Intelligent control method and system for soil state of farmland

Abstract

The present invention discloses an intelligent control method and system for the soil state of farmland. The present invention relates to the technical field of intelligent control methods. Based on a soil region to be detected, a corresponding soil layer map and multiple soil data are determined. The soil state of the soil region to be detected is determined based on the soil layer map, multiple soil data, and previous state data of the soil region to be detected, thereby improving the accuracy of the soil state of the soil region to be detected. Based on two adjacent soil regions to be detected, a transitional soil region is determined and the state of the transitional soil region is marked. Abnormal soil states are determined based on the location and soil state of each soil region to be detected and the state of the corresponding transitional soil region. Based on the tracing of the abnormal soil state, the corresponding abnormal soil region is determined. Intelligent control measures are determined based on the location, abnormal soil state, and soil state of the adjacent soil region to be detected, thereby improving the accuracy of the intelligent control measures.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent control method, and particularly relates to an intelligent control method and system for soil state of farmland. BACKGROUND

[0002] With the development of science and technology, farmland is generally located in a crop planting scene and cultivates corresponding crops. The farmland presents a corresponding remote sensing image under the detection of a satellite. In the prior art, the actual area of the farmland is determined based on the recognition of the remote sensing image, and the corresponding soil surface layer is marked according to the recognition of the actual area of the farmland. The soil state of the to-be-detected soil area is not evaluated, which affects the accuracy of the soil state of the to-be-detected soil area, and the intelligent control cannot be performed on the abnormal soil area. SUMMARY

[0003] The present application provides an intelligent control method and system for soil state of farmland to overcome the shortcomings of the prior art.

[0004] The present application provides an intelligent control method for soil state of farmland, which comprises the following steps: determining the actual area of the farmland according to the remote sensing image of the farmland, and marking the position and irrigation path of each crop in the actual area of the farmland; determining the soil distribution map of the farmland according to the position and irrigation path of each crop, determining a plurality of to-be-detected soil areas based on the soil distribution map of the farmland and the position of each crop; in each to-be-detected soil area, determining a corresponding soil level map and a plurality of soil data based on the to-be-detected soil area, determining the soil state of the to-be-detected soil area according to the soil level map, the plurality of soil data and the past state data of the to-be-detected soil area; determining a transition soil area based on two adjacent to-be-detected soil areas, and marking the state of the transition soil area, determining an abnormal soil state based on the position of each to-be-detected soil area, the soil state and the state of the corresponding transition soil area; determining a corresponding abnormal soil area based on the trace of the abnormal soil state, determining an intelligent control measure according to the position of the abnormal soil area, the abnormal soil state and the soil state of the adjacent to-be-detected soil area, and intelligently controlling the abnormal soil area.

[0005] The present application provides an intelligent control system for soil state of farmland, which is applied to the intelligent control method for soil state of farmland described above, and comprises:

[0006] A crop module is configured to determine the actual area of the farmland according to the remote sensing image of the farmland, and mark the position and irrigation path of each crop in the actual area of the farmland.

[0007] The soil region to be detected module is used for determining a soil distribution map of the farmland according to positions of the crops and an irrigation path, determining a plurality of soil regions to be detected based on the soil distribution map of the farmland and the positions of the crops;

[0008] The first state module is used for determining a corresponding soil level map and a plurality of soil data in each soil region to be detected based on the soil region to be detected, determining a soil state of the soil region to be detected according to the soil level map, the plurality of soil data and past state data of the soil region to be detected;

[0009] The second state module is used for determining a transition soil region based on two adjacent soil regions to be detected, marking a state of the transition soil region, and determining an abnormal soil state based on positions of the soil regions to be detected, the soil states and the states of the corresponding transition soil regions;

[0010] The intelligent control module is used for determining a corresponding abnormal soil region based on a trace of the abnormal soil state, determining an intelligent control measure according to the position of the abnormal soil region, the abnormal soil state and soil states of adjacent soil regions to be detected, and intelligently controlling the abnormal soil region.

[0011] Compared with the prior art, the present application has the following advantages:

[0012] In the embodiment of the present application, the method in the embodiment of the present application is used to determine a corresponding soil level map and a plurality of soil data in each soil region to be detected based on the soil region to be detected, determine a soil state of the soil region to be detected according to the soil level map, the plurality of soil data and past state data of the soil region to be detected, introduce a plurality of soil regions to be detected, and comprehensively consider the soil level map, the plurality of soil data and the past state data of the soil region to be detected, thereby improving the accuracy of the soil state of the soil region to be detected.

[0013] Therefore, the transition soil region is determined based on two adjacent soil regions to be detected, the state of the transition soil region is marked, the abnormal soil state is determined based on the positions of the soil regions to be detected, the soil states and the states of the corresponding transition soil regions, the corresponding abnormal soil region is determined based on the trace of the abnormal soil state, the intelligent control measure is determined according to the position of the abnormal soil region, the abnormal soil state and the soil states of adjacent soil regions to be detected, the accuracy of the abnormal soil state is ensured, the overall consideration of the position of the abnormal soil region, the abnormal soil state and the soil states of adjacent soil regions to be detected is realized, the accuracy of the intelligent control measure is improved, and the intelligent control of the abnormal soil region is realized. BRIEF DESCRIPTION OF DRAWINGS

[0014] Figure 1is a flowchart of an intelligent control method for a soil state of a farmland in an embodiment of the present application;

[0015] Figure 2 is a flowchart of step S11 in the intelligent control method for the soil state of the farmland in the embodiment of the present application;

[0016] Figure 3 is a flowchart of step S12 in the intelligent control method for the soil state of the farmland in the embodiment of the present application;

[0017] Figure 4 is a flowchart of step S13 in the intelligent control method for the soil state of the farmland in the embodiment of the present application;

[0018] Figure 5 is a flowchart of step S14 in the intelligent control method for the soil state of the farmland in the embodiment of the present application;

[0019] Figure 6 is a flowchart of step S15 in the intelligent control method for the soil state of the farmland in the embodiment of the present application;

[0020] Figure 7 is a structural composition diagram of the intelligent control system for the soil state of the farmland in the embodiment of the present application. DETAILED DESCRIPTION

[0021] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application.

[0022] Please refer to Figures 1 to 7 An intelligent control method for a soil state of a farmland is applied to an intelligent control scenario of the soil state. The intelligent control method for the soil state of the farmland comprises the following steps.

[0023] Step S11: determining an actual area of the farmland according to a remote sensing image of the farmland, and marking positions of each crop and an irrigation path in the actual area of the farmland;

[0024] Step S12: determining a soil distribution map of the farmland according to the positions of each crop and the irrigation path, and determining a plurality of to-be-detected soil areas based on the soil distribution map of the farmland and the positions of each crop;

[0025] Step S13: in each to-be-detected soil area, determining a corresponding soil level map and a plurality of soil data based on the to-be-detected soil area, and determining a soil state of the to-be-detected soil area according to the soil level map, the plurality of soil data, and past state data of the to-be-detected soil area;

[0026] Step S14: determining a transition soil region based on the two adjacent to-be-detected soil regions, marking the state of the transition soil region, and determining an abnormal soil state based on the position of each to-be-detected soil region, the soil state, and the state of the corresponding transition soil region;

[0027] Step S15: determining a corresponding abnormal soil region based on the tracing of the abnormal soil state, determining an intelligent control measure according to the position of the abnormal soil region, the abnormal soil state, and the soil state of the adjacent to-be-detected soil region, and intelligently controlling the abnormal soil region;

[0028] Reference Figure 2 In step S11, the specific steps are as follows:

[0029] S111: collecting a remote sensing image of the farmland, determining an actual distribution map of the farmland based on preprocessing of the remote sensing image of the farmland, and determining an actual region of the farmland according to region recognition of the actual distribution map of the farmland;

[0030] S112: in the actual region of the farmland, triggering detection of the actual region of the farmland by the unmanned aerial vehicle based on the position of the actual region of the farmland, and outputting a corresponding actual image, determining the position of each crop according to the actual image, the region form of the actual region of the farmland, and the response signal of the unmanned aerial vehicle to the crop;

[0031] S113: determining a pipeline path of the water pipe based on the recognition of the actual region of the farmland, and determining an irrigation path of each crop according to the pipeline path of the water pipe and the position of each crop.

[0032] In the embodiment of the present application, the remote sensing image of the farmland is collected, the actual distribution map of the farmland is determined based on preprocessing of the remote sensing image of the farmland, and the actual region of the farmland is determined according to region recognition of the actual distribution map of the farmland, thereby improving the accuracy of the actual region of the farmland.

[0033] At this time, the remote sensing image is image data obtained by observing the ground from the air or space through sensors installed on platforms such as unmanned aerial vehicles, satellites, or unmanned aerial vehicles; optionally, a multispectral, hyperspectral, or RGB camera is used; the advantage is that the resolution is very high, the maneuverability is flexible, data can be obtained on demand, and the cost is relatively low compared to aerial remote sensing; the disadvantage is that the single coverage range is small, and it is limited by weather and flight regulations; in the remote sensing image, it is not only a color photo, but more importantly, it contains information of different spectral bands (such as red light, green light, blue light, near-infrared, short-wave infrared, etc.), which is crucial for analyzing vegetation growth, soil moisture, and pest control.

[0034] Remote sensing images will be disturbed by various factors during the acquisition process, resulting in image quality degradation or information distortion; preprocessing is to eliminate or weaken these disturbances through various technical means to improve image quality and prepare for subsequent interpretation and analysis; the main steps of preprocessing usually include: radiometric correction: eliminate the influence of sensor performance unevenness, atmospheric scattering and absorption, etc. on image brightness value, and convert the gray value on the image to the actual reflectivity or radiometric brightness of the ground object; geometric correction: eliminate the geometric distortion of the image caused by factors such as sensor viewing angle, earth curvature and terrain undulation during imaging, and make the pixel position on the image correspond to the actual ground coordinate; image enhancement: adjust the contrast, brightness and other parameters of the image to make it clearer and the features more prominent, which is convenient for human eye observation or computer recognition.

[0035] After preprocessing, the obtained image is no longer the original, distorted or distorted data, but an image with accurate geometric position and more realistic radiation information. This image can be regarded as a "true portrayal" of farmland and its surrounding environment, called "actual distribution map", which contains accurate position and shape information of farmland, roads, houses, rivers and other elements.

[0036] According to the area recognition of the actual distribution map of farmland, the actual area of farmland is determined, and the image processing and computer vision technology are used to automatically or semi-automatically recognize the interested area or ground object from the actual distribution map after preprocessing, which is usually based on the pixel features (such as color, texture, shape) or spatial relationship of the image; at this time, for the areas with obvious contrast (such as vegetation and non-vegetation), threshold segmentation can be set to clearly delineate the spatial range occupied by the specific ground object "farmland" in the entire actual distribution map, which is usually represented by a vector boundary (polygon), which accurately outlines the boundary of the farmland, which excludes the non-cultivation areas such as roads and houses around the farmland.

[0037] Further, in the actual area of farmland, the detection of the actual area of farmland by the unmanned aerial vehicle is triggered based on the position of the actual area of farmland, and the unmanned aerial vehicle is started to perform the detection task; this start signal usually comes from a control center or an automatic system; the system will be according to the preset schedule (for example, once a week), a specific event (for example, after receiving the processing result of S111 step).

[0038] The system needs to know the exact geographic coordinate range of the actual area of farmland (for example, the vector boundary obtained by S111 step, containing its minimum / maximum latitude and longitude); the unmanned aerial vehicle control software will use these coordinates to plan the flight path to ensure that the detection range completely covers the actual area of farmland, avoiding omission or flying out of the boundary; the detection is not just flying over the farmland, it usually has a preset flight plan, including:

[0039] Flight altitude: set according to the required image resolution and sensor type; flight path planning: according to the actual area shape and size of the farmland, plan one or more parallel or optimal routes to ensure coverage of the entire area without overlap or controllable overlap rate; sensor configuration: determine which sensors the UAV carries; high-resolution RGB camera (for visible light images), multispectral camera (for obtaining different waveband information such as vegetation index), thermal imaging camera (for monitoring crop water stress or pest and disease heat anomalies), etc.; data acquisition frequency: set the frequency of camera shooting or sensor data acquisition.

[0040] Optionally, the actual area of the farmland is an irregular polygon located between north latitude 40.123 degrees, east longitude 116.456 degrees and north latitude 40.124 degrees, east longitude 116.458 degrees; the system is set to trigger detection at 10:00 am every day; after the control software receives the instruction, it reads the coordinate boundary of this area and automatically plans a parallel route from the northwest corner to the southeast corner of the farmland, with a flight altitude of 30 meters, carrying a high-resolution RGB camera and a multispectral camera; the UAV automatically takes off, flies and collects data according to the planned route.

[0041] The original or preliminary processed image data collected by the UAV during the detection process; depending on the sensors carried, these images can be: RGB visible light images: color photos that can be seen by the naked eye, which can directly observe the growth of crops, color, whether there is lodging, etc.; multispectral images: contain information of red, green, blue and near-infrared (NIR) and other specific wavebands, which are very useful for calculating vegetation index (such as NDVI, EVI), which can reflect the chlorophyll content, biomass, water stress of crops, etc., these data are usually transmitted to the ground station or cloud server in real time or quasi-real time.

[0042] Optionally, the UAV flies according to the planned flight, the RGB camera takes a photo every second, and the multispectral camera synchronously collects waveband data; after the flight is completed, the UAV transmits the hundreds of RGB photos, multispectral data and GPS positioning information back to the ground station; the ground station software preliminarily splices these photos into an orthomosaic image covering the entire farmland area, and the multispectral data generates false color images or data cubes of each waveband, which are the "corresponding actual images" or data.

[0043] The actual image, the actual area shape of the farmland and the response signal of the UAV to the crops are introduced; the actual area shape of the farmland refers to the shape and structure of the farmland boundary determined in step S111; understanding the overall layout of the farmland helps to locate the crops in the image and to distinguish the crops from the vegetation outside the farmland boundary.

[0044] Identify individual crop plants or crop patches from complex images or data and determine their precise location in the farm coordinate system (usually latitude and longitude or coordinates relative to a reference point in the farm); common methods include: computer vision-based image processing: such as thresholding, edge detection, morphological operations, feature extraction (such as color, texture, shape), etc., suitable for structured farmland (such as regularly planted corn, vegetables); or deep learning models: such as convolutional neural networks (CNN) and their variants (such as YOLO, SSD for object detection; U-Net, MaskR-CNN for instance segmentation), which require a large amount of labeled data for training, but can handle more complex scenarios and more subtle features, for example, training a U-Net model to segment RGB and multispectral images to identify the location and boundary of each corn plant.

[0045] Specifically, given an RGB orthoimage map and multispectral data covering the farmland; the actual area of the farmland tells that it is a roughly rectangular area with a straight road in the middle; use a pre-trained deep learning instance segmentation model (such as a U-Net model trained on a large corn dataset) to process the RGB image; the model will analyze each part of the image and identify green pixel regions that match the shape of a corn plant; for each identified region (i.e. each corn plant), the model will output a bounding box or pixel-level mask and record the center point coordinates of the region (e.g. X, Y coordinates relative to the southwest corner of the farm, or directly latitude and longitude); at the same time, the model will use NDVI values to assist in judgment and exclude areas where green is weeds; finally, a list is obtained, listing the precise location of each corn plant (or each small cluster) in the farm.

[0046] Therefore, based on the identification of the actual area of the farmland, the pipe path of the water pipe is determined, and the irrigation path of each crop is determined according to the pipe path of the water pipe and the location of each crop, which is compatible with the overall consideration of the pipe path of the water pipe and the location of each crop, and improves the accuracy of the irrigation path of each crop.

[0047] At this time, the actual area of the farmland is identified, and the farm boundary, internal features (such as roads, dikes, existing buildings, etc.) and terrain information (for example, a digital elevation model DEM obtained by unmanned aerial vehicle LiDAR scanning) determined in steps S111 and S112 are used. These information together constitute the geographical spatial layout of the farmland.

[0048] One or more water pipe laying routes connecting the water source (such as irrigation pumping station, water reservoir) and covering the entire farmland irrigation area need to be found, and multiple factors need to be considered when determining the path: efficiency and coverage, cost minimization, terrain factors, maintainability and water source location.

[0049] Optionally, the boundaries of the field are known, there is a north-south road in the middle, and the water source is located at the northwest corner of the field; when planning the pipeline path, first consider laying the main pipeline along the north-south road, because the road is usually flat and easy to construct; the main pipeline extends southward from the northwest corner water source along the road; then, in the areas that need to be irrigated (such as corn planting areas), branch pipelines are vertically drawn from the main pipeline and extend into the field; if there is a low-lying area in the middle of the field, consider setting up a water storage point there; the entire planning process can be completed in GIS software, visualizing the field boundaries, roads, water source points, planned main pipelines, and branch pipelines, and calculating the total length and cost; the final main pipeline is roughly along the road, while the branch pipelines are distributed in a grid or radial pattern to the corn planting areas.

[0050] The pipeline path of the water pipe and the location of each crop are introduced, and the location of each crop is the output of step S112, and the precise location of each plant (or each small cluster) is known; for each plant, the nearest pipeline point (which is a valve on the main pipeline, or a branch point of the branch pipeline, or even the location of a drip irrigation belt or sprinkler) needs to be found, which can be achieved through spatial analysis, such as calculating the distance from the crop location to all pipeline nodes and finding the minimum value; then, determine the valve or controller that controls the nearest pipeline point (or irrigation unit), so that a mapping relationship is established: crop A is controlled by valve X, and crop B is supplied by drip irrigation belt Y section; when the system needs to irrigate a specific area or specific crop, it needs to know which valves to open, which pipelines the water will flow through, and ultimately reach the target area; if precision irrigation (such as drip irrigation) is used, this directly corresponds to a specific dripper or drip irrigation section.

[0051] Reference Figure 3 In step S12, the specific steps are as follows:

[0052] S121: Collect the irrigation path of each crop, and determine the irrigation location of each crop based on the traversal of the irrigation path of each crop; determine the soil distribution map of the farmland according to the location of each crop and the irrigation location of each crop;

[0053] S122: In the soil distribution map of the farmland, mark the location of each crop, determine the corresponding soil area of each crop according to the location of each crop, the type of each crop, and the crop soil database, and determine a plurality of to-be-detected soil areas based on the area shape and location of the corresponding soil area of each crop. At this time, the same type of crop exists in the same to-be-detected soil area.

[0054] In an embodiment of the present application, the irrigation paths of various crops are collected, and calculations and analyses are performed on the digitized path data. Each irrigation path needs to be followed, starting from the water source, passing through various pipe sections and valves, and finally reaching the root zone of the crop. The specific location of the water flow when it reaches the root zone of the crop is accurately determined. This is usually the location of the crop itself (i.e., the coordinates determined in S112). However, more deeply, it can also be understood as determining the precise geographic coordinates of the "point" or "small area" where the irrigation water affects the root zone of the crop. This location is a key reference point for soil condition detection and subsequent irrigation control.

[0055] The most straightforward approach is to directly reference the crop position coordinates determined in S112, but more refined factors can also be considered, such as: Drip point location: For a drip irrigation system, the irrigation location is the coordinate of the point at which the end of the dripper or drip tape is closest to the center of the crop root zone;

[0056] Center of influence: If the irrigation method is sprinkler irrigation, the irrigation position can be considered as the coordinates of the center point of the influence area under the sprinkler;

[0057] Path end coordinates: In many cases, the end point coordinates of the irrigation path data directly represent the irrigation location.

[0058] Specifically, suppose there is a corn numbered "Corn-007" in the farmland; through S113, we know that it is supplied with water by a drip irrigation tape section controlled by valve V3 on the main pipeline; then, the collected irrigation path information of "Corn-007" is as follows: Path ID: Path_007; Starting point: water source pump station (coordinates: X=100, Y=100); Passing nodes: [main pipeline section A (starting point: X=100, Y=100; end point: X=200, Y=100), valve V3 (coordinates: X=200, Y=100), branch pipeline section B (starting point: X=200, Y=100; end point: X=220, Y=150), drip irrigation tape section C (starting point: X=220, Y=150; end point: X=230, Y=150)]; end point: center of the root zone of Corn-007 (coordinates: X=230, Y=150).

[0059] Traversing the data of path Path_007, we found that the coordinates of the end point of the path are (X=230, Y=150). This coordinate point is the irrigation position of "Corn-007". It is the coordinate of the point where the C end of the drip irrigation belt is closest to the center of the root zone of Corn-007, or the coordinate of the center of the root zone of Corn-007 itself.

[0060] The soil distribution map indicates the areas where crops grow and the areas where water is provided for these crops, laying the foundation for subsequent understanding of the relationship between soil conditions and crop growth, which is usually a geographic spatial data layer (e.g. created in GIS software).

[0061] Visualizing these points / small areas on the map and distinguishing different crops or irrigation areas with different colors or symbols forms a preliminary soil distribution map, which intuitively shows the layout of crop planting in the farmland and how irrigation facilities serve these crops. This soil distribution map can be further refined, for example, an influence radius or influence polygon can be drawn around each irrigation location (or crop location) to represent the importance of the water and nutrient conditions of the soil in that area to the growth of a particular crop. These influence areas can be overlaid on the map to form more complex soil management units.

[0062] Specifically, for "corn-007", it is known that its location is (X=230, Y=150) and the irrigation location is also (X=230, Y=150); on the soil distribution map, a point is marked at this coordinate point and noted as "corn-007"; then, this process is repeated for all corns (corn-001 to corn-N) in the field; finally, the soil distribution map will be filled with points representing different corns and their irrigation locations; if there are other crops (such as tomatoes) in the field, their points will be marked with different colors or symbols, and this map clearly shows the spatial distribution of different crops and their irrigation points in the entire farmland.

[0063] Further, on the soil distribution map (a map containing crop locations and irrigation locations) generated in S121, the exact coordinate points of each plant (or each cluster) of crop are clearly marked, and this marking is not just a dot, but usually accompanied by basic information of the crop, such as crop ID, type, etc.; the system will read the crop location data output by S112 and S121 and generate corresponding point features on the map layer; each point can have different colors or icons to distinguish crop types; clear marking makes subsequent analysis and processing more intuitive and convenient, allowing operators or algorithms to see the distribution of crops in the field at a glance, providing a clear starting point for the next step of delineating soil areas.

[0064] Specifically, on the soil distribution map generated in S121, there are already a bunch of points representing corns and their irrigation locations; now, in sub-step 1 of S122, each point is clearly marked on the map with the label "corn" or a green corn icon, ensuring that each point clearly represents a corn plant; if there are other crops in the field, such as rows of tomatoes, their location points will also be marked, represented by red tomato icons.

[0065] For each plant (or each cluster) of crops, a soil range that it depends on is demarcated, which is not randomly demarcated, but is based on the following three key information: a. Crop location: the specific coordinates of the crop in the field, which is the center or core of the region; b. Crop species: different species of crops have different root depth, breadth, and demand range for water and nutrients, for example, deep-rooted crops (such as corn) require a larger and deeper soil area than shallow-rooted crops (such as spinach); c. Crop soil database: a key resource that stores the soil conditions (such as optimal pH value, humidity range, nutrient demand) required by various crops at a specific growth stage and its typical root distribution pattern (such as range radius, depth); the database tells that for a crop such as "corn", the soil state within a certain range (such as 1 meter in diameter) around its location point is crucial for its growth.

[0066] Read the crop species information, query the database to obtain the root influence range parameters of the crop species (for example, a circular area with a radius of 0.5 meters; or an elliptical area with a major axis of 1 meter and a minor axis of 0.5 meters); then, with the crop location point as the center, draw the corresponding area shape on the map according to these parameters, and this area represents the key soil range that the crop depends on for growth; the output is to generate a corresponding soil area polygon (or geometric shape such as a circle) for each plant (or each cluster) of crops in the field.

[0067] Specifically, taking the location (X230, Y150) of corn-007 as the center, querying the crop soil database, it is found that corn needs a circular soil area with a diameter of about 1 meter to ensure normal growth of root system and absorption of water and nutrients; therefore, a circle with a radius of 0.5 meters is drawn on the map with (X230, Y150) as the center, and the soil area within this circle is the soil area corresponding to corn-007; repeating this process for all corns in the field will result in many circular areas around the location points of corns on the map.

[0068] The numerous small soil areas demarcated for individual crops in the previous step are merged and optimized to form several larger and more regular "to-be-detected soil areas", which has the advantage of not needing to detect each small area separately, but can be detected in combination to improve efficiency; adjacent areas of the same crop are merged to facilitate the formulation of detection plans and subsequent control measures.

[0069] Scan all single crop soil areas, find those adjacent to each other and the same crop, and merge them into a larger polygon, which requires multiple iterations until it can no longer be merged according to the rules; finally, a number of polygon areas with clear boundaries will be formed on the farmland soil distribution map, each containing only one crop; output multiple soil areas to be detected, each defined by a polygon and clearly marked with the crop contained.

[0070] Reference Figure 4 In step S13, the specific steps are:

[0071] S131: Real-time monitoring of each soil area to be detected, determining multiple soil data based on the detection of each soil area to be detected, and marking the soil detection position corresponding to each soil data;

[0072] S132: Collecting the maturity of each crop, determining the soil level map according to the multiple soil data, the corresponding soil detection position and the maturity corresponding to the crop, at this time, the soil level map presents the level state of the soil at different height layers, and marks the burial depth of each crop in the soil area to be detected;

[0073] S133: Determining the first sub-soil state coefficient according to the soil level map and the burial depth of each crop in the soil area to be detected, determining the second sub-soil state coefficient according to the soil level map and the past state data of the soil area to be detected, and determining the soil state of the soil area to be detected based on the first sub-soil state coefficient, the second sub-soil state coefficient and the soil state mapping relationship.

[0074] In the embodiments of the present application, in the previous example, there are corn area A, tomato area B and potato area C; monitoring needs to be carried out continuously or at fixed intervals; the frequency depends on the growth stage of the crop, the speed of environmental change and the available resources; optionally, unmanned aerial vehicles are used, which fly over these areas to be detected according to the predetermined route at regular intervals, and sample at the preset detection points, for example, the unmanned aerial vehicles fly over every day in the afternoon, and stay for 1 minute at 4 corners and the center point of area A for detection; the collected raw data needs to be transmitted to the central processing system or cloud platform in real time or quasi-real time, and stored for subsequent processing and analysis; data transmission can be carried out through wireless network (Wi-Fi, LoRa, NB-IoT, 4G / 5G).

[0075] The soil data items to be recorded are determined according to the detection purpose and available sensors. Common soil data includes: soil moisture content: volumetric water content (%) or mass water content (%); soil temperature: Celsius (°C); soil electrical conductivity (EC): reflecting salt content, usually in units of dS / m; soil pH value: reflecting the acid-base degree; soil nutrient content: such as nitrate nitrogen (NO3-N), ammonium nitrogen (NH4-N), available phosphorus (Olsen-P), available potassium (K), etc., usually in units of mg / kg or ppm; soil bulk density: g / cm³, reflecting the soil compactness; soil oxidation-reduction potential (Eh): reflecting the soil aeration condition, in units of mV.

[0076] At the same time of collecting data, the precise position of the data collection point must be recorded, which can be achieved by the following ways: the UAV usually has a GPS, which can record the longitude and latitude coordinates of each detection point; the soil data items obtained in the previous step are bound with the corresponding collection position coordinates, for example, the data of water content 25% must be clear that it is measured at which specific point (such as P1(11,11)) in the corn area A; the position information is stored in the database together with the soil data to form a "position-data" pair, for example: (P1(11,11), water content: 25%, temperature: 22C, EC: 0.3dS / m).

[0077] Specifically, the system identifies that the corn area A is the object to be monitored; it is decided to use 4 fixed sensors deployed in the area A (located at the center point P1(11,11) and P2(10,10), P3(12,10), P4(10,12), P5(12,12) near the four corners); the sensors are set to automatically collect data every 2 hours and send the data to the management center of the farm through the LoRa network.

[0078] At a certain time point (for example, 10 o'clock in the morning), the 5 sensors collect data at the same time; the P1(11,11) sensor reading: the voltage corresponds to the water content of about 25%, the temperature sensor reading 22C, the EC sensor reading corresponds to 0.3dS / m; the P2(10,10) sensor reading: water content 28%, temperature 21C, EC 0.35dS / m; the P3(12,10) sensor reading: water content 23%, temperature 23C, EC 0.28dS / m; the P4(10,12) sensor reading: water content 26%, temperature 22C, EC 0.32dS / m; the P5(12,12) sensor reading: water content 24%, temperature 22.5C, EC 0.30dS / m; these raw readings are converted into standardized soil data through simple filtering and calibration algorithms built-in the sensors.

[0079] After the system receives the sensor data, it automatically associates the data with its fixed location coordinates; the stored data records are as follows: (P1(11,11), moisture content: 25%, temperature: 22C, EC: 0.3 dS / m); (P2(10,10), moisture content: 28%, temperature: 21C, EC: 0.35 dS / m); (P3(12,10), moisture content: 23%, temperature: 23C, EC: 0.28 dS / m); (P4(10,12), moisture content: 26%, temperature: 22C, EC: 0.32 dS / m); (P5(12,12), moisture content: 24%, temperature: 22.5C, EC: 0.30 dS / m).

[0080] Further, the maturity of each crop is collected, and a soil level map is determined based on the multiple soil data, corresponding soil detection positions, and the maturity corresponding to the crops. At this time, the soil level map presents the level state of the soil at different height levels and marks the burial depth of each crop in the soil area to be detected, improving the accuracy of the soil level map.

[0081] At this time, the maturity of each crop is collected, and the collection of maturity can have multiple ways: visual recognition: using a camera (fixed or carried by a drone) to take pictures of the crops, and analyzing the color, size, shape, and other features through image processing algorithms to judge the maturity, for example, analyzing the fullness and color change of corn ears; spectral analysis: using multispectral or hyperspectral sensors to obtain the reflectance spectrum of crops, and crops at different maturity stages have characteristic performance at specific wavebands.

[0082] The collected maturity information needs to be converted into quantifiable indicators that the system can process, for example, a maturity coefficient of 0 to 1 can be set: 0: just emerged; 0.3: seedling stage; 0.6: growth period; 0.8: approaching maturity; 1.0: complete maturity; optionally, assuming that in area A, through visual recognition algorithm, the system judges that the current maturity coefficient of corn is 0.7, which is in the transition stage from growth period to approaching maturity period.

[0083] The soil data (such as moisture, temperature, EC value) collected in S131 with position information is combined with the current crop maturity information, for example, the soil data of area A is as follows: P1(11,11): moisture 25%, temperature 22C, EC 0.3; P2(10,10): moisture 28%, temperature 21C, EC 0.35; P3(12,10): moisture 23%, temperature 23C, EC 0.28; P4(10,12): moisture 26%, temperature 22C, EC 0.32; P5(12,12): moisture 24%, temperature 22.5C, EC 0.30; crop maturity: corn, maturity coefficient 0.7.

[0084] Soil properties typically vary with depth. Although the S131 only detects the surface or shallow layers, the "soil layer map" needs to reflect this vertical variation. Based on historical data, soil types, or crop root distribution models for the area, the system assumes a vertical variation pattern in soil properties. For example, it assumes that 0-20 cm below the surface is the cultivated layer, 20-40 cm is the transition layer, and below 40 cm is the bottom layer, and that the moisture and nutrients in the bottom layer are generally low. The integrated data (taking into account maturity and layer variations) is visualized and structured. The map is no longer a two-dimensional plane map, but a three-dimensional space or a layered two-dimensional map. The three-dimensional space can be a three-dimensional grid map with the XY axis representing the surface position and the Z axis representing the depth. Each grid point stores the soil properties (moisture, temperature, EC) at that location and depth. The layered two-dimensional map can be several two-dimensional maps, each representing the soil property distribution map (such as moisture distribution map and EC distribution map) at different depth layers (such as 0-20 cm, 20-40 cm, and >40 cm).

[0085] It is represented by a structured data set, including location coordinates, depth, and soil property values. For example, for area A, the following simplified "soil layer map" data fragment is generated (only the surface and the assumed bottom layer): Cultivated layer (0-20cm): P1 (11,11,0-20): moisture 25%, temperature 22C, EC0.3; P2 (10,10,0-20): moisture 28%, temperature 21C, EC0.35; Bottom layer (>40cm): P1 (11,11,>40): moisture ~15% (25%-10%), temperature assumed unchanged at 22C, EC assumed unchanged at 0.3; P2 (10,10,>40): moisture ~18%, temperature 21C, EC0.35. This map also implies information about crop maturity of 0.7, which is used in subsequent analysis to assess whether these soil conditions are suitable for the current growth stage.

[0086] The burial depth of crops (i.e., the seed sowing depth or the starting depth of the main root distribution) is usually a relatively fixed value, or can be obtained through planting records; the system can store typical sowing depths for different crop types, for example, the sowing depth of corn is usually 5-10 cm; or, the precise sowing depth is recorded at the time of planting; the obtained burial depth information is associated with the soil layer map, which usually means adding a mark to the crop in the area in the soil layer map to indicate the approximate depth at which its roots begin to grow. For example, in the soil layer map of area A, add a note: "Main root starting depth of corn: about 8 cm", which helps to understand the impact of soil surface and deep conditions on crop roots and absorption.

[0087] Specifically, the soil area A to be detected mainly plants corn; through the multi-spectral camera carried by the unmanned aerial vehicle to shoot the area A, the image processing algorithm analyzes and outputs the current maturity coefficient of corn as 0.7 (transition period from growth to maturity).

[0088] The system retrieves S131 the stored soil data (moisture, temperature, EC value of P1-P5) of the area A; the system applies a preset soil vertical stratification model: 0-15cm is the surface layer (tillage layer), 15-40cm is the middle layer, and >40cm is the bottom layer; it is assumed that through multi-layer sensors or interpolation, the data of each layer is obtained, for example, P1(11,11) has 25% moisture in the surface layer (0-15cm), 22% moisture in the middle layer (15-40cm), and 18% moisture in the bottom layer (>40cm); the temperature and EC value also have corresponding changes; the system integrates the multi-layer data of all points to generate a three-dimensional soil property distribution map, which can show the moisture, temperature, and EC value distribution of different positions and depths in the area A; at the same time, the current maturity coefficient 0.7 is implicitly or explicitly associated with the map.

[0089] The system queries the database and learns that the typical sowing depth of the corn variety planted in the area is 8cm; therefore, on the generated soil stratification map (whether in a visual interface or a data structure), it is explicitly marked: “Area A: corn, root initiation depth about 8cm”; after S132 processing, a detailed description of the soil area A to be detected is obtained:

[0090] Soil stratification map: a three-dimensional or stratified data structure / visualization map that shows the distribution of soil moisture, temperature, and EC value from the surface to the deep layer (such as 0-15cm, 15-40cm, >40cm) in the area A, for example, it can be seen that the surface soil moisture is relatively uniform in the range of 10-12, 10-12, and is about 23%-28%, while the bottom layer moisture is generally low, about 15%-20%;

[0091] Crop maturity: the current maturity coefficient of corn in the area A is 0.7; root information: corn roots mainly grow from about 8cm below the surface; this “soil stratification map” combines spatial (position), depth, soil property, and crop growth stage information, providing extremely rich basis for the next step (S133) of evaluating the soil state.

[0092] Therefore, the first sub-soil state coefficient is determined according to the soil layer chart and the embedding depth of each crop in the to-be-detected soil region, the second sub-soil state coefficient is determined according to the soil layer chart and the past state data of the to-be-detected soil region, and the soil state of the to-be-detected soil region is determined based on the first sub-soil state coefficient, the second sub-soil state coefficient and the soil state mapping relationship, so that the matching of the first sub-soil state coefficient, the second sub-soil state coefficient and the soil state mapping relationship is realized, the accuracy of the soil state of the to-be-detected soil region is improved, meanwhile, the plurality of to-be-detected soil regions are introduced, the overall consideration of the soil layer chart, the plurality of soil data and the past state data of the to-be-detected soil region is compatible, and the accuracy of the soil state of the to-be-detected soil region is improved.

[0093] At this time, the soil data related to the crop root activity layer is extracted from the soil layer chart generated in S132, for example, for the corn region A, the root mainly activities in the depth range of 0-40 cm (assuming), then the data such as water, temperature, EC value in this depth range needs to be extracted; at the same time, the embedding depth (sowing depth) of the crop and the approximate vertical range of the root system need to be determined.

[0094] The system needs to call the crop soil database to obtain the demand of corn at the current maturity (determined as 0.7 in S132, that is, close to the mature period) for the soil state of the root activity layer, for example, the database indicates that the ideal water range of the mature corn is 60%-80% of the field water capacity, the suitable root zone temperature is 20-25°C, and the suitable EC value is lower than 0.8 dS / m.

[0095] The actual detected soil state is compared with the ideal demand of the crop to calculate a deviation degree or a satisfaction degree index, which can be a simple ratio or a weighted calculation;

[0096] Water deviation degree: assuming that the average root zone water content detected is 25% (assuming that the field water capacity is 30%, then it accounts for 83%), and the ideal is 60%-80% (that is, 18%-24%); the actual value 83% exceeds the upper limit of the ideal, and the deviation degree is (83%-80%) / (80%-60%)=0.15 (15% exceeds); temperature deviation degree: assuming that the average root zone temperature detected is 22.2°C, and the ideal is 20-25°C; the temperature is within the ideal range, and the deviation degree is 0; EC deviation degree: assuming that the average root zone EC value detected is 0.32 dS / m, and the ideal is lower than 0.8 dS / m; the EC value is far lower than the upper limit, and the deviation degree can be considered as 0 (or a very small value).

[0097] The deviation / satisfaction of each soil property (water, temperature, EC, etc.) is weighted and integrated to obtain the first sub-soil state coefficient; the weight can be set according to the sensitivity of the current stage of the crop to each index, for example, mature corn is more sensitive to water.

[0098] Assuming the weight is: water 0.6, temperature 0.3, EC 0.1; the first sub-coefficient = (water deviation * 0.6) + (temperature deviation * 0.3) + (EC deviation * 0.1) = (0.15 * 0.6) + (0 * 0.3) + (0 * 0.1) = 0.09, this coefficient 0.09 indicates that the current root zone soil state deviates from the crop demand to a lower degree, but the water is slightly high.

[0099] Extract the soil state data (such as average water, temperature, EC value) at the current time from the soil layer chart; at the same time, extract the same kind of data (water, temperature, EC value) of the soil area to be detected (area A) in the past period (for example, in the past week, a month) from the historical database.

[0100] Analyze the difference between the current data and the historical data, calculate the change trend or volatility; at the same time, the change trend and volatility are introduced; change trend: compare the current average water (25%) with the average water in the past week (assuming 22%), calculate the change rate = (25%-22%) / 22%≈0.136 (13.6% increase); volatility: calculate the standard deviation of soil water in the past period to see if the data is stable; according to the change trend and volatility, evaluate the stability of the current soil state relative to history, or potential risks, for example, continuous rise in water indicates drainage problems or future waterlogging risk; temperature fluctuation affects crop growth.

[0101] Quantify the change trend, volatility and other factors, and combine a preset threshold or scoring rule to obtain the second sub-soil state coefficient, which reflects the dynamic change and stability of the soil state; assuming a simple rule is set: if the change rate of any index (water, temperature, EC) exceeds 10%, the second sub-coefficient increases by a fixed value (such as 0.2); the water change rate is 13.6%>10%, so the second sub-coefficient = base value (assuming 0) + 0.2 = 0.2, this coefficient 0.2 indicates that the current soil state has a more obvious change compared to history, and there is a certain dynamic risk.

[0102] Determine the soil state of the to-be-detected soil region based on the first sub-soil state coefficient, the second sub-soil state coefficient, and the soil state mapping relationship. At this time, the first sub-coefficient (0.09) obtained in the first step and the second sub-coefficient (0.2) obtained in the second step are integrated, which can be a simple addition or a more complex weighted integration; the integrated coefficient = the first sub-coefficient + the second sub-coefficient = 0.09 + 0.2 = 0.29; the system presets a "soil state mapping relationship", which is a lookup table, a classification rule set, or a more complex model, which defines the corresponding relationship between the integrated coefficient and the final soil state description, for example, the mapping relationship is as follows: integrated coefficient < 0.1: state excellent; 0.1 ≤ integrated coefficient < 0.3: state good; 0.3 ≤ integrated coefficient < 0.6: state general; 0.6 ≤ integrated coefficient < 0.8: state poor; integrated coefficient ≥ 0.8: state bad.

[0103] At this time, according to the integrated coefficient 0.29, the corresponding soil state is found in the mapping relationship; 0.29 falls within the interval 0.1 ≤ integrated coefficient < 0.3, therefore, the final soil state of the to-be-detected soil region A is determined to be "good", although the overall state is "good", but the system records that the water content is slightly high (the main source of the first sub-coefficient 0.09) and there is an upward trend in the recent period (the main source of the second sub-coefficient 0.2), which can be used as reference details when making decisions.

[0104] Reference Figure 5 In step S14, the specific steps are as follows:

[0105] S141: Mark two adjacent to-be-detected soil regions in the soil distribution map of the farmland, determine the corresponding transition soil region according to the positions and region shapes of the two adjacent to-be-detected soil regions, and the transition soil region is between the two adjacent to-be-detected soil regions;

[0106] S142: Collect a plurality of soil data corresponding to the transition soil region based on the detection of the transition soil region, determine the state of the transition soil region according to the plurality of soil data and the region shape of the transition soil region, sort the positions of each to-be-detected soil region and the corresponding transition soil region, and record the soil state of each to-be-detected soil region and the state of the transition soil region in sequence;

[0107] S143: Construct a corresponding state gradient map based on the soil state of each to-be-detected soil region and the state of the transition soil region, determine an abnormal state position according to the detection of the state gradient map, and determine an abnormal soil state according to the tracing of the abnormal state position. The abnormal soil state has a state change amount that exceeds a preset state change amount threshold from the soil state of the surrounding detected soil region.

[0108] In the embodiments of the present application, the system needs to obtain the farmland soil distribution map generated in the previous steps (S12, S13), which has marked a plurality of to-be-detected soil regions (determined in S12) and their approximate positions and boundaries; the system needs to analyze the spatial relationship between these to-be-detected soil regions, which is usually achieved by calculating the distance between regions or the length of shared boundaries; two regions can be considered adjacent if they share a boundary or the center points are very close (for example, less than a certain preset threshold, such as 10 meters); according to the research purpose or preset rules, the system will select one or more pairs of adjacent to-be-detected soil regions for marking; the selection criteria include: large differences in region status (for example, one region is fertile and the other is barren), special region morphology (for example, tortuous boundaries), or located in key positions (for example, near irrigation channels or roads); marking can be achieved by highlighting the two regions on the map and assigning them specific identifiers (such as region A and region B).

[0109] The system extracts the precise boundary coordinates of the two marked adjacent regions (such as region A and region B) and their respective geometric shape features (for example, rectangular, circular, polygonal, straight or curved boundaries, etc.); the system focuses on analyzing the boundary line or the closest boundary segment shared by the two regions; understand the direction, length and shape of the boundary line; based on the analysis of the junction, the system needs to define a buffer region between region A and region B, which is the transition soil region;

[0110] According to the shape of the boundary line or the importance of the region, the width of the buffer zone is dynamically adjusted, for example, in places where the boundary is tortuous or steep, the buffer zone can be wider; in places where the boundary is straight, it can be narrower; if the boundaries of the two regions are not directly connected (for example, there is a small gap in the middle), the transition region can be defined as the region connecting the closest boundary points of the two regions.

[0111] The boundaries of this transition soil region are precisely delineated on the soil distribution map, and it is assigned an identifier (such as transition region AB); this region is logically completely between region A and region B and does not contain the main part of either region.

[0112] Further, based on the detection of the transition soil region, a plurality of soil data corresponding to the transition soil region are collected, the state of the transition soil region is determined according to the plurality of soil data and the region morphology of the transition soil region, the positions of the to-be-detected soil regions and the corresponding transition soil regions are sorted, and the soil states of the to-be-detected soil regions and the states of the transition soil regions are recorded in sequence, introducing the soil states of the to-be-detected soil regions and the states of the transition soil regions.

[0113] At this time, the system first needs to determine which soil data needs to be collected in the transition soil area, which is usually based on pre-set monitoring indicators, such as soil moisture, soil temperature, pH value, electrical conductivity (EC, reflecting salt or nutrients), soil compaction, etc.; the density and depth of detection depend on the size and shape of the transition area and research needs, for example, for a long and narrow transition zone, points need to be placed at equal intervals along the line; for a slightly larger transition area, points need to be placed in a grid.

[0114] Data is collected in the transition soil area using sensor networks deployed in farmland (such as buried sensors, surface mobile sensors), unmanned aerial vehicles equipped with remote sensing equipment (such as thermal imaging, multispectral), or manual sampling and laboratory analysis, etc. The data needs to accurately record its collection location (latitude and longitude or coordinates relative to the farmland coordinate system); the collected data is associated with the corresponding collection location to form a data set of the transition soil area, for example, point 1 (coordinates X1, Y1) has a humidity of 30%, point 2 (coordinates X2, Y2) has a pH value of 6.5, etc.

[0115] The system needs to arrange the detected soil area and its adjacent transition soil area according to a certain logical order; the basis for sorting can be: geographical location (for example, from north to south, from west to east), or relative distance from a reference point (such as the entrance of the farmland, the starting point of the irrigation main pipe), or according to the order marked in S141; the purpose of sorting is to ensure consistency and systematicness of subsequent data recording and gradient analysis (S143).

[0116] According to the sorted order, the system records the soil state of each detected soil area (from S133) and its adjacent transition soil area (from sub-step 2 of this step) in turn; the record can be stored in a database or a structured report can be generated; the record needs to clearly mark the region number, transition region number and their location information.

[0117] Specifically, assume that in S141, region A and region B have been marked and the transition area AB (a 3-meter-wide, 50-meter-long strip) between them has been determined; the system decides to set a detection point every 5 meters along the length direction (east-west) in the transition area AB, a total of 11 points (including the starting point and the ending point); at each point, a sensor is buried and soil humidity is measured; assume that the measured data is as follows (simplified): point 1 (the westernmost end, close to A): humidity 35%; 2: humidity 34%; … point 6 (middle point): humidity 28%; … point 11 (the easternmost end, close to B): humidity 22%; after data collection, the system records the coordinates of each point and the corresponding humidity value.

[0118] The system analyzes the humidity data of these 11 points; it can be seen that the humidity shows a clear downward trend from west to east (from A to B); by calculating the average value (for example, about 30%), and drawing a graph of humidity change with distance, the system finds that the humidity decreases most rapidly near the middle point; combined with the fact that the transition area AB is a strip, the system judges that the overall state of the transition area AB is "humidity gradient significantly decreased area"; it can be further quantified, for example, calculate the humidity gradient (for example, (35%-22%) / 50 meters=0.26% / meter), or give a state coefficient (for example, based on the average value and gradient of humidity, the score is 0.7, indicating that the humidity is moderate but changes sharply).

[0119] The system decides to sort in the order from west to east (from area A to area B); the sorting sequence is: area A≥ transition area AB≥ area B; the system records in turn: area A: soil state coefficient 0.85 (from S133); transition area AB: state description "humidity gradient significantly decreased area", state coefficient 0.7 (from this step sub-step 2); area B: soil state coefficient 0.6 (from S133), these records are stored, ready for analysis in the next step S143.

[0120] Therefore, based on the soil state of each to-be-detected soil area and the state of the transition soil area, a corresponding state gradient graph is constructed, the abnormal state position is determined according to the detection of the state gradient graph, and the abnormal soil state is determined according to the tracing of the abnormal state position. The abnormal soil state and the state change amount of the soil state of the surrounding to-be-detected soil area exceed the preset state change amount threshold, ensuring the accuracy of the abnormal soil state.

[0121] At this time, the system has obtained the state information of all to-be-detected soil areas and their adjacent transition areas from S133 and S142, which are numerical state coefficients (such as 0.85, 0.7, 0.6) and classified state descriptions (such as good, general, and poor); in order to construct the gradient graph, it is usually necessary to convert these information into numerical values that can be represented continuously or semi-continuously in space, ensuring that each state value is associated with its corresponding geographical spatial position; the to-be-detected areas have clear boundaries, and the transition areas are located between them; the system needs to map the state values to these areas.

[0122] Use geographic information system (GIS) software or custom drawing algorithm to draw the state values on the farmland map; for to-be-detected areas, you can fill the entire area; for transition areas, you can draw their state values (an average value or a representative value) at the center point or along the line of the transition area; connecting these points, you can generate a curve or surface representing the change of state, which is the state gradient graph.

[0123] Optionally, the system plots the state coefficient of region A 0.85, the state coefficient of transition region AB 0.7, and the state coefficient of region B 0.6 on the farmland map according to their positions (region A is in the west, transition region AB is in the middle, and region B is in the east); connecting these three points, a curve gradually declining from west to east is obtained, which is the state gradient graph, and this curve intuitively shows the trend of change of the soil state (in this case, mainly reflecting humidity) from region A to region B.

[0124] The system analyzes the rate of change or change pattern of the state value on the state gradient graph; under normal circumstances, the state change between adjacent regions should be relatively gentle or conform to a certain pattern (for example, uniformly declining from high to low); the system calculates the state change between two adjacent points (representing two adjacent to-be-detected regions or a to-be-detected region and a transition region) on the state gradient graph, which can be simply obtained by subtracting the value of the former point from the value of the latter point (pay attention to the direction, and take the absolute value); the system compares the calculated change with a pre-set "state change threshold", which is set according to experience, historical data or crop growth requirements, and represents an acceptable state change range; if the change at a certain position (usually a transition region or the boundary of the to-be-detected region adjacent thereto) exceeds the threshold, the system marks it as a potential abnormal state position.

[0125] Once the abnormal state position (usually a certain transition region or the boundary of the to-be-detected region adjacent thereto) is determined, the system needs to "trace back" to the soil itself corresponding to the position; the system reviews the detailed soil data (humidity, temperature, pH, EC, etc.) collected and analyzed for the transition region in S132 and S142, as well as the calculated state description, which are the specific manifestations of the abnormal state; the traced back detailed soil data and state description are defined as "abnormal soil state", which is not only a "abnormal" label, but also a description containing specific information, for example, "transition region AB has an abnormality, which is manifested as a sharp decrease in soil humidity (gradient 0.26% / m) and a high pH value (average 7.8)".

[0126] Specifically, if the transition region AB is marked as an abnormal position, the system reviews the analysis result of the transition region AB in S142: a significant decrease in humidity gradient, a state coefficient of 0.7; combined with the states of regions A and B (assuming that region A has good humidity and region B has general humidity), the system can determine that the abnormal soil state is: "in the transition region AB, the soil humidity shows an abnormally rapid downward trend, accompanied by a high pH value", which is more instructive than a simple "abnormality".

[0127] Reference Figure 6 In step S15, the specific steps are as follows:

[0128] S151: Collect an abnormal soil state, and mark a corresponding abnormal position according to the traceability of the abnormal soil state, take the abnormal position as an abnormal soil region, and mark the position of the abnormal soil region;

[0129] S152: Determine a first abnormal state coefficient according to the position of the abnormal soil region and the abnormal soil state, and determine a second abnormal state coefficient according to the abnormal soil state of the abnormal soil region and the soil state of a neighboring soil region to be detected;

[0130] S153: Determine an intelligent control measure according to the first abnormal state coefficient, the second abnormal state coefficient, and a control measure mapping relationship, at this time, the intelligent control measure covers a state repair event for the abnormal soil region, and the corresponding state repair event is determined according to the maturity of the crop corresponding to the abnormal soil region, the current time, and the abnormal soil state, and the intelligent control of the abnormal soil region is triggered based on the state repair event.

[0131] In the embodiment of the present application, the abnormal soil state is collected, S143 determines the specific position of the abnormal state in the farmland at the same time, and the position information is a geographic coordinate (latitude and longitude), a grid number (for example, after the farmland is divided into 10x10 grids, the grid in the 5th row and the 7th column), or a region identifier, this step is to use the traceability information provided by S143 to accurately associate the abnormal state with the actual position where the abnormal state occurs, and to clearly mark in the digital map or record of the farmland.

[0132] When the soil is detected, the entire farmland is usually divided into a plurality of "soil regions to be detected" (regular grids or irregular blocks), this step is to attribute the abnormal position (a point or a small range) just marked to the "soil region to be detected" where it is located; once the attribution is determined, the "soil region to be detected" is redefined as an "abnormal soil region", which means that it is marked as a region that needs special attention and processing; at the same time, the identifier of the "abnormal soil region" or the geographic range covered thereby needs to be clearly recorded.

[0133] Further, the first abnormal state coefficient is determined according to the position of the abnormal soil region and the abnormal soil state, and the second abnormal state coefficient is determined according to the abnormal soil state of the abnormal soil region and the soil state of the neighboring soil region to be detected, and the first abnormal state coefficient and the second abnormal state coefficient are introduced.

[0134] At this time, the severity of the abnormal soil region itself needs to be quantified, which needs to consider two aspects: a) the abnormal soil state: this refers to the specific nature and degree of the abnormality, for example, is it extremely dry, severely salinized, or severely lacking in a certain key nutrient (such as nitrogen, phosphorus, potassium); the severity of the state (such as how low the water content is, how high the salt concentration is) is a key factor in calculating the coefficient; b) the location of the abnormal soil region: in some cases, the location of the region will affect its priority or the ease of treatment, for example, a region located in the core planting area of a farmland, near a water source or a road, needs a faster response or easier implementation of control measures than a region located in the edge, a remote corner; the location factor can be used to adjust the coefficient to reflect the importance or urgency of the location.

[0135] The determination of the coefficient is usually based on pre-set rules, empirical models or machine learning algorithms, for example, a scoring system can be established: the more the water content is below the threshold, the higher the score; the higher the degree of salinization, the higher the score; the more central the location, the higher the score; the weighted sum of these scores or the conversion through a certain function gives the first abnormal state coefficient, which represents the severity and urgency of the problem of the abnormal region itself.

[0136] Specifically, the abnormal soil region is plot 7, and the abnormal state is extremely low soil moisture (0.15), which is severe; now calculate the first abnormal state coefficient: a) abnormal state: the system sets that soil moisture below 0.2 is abnormal, and each 0.05 lower is one level more severe; 0.15 is 0.05 lower than 0.2, which belongs to the first level of abnormality; another 0.05 lower (i.e. 0.15) to 0.10 belongs to the second level; another 0.05 lower to 0.05 belongs to the third level; 0.15 belongs to the second level of abnormality; the system also sets that for the second level of abnormality, the basic score is 60 points; b) abnormal location: plot 7 is located in the central area of the farmland, which is the main planting area; the system sets that the location weight of the core area is 1.2; c) calculate the first abnormal state coefficient: basic score (60 points) * location weight (1.2) = 72 points; therefore, the first abnormal state coefficient is 72, which indicates that the drought problem of plot 7 is not only very severe (second level) but also needs to be highly concerned because it is located in the core area.

[0137] To quantify the impact of the abnormal region on the surrounding environment, or to evaluate the degree of isolation of the abnormality, it needs to compare: a) the state of the abnormal soil region: that is, the known abnormal state (such as the soil moisture of plot 7 is 0.15); b) the state of the adjacent soil region to be detected: the system needs to obtain the current soil state (especially the indicators related to the abnormal state, such as soil moisture) of several to-be-detected regions (for example, plot 6, plot 8, plot 10) that are adjacent to the abnormal soil region (plot 7); assume that the moisture of plot 6 is 0.35, the moisture of plot 8 is 0.38, and the moisture of plot 10 is 0.32.

[0138] The determination of the coefficient is usually based on the difference between the abnormal state and the adjacent state; the greater the difference, the more isolated the abnormality, or the clearer the boundary of its influence, or the lower the risk of spread (depending on the specific application scenario); on the contrary, if the state of the adjacent area is also very poor, it means that the problem area is larger, or the abnormality has a tendency to spread, and this coefficient can reflect the boundary characteristics of the abnormality or the potential risk of chain reaction.

[0139] Specifically, the humidity of the abnormal soil area field 7 is 0.15, and the humidities of the adjacent fields 6, 8, and 10 are 0.35, 0.38, and 0.32, respectively; a) abnormal state: field 7 humidity 0.15; b) adjacent state: average humidity = (0.35 + 0.38 + 0.32) / 3 = 0.35; c) calculate the second abnormal state coefficient: the system is set to calculate the absolute value of the difference between the abnormal area state and the average state of the adjacent area, and multiply it by a basic score (for example, a basic score of 50); the absolute value of the difference is |0.15-0.35|=0.20; the system sets a conversion function, such as multiplying the difference by 250 (i.e. 0.20*250=50); or more simply, the difference can be directly multiplied by a larger base, such as 0.20*250=50; or set a nonlinear function, such as squaring the difference and multiplying it by a certain coefficient; assuming the method of multiplying the difference by 250 is adopted, the second abnormal state coefficient is 50, which indicates that the soil humidity of field 7 is significantly different from its surroundings (0.2 difference), and this difference is quantified as 50 points, reflecting that the boundary of the abnormality is very clear, or the surrounding area is relatively healthy.

[0140] Therefore, according to the first abnormal state coefficient, the second abnormal state coefficient, and the control measure mapping relationship, the intelligent control measure is determined, at this time, the intelligent control measure covers the state repair event of the abnormal soil area, and the corresponding state repair event is determined according to the maturity of the crop corresponding to the abnormal soil area, the current time, and the abnormal soil state, and the intelligent control of the abnormal soil area is triggered based on the state repair event, which takes into account the maturity of the crop corresponding to the abnormal soil area, the current time, and the abnormal soil state, improves the accuracy of the state repair event, and at the same time, ensures the accuracy of the abnormal soil state, realizes the overall consideration of the position of the abnormal soil area, the abnormal soil state, and the soil state of the adjacent soil area to be detected, improves the accuracy of the intelligent control measure, and realizes the intelligent control of the abnormal soil area.

[0141] At this time, the system needs to integrate the two coefficients calculated in S152, and combine a pre-defined "control measure mapping relationship" (which can be regarded as a rule base, decision tree, or machine learning model) to select the most appropriate control measure.

[0142] The system combines the first anomaly state coefficient (reflecting the severity and location importance of the anomaly itself) and the second anomaly state coefficient (reflecting the difference degree of the anomaly from the surrounding), for example, if the first coefficient is very high (indicating that the anomaly is very serious and location critical), even if the second coefficient is not high, strong measures need to be taken immediately; conversely, if the second coefficient is very high (indicating a large difference from the surrounding), attention needs to be paid to the boundary processing of the anomaly.

[0143] Control measure mapping relationship This is a core database or logic module that defines which control measures should be taken under different coefficient combinations, for example: IF (first coefficient > 80 AND second coefficient > 40) THEN measure = "emergency irrigation + foliage spraying regulator"; IF (first coefficient > 60 AND second coefficient <= 30) THEN measure = "standard irrigation"; IF (first coefficient <= 50 AND second coefficient > 50) THEN measure = "boundary monitoring + fine-tuning irrigation"; IF (first coefficient <= 40 AND second coefficient <= 40) THEN measure = "record observation, no intervention for the time being"; The system looks up or calculates a preliminary intelligent control measure in the mapping relationship according to the current coefficient value, which is usually a combination of one or more operations aimed at repairing the abnormal soil state.

[0144] Intelligent control measures cover state repair events for abnormal soil areas; state repair events refer to actions that can directly change or improve the state of abnormal soil areas, such as: irrigation: increase soil moisture; fertilization: supplement soil nutrients; adjust pH: improve soil acidity and alkalinity; spread improver: improve soil structure; drainage: reduce soil moisture (if the anomaly is over-wet); spread pesticides / herbicides: treat diseases, pests, and weeds caused by soil problems.

[0145] Refine, adjust, and specify the preliminary control measures to make them more accurate and adaptive to the current actual situation, which introduces three key context information:

[0146] Maturity of the crop: The growth stage of the crop (such as seedling stage, growth stage, flowering stage, fruiting stage, and maturity stage) greatly affects its water and nutrient needs and sensitivity, for example: lack of water in the seedling stage leads to seedling death, requiring rapid and sufficient irrigation; lack of water in the flowering stage leads to poor pollination, requiring stable water supply; lack of water in the maturity stage (such as grain filling stage) affects yield and quality, requiring continuous water supply; water needs to be controlled to enhance flavor in the maturity stage (such as before fruit picking), even if the soil is slightly dry, do not irrigate immediately.

[0147] Current time: The time of day can affect the effectiveness and efficiency of control measures, for example: Irrigation: usually done in the morning or late afternoon to reduce evaporation losses and avoid over-wetting at night which can lead to disease; if done at noon, water evaporates quickly and the effect is poor; Fertilization: certain fertilizers work better when applied at specific times, or need to be avoided during periods of high temperature.

[0148] Abnormal soil state: While S152 has determined the abnormal state, more precise state information is needed here to fine-tune the remediation event, for example, is it water deficit, or nitrogen deficit, or both, and to what extent is the water deficit (mild, moderate, severe), which affects the irrigation volume, the type of fertilizer, and the amount of fertilizer.

[0149] Based on the above information, the system translates the preliminary control measures into specific, executable state remediation events; the system generates control instructions based on the specific state remediation events determined in the previous step, and triggers these devices to perform corresponding operations through the interface with the farmland automation equipment (such as irrigation systems, fertilizer machines, environmental monitoring equipment, etc.); At the same time, the system sends instructions to the automated irrigation controller: "Start the drip irrigation system of field 7 at 4 pm, run for 1.5 hours, flow rate to 100 cubic meters / ha;" The system also continuously monitors the state changes of the abnormal soil area (such as soil moisture sensor readings) to evaluate the effectiveness of the control measures and make adjustments or trigger subsequent measures if necessary.

[0150] Please refer to Figure 7 , Figure 7 is a structural composition diagram of the intelligent control system of the soil state of the farmland in the embodiment of the application; the intelligent control system of the soil state of the farmland comprises:

[0151] The crop module 21 is used to determine the actual area of the farmland according to the remote sensing image of the farmland, and mark the positions and irrigation paths of each crop in the actual area of the farmland;

[0152] The soil area to be detected module 22 is used to determine a soil distribution map of the farmland according to the positions and irrigation paths of each crop, and determine a plurality of soil areas to be detected based on the soil distribution map of the farmland and the positions of each crop;

[0153] The first state module 23 is used to determine a corresponding soil level map and a plurality of soil data in each soil area to be detected based on the soil area to be detected, determine the soil state of the soil area to be detected according to the soil level map, the plurality of soil data, and the past state data of the soil area to be detected;

[0154] The second state module 24 is configured to determine a transition soil region based on two adjacent to-be-detected soil regions, mark a state of the transition soil region, and determine an abnormal soil state based on a position of each to-be-detected soil region, a soil state, and the state of the corresponding transition soil region.

[0155] The intelligent control module 25 is configured to determine a corresponding abnormal soil region based on a trace of the abnormal soil state, determine an intelligent control measure based on a position of the abnormal soil region, the abnormal soil state, and a soil state of an adjacent to-be-detected soil region, and perform intelligent control on the abnormal soil region.

[0156] Any combination of the technical features in the above embodiments is possible. In order to make the description concise, not all combinations of the technical features in the above embodiments are described, however, as long as the combinations of the technical features do not exist, they should be considered as the scope of the present disclosure.

Claims

1. An intelligent control method for soil conditions in farmland, characterized in that: include: Determine the actual area of ​​the farmland based on the remote sensing image of the farmland, and mark the location of each crop and the irrigation path in the actual area of ​​the farmland; Determine a soil distribution map of the farmland according to the location of each crop and the irrigation path, and determine a plurality of soil areas to be tested based on the soil distribution map of the farmland and the location of each crop; In each soil area to be detected, a corresponding soil level map and multiple soil data are determined based on the soil area to be detected, and the soil state of the soil area to be detected is determined according to the soil level map, the multiple soil data and the previous state data of the soil area to be detected, including: real-time monitoring of each soil area to be detected, determining multiple soil data based on the detection of each soil area to be detected, and marking the soil detection position corresponding to each soil data; collecting the maturity of each crop, determining a soil level map based on the multiple soil data, the corresponding soil detection position and the maturity corresponding to the crop, at this time, the soil level map presents the hierarchical state of the soil at different height layers, and marks the burial depth of each crop in the soil area to be detected; determining a first sub-soil state coefficient according to the soil level map and the burial depth of each crop in the soil area to be detected, determining a second sub-soil state coefficient according to the soil level map and the previous state data of the soil area to be detected, and determining the soil state of the soil area to be detected based on the first sub-soil state coefficient, the second sub-soil state coefficient and the soil state mapping relationship; Determine a transition soil area based on two adjacent soil areas to be detected, mark the state of the transition soil area, and determine an abnormal soil state based on the position, soil state and state of each soil area to be detected, including: marking two adjacent soil areas to be detected in a soil distribution map of the farmland, determining a corresponding transition soil area based on the position and regional morphology of the two adjacent soil areas to be detected, wherein the transition soil area is located between the two adjacent soil areas to be detected; collect corresponding multiple soil data based on the detection of the transition soil area, determine the state of the transition soil area based on the multiple soil data and the regional morphology of the transition soil area, sort the positions of each soil area to be detected and the corresponding transition soil area, and record the soil state of each soil area to be detected and the state of the transition soil area in sequence; construct a corresponding state gradient map based on the soil state of each soil area to be detected and the state of the transition soil area, determine the abnormal state position based on the detection of the state gradient map, determine the abnormal soil state based on the tracing of the abnormal state position, and the state change amount between the abnormal soil state and the soil state of the surrounding detection soil area exceeds a preset state change amount threshold; Based on the tracing of abnormal soil conditions, the corresponding abnormal soil area is determined, and intelligent control measures are determined according to the location of the abnormal soil area, the abnormal soil condition and the soil condition of the adjacent soil area to be detected, and the abnormal soil area is intelligently controlled, including: determining a first abnormal condition coefficient according to the location and abnormal soil condition of the abnormal soil area, and determining a second abnormal condition coefficient according to the abnormal soil condition of the abnormal soil area and the soil condition of the adjacent soil area to be detected; determining intelligent control measures according to the mapping relationship between the first abnormal condition coefficient, the second abnormal condition coefficient and the control measures. At this time, the intelligent control measures cover the condition repair events of the abnormal soil area, and determine the corresponding condition repair events according to the maturity of the crops corresponding to the abnormal soil area, the current time and the abnormal soil condition, and trigger the intelligent control of the abnormal soil area based on the condition repair events.

2. The intelligent control method for soil conditions of farmland according to claim 1, characterized in that: The method of determining the actual area of ​​the farmland based on the remote sensing image of the farmland and marking the location and irrigation path of each crop in the actual area of ​​the farmland includes: Collecting remote sensing images of farmland, determining an actual distribution map of the farmland based on pre-processing of the remote sensing images of the farmland, and determining an actual area of ​​the farmland based on area recognition of the actual distribution map of the farmland; In the actual area of ​​the farmland, the drone is triggered to detect the actual area of ​​the farmland based on the position of the actual area of ​​the farmland, and a corresponding actual image is output. The position of each crop is determined according to the actual image, the regional morphology of the actual area of ​​the farmland, and the response signal of the drone to the crop; The pipeline path of the water pipe is determined based on the recognition of the actual area of ​​the farmland, and the irrigation path of each crop is determined according to the pipeline path of the water pipe and the position of each crop.

3. The intelligent control method for soil conditions of farmland according to claim 1, characterized in that: The method of determining a soil distribution map of the farmland according to the location of each crop and the irrigation path, and determining a plurality of soil areas to be detected based on the soil distribution map of the farmland and the location of each crop, includes: Collecting irrigation paths for each crop, and determining irrigation positions for each crop based on traversal of the irrigation paths for each crop, and determining a soil distribution map of the farmland based on the positions of each crop and the irrigation positions of each crop; In the soil distribution map of the farmland, the position of each crop is marked, and the soil area corresponding to each crop is determined according to the position of each crop, the type of each crop and the crop soil database. Based on the regional morphology and position of the soil area corresponding to each crop, multiple soil areas to be detected are determined. At this time, the same type of crops exist in the same soil area to be detected.

4. The intelligent control method for soil conditions of farmland according to claim 1, characterized in that: The method of determining a corresponding abnormal soil area based on tracing the abnormal soil state, determining intelligent control measures according to the location of the abnormal soil area, the abnormal soil state, and the soil state of the adjacent soil area to be detected, and performing intelligent control on the abnormal soil area includes: Abnormal soil conditions are collected, and corresponding abnormal positions are marked according to the tracing of the abnormal soil conditions. The area to be detected corresponding to the abnormal position is used as the abnormal soil area, and the position of the abnormal soil area is marked.

5. An intelligent control system for soil conditions in farmland, characterized in that: The intelligent control system for the soil state of a farmland is applied to the intelligent control method for the soil state of a farmland as claimed in any one of claims 1 to 4, and the intelligent control system for the soil state of a farmland comprises: A crop module is used to determine the actual area of ​​the farmland based on the remote sensing image of the farmland, and mark the location and irrigation path of each crop in the actual area of ​​the farmland; A soil region module to be detected is used to determine a soil distribution map of the farmland according to the location of each crop and the irrigation path, and to determine multiple soil regions to be detected based on the soil distribution map of the farmland and the location of each crop; A first state module is configured to determine, in each soil area to be detected, a corresponding soil layer map and a plurality of soil data based on the soil area to be detected, and determine the soil state of the soil area to be detected based on the soil layer map, the plurality of soil data, and the previous state data of the soil area to be detected, including: real-time monitoring of each soil area to be detected, determining a plurality of soil data based on the detection of each soil area to be detected, and marking the soil detection position corresponding to each soil data; collecting the maturity of each crop, determining a soil layer map based on the plurality of soil data, the corresponding soil detection position, and the maturity corresponding to the crop, wherein the soil layer map presents the hierarchical state of the soil at different height layers, and marks the burial depth of each crop in the soil area to be detected; determining a first sub-soil state coefficient based on the soil layer map and the burial depth of each crop in the soil area to be detected, determining a second sub-soil state coefficient based on the soil layer map and the previous state data of the soil area to be detected, and determining the soil state of the soil area to be detected based on the first sub-soil state coefficient, the second sub-soil state coefficient, and the soil state mapping relationship; The second state module is used to determine the transition soil area based on two adjacent soil areas to be detected, mark the state of the transition soil area, and determine the abnormal soil state based on the position, soil state and state of each soil area to be detected. The module includes: marking two adjacent soil areas to be detected in the soil distribution map of the farmland, determining the corresponding transition soil area according to the position and regional morphology of the two adjacent soil areas to be detected, and the transition soil area is located between the two adjacent soil areas to be detected; collecting corresponding multiple soil data based on the detection of the transition soil area, determining the state of the transition soil area according to the multiple soil data and the regional morphology of the transition soil area, sorting the positions of each soil area to be detected and the corresponding transition soil area, and recording the soil state of each soil area to be detected and the state of the transition soil area in sequence; constructing a corresponding state gradient map based on the soil state of each soil area to be detected and the state of the transition soil area, determining the abnormal state position according to the detection of the state gradient map, determining the abnormal soil state according to the tracing of the abnormal state position, and the state change between the abnormal soil state and the soil state of the surrounding detection soil areas exceeds a preset state change threshold; An intelligent control module is used to determine the corresponding abnormal soil area based on the tracing of abnormal soil conditions, determine intelligent control measures according to the location of the abnormal soil area, the abnormal soil condition and the soil condition of the adjacent soil area to be detected, and perform intelligent control on the abnormal soil area, including: determining a first abnormal condition coefficient according to the location and abnormal soil condition of the abnormal soil area, determining a second abnormal condition coefficient according to the abnormal soil condition of the abnormal soil area and the soil condition of the adjacent soil area to be detected; determining the intelligent control measures according to the mapping relationship between the first abnormal condition coefficient, the second abnormal condition coefficient and the control measures. At this time, the intelligent control measures cover the state repair event of the abnormal soil area, and determine the corresponding state repair event according to the maturity of the crops corresponding to the abnormal soil area, the current time and the abnormal soil condition, and trigger the intelligent control of the abnormal soil area based on the state repair event.

Images