the recognition results, the intelligent analysis module automatically uses machine learning algorithms to perform statistical analysis and processing to calculate the regional drought index and the corresponding drought level of the patrol area.

[0089] The detailed implementation process of the drought inspection method is as follows: S1. User login and navigation: S1.1 System login: The inspector starts the drought inspection system on the mobile terminal 1. For the first use, the inspector needs to register an account and enter the name, mobile phone number, ID number, unit, etc. Registered users can directly enter the user name and password to log in. Biometric login methods such as fingerprint recognition and face recognition are supported.

[0090] S1.2 Identity verification: The system verifies the inspector’s identity and authority, confirms his / her inspection qualifications and responsibility areas. The inspector’s current location is confirmed through GPS positioning to verify whether he / she is in the authorized area.

[0091] S1.3 Task selection: The inspector selects the inspection type (township agricultural technician or reservoir inspector) on the main interface, and the system displays a list of currently executable inspection tasks, including task name, inspection area, deadline and other information.

[0092] S1.4 Navigation and positioning: After selecting a specific inspection task, the system automatically loads a map of the corresponding area, showing the inspection route and inspection points. The GPS navigation function is used to guide inspectors to the designated paddy field area, supporting voice broadcast and path planning optimization.

[0093] S2. Field data collection: S2.1 Position verification: After the inspector arrives at the target inspection point, the system verifies whether the current position is within the specified inspection range through GPS positioning. The camera function will only be activated within the legal range. If it is out of range, the system will automatically disable the camera function and prompt the inspector to adjust the position.

[0094] S2.2 Environmental information recording: Before taking photos, inspectors need to record the specific growth period information of the current rice in the system, including tillering period, jointing period, heading period, filling period, maturity period, etc. The system also automatically records GPS coordinates, shooting time, weather conditions and other environmental information.

[0095] S2.3 Image acquisition: Start the camera to take images of paddy fields that reflect the degree of drought in paddy fields. The system requires inspectors to take 3-5 photos from different angles to ensure that the drought conditions in the fields can be fully reflected. During shooting, the system automatically performs image quality detection, including clarity, lighting conditions, rationality of composition, etc. Unqualified photos are required to be retaken.

[0096] S2.4 Supplementary Information: The inspector can add a text description to record special circumstances or other information that needs to be explained. The system supports speech-to-text function for convenient on-site input.

[0097] S3. Intelligent Recognition and Confirmation: S3.1 Automatic Recognition: After receiving the uploaded image data, spatio-temporal information, and growth stage characteristics, the intelligent recognition module inputs them into the drought level diagnosis model at the inspection point for recognition, and finally outputs the recognition results of four drought levels.

[0098] S3.2 Result Display: The system displays the recognized drought levels (normal water layer, soil exposure, water layer absence, field surface cracks) and the corresponding confidence scores on Mobile Terminal 1. At the same time, the annotations of key feature areas are displayed to help the inspector understand the judgment basis of the algorithm.

[0099] S3.3 Manual Confirmation: The inspector can view the recognition results and choose to confirm or correct them. If the recognition result does not match the actual situation, the inspector can manually correct it to the correct drought level and add an explanation for the correction reason. These correction data will be used for the continuous optimization of the model.

[0100] S3.4 Data Upload: After confirmation, the system uploads the image data and recognition results to Server 2 via the network. Compressed transmission and resume interrupted transfer technologies are used to ensure reliable transmission even in an environment with poor network conditions.

[0101] S4. Intelligent Analysis and Diagnosis: S4.1 Data Reception and Processing: After Server 2 receives the drought level corresponding to each inspection point, the data processing module first performs data integrity verification and format standardization processing. The processed image data is stored in the distributed file system, and the structured data is written into the database.

[0102] S4.2 Intelligent Analysis: The intelligent analysis module performs batch analysis on the newly added data regularly (every hour or every day). The K-means clustering algorithm is used to identify the spatial distribution pattern of drought conditions, and the area statistics and level distribution of each region are calculated.

[0103] S4.3 Index Calculation: Calculate the drought degree index of each region based on the analysis results, considering the drought level weight, area ratio, and growth stage adjustment coefficient.

[0104] S4.4 Visualization Display: Generate a heat map of the drought situation distribution in regional paddy fields, using different colors and patterns to represent different drought levels. At the same time, generate statistical charts to provide intuitive data support for decision-makers.

[0105] Example 4 Taking the drought diagnosis and monitoring of rice heading period in a certain county as an example, the actual application effect of the system is explained in detail: Case background: The county has a rice planting area of about 150,000 mu. In 2022, during the heading period, there has been no effective rainfall for 20 consecutive days. It is urgent to accurately assess the drought situation in order to formulate drought prevention and relief measures.

[0106] Implementation process: 1. Inspection deployment: Assign inspection tasks to 25 township agricultural technicians and 8 reservoir inspectors through server 3, covering 128 key plots in the county. Each plot is set up with 2-3 inspection points, totaling 315 inspection points.

[0107] 2. On-site inspection: The inspectors used the drought inspection system on mobile terminal 1 to complete all inspection tasks within 2 days, collecting 945 paddy field images and recording detailed growth period and environmental information. The system verified the authenticity of the inspection through GPS track verification, and the inspection coverage rate reached 100%.

[0108] 3. Intelligent identification: MASPN automatically identified 945 images, and the results showed that normal water layers accounted for 30%, soil exposure accounted for 40%, water layer loss accounted for 20%, and field cracks accounted for 10%. The inspectors confirmed the identification results, with an accuracy rate of 91.2%.

[0109] 4. Intelligent analysis: The intelligent analysis module conducts in-depth analysis of the data and obtains the regional drought index DI = (0×0.3 + 1×0.4 + 2×0.2 + 3×0.1) ×1.5 = 1.1 ×1.5 = 1.65. According to the judgment criteria, the region is severely droughty. The generated drought distribution heat map shows that the drought is most serious in the northeast and central regions, while the areas near the reservoir in the southwest are relatively mild. Based on this, the drought emergency plan was quickly launched, and water resources were allocated in time, effectively alleviating the impact of drought on rice production. Compared with traditional manual estimation, the drought information provided by the system is more accurate, timely and objective, which significantly improves the efficiency and effectiveness of drought prevention and relief work.

[0110] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0111] In this article, specific examples are used to elaborate on the principles and implementation modes of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation modes and application scopes. To sum up, the content of this specification should not be construed as a limitation on the present invention.

Claims

1. An intelligent diagnosis method for drought conditions in paddy fields, characterized in that, The method includes: Navigating to each inspection point within the inspection area to obtain paddy field images corresponding to each inspection point; Obtaining spatio-temporal acquisition data and rice growth stages corresponding to each inspection point; the spatio-temporal acquisition data includes location information, image acquisition time, and meteorological data; Respectively inputting the paddy field images, the spatio-temporal acquisition data, and the rice growth stages corresponding to each inspection point into an inspection point drought level diagnosis model for identification to obtain the drought level corresponding to each inspection point; Determining the area ratio corresponding to each drought level in the inspection area based on the drought levels corresponding to each inspection point; Calculating the drought level corresponding to the inspection area based on the area ratios corresponding to each drought level in the inspection area.

2. The intelligent diagnosis method for drought conditions in paddy fields according to claim 1, wherein The method further includes: Using the paddy field images, spatio-temporal acquisition data, and rice growth stages corresponding to historical inspection points as a data input set, and using the drought levels corresponding to historical inspection points as a data output set, training with a multi-modal agricultural scenario perception network MASPN to obtain an inspection point drought level diagnosis model.

3. The intelligent diagnosis method for drought conditions in paddy fields according to claim 2, characterized in that The process of using the paddy field images, spatio-temporal acquisition data, and rice growth stages corresponding to historical inspection points as a data input set, using the drought levels corresponding to historical inspection points as a data output set, and training with a multi-modal agricultural scenario perception network MASPN to obtain an inspection point drought level diagnosis model specifically includes: Performing feature extraction on the paddy field images corresponding to historical inspection points based on the ConvNeXt-V2 deep convolutional neural network architecture to obtain image visual feature vectors; Performing feature extraction on the spatio-temporal acquisition data corresponding to historical inspection points to obtain spatio-temporal feature vectors; Performing feature extraction based on the water demand characteristics of the rice growth stages corresponding to historical inspection points to obtain growth stage feature vectors; Adopting an adaptive cross-modal attention mechanism to train and calculate the total loss value based on the image visual feature vectors, the spatio-temporal feature vectors, the growth stage feature vectors, and the drought levels corresponding to historical inspection points; Judging whether the total loss value meets the set requirements; if it meets the set requirements, constructing an inspection point drought level diagnosis model based on the weights corresponding to the three feature vectors; if it does not meet the set requirements, adjusting the weights corresponding to the three feature vectors through the backpropagation algorithm.

4. The intelligent diagnosis method for drought situation of paddy fields according to claim 2, characterized in that The method further includes: Performing spatial clustering analysis on the drought data corresponding to each inspection point within the inspection area to generate a spatial distribution heat map and hot spots corresponding to the inspection area; the drought data includes longitude and latitude, the drought levels corresponding to each inspection point, and the encoding of the growth stage.

5. The intelligent diagnosis method for drought conditions in paddy fields according to claim 2, wherein, The method further includes: Adopting a method combining inverse distance weighted interpolation and Kriging interpolation to perform spatial interpolation on the drought levels corresponding to each inspection point to generate a continuous drought situation distribution surface.

6. An intelligent drought diagnosis system for paddy fields, characterized in that, The system includes: A mobile terminal with a drought inspection system, a server, and a service end; the mobile terminal is connected to the service end through the server; The mobile terminal includes: a camera and a GPS positioning module; the camera is used to obtain paddy field images corresponding to each inspection point; the GPS positioning module is used to position each inspection point to obtain location information; The drought inspection system includes: a data acquisition module and an intelligent identification module; The data acquisition module is used to acquire paddy field images, spatio-temporal acquisition data, and rice growth stages corresponding to each inspection point; the spatio-temporal acquisition data includes location information, image acquisition time, and meteorological data; The intelligent recognition module is used to input the paddy field images, the spatio-temporal acquisition data, and the rice growth stages corresponding to each inspection point into the inspection point drought level diagnosis model for recognition, and obtain the drought levels corresponding to each inspection point; The server includes: an intelligent analysis module; The intelligent analysis module is used to receive the drought levels corresponding to each inspection point sent by the mobile terminal; and determine the area ratio corresponding to each drought level in the inspection area based on the drought levels corresponding to each inspection point; calculate the drought level corresponding to the inspection area based on the area ratio corresponding to each drought level in the inspection area; The server side includes: a data display module; The data display module is used to display the drought levels corresponding to each inspection point sent by the server and / or the drought level corresponding to the inspection area.

7. The intelligent drought diagnosis system for paddy fields according to claim 6, wherein The drought inspection system further includes: A user management module, which is used to integrate functions such as inspection personnel type selection, personal information management, and inspection task management; A map navigation module, which is used to implement paddy field inspection location positioning, inspection point marking, inspection range query, inspection map zooming, and inspection path planning.

8. The intelligent rice paddy drought diagnosis system according to claim 6, wherein, The server further includes: A data processing module, which is used to receive the paddy field images uploaded by the mobile terminal, perform format standardization and quality inspection processing, and use them as the data input set for subsequent optimization of the inspection point drought level diagnosis model; A storage module, which is used to adopt a distributed storage architecture and use MongoDB to store the processed data.

9. The intelligent rice paddy drought diagnosis system according to claim 6, characterized in that, The intelligent analysis module is further used to perform spatial clustering analysis on the drought situation data corresponding to each inspection point in the inspection area, generate a spatial distribution heat map and hot spots corresponding to the inspection area, and send them to the server side for display; the drought situation data includes longitude and latitude, the drought levels corresponding to each inspection point, and the encoding of the growth stage.

Images