Farmland intelligent planting method and system based on Internet of Things

Abstract

The invention discloses an intelligent farmland planting method and system based on the Internet of Things, and relates to the technical field of the Internet of Things, and the method comprises the steps: marking the maturity degree of each crop; in the Internet of Things, a sub-monitoring area is determined according to the ripening degree of each crop and the position of each crop, abnormal growth characteristics of the crops in the sub-monitoring area are marked, and a planting dynamic graph of the farmland is determined based on each abnormal growth characteristic and the ripening degree of each crop. The accuracy of the planting dynamic graph of the farmland is improved; therefore, an abnormal planting area is determined according to each abnormal growth characteristic and the type of the crop, and a plurality of intelligent planting nodes and corresponding planting optimization measures are marked; the planting optimization event of each intelligent planting node is constructed based on each intelligent planting node and the Internet of Things, the accuracy of the planting optimization event of each intelligent planting node is improved, and the dynamic optimization state diagram of each abnormal growth feature is generated.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) technology, and more particularly to an intelligent planting method and system for farmland based on IoT. Background Technology

[0002] With the development of technology, farmland is now arranged in continuous sheets with multiple planting areas, each containing at least one crop. Generally, multiple different types of crops may also exist in the same planting area. In existing technologies, monitoring the farmland, observing the growth of each crop in each planting area, detecting anomalies in each planting area, and marking abnormal growth characteristics do not take into account the maturity of each crop. This affects the accuracy of the farmland planting dynamic map and makes it impossible to generate a dynamic optimization state map of each abnormal growth characteristic. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides an intelligent planting method and system for farmland based on the Internet of Things.

[0004] This invention provides an intelligent planting method for farmland based on the Internet of Things (IoT), comprising: determining multiple planting areas according to a farmland distribution map; determining the types and distribution of crops in each planting area based on detection of the planting area; determining a detection path for each planting area based on the distribution of crops and the regional morphology of the planting area, and detecting each crop according to the detection path and the IoT control of the farmland to mark the maturity of each crop; determining sub-monitoring areas in the IoT based on the maturity and location of each crop, and marking abnormal growth characteristics of crops in the sub-monitoring areas, and determining a dynamic planting map of the farmland based on the abnormal growth characteristics and the maturity of each crop; determining abnormal planting areas in the dynamic planting map of the farmland based on the abnormal growth characteristics and the types of crops, and marking multiple intelligent planting nodes and corresponding planting optimization measures; constructing planting optimization events for each intelligent planting node based on each intelligent planting node and the IoT, and recording the dynamic optimization state map of each abnormal growth characteristic.

[0005] This invention provides an intelligent planting system for farmland based on the Internet of Things (IoT). This intelligent planting system is applied to the aforementioned intelligent planting method for farmland based on the IoT. The intelligent planting system for farmland based on the IoT includes:

[0006] The crop module is used to determine multiple planting areas based on the farmland distribution map. Within each planting area, the corresponding crop types and their distribution are determined based on the detection of the planting area.

[0007] The maturity module is used to determine the detection path of the planting area based on the distribution of each crop and the regional morphology of the planting area, and to detect each crop according to the detection path and the Internet of Things control corresponding to the farmland, so as to mark the maturity of each crop.

[0008] The planting dynamic map module is used in this Internet of Things to determine sub-monitoring areas based on the maturity and location of each crop, mark abnormal growth characteristics of crops in the sub-monitoring areas, and determine the planting dynamic map of the farmland based on each abnormal growth characteristic and the maturity of each crop.

[0009] The planting optimization measures module is used to identify abnormal planting areas in the dynamic planting map of farmland based on various abnormal growth characteristics and crop types, and to mark multiple intelligent planting nodes and corresponding planting optimization measures.

[0010] The dynamic optimization state graph module is used to construct planting optimization events for each intelligent planting node based on each intelligent planting node and the Internet of Things, and to record the dynamic optimization state graph of each abnormal growth characteristic.

[0011] Compared with the prior art, the beneficial effects of the present invention are:

[0012] In this embodiment of the invention, the method is used to determine sub-monitoring areas based on the maturity and location of each crop, and to mark abnormal growth characteristics of crops in the sub-monitoring areas. Based on these abnormal growth characteristics and the maturity of each crop, a dynamic planting map of the farmland is determined. This method incorporates the maturity of each crop, taking into account both abnormal growth characteristics and the overall maturity of each crop, thereby improving the accuracy of the dynamic planting map of the farmland.

[0013] Therefore, in the dynamic map of farmland planting, abnormal planting areas are identified based on various abnormal growth characteristics and crop types, and multiple intelligent planting nodes and corresponding planting optimization measures are marked. Based on each intelligent planting node and the Internet of Things, planting optimization events for each intelligent planting node are constructed, and dynamic optimization state diagrams for each abnormal growth characteristic are recorded. Planting optimization measures are implemented for multiple intelligent planting nodes to promote intelligent optimization of multiple intelligent planting nodes, improve the accuracy of planting optimization events for each intelligent planting node, and generate dynamic optimization state diagrams for each abnormal growth characteristic. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating the intelligent planting method for farmland based on the Internet of Things in an embodiment of the present invention.

[0015] Figure 2 This is a flowchart illustrating step S11 of the intelligent planting method for farmland based on the Internet of Things in this embodiment of the invention.

[0016] Figure 3 This is a flowchart illustrating step S12 in the intelligent planting method for farmland based on the Internet of Things in this embodiment of the invention.

[0017] Figure 4 This is a flowchart illustrating step S13 in the intelligent planting method for farmland based on the Internet of Things in this embodiment of the invention.

[0018] Figure 5 This is a flowchart illustrating step S14 of the intelligent planting method for farmland based on the Internet of Things in this embodiment of the invention.

[0019] Figure 6 This is a flowchart illustrating step S15 of the intelligent planting method for farmland based on the Internet of Things in this embodiment of the invention.

[0020] Figure 7 This is a schematic diagram of the structural composition of an intelligent planting system for farmland based on the Internet of Things in an embodiment of the present invention. Detailed Implementation

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

[0022] Please see Figures 1 to 7 An intelligent planting method for farmland based on the Internet of Things (IoT) is applied to intelligent planting scenarios in farmland. The intelligent planting method for farmland based on the IoT includes:

[0023] Step S11: Determine multiple planting areas based on the farmland distribution map. In each planting area, determine the corresponding crop types and the distribution of each crop based on the detection of the planting area.

[0024] Step S12: Based on the distribution of each crop and the regional morphology of the planting area, determine the detection path of the planting area, and detect each crop according to the detection path and the IoT control of the farmland to mark the maturity of each crop.

[0025] Step S13: In this Internet of Things, sub-monitoring areas are determined based on the maturity and location of each crop, and abnormal growth characteristics of crops in the sub-monitoring areas are marked. Based on the abnormal growth characteristics and maturity of each crop, a dynamic planting map of the farmland is determined.

[0026] Step S14: In the dynamic map of farmland planting, determine the abnormal planting area based on various abnormal growth characteristics and crop types, and mark multiple intelligent planting nodes and corresponding planting optimization measures.

[0027] Step S15: Based on each intelligent planting node and the Internet of Things, construct planting optimization events for each intelligent planting node, and record the dynamic optimization state diagram of each abnormal growth characteristic.

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

[0029] S111: Collect a distribution map of farmland, determine the location and type of multiple crops based on the detection of the farmland distribution map, and determine multiple planting areas based on the location and type of multiple crops and the regional boundary lines of the farmland distribution map. At this time, each planting area contains at least two different types of crops and no more than four different types of crops.

[0030] S112: Real-time monitoring of each planting area; determining the area detection method for each planting area based on its location, shape, and area; and triggering online detection of each planting area along the detection method to mark the type and location of each crop.

[0031] S113: Determine a first distribution map of the planting area based on the type and location of each crop, determine a second distribution map of the planting area based on the quantity and location of each crop, and determine the distribution of each crop based on the synthesis of the first and second distribution maps.

[0032] In the embodiments of this application, a distribution map of farmland is collected, and at the same time, a drone equipped with a multispectral, hyperspectral, or ordinary camera is used to take low-altitude photos; high-resolution images can be obtained, which can more clearly identify crop types (different crops have different reflectance characteristics under different spectra); the collected map needs to have a unified coordinate system for subsequent accurate positioning; the map should contain geometric information (points, lines, and surfaces) and attribute information (such as plot numbers).

[0033] On the collected distribution map, the specific growth location of different crops is accurately identified, and their respective crop types are determined. At this point, the collected images (especially drone or satellite images) are processed. Using visual features such as color, texture, and shape, image segmentation is performed using algorithms (such as convolutional neural networks, CNNs) to divide the image into different regions, each corresponding to a different crop. For example, based on multispectral information, corn is clearly different from soybeans in specific wavelengths. At the same time, ground sensor data distributed in the farmland (such as soil moisture sensors and temperature sensors) are combined. Although these do not directly identify crop types, they can serve as auxiliary information. For example, a certain crop typically grows within a specific soil moisture range, which can help verify the image recognition results.

[0034] The identified crop location and type information is combined with the overall boundary of the farmland and its internal natural or artificial boundaries (such as field ridges, roads, and irrigation ditches) to delineate relatively independent planting areas for management. At this point, GIS software or related algorithms are used to overlay the crop distribution polygons obtained in step two with linear or planar elements such as farmland boundary lines, internal roads, and irrigation ditches for analysis. Simultaneously, based on the crop type distribution, location proximity, and the barrier effect of boundary lines, for example, areas enclosed by the same field ridge that contain multiple crops are designated as one planting area. In addition, clear delineation rules are established, namely, each area must contain at least two and at most four different types of crops.

[0035] Furthermore, it is crucial to maintain continuous monitoring of crop conditions in each planting area to promptly detect changes (such as crop growth, pest and disease occurrence, and environmental changes). This typically does not mean that people constantly monitor each area, but rather that automated and continuous data collection is achieved using IoT devices. This can include: video surveillance: installing cameras at key locations; environmental sensors: deploying sensors for soil moisture, temperature, light intensity, and air temperature and humidity to continuously collect environmental data, as changes in these data often foreshadow changes in crop conditions; and drone inspections: periodically or irregularly using drones equipped with multispectral or hyperspectral cameras to conduct aerial photography and acquire image data of the overall crop growth status.

[0036] Maintaining continuous monitoring of crop status in each planting area is crucial for timely detection of changes (such as crop growth, pest and disease occurrence, water and fertilizer deficiencies, etc.) to provide a basis for subsequent dynamic adjustments and precision management. This typically does not mean constant manual monitoring, but rather the use of IoT devices for automated, periodic data collection. Examples include cameras deployed in the fields (taking photos or videos at regular intervals), soil sensors (reading temperature, humidity, and pH values ​​at regular intervals), and weather stations (recording weather data in real time). These devices transmit data to the central management system in real-time or near real-time.

[0037] The system incorporates the location, shape, and area of ​​each planting region. Based on preset rules and these three factors, it automatically or semi-automatically selects the detection method. For example, it selects the type of detection equipment: drone, ground mobile robot, fixed camera, manual inspection, etc.; the type of sensor: hyperspectral camera (for precise identification of crop type and health status), ordinary RGB camera (for identification of pests and diseases, and growth), soil sensor (for detection of soil conditions), etc.; and it plans the detection path: whether it is a straight-line scan, a zigzag scan, or intelligent planning based on crop distribution.

[0038] Execute the detection plan determined in the previous step to actually collect crop data in the planting area; this can be triggered by a timer (e.g., a detection is performed once every morning at 8:00) or by an event (e.g., a detection is triggered when the sensor detects that the soil moisture is below a threshold).

[0039] Based on the determined detection method, the equipment in the control system is operated; for example, if a drone is selected, the flight path is planned and the flight mission is initiated, and the camera or sensor on the drone begins to work; the data can be quickly transmitted to the processing center for analysis after collection.

[0040] The detected information is structured to clearly record the current time, location, and type of crop found. This is a crucial step in data processing, providing a foundation for subsequent analysis (such as growth status assessment and pest and disease location). At this point, the raw data collected by the detection equipment (such as photos, videos, and sensor readings) is transmitted to the backend system. The image recognition model is used to analyze the photos or videos to identify the crop type. GPS data or the relative position information of the sensors in the field is used to determine the specific location of the crop. The system associates the identified crop type with its location information and marks it on a digital farmland map.

[0041] Therefore, a first distribution map of planting areas is determined based on the type and location of each crop, and a second distribution map of planting areas is determined based on the quantity and location of each crop. The distribution of each crop is determined by combining the first and second distribution maps.

[0042] At this point, a first distribution map of the planting area is created, visually showing where different types of crops "grow" within the planting area. This map focuses more on "species distribution" and is derived from the real-time monitoring results of steps S111 and S112. For example, images captured by drone cameras, data collected by ground sensors, or on-site observation records by humans (or robots in the future). Simultaneously, the collected data (such as GPS coordinates and image pixel positions) is associated with crop type information. For example, corn is identified at a certain coordinate point (X1, Y1), and soybeans are identified at a certain coordinate point (X2, Y2). This information is marked on a digital map, with different colors or symbols representing different crops. A layer or image with geographic coordinates is output. For example, a two-dimensional map of a farmland, on which the positions of all corn plants are marked in red and the positions of all soybean plants are marked in blue; or a raster map, where each pixel is assigned a crop type code.

[0043] Create a second distribution map of the planting area, showing the "density" or "quantity distribution" of crops within the area. This map focuses more on "quantity distribution" or "density distribution." In addition to crop type and location information, it is also necessary to obtain the number of individual crops at that location. For example, the number of corn plants in a small area can be counted using image processing algorithms, or the number of plants can be detected by a sensor array. At this point, on the digital map, the density is represented by different shades of color or numerical values ​​based on the number of crops counted at each location point or small area (e.g., a grid cell). For example, a region with many corn plants is represented by dark green, and a region with few is represented by light green. Similarly, soybeans can be represented by a series of blue colors. Output a heatmap, where darker colors indicate a greater number of crops in the area; or a contour map, showing the contour lines of crop quantity; or a raster map, where each pixel is assigned a crop quantity value.

[0044] The first distribution map (species distribution) and the second distribution map (quantity distribution) are overlaid or correlated. For example, the color intensity of the second distribution map can be overlaid on the first distribution map so that the final map not only shows where corn is, but also which areas have more corn and which areas have less corn. Alternatively, a new layer can be generated in which each location contains not only the crop species, but also the quantity or density level of that crop species. The final distribution can be a composite layer or a series of charts (such as classification statistics charts, density distribution maps). This result will be used in step S14 to help identify abnormal growth areas.

[0045] Specifically, suppose there is a small rectangular experimental field with an area of ​​100 square meters; according to S111 and S112, the following operations were performed:

[0046] Determining the first distribution map (species distribution): Aerial photography of the field was conducted using a drone, and the images were analyzed using image recognition algorithms. It was found that two crops, corn and soybeans, were planted in the field. There was a corn patch in the upper left corner of the field (location A), a small soybean patch at location B, a corn patch at location C, and a large soybean patch at location D. This information was used to create the first distribution map: On the map, the locations of all corn plants (areas A and C) were marked in red, and the locations of all soybean plants (areas B and D) were marked in blue. This map clearly shows that "corn is in the upper left and middle, and soybeans are in the upper right and lower right."

[0047] Determine the second distribution map (quantity distribution): Count the number of each crop in each small area; for example, using ground sensors or manual counting, we find: Area A: 10 corn plants; Area B: 5 soybean plants; Area C: 15 corn plants; Area D: 25 soybean plants; Plot this quantity information into a second distribution map: On the map, use dark green to mark the area with a large number of corn plants (Area C), and light green to mark the area with a small number of corn plants (Area A); use dark blue to mark the area with a large number of soybean plants (Area D), and light blue to mark the area with a small number of soybean plants (Area B). This map shows that "corn is most densely distributed in Area C, and soybeans are most densely distributed in Area D."

[0048] The first and second distribution maps are combined to form the final distribution map. In the final distribution map, you will see: Top corner (A): labeled "corn", light color (because of low quantity); Top right corner (B): labeled "soybean", light color (because of low quantity); Middle (C): labeled "corn", dark color (because of high quantity); Bottom right corner (D): labeled "soybean", dark color (because of high quantity).

[0049] refer to Figure 3 In step S12, the specific steps are as follows:

[0050] S121: Collect the distribution information of each crop, determine the first sub-path based on the distribution information and types of crops, and determine the second sub-path based on the distribution information and regional morphology of the planting area.

[0051] S122: Collect farmland database, and match the corresponding detection path mapping relationship based on farmland database, planting area and crop type, and determine the detection path for the planting area according to the first sub-path, second sub-path and detection path mapping relationship;

[0052] S123: Based on the detection path of the planting area and the matching of the Internet of Things, the corresponding detection drone is determined. The detection drone detects the planting area along the detection path of the planting area and collects multiple features of the crop during the flight. Based on the multiple features of the crop, the maturity of each crop is determined.

[0053] In the embodiments of this application, the distribution of various crops is collected, and the planning of the first sub-path focuses more on the characteristics of the "crop type" itself. Different crops require different detection frequencies, detection priorities, or have special requirements for detection equipment. The planning will consider: crop importance / value: high-value crops need to be detected preferentially or more intensively; growth cycle / maturation speed: crops with fast maturity need to be detected more frequently; susceptibility: crops that are prone to diseases and pests or are affected by environmental factors need to be monitored in a focused manner; detection difficulty: some crops (such as tall crops and dense crops) require specific sensors or flight altitudes. Based on these considerations, the system will generate a preliminary path suggestion, prioritizing the coverage of areas containing specific types of crops, or planning according to the distribution patterns of specific types of crops.

[0054] Specifically, assuming the system database records: corn matures quickly and is susceptible to corn borer in the current season, requiring close monitoring; rice matures relatively stably, but lodging needs to be monitored later; soybeans have moderate value and are monitored less frequently; corn borer outbreaks mainly occur on corn; based on this information, when planning the first sub-path, the system will: prioritize covering the corn area because pests and diseases and maturity need to be monitored closely; within the corn area, more dense sampling points will be planned, or drones will be required to lower their flight altitude to obtain clearer images; for the rice area, the path planning will be relatively sparse, but will ensure coverage of the edges and center; the first sub-path will roughly resemble a "Z" shape, prioritizing the passage through the upper left and lower left corn areas, and then covering the upper right and lower right areas.

[0055] The first sub-path is illustrated as follows: The starting point is the corn area in the upper left corner. It crosses the corn area horizontally to the upper right corner, then goes down through the soybean area in the upper right corner to the lower right corner, then crosses the rice area in the lower right corner to the lower right boundary, and finally goes up through the corn area in the lower left corner back to the vicinity of the starting point. This path is clearly biased towards corn coverage.

[0056] The planning of the second sub-path focuses more on the geometry and physical characteristics of the "planting area" itself; the goal is to plan a physically feasible and efficient path; considerations include: area boundaries: how to effectively cover the entire area and avoid missing corners; internal obstacles: whether there are ditches, field ridges, small buildings, etc. that need to be avoided; terrain undulation: if there is a slope, the path needs to adapt to the terrain; flight efficiency: how to reduce the number of turns and make the path as straight or regular as possible (such as checkerboard or serpentine); coverage uniformity: ensuring that the entire area can be detected and avoiding some parts being ignored due to the shape of the area; based on these considerations, the system will generate another preliminary path suggestion, focusing on geometric coverage and physical feasibility.

[0057] Specifically, suppose planting area A is a rectangle, but there is a north-south irrigation ditch in the middle, roughly dividing the area into east and west parts; drones are not allowed to fly over the irrigation ditch; based on this information, when planning the second sub-path, the system will: adopt a "serpentine" or "chessboard" path to ensure uniform coverage; the path needs to bypass the irrigation ditch; for example, it can first complete the coverage of the eastern half (from north to south, shuttling back and forth), then fly to the northern end of the western half, and then complete the coverage of the western half (from north to south, shuttling back and forth); the path will minimize hovering or turning over the irrigation ditch.

[0058] The second sub-path is illustrated as follows: Starting at the northern end of the eastern half, it flies south in a zigzag pattern to cover the eastern half, then flies laterally to the northern end of the western half, and then flies south in a zigzag pattern to cover the western half. This path mainly takes into account the geometry of the area and obstacles.

[0059] Furthermore, a farmland database is collected, which includes historical detection data, crop characteristic data, regional environmental data, and equipment performance data. Based on the specific conditions of the current planting area (regional morphology, divided sub-regions) and the main types of crops planted, the system searches or calculates a "detection path mapping relationship" in the database. This relationship can be understood as a set of rules or an algorithm model that describes what path strategy should be adopted under specific conditions (such as a certain crop, a certain area shape, or the presence of a certain obstacle). This mapping relationship is not pre-coded but is trained using machine learning and other methods based on historical data in the database, and can dynamically generate path strategies based on input conditions.

[0060] Specifically, combining database information and planting area A (corn + soybean, rectangular, with irrigation canals and groves): the system will query / calculate: for a combination of "rectangular area, main crops are corn and soybean, with north-south irrigation canals and groves in the southeast corner", the system will detect the path mapping relationship: "For this type of area, the first sub-path based on crop type (focusing on corn area) will be used as the main path framework, but when encountering irrigation canals, the second sub-path based on the area morphology will be used as a detour strategy; at the same time, when detecting soybean area, the path interval should be appropriately reduced (because soybean density is medium and more detailed coverage is required), while it can be appropriately relaxed when detecting corn area; grove areas are excluded in the path planning stage."

[0061] The system combines the two sub-paths generated in S121 with the path mapping relationship (including specific rules, weights, constraints, etc.) obtained from action two to generate a specific, executable final detection path. At this point, the advantages of the two sub-paths are combined. For example, the starting and ending points of the first sub-path are used, but the geometric processing of the second sub-path is incorporated into key areas (such as corn-concentrated areas or near obstacles). Path parameters, such as flight speed, turning radius, sensor height, and coverage interval, are adjusted according to the mapping relationship to ensure that the final path meets all physical constraints (does not hit obstacles, does not fly out of the boundary) and task constraints (covers all target crops, meets detection frequency requirements).

[0062] Specifically, combining the first sub-path (focusing on corn, starting from the upper left cornfield) and the second sub-path (a serpentine path bypassing the irrigation ditch), and applying a mapping relationship (prioritizing the corn frame, bypassing the irrigation ditch, and reducing the interval in the soybean field): the final determined detection path is as follows: the starting point is set at the edge of the upper left cornfield; first, a dense serpentine scan of the upper left cornfield is performed (the interval is set according to the characteristics of corn); after reaching the lower left cornfield, the bypass strategy of the second sub-path is adopted, and a serpentine scan of the eastern half of the cornfield is performed to ensure that the irrigation ditch is bypassed; after completing the eastern half, the path flies laterally to the western half; in the western half, the lower left cornfield is scanned first (using corn path parameters); then the upper right soybean field is scanned, at which point the path interval is reduced, and a zigzag or denser grid scan is used to ensure coverage; finally, the lower right soybean field is scanned, also using a scan method with reduced intervals; throughout the process, the small grove in the southeast corner is always avoided; the final path will generate a specific sequence of latitude and longitude coordinates, including instructions such as flight altitude and speed for each point.

[0063] Therefore, the corresponding detection drone is determined based on the detection path of the planting area and the matching of the Internet of Things. The detection drone detects the planting area along the detection path and collects multiple features of the crop during the flight. Based on the multiple features of the crop, the maturity of each crop is determined.

[0064] At this point, the system needs to select the most suitable drone based on the following factors: the characteristics of the detection path: the length and complexity of the path (whether frequent turns or detours are required), and the coverage area; for example, a long-distance, mainly straight path is suitable for a drone with long endurance and high speed; while a complex path that requires precise operation requires a drone with good maneuverability and strong load capacity.

[0065] The characteristics to be collected determine which sensors the drone needs to carry (such as high-definition cameras, multispectral sensors, thermal imagers, etc.). For example, detecting corn pollination requires a high-resolution visible light camera, while detecting nitrogen content in soybean leaves requires a multispectral sensor. Simultaneously, the drone's real-time status (obtained via the Internet of Things) is crucial: its current location, battery level, payload, and whether it is performing other tasks. The system needs to select a usable drone with sufficient power, located relatively close to the starting point, and equipped with the necessary sensors. The IoT platform integrates the real-time status information of all drones and, based on path and task requirements, matches and schedules them using algorithms, issuing task commands.

[0066] The drone flies autonomously based on the precise coordinate sequence and flight parameters (altitude, speed) determined by S122; the navigation system (GPS, RTK, etc.) ensures that it moves precisely along the planned path; the drone controls the onboard sensors to collect data during flight according to a preset program; for example, in a densely scanned corn area, the camera takes pictures at a lower flight altitude, slower speed, and smaller overlap; when bypassing irrigation canals, it needs to slightly increase altitude to ensure a safe distance; in soybean areas, multispectral sensors simultaneously collect spectral data in different bands; the collected data (images, spectra, etc.) is transmitted back to the IoT platform or ground station in real time or near real time via wireless networks (such as 4G / 5G, Wi-Fi).

[0067] The sensors on the drone collect raw data (such as pixel values ​​and spectral values). The system needs to extract multiple features that can reflect crop maturity from this raw data. Common features include: image features: color (such as the color of corn ears changing from green to yellow / red, and the color change of soybean leaves), morphology (such as the fullness of corn ears, and the fullness and color of soybean pods), texture (the reflection of the waxy layer of leaves); spectral features: different crops have different reflection / absorption characteristics of specific spectral bands at different growth stages (such as NDVI value, red edge position, absorption peaks of specific pigments, etc.); other sensor features: such as thermal imaging (leaf temperature is related to water stress, indirectly reflecting the growth status).

[0068] The system needs to perform correlation analysis between the extracted features and the crop type (corn or soybean). This usually relies on a pre-trained machine learning model or a rule base based on expert knowledge. For example, the model learns that "for corn variety X, when the average RGB value of the ear region changes from (100, 150, 50) to (180, 160, 100), the maturity reaches 80%"; or "for soybean variety Y, when the NDVI value decreases from 0.75 to 0.65 and the reflectance at a specific wavelength increases, it indicates that the crop is close to maturity". Based on the results of feature analysis, the system gives a maturity score or grade (such as percentage, or "unripe", "harvestable", "overripe", etc.) for each detection point (or each small area).

[0069] refer to Figure 4 In step S13, the specific steps are as follows:

[0070] S131: Dynamically monitor each planting area based on the control of detection drones through the Internet of Things, collect the location of each crop, determine the preliminary monitoring area based on the location, type and maturity of each crop, and determine the sub-monitoring area based on the preliminary monitoring area and the regional morphology of the planting area.

[0071] S132: In this sub-monitoring area, multiple abnormal growth three-dimensional spaces are determined based on the anomaly detection of the sub-monitoring area. Each abnormal growth three-dimensional space corresponds to at least one crop. The abnormal growth characteristics of the crop are determined based on the three-dimensional detection of each abnormal growth three-dimensional space by the detection drone.

[0072] S133: Construct an abnormal dynamic map of crops based on their abnormal growth characteristics, corresponding maturity levels and locations, and determine the planting dynamic map of farmland based on the abnormal dynamic maps of each crop, each planting area and time.

[0073] In the embodiments of this application, the Internet of Things (IoT) system (such as a server in a farm control center) continuously sends instructions to the detection drones deployed over area A. These instructions include maintaining patrol status, flying along a predetermined path, and returning to base for charging at scheduled times. The drones maintain a connection with the IoT system via wireless networks (such as 4G / 5G or Wi-Fi), report their own status (battery level, GPS location, flight attitude, etc.) in real time, and receive new tasks.

[0074] During dynamic monitoring, the sensors carried by the drone (such as high-resolution cameras, LiDAR, thermal imagers, etc.) continuously scan the farmland below. Through image recognition algorithms, point cloud processing, or thermal imaging analysis, the system can continuously update or verify the precise location coordinates of each crop. By comparing this location information with the location information obtained in the previous steps, it can detect whether new crops have appeared (such as weeds) or whether the crop location has changed significantly.

[0075] Based on the collected data, the system filters out the parts that need to be focused on. For example, it can select crops that are at a specific stage of maturity (such as "semi-mature" or "nearly mature") or special types of crops (such as experimental varieties). Then, the system delineates one or more "preliminary monitoring areas" with these selected crops as the center or boundary. This area is usually a simple geometric shape (such as a circle or rectangle) that contains one or more target crops and reserves a certain amount of surrounding space for more detailed detection later.

[0076] The initial monitoring area is a simple geometric shape, but the actual boundaries and crop distribution of farmland are more complex. In order to make subsequent drone detection more accurate and efficient, the initial monitoring area needs to be further refined into "sub-monitoring areas" that are more realistic and easier for drones to fly. This requires taking into account the overall shape of the planting area (such as whether the plot is rectangular, irregular, has field ridges, or has obstacles). The design of sub-monitoring areas involves dividing the initial area into smaller grids, delineating scanning zones along the direction of crop rows, or planning safe flight paths to avoid obstacles.

[0077] A rule is set: "Mark all corn plants with a maturity level of 'semi-mature' and located within 10 meters of the irrigation ditch as objects of interest." Based on the corn location and maturity data obtained in the previous steps, as well as the location information just collected, the system filters out corn plants that meet the criteria. Then, the system sets a circular area with a radius of 2 meters or a square area with a side length of 4 meters centered on these plants, defining these areas as "preliminary monitoring areas." At the same time, the small cluster of non-corn plants previously discovered is also marked, and a preliminary monitoring area is defined for them.

[0078] Specifically, for the initial monitoring area near the irrigation canal in the middle of Region A (assuming it is circular), considering that drones usually fly in a straight line or along a specific path, the system will transform it into a rectangular "sub-monitoring area" covering the circular area, or plan several parallel flight paths through the area; for the initial monitoring area near the field ridge, the design of the sub-monitoring area will automatically avoid the field ridge to ensure the safe flight of the drone; finally, the system generates detailed flight plans for these initial monitoring areas, namely the specific "sub-monitoring areas" and the corresponding detection paths.

[0079] Furthermore, within this sub-monitoring area, the drone does not simply take photos or scan, but combines data from multiple sensors (such as high-resolution visible light cameras, multispectral / hyperspectral cameras, thermal imagers, etc.) with preset anomaly detection algorithms. The algorithm compares the currently collected data with the "standard model" of normally growing corn in the area or historical data from the same period. For example, visible light image analysis may detect that the leaves of a corn plant are abnormally yellow or have spots, which contrasts with the surrounding healthy green leaves.

[0080] When the system identifies points or small areas that significantly deviate from the normal range, it marks them as "abnormal growth three-dimensional spaces." These "abnormal growth three-dimensional spaces" do not refer to a single point on the ground, but rather to areas in three-dimensional space from the ground to the top of the crop that are associated with the abnormality. For example, a yellowing leaf corresponds to a small three-dimensional space, and an area eaten by insects corresponds to another three-dimensional space. The system will identify three such "abnormal growth three-dimensional spaces," which correspond to: an area with a yellowing leaf, an area with several stunted plants, and an area with obvious lesions on a leaf.

[0081] Once the three-dimensional space of the abnormal growth is identified, the drone will conduct a more detailed "three-dimensional inspection" of these specific areas. This means that the drone will: lower its flight altitude to obtain higher resolution images; adjust the camera angle to observe the abnormal area from different sides; activate specific sensor modes, such as hyperspectral scanning, to obtain more detailed biochemical information; and take repeated shots to confirm that the anomaly is not an instantaneous phenomenon.

[0082] Through these detailed tests, the system can extract specific abnormal growth characteristics: For yellowing leaf areas: characteristics include "yellowish color", "reduced chlorophyll content (estimated through multispectral data)" and "normal leaf integrity"; For stunted plant areas: characteristics include "shorter plant height (20% lower than average)", "fewer leaves", and "thin stem diameter"; For lesion areas: characteristics include "circular brown spots on leaves", "yellow halos around the spots", and "high spot density".

[0083] Therefore, based on the abnormal growth characteristics, corresponding maturity levels, and locations of crops, anomaly dynamic maps of crops are constructed. Then, based on the abnormal dynamic maps of each crop, each planting area, and each time period, the planting dynamic map of the farmland is determined, improving the accuracy of the planting dynamic map. Simultaneously, the maturity levels of each crop are incorporated, accommodating a holistic consideration of both abnormal growth characteristics and the maturity levels of each crop, further enhancing the accuracy of the planting dynamic map.

[0084] At this point, the abnormal growth characteristics of crops, their corresponding maturity levels and locations are introduced. Abnormal growth characteristics: Space 1 - leaves turn yellow and chlorophyll content is low; Space 2 - plants are short and have abnormal height; Space 3 - leaves have dense lesions; Corresponding maturity level: the overall maturity level of the corn area where these abnormal points are located is "semi-mature"; Location: Space 1 - (15,23); Space 2 - (18,25); Space 3 - (16,24).

[0085] The system integrates the above information into a data structure or database, usually associated with a timestamp (e.g., detection date: July 5, 2025, 10:30); and creates a record for each anomaly, including its ID, location coordinates, crop type (corn), overall maturity (semi-mature), anomaly characteristic description, and anomaly type (e.g., malnutrition, growth retardation, disease).

[0086] Meanwhile, the system can mark these anomalies on a digital map or image of the farmland; for example, different colors or shapes of markers can be used to represent different types of anomalies: a yellow asterisk is marked at position (15,23) on the map, with the annotation "yellow leaves, semi-mature"; a blue circle is marked at position (18,25) on the map, with the annotation "stunted, semi-mature"; a red square is marked at position (16,24) on the map, with the annotation "lesion, semi-mature". This visualization, together with the data records behind it, constitutes a "dynamic map of crop anomalies"; it shows the specific anomalies of corn crops in region A at a specific time point (10:30 on July 5) and their correlation with maturity.

[0087] Anomaly dynamic maps for each crop, planting area, and time are introduced. Anomaly dynamic maps for each crop include: the anomaly dynamic map of area A generated above, and anomaly dynamic maps of other areas (such as the tomato area in area B) that were detected (containing different anomaly types, such as water shortage, pests, etc.); planting area: the entire farmland is divided into area A (corn), area B (tomato), area C (wheat), etc.; time: the time point when all anomaly detections occurred (e.g., 10:30 on July 5).

[0088] The system integrates all "abnormal dynamic maps of crops" generated from different planting areas at different times; the system takes the time dimension into account; if abnormal dynamic maps were also detected and generated previously (e.g., on July 3), the system can overlay or compare and analyze the maps from different time points; the final "planting dynamic map of farmland" is usually a comprehensive visualization interface, an interactive map or dashboard; it will display: an overview of the crop distribution of the entire farmland (from S113); the overall maturity of crops in each area (from S12); and all detected anomalies, categorized and displayed by region, time, and anomaly type.

[0089] refer to Figure 5 In step S14, the specific steps are as follows:

[0090] S141: Real-time monitoring of farmland planting dynamics, marking the location of each abnormal growth feature, and determining the first region coefficient based on the location and type of each abnormal growth feature.

[0091] S142: Determine the second regional coefficient based on the location of each abnormal growth characteristic and the type of crop; determine the abnormal planting area based on the matching of the farmland planting dynamic map, the first regional coefficient and the second regional coefficient;

[0092] S143: In the abnormal planting area, the location to be optimized is determined based on the detection of the abnormal planting area. The corresponding intelligent planting node is determined according to the location to be optimized, the abnormal growth characteristics and the type of crop. Multiple intelligent planting nodes are collected. The corresponding planting optimization measures are determined based on the matching of multiple intelligent planting nodes and farmland database.

[0093] In the embodiments of this application, the system continuously accesses and displays a dynamic map of farmland planting generated by S133, which includes timestamps. This map not only shows the distribution and maturity of corn (e.g., represented by different shades of color), but also uses special markers (e.g., red circles, different shapes) to update and display detected abnormal points (A, B, C, D, E) and their types (yellow represents yellowing, blue represents stunted growth, and green represents spots) in real time. Monitoring means that the system not only looks at the map, but also performs simple preliminary analysis, such as changes in the number of abnormal points and the location of newly added abnormal points.

[0094] On the dynamic graph, each confirmed anomalous growth feature is assigned precise geographic coordinates or grid indexes, which ensures that subsequent analysis can be based on spatial relationships; the marking is not just for display, but also for the structured storage of location information for computation.

[0095] Specifically, the dynamic graph currently displayed on the administrator or system interface clearly marks five abnormal points: A, B, C, D, and E, with small labels next to them indicating "yellowing leaves," "stunted plant," or "leaf spots." The system notices that point E was added today, and that A and B are adjacent. The system records the following in its internal database: Abnormal A: coordinates (3,4), type: yellowing leaves; Abnormal B: coordinates (3,5), type: yellowing leaves; Abnormal C: coordinates (7,8), type: stunted plant; Abnormal D: coordinates (7,9), type: stunted plant; Abnormal E: coordinates (8,8), type: leaf spots.

[0096] The "First Regional Coefficient" aims to quantify the spatial clustering and potential impact range of anomalies. Calculation methods can vary, but the core idea is to combine location (spatial relationships) and anomaly type (severity / spreadability). In this case, the distance between anomalies of the same or similar types is calculated; the closer the distance, the higher the coefficient, indicating strong spatial clustering. Alternatively, the number of similar anomalies within a certain radius around a given anomaly can be calculated. Different anomaly types are assigned different weights; for example, pests and diseases (such as disease X) have a higher weight than simple nutrient deficiencies (such as nitrogen deficiency) because they spread faster and have a greater impact. Alternatively, an anomaly may have a higher weight if it occurs during a crop's critical growth period.

[0097] The calculation is as follows:

[0098] Simple clustering degree calculation: For each anomaly type, calculate the average distance between all its anomaly points; the smaller the average distance, the higher the clustering degree, and the greater the coefficient value contributed by that type; for example, the average distance of all "yellow leaves" points (A, B) is sqrt((3-3)²+(4-5)²)=1; the average distance of all "stunted plants" points (C, D) is sqrt((7-7)²+(8-9)²)=1; the clustering degree of the "leaf spots" point (E) is the lowest because it has no similar neighbors, so the distance is infinite or set to the maximum value.

[0099] Specifically, the system calculations revealed that the anomalies C and D of "stunted growth" were very close (distance 1), and "stunted growth" was designated by the system as a more noteworthy anomaly type (with higher weight) than "yellowed leaves." Although "yellowed leaves" also clustered (distance 1 between A and B), its weight was lower. "Leaf spots" E was isolated. Therefore, the first regional coefficient calculated by the system mainly reflected the clustering and importance of the two points C and D, and its value was relatively high (e.g., a calculation result of 0.14). This coefficient indicates that, in terms of spatial distribution, the "small group" formed by C and D requires priority attention.

[0100] By monitoring dynamic graphs, marking locations, and calculating the first regional coefficient, a preliminary quantification of the spatial distribution characteristics and potential impact range of abnormal growth was achieved. This coefficient helps the system and managers quickly identify which anomalies are not isolated points, but areas that form problem "hot spots," providing important spatial basis for the next step of more precise regional division in combination with crop types (S142). In the example, the calculation results of the first regional coefficient indicate that the area formed by the two "stunted plants" points located at (7,8) and (7,9) has a higher spatial clustering and potential impact than other anomaly points.

[0101] Furthermore, a second regional coefficient is determined based on the location of each abnormal growth characteristic and the type of crop; abnormal planting areas are identified based on the matching of the farmland planting dynamic map, the first regional coefficient, and the second regional coefficient, thereby improving the accuracy of abnormal planting areas.

[0102] At this point, each abnormal location is associated with its corresponding crop type; for example, yellowing leaves at location A correspond to corn; assess crop sensitivity: the system needs to have built-in or learn the sensitivity of different crops to different types of abnormalities; for example, corn is more sensitive to nitrogen deficiency (yellowing leaves) during the grain-filling stage, while soybeans are more sensitive to a certain disease, and this sensitivity can be quantified into a weight value.

[0103] At this point, for each anomalous location, a "crop correlation coefficient" is calculated based on its anomalous type and the sensitivity weight of the corresponding crop. This coefficient reflects the potential impact of the anomalous point on the crops in that location. The calculation method can take into account the clustering of similar anomalous points (and the same type of crop) around the anomalous point, and combine it with the sensitivity weight. For example, if the yellowing of leaves occurs on corn, which has a high demand for nitrogen fertilizer, the coefficient will be higher than that on soybeans, which have the same yellowing of leaves but are relatively tolerant of poor soil.

[0104] The system introduces a dynamic map of farmland planting, a first regional coefficient, and a second regional coefficient. The system needs to match and integrate the first regional coefficient and the second regional coefficient. This usually means that the ultimate importance of a location or a small area is determined by both spatial clustering (first coefficient) and crop correlation (second coefficient). For example, if an area in S141 has a high first coefficient due to spatial clustering, and the crop species in that area are also more sensitive to anomalous species (the second coefficient is also high), then this area is likely to be identified as an "abnormal planting area".

[0105] Based on the integrated results, the system delineates specific "abnormal planting areas" on the planting dynamic map. These can be one or more continuous grid areas or irregular polygonal areas. The criteria for delineation can be that the integrated coefficient exceeds a preset threshold, or that the area is classified according to the coefficient (e.g., high-risk area, medium-risk area). During the delineation process, other information in the planting dynamic map (such as crop maturity and planting density) is also considered. For example, if the crops in an area are close to maturity, even if some anomalies are detected, its priority as an "abnormal planting area" will be reduced. The system outputs a clear list of "abnormal planting areas" and their boundaries.

[0106] Specifically, assuming the following outliers and their first regional coefficients are calculated in S141 (assuming the coefficient range is 0-1, the higher the value, the higher the spatial clustering and the higher the initial importance): Location A: (3,4) - Outlier: Yellowing leaves - First coefficient: 0.2; Location B: (3,5) - Outlier: Yellowing leaves - First coefficient: 0.2; Location C: (7,8) - Outlier: Stunted plants - First coefficient: 0.5; Location D: (7,9) - Outlier: Stunted plants - First coefficient: 0.5; Location E: (9,9) - Outlier: Leaf spots - First coefficient: 0.1.

[0107] Assume the system has built-in sensitivity weights for corn to different anomalies (higher values ​​indicate greater sensitivity): yellowing leaves: 0.6, stunted growth: 0.8, leaf spots: 0.4; Calculation method: consider the number of similar anomalies at the given point and its surroundings (e.g., neighbors with a Manhattan distance of 1), multiply by the sensitivity weight; Location A: (3,4) - yellowing leaves; Neighbor B also has yellowing leaves; Number of similar neighbors = 1; Second coefficient = 1 * 0.6 = 0.6; Location B: (3,5) - yellowing leaves; Neighbor A also... The leaves are yellow; number of similar neighbors = 1; second coefficient = 1 * 0.6 = 0.6; position C: (7,8) - plant is short; neighbor D is also a short plant; number of similar neighbors = 1; second coefficient = 1 * 0.8 = 0.8; position D: (7,9) - plant is short; neighbor C is also a short plant; number of similar neighbors = 1; second coefficient = 1 * 0.8 = 0.8; position E: (9,9) - leaf spot; isolated point; number of similar neighbors = 0; second coefficient = 0 * 0.4 = 0.0.

[0108] Assuming a simple weighted summation model is used to combine the two coefficients, with weights of 0.4 (spatial) and 0.6 (crop): Combined coefficient = 0.4 * first coefficient + 0.6 * second coefficient; Calculate the combined coefficients: A: 0.4 * 0.2 + 0.6 * 0.6 = 0.08 + 0.36 = 0.44; B: 0.4 * 0.2 + 0.6 * 0.6 = 0.08 + 0.36 = 0.44; C: 0.4 * 0.5 + 0.6 * 0.8 = 0.20 + 0.48 = 0.68;

[0109] D: 0.4 + 0.5 + 0.6 + 0.8 = 0.20 + 0.48 = 0.68; E: 0.4 + 0.1 + 0.6 + 0.0 = 0.04 + 0.00 = 0.04; Set a threshold, such as 0.5; Points with a comprehensive coefficient greater than 0.5 will be considered as part of the "abnormal planting area"; A: 0.44 < 0.5; B: 0.44 < 0.5; C: 0.68 > 0.5; D: 0.68 > 0.5; E: 0.04 < 0.5; Although the second coefficient (0.6) of A and B is not low, the first coefficient (spatial clustering) is low, and the comprehensive coefficient does not reach the threshold; C and D have a comprehensive coefficient that far exceeds the threshold because of strong spatial clustering (first coefficient 0.5) and high crop sensitivity (second coefficient 0.8).

[0110] The system will check the positions of C and D on the planting dynamic map. They are adjacent ((7,8) and (7,9)) and surrounded by healthy corn. Therefore, the system will define these two points and the area directly connected to them as an "abnormal planting area", for example, named "region X", whose boundary is the area containing the two grids (7,8) and (7,9). The "abnormal planting area" X is determined to be located in grids (7,8) and (7,9).

[0111] Therefore, in this abnormal planting area, the location to be optimized is determined based on the detection of the abnormal planting area, and the corresponding intelligent planting node is determined according to the location to be optimized, the abnormal growth characteristics and the type of crop. Multiple intelligent planting nodes are collected, and the corresponding planting optimization measures are determined based on the matching of multiple intelligent planting nodes and farmland database.

[0112] At this point, review all the outliers detected in region X and their details; in this example, the plant stunting anomalies are located at positions (7,8) and (7,9); for each outlier in region X, determine whether it needs immediate or priority optimization; since plant stunting usually means stunted growth and affects yield, these two positions are marked as "positions to be optimized"; refine these positions to specific coordinates or plot units; for example, position 1 to be optimized: (7,8); position 2 to be optimized: (7,9); output a list of positions to be optimized, such as [(7,8), (7,9)].

[0113] The system collects a list of locations to be optimized, along with the abnormal growth characteristics (stunted plants) and crop type (corn) for each location. For each location to be optimized, its coordinates, abnormal characteristics (stunted plants), and crop type (corn) are extracted. Based on the built-in optimization model or machine learning algorithm, and combined with the specific circumstances of the location to be optimized, the system generates preliminary, intelligent planting decision suggestions. These suggestions are operational units with a certain degree of universality for specific problems and are called intelligent planting nodes.

[0114] For example: Node 1 (for (7,8)): {"Location":(7,8),"Crop":"Corn","Abnormality":"Stunted Plants","Suggested Operation Type":"Soil Sampling and Analysis","Suggested Parameters":"Detect Nitrogen, Phosphorus, and Potassium Content and pH Value"}; Node 2 (for (7,9)): {"Location":(7,9),"Crop":"Corn","Abnormality":"Stunted Plants","Suggested Operation Type":"Micro-irrigation and Fertilization","Suggested Parameters":{"Nitrogen Fertilizer":"10ml High Nitrogen Solution","Phosphorus and Potassium Fertilizer": The system considers that the stunted growth of plants in (7,8) is related to soil nutrient deficiency, and therefore suggests taking soil samples first. The abnormality in (7,9) is more likely to be due to water or early nutrient problems, and it is closer to the irrigation facility, so it suggests direct micro-irrigation and fertilization. This distinction is based on historical data, crop growth models or more refined sensor data. These decision suggestions are encapsulated into structured intelligent planting nodes, and multiple intelligent planting node lists are output, such as [node1, node2].

[0115] Collect all the intelligent planting nodes generated for the abnormal planting area to form a complete node set with an optimized solution for the area; store this node set in a structured form or pass it to the next processing unit; output a set containing all relevant intelligent planting nodes, for example, {"area":"X","node list":[node1,node2]}.

[0116] Based on the matching of multiple intelligent planting nodes and farmland databases, corresponding planting optimization measures are determined. At this time, the suggested operation types and suggested parameters in each intelligent planting node are matched and refined with the information in the farmland database. The database information includes: resource information: current farm inventory (fertilizer type, quantity, pesticides), agricultural machinery and equipment (type, location, status), labor arrangement, etc.; environmental information: real-time weather data (temperature, humidity, precipitation probability), soil historical data (average pH value of the area, fertilization records of previous years); operation specifications: standard procedures, safety specifications, and best practices for different operations (such as fertilization and pesticide application).

[0117] At this point, for node 1's "Soil Sampling Analysis": Check the database: Are there available soil sampling tools? Is the sampler available? Is the nearest laboratory analysis service available? Matching result: Sampling tools are confirmed to be available. Arrange for technician Wang to perform sampling in the afternoon, and contact the cooperating agricultural laboratory for priority analysis. For node 1's "Soil Sampling Analysis": Check the database: Are there available soil sampling tools? Is the sampler available? Is the nearest laboratory analysis service available? Matching result: Sampling tools are confirmed to be available. Arrange for technician Wang to perform sampling in the afternoon, and contact the cooperating... The agricultural laboratory conducts priority analysis; the matched and refined operational recommendations are transformed into specific and actionable action plans, namely planting optimization measures; at this time, measure 1: at 14:00 this afternoon, technician Wang will collect soil samples at location (7,8) and send them to XX laboratory for nitrogen, phosphorus, potassium and pH value testing; measure 2: at 15:00 this afternoon, use the micro-irrigation system to deliver 10ml of high nitrogen solution A and 5ml of balanced solution B to location (7,9), and control the total irrigation volume to about 50ml; output a specific list of planting optimization measures, such as [measure 1, measure 2].

[0118] Specifically, a list of specific planting optimization measures is provided, such as [Measure 1, Measure 2]. The system confirms that both (7,8) and (7,9) need optimization. The intelligent planting nodes are determined as follows: For (7,8), the system analysis indicates that the soil conditions need to be understood first, and the node "Sampling soil samples from corn at (7,8) and testing nitrogen, phosphorus, potassium and pH" is generated. For (7,9), the system analysis indicates that there is a lack of water or early fertilizer, and micro-irrigation is convenient, so the node "Micro-irrigating corn at (7,9) and applying a specific ratio of nitrogen fertilizer and balanced fertilizer" is generated.

[0119] Collect these two nodes; query the database and find that the farm has soil sampling tools, Technician Wang is available, and the laboratory is cooperating well; therefore, generate the optimization measure: "At 14:00 this afternoon, Technician Wang will take samples at (7,8) for testing"; query the database and find that the farm has high nitrogen and balanced solutions in stock, the micro-irrigation system is working normally at (7,9), and the weather is sunny; therefore, generate the optimization measure: "At 15:00 this afternoon, use the micro-irrigation system to accurately deliver 10ml of high nitrogen solution A and 5ml of balanced solution B to (7,9)".

[0120] First, the system precisely locates the points requiring intervention (locations to be optimized) within the abnormal area. Then, based on the location, anomalies, and crop information, it generates intelligent preliminary suggestions (intelligent planting nodes). Next, these suggestions are combined with the farm's actual resources and environmental conditions (matching with the database). Finally, it outputs clear, actionable, and site-specific planting optimization measures. This process ensures that the intervention measures are both targeted and feasible, making it a key link in achieving precision agricultural management. In this example, the system not only identified the problem but also provided differentiated and intelligent solutions based on diagnosis before treatment (sampling analysis vs. direct fertilization).

[0121] refer to Figure 6 In step S15, the specific steps are as follows:

[0122] S151: Collect IoT data of farmland and trigger corresponding detection drones based on the IoT data of farmland. Based on the detection drones, detect the execution of planting optimization measures corresponding to intelligent planting nodes in real time, so as to collect multiple execution images of planting optimization measures.

[0123] S152: Determine multiple sub-planting optimization events based on the recognition of multiple execution images, and construct planting optimization events for each intelligent planting node based on the synthesis of multiple sub-planting optimization events and their corresponding execution times;

[0124] S153: In this planting optimization event, state images of each abnormal growth feature at different times are collected, and each state image is sorted according to time. Based on the sequential comparison of each state image, multiple changes of each abnormal growth feature are determined, and the dynamic optimization state map of each abnormal growth feature is determined according to the multiple changes, the corresponding abnormal growth features, and the type of crop.

[0125] In the embodiments of this application, the Internet of Things (IoT) of farmland is collected, and the corresponding detection drone is triggered based on the IoT of farmland. The detection drone detects in real time the execution of planting optimization measures corresponding to the intelligent planting nodes, so as to collect multiple execution images of planting optimization measures.

[0126] At this time, the system collects data from the IoT nodes in the farmland. Simultaneously, it continuously receives data from the IoT nodes in the farmland, such as sensor readings (soil moisture, temperature, light intensity, etc.), equipment status (whether the water pump is on, whether the valve is open), and task scheduling information. This information is used to determine whether a new detection task needs to be started, or to respond to a task that has already been started.

[0127] When the system confirms that it needs to monitor the execution of a certain "planting optimization measure" (for example, when Technician Wang starts the sampling task at (7,8), or when the micro-irrigation system starts delivering fertilizer to (7,9), or according to the preset monitoring plan, the system will send instructions to the predetermined detection drone through the IoT network; the instructions include: target area (such as (7,8) or (7,9)), flight path, tasks to be performed (such as hovering and shooting, scanning along a specific route), image acquisition frequency, etc.

[0128] After receiving instructions, the drone flies to the designated area. During the execution of "planting optimization measures" (e.g., when Technician Wang is taking samples, or when the micro-irrigation system is working), the drone performs real-time detection according to instructions. This includes: taking high-definition photos, recording videos, and using multispectral / thermal imaging cameras to obtain crop status information. The drone transmits multiple "execution images" (e.g., photos of Technician Wang operating at the sampling point, photos of the micro-irrigation pipe outlet, and photos of the soil surface in the fertilization area) back to the ground station or cloud platform in real time or periodically. It outputs multiple execution images related to each "planting optimization measure".

[0129] Furthermore, multiple sub-planting optimization events are determined based on the recognition of multiple execution images, and planting optimization events for each intelligent planting node are constructed by synthesizing multiple sub-planting optimization events and their corresponding execution times.

[0130] At this point, in S151, the drone has already captured a series of images related to the "implementation of planting optimization measures." These images include continuous footage taken by the drone along its execution path or scenes taken while hovering at specific locations. Now, these images need to be intelligently analyzed. The goal of the analysis is to identify which specific "planting optimization measures" are being implemented, using image recognition or computer vision technologies. For example, it can identify: personnel / equipment: whether there are human operators (such as spraying pesticides or applying fertilizer) or specific agricultural machinery (such as the activation of automatic sprinkler systems); operational behavior: identifying specific operational actions; for example, identifying "spraying pesticides" or "foliar fertilization" by recognizing the mist sprayed from the nozzles; identifying "irrigation" by recognizing liquid traces on the ground or the operating status of equipment; identifying "fertilizer application" by recognizing granular material scattered on the ground; materials / tools: sometimes even identifying the specific type of material used (e.g., fertilizer bags of different colors, specific types of pesticide sprayers).

[0131] When image analysis confirms that a specific optimization measure is being implemented, the system records it as a "sub-planting optimization event." This event is atomic and represents a minimal, identifiable unit of operation. For example, "foliar spraying of potassium dihydrogen phosphate in the northwest corner of plot A" is a sub-event. This stage outputs a series of sub-event lists with preliminary descriptions, such as: Sub-event 1: [Timestamp 1] - Image shows that foliar spraying operation was detected in the northwest corner of plot A, suspected to be potassium dihydrogen phosphate spraying; Sub-event 2: [Timestamp 2] - Image shows that irrigation equipment was detected to be activated in the middle of plot A, and drip irrigation is in progress.

[0132] Each "sub-planting optimization event" identified by the image carries a timestamp of its occurrence (this timestamp usually comes from the drone's GPS or internal clock, recorded during the S151's capture). This is crucial for organizing these scattered events. Simultaneously, the image typically also contains geographic location information (again from the drone's GPS), allowing us to determine which specific "intelligent planting node" or its corresponding "planting area" the sub-event occurred in. For example, sub-event 1 and sub-event 2 both occur within the planting node "Plot A." The system merges and organizes multiple "sub-planting optimization events" belonging to the same "intelligent planting node" and occurring within similar time periods. This is not merely a simple listing, but rather the construction of a logical and chronologically ordered complete event flow. For example:

[0133] If multiple sub-events such as "person approach detected", "sprayer bottle removed detected", "foliar spraying detected", and "person departure detected" occur consecutively in the northwest corner of plot A within a short period of time, the system will combine them into a more macro-level "planting optimization event" that "foliar fertilization optimization measures were implemented in the northwest corner of plot A".

[0134] If irrigation and fertilization occur in different areas of plot A, the system will record "Plot A implemented two optimization measures today: irrigation and fertilization", and record the start and end times and specific locations of each measure.

[0135] Ultimately, for each "intelligent planting node" (such as plot A and plot B), the system will construct one or more "planting optimization events"; each event includes: the planting node to which it belongs; the event type (such as irrigation, fertilization, spraying, weeding, etc.); the execution time range (start time and end time); the execution location (a more precise area, if the sub-events are distributed across multiple small areas); and a list of included sub-events (optional, for detailed traceability).

[0136] The output stage will output a structured "Planting Optimization Event" record; for example, for plot A, there will be a record: Planting Optimization Event: Comprehensive Optimization Measures for Plot A Today; Start Time: [Timestamp 1]; End Time: [Timestamp N]; Includes sub-events: Sub-event 1: [Timestamp 1] - Northwest corner of plot A, foliar spraying of potassium dihydrogen phosphate; Sub-event 2: [Timestamp 2] - Center of plot A, start drip irrigation system.

[0137] Optionally, the final constructed "planting optimization event" is as follows:

[0138] Event 1: Foliar fertilization optimization measures for cornfield A: Node: Cornfield A; Type: Foliar fertilization; Time: 10:15-10:20 AM; Location: Northwest corner; Sub-event: [10:15 AM] - Foliar spraying operation detected in the northwest corner of cornfield A, suspected to be potassium dihydrogen phosphate spraying;

[0139] Event 2: Irrigation Optimization Measures for Corn Field A: Node: Corn Field A; Type: Irrigation; Time: 2:30 PM - 2:45 PM; Location: Entire Field;

[0140] Sub-events: [2:30 PM] - Drip irrigation equipment was detected starting in the middle of cornfield A and is currently irrigating; [2:35 PM] - Drip irrigation equipment was detected starting in the eastern part of cornfield A and is currently irrigating.

[0141] Therefore, in this planting optimization event, state images of various abnormal growth characteristics at different times are collected, and the state images are sorted according to time. Based on the sequential comparison of each state image, multiple changing parts of each abnormal growth characteristic are determined. Based on the multiple changing parts, the corresponding abnormal growth characteristics, and the type of crop, a dynamic optimization state diagram of each abnormal growth characteristic is determined. This achieves a holistic consideration of multiple changing parts, the corresponding abnormal growth characteristics, and the type of crop. At the same time, it improves the accuracy of planting optimization events at each intelligent planting node and generates a dynamic optimization state diagram of each abnormal growth characteristic.

[0142] At this point, "different times" typically include: baseline time point: before implementing optimization measures, a detailed image acquisition of the abnormal area is conducted as a benchmark for comparison; key time points: after the implementation of optimization measures, several key time points are selected for repeated image acquisition based on the crop growth cycle and the type of measure; for example, for fertilization or irrigation measures, time points such as 1 day, 3 days, 7 days, and 14 days after the implementation of the measure are selected; dynamic monitoring: if conditions permit, near real-time and higher frequency monitoring can also be carried out, but this is usually more costly.

[0143] It is necessary to ensure that each image captured focuses on the previously identified area of ​​"abnormal growth characteristics". For example, if corn plants in a certain area were previously identified as having "yellowing leaves" (nitrogen deficiency), then subsequent image captures need to ensure that this specific area and plant can be clearly captured. At this time, these status images can be captured by various methods such as fixed cameras, drone patrols, and field mobile robots. The images should maintain a consistent shooting angle, lighting conditions, and resolution for subsequent comparison.

[0144] All collected state images are arranged into a time series according to their corresponding collection timestamps, from earliest to latest. For example, the sorted image sequence is: [T0_baseline image, T1_image 1 day after the measure, T2_image 3 days after the measure, T3_image 7 days after the measure, T4_image 14 days after the measure]. Each image is assigned a time index in the sequence for easy reference and comparison later. For example, T0 represents the baseline, T1 represents the first time point, and so on.

[0145] The sorted time series images are compared pairwise, usually by comparing images at adjacent time points (T0 vs T1, T1 vs T2, T2 vs T3, etc.), or by comparing images across time points (T0 vs T3, T1 vs T4, etc.).

[0146] If the image has semantic segmentation results (such as segmenting the image into "healthy leaves" and "yellowed leaves" regions), then the category and area changes of the same region at different time points can be directly compared. By comparison, specific changes in specific abnormal growth feature regions can be identified. For example: color changes: whether the yellowed area of ​​the leaves has decreased or new green has appeared; area changes: whether the area of ​​the yellowed leaves has expanded or shrunk; morphological changes: whether there have been changes in plant height and number of leaves; new anomalies: whether new abnormal features have appeared. These changes are recorded to form a description of the "changed parts". For example, "Compared to T0, the area of ​​the yellowed area of ​​the leaves decreased by 15% in T1", "Compared to T1, new green spots appeared in the center of the yellowed area of ​​the leaves in T2".

[0147] Associate the "changed parts" identified in the previous step with the nature of the abnormal growth characteristic itself (e.g., "yellowing leaves"), the crop type (e.g., "corn"), and the optimization measures taken (e.g., "foliar spraying of potassium dihydrogen phosphate"). Evaluate the effect: Combine crop growth knowledge to determine whether these changes are developing in the expected and healthy direction. For example, for the abnormality of "yellowing leaves" (nitrogen deficiency), if the changed parts show that the yellow area is reduced and the new leaves are green, then it can be determined that the optimization measures (fertilization) are effective; conversely, if the yellow area expands, it is ineffective or needs to be adjusted.

[0148] Generate dynamic graphs: Visualize state changes in a time series, which can be represented as: Time series image sets: Directly display images sorted by time, allowing users to intuitively feel the changes; Change heatmaps: Overlay heatmaps of color and area changes on images; Trend graphs: Plot the changes of quantitative indicators (such as the percentage of yellowed areas, plant height) over time into curves; Text / symbol reports: Describe key changes and conclusions in text.

[0149] Please see Figure 7 , Figure 7 This is a schematic diagram of the structural composition of an intelligent planting system for farmland based on the Internet of Things (IoT) according to an embodiment of the present invention; the intelligent planting system for farmland based on the IoT includes:

[0150] The crop module 21 is used to determine multiple planting areas based on the distribution map of farmland. In each planting area, the corresponding crop type and distribution of each crop are determined based on the detection of the planting area.

[0151] The maturity module 22 is used to determine the detection path of the planting area based on the distribution of each crop and the regional morphology of the planting area, and to detect each crop according to the detection path and the Internet of Things control corresponding to the farmland, so as to mark the maturity of each crop.

[0152] The planting dynamic map module 23 is used in the Internet of Things to determine sub-monitoring areas based on the maturity and location of each crop, mark abnormal growth characteristics of crops in the sub-monitoring areas, and determine the planting dynamic map of the farmland based on each abnormal growth characteristic and the maturity of each crop.

[0153] The planting optimization measures module 24 is used to determine abnormal planting areas in the planting dynamic map of farmland based on various abnormal growth characteristics and crop types, and to mark multiple intelligent planting nodes and corresponding planting optimization measures.

[0154] The dynamic optimization state diagram module 25 is used to construct planting optimization events for each intelligent planting node based on each intelligent planting node and the Internet of Things, and to record the dynamic optimization state diagram of each abnormal growth characteristic.

[0155] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A smart planting method for farmland based on the Internet of Things, characterized in that, include: Multiple planting areas are determined based on the farmland distribution map. Within each planting area, the types of crops and their distribution are determined based on the monitoring of the planting area. Based on the distribution of various crops and the regional morphology of the planting area, the detection path of the planting area is determined, and each crop is detected according to the detection path and the Internet of Things control corresponding to the farmland to mark the maturity of each crop. In this Internet of Things, sub-monitoring areas are determined based on the maturity and location of each crop, and abnormal growth characteristics of crops in the sub-monitoring areas are marked. Based on the abnormal growth characteristics and the maturity of each crop, a dynamic map of farmland planting is determined. In the dynamic map of farmland planting, abnormal planting areas are identified based on various abnormal growth characteristics and crop types, and multiple intelligent planting nodes and corresponding planting optimization measures are marked. Based on each intelligent planting node and the Internet of Things, planting optimization events for each intelligent planting node are constructed, and dynamic optimization state diagrams of each abnormal growth characteristic are recorded.

2. The intelligent planting method for farmland based on the Internet of Things according to claim 1, characterized in that, The process involves determining multiple planting areas based on a farmland distribution map, and within each planting area, determining the corresponding crop types and their distribution based on area monitoring. This includes: Collect farmland distribution maps, determine the location and types of multiple crops based on the detection of farmland distribution maps, and determine multiple planting areas based on the location and types of multiple crops and the regional boundary lines of farmland distribution maps. At this time, each planting area contains at least two different types of crops and no more than four different types of crops. Real-time monitoring of each planting area; determining the area detection method for each planting area based on its location, shape, and area; and triggering online detection of each planting area along the area detection method to mark the type and location of each crop. A first distribution map of planting areas is determined based on the type and location of each crop, and a second distribution map of planting areas is determined based on the quantity and location of each crop. The distribution of each crop is determined by combining the first and second distribution maps.

3. The intelligent planting method for farmland based on the Internet of Things according to claim 1, characterized in that, The detection path for each planting area is determined based on the distribution of various crops and the regional morphology of the planting area. Then, based on this detection path and the IoT control system corresponding to the farmland, each crop is detected to mark its maturity level. This includes: Collect the distribution information of each crop, determine the first sub-path based on the distribution information and types of crops, and determine the second sub-path based on the distribution information and regional morphology of the planting area. Collect farmland database, and match the corresponding detection path mapping relationship based on farmland database, planting area and crop type. Determine the detection path for the planting area according to the first sub-path, second sub-path and detection path mapping relationship. The corresponding detection drone is determined based on the detection path of the planting area and the matching of the Internet of Things. The detection drone detects the planting area along the detection path and collects multiple features of the crop during the flight. Based on the multiple features of the crop, the maturity of each crop is determined.

4. The intelligent planting method for farmland based on the Internet of Things according to claim 1, characterized in that, In this Internet of Things (IoT), sub-monitoring areas are determined based on the maturity and location of each crop, and abnormal growth characteristics of crops within these sub-monitoring areas are marked. A dynamic planting map of the farmland is then determined based on these abnormal growth characteristics and the maturity of each crop, including: Dynamic monitoring of various planting areas is achieved by using the Internet of Things to control and manage detection drones, collecting the location of each crop, determining the initial monitoring area based on the location, type and maturity of each crop, and determining the sub-monitoring area based on the initial monitoring area and the regional morphology of the planting area.

5. The intelligent planting method for farmland based on the Internet of Things according to claim 4, characterized in that, In this Internet of Things (IoT), sub-monitoring areas are determined based on the maturity and location of each crop, and abnormal growth characteristics of crops within these sub-monitoring areas are marked. A dynamic planting map of the farmland is then determined based on these abnormal growth characteristics and the maturity of each crop. The method also includes: In this sub-monitoring area, multiple abnormal growth three-dimensional spaces are identified based on the anomaly detection of the sub-monitoring area. Each abnormal growth three-dimensional space corresponds to at least one crop. The abnormal growth characteristics of the crop are determined based on the three-dimensional detection of each abnormal growth three-dimensional space by the detection drone. Anomaly dynamic maps of crops are constructed based on their abnormal growth characteristics, corresponding maturity levels, and locations. Planting dynamic maps of farmland are then determined based on the abnormal dynamic maps of each crop, as well as the planting areas and times.

6. The intelligent planting method for farmland based on the Internet of Things according to claim 1, characterized in that, In the dynamic planting map of the farmland, abnormal planting areas are identified based on various abnormal growth characteristics and crop types, and multiple intelligent planting nodes and corresponding planting optimization measures are marked, including: Real-time monitoring of farmland planting dynamics map, marking the location of each abnormal growth feature, and determining the first region coefficient based on the location and type of each abnormal growth feature; The second regional coefficient is determined based on the location of each abnormal growth characteristic and the type of crop; the abnormal planting area is determined based on the matching of the planting dynamic map of the farmland, the first regional coefficient and the second regional coefficient.

7. The intelligent planting method for farmland based on the Internet of Things according to claim 6, characterized in that, The aforementioned dynamic planting map of farmland identifies abnormal planting areas based on various abnormal growth characteristics and crop types, and marks multiple intelligent planting nodes and corresponding planting optimization measures. It also includes: In the abnormal planting area, the location to be optimized is determined based on the detection of the abnormal planting area. The corresponding intelligent planting node is determined according to the location to be optimized, the abnormal growth characteristics and the type of crop. Multiple intelligent planting nodes are collected. The corresponding planting optimization measures are determined by matching multiple intelligent planting nodes with the farmland database.

8. The intelligent planting method for farmland based on the Internet of Things according to claim 1, characterized in that, The process involves constructing planting optimization events for each intelligent planting node based on the Internet of Things (IoT) and recording dynamic optimization state diagrams for various abnormal growth characteristics, including: The system collects IoT data from farmland and triggers corresponding detection drones based on this data. These drones then monitor the execution of planting optimization measures at intelligent planting nodes in real time, collecting multiple execution images of these measures. Multiple sub-planting optimization events are determined based on the recognition of multiple execution images, and planting optimization events for each intelligent planting node are constructed by synthesizing multiple sub-planting optimization events and their corresponding execution times.

9. The intelligent planting method for farmland based on the Internet of Things according to claim 8, characterized in that, The method of constructing planting optimization events for each intelligent planting node based on each intelligent planting node and the Internet of Things, and recording a dynamic optimization state diagram of each abnormal growth characteristic, also includes: In this planting optimization event, state images of various abnormal growth features at different times are collected and sorted according to time. Based on the sequential comparison of each state image, multiple changes of each abnormal growth feature are determined. Based on the multiple changes, the corresponding abnormal growth features, and the type of crop, the dynamic optimization state map of each abnormal growth feature is determined.

10. An intelligent planting system for farmland based on the Internet of Things, characterized in that, The IoT-based intelligent planting system for farmland is applied to the IoT-based intelligent planting method for farmland as described in any one of claims 1-9, wherein the IoT-based intelligent planting system for farmland includes: The crop module is used to determine multiple planting areas based on the farmland distribution map. Within each planting area, the corresponding crop types and their distribution are determined based on the detection of the planting area. The maturity module is used to determine the detection path of the planting area based on the distribution of each crop and the regional morphology of the planting area, and to detect each crop according to the detection path and the Internet of Things control corresponding to the farmland, so as to mark the maturity of each crop. The planting dynamic map module is used in this Internet of Things to determine sub-monitoring areas based on the maturity and location of each crop, mark abnormal growth characteristics of crops in the sub-monitoring areas, and determine the planting dynamic map of the farmland based on each abnormal growth characteristic and the maturity of each crop. The planting optimization measures module is used to identify abnormal planting areas in the dynamic planting map of farmland based on various abnormal growth characteristics and crop types, and to mark multiple intelligent planting nodes and corresponding planting optimization measures. The dynamic optimization state graph module is used to construct planting optimization events for each intelligent planting node based on each intelligent planting node and the Internet of Things, and to record the dynamic optimization state graph of each abnormal growth characteristic.

Images