A sowing method for improving the uniformity of sowing of light forage grass seeds

Abstract

This invention discloses a sowing method for improving the uniformity of light forage seed dispersal, relating to the field of intelligent agricultural sowing technology. The method includes: S1, obtaining soil moisture weight and temperature influence factor data for the target plot from a preset database; constructing an initial distribution map by integrating soil moisture boundaries and temperature distribution differences; calibrating the data according to regional crop characteristics to determine the basic influence parameters of seed drift within the plot; S2, based on the basic influence parameters of seed drift, combined with terrain undulation constraints and drift influence gradients, dividing the plot into regions; and adjusting the regional overlap ratio through boundary dynamic correction technology to obtain a preliminary zoning layout. This sowing method for improving the uniformity of light forage seed dispersal significantly enhances the accuracy and efficiency of large-area grassland sowing operations and is suitable for high-precision forage planting scenarios such as degraded grassland restoration and ecological engineering construction.

Description

Technical Field

[0001] This invention relates to the field of intelligent agricultural sowing technology, specifically to a sowing method for improving the uniformity of sowing lightweight forage seeds. Background Technology

[0002] In current grassland ecological restoration and livestock development, forage planting plays a crucial role as a fundamental link in improving grassland productivity and ecological carrying capacity. Especially in large-scale restoration operations of degraded grasslands, the ability to accurately achieve the expected forage yield target has become a key indicator for measuring planting quality and project effectiveness.

[0003] Traditional forage seed sowing typically employs terrestrial forage seeders. While these machines offer efficiency advantages in flat, accessible areas, they suffer from limitations such as operational constraints, difficulty in route planning, and the risk of repeated or missed sowing when operating in mountainous, sloping, or large, irregularly shaped plots. To improve sowing efficiency, drones have been introduced for aerial sowing in recent years. Drones offer advantages such as rapid seed distribution, adaptability to complex terrain, and labor savings, making them a popular new method for large-scale sowing. However, due to the lightweight, small-particle, and easily drifting characteristics of forage seeds, drone sowing is prone to secondary diffusion, deviation, and even accumulation of seeds due to the combined effects of the downwash from the drone rotor and high-altitude wind disturbances. This results in an uneven distribution of seeds on the ground, characterized by "localized high density and large sparse areas," leading to uneven germination, significant growth differences, and drastic yield fluctuations. This uneven sowing phenomenon is particularly severe in large-scale sowing tasks and has become a major technical bottleneck restricting the widespread application of drone sowing technology. Furthermore, even with the same sowing density, the germination and survival rates of seeds can vary significantly across different regions due to differences in microenvironments such as soil moisture, temperature, and topographic relief. This means that forage seed sowing operations not only face the problem of "unstable physical landing points" but also need to overcome the challenge of "uncertain seedling emergence." Traditionally experience-based sowing parameters are insufficient to guarantee the accuracy and uniformity of sowing on a large scale. Summary of the Invention

[0004] The purpose of this invention is to provide a sowing method for improving the uniformity of light forage seed dispersal, thereby solving the problems existing in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a sowing method for improving the uniformity of light forage seed sowing, comprising:

[0006] S1. Obtain soil moisture weight and temperature influence factor data of the target plot from the preset database, construct an initial distribution map by integrating soil moisture boundary and temperature distribution differences, perform data calibration for regional crop characteristics, and determine the basic influence parameters of seed drift within the plot;

[0007] S2. Based on the basic influence parameters of seed drift, combined with terrain undulation constraints and drift influence gradient, the plots are divided into regions. The region overlap ratio is adjusted through boundary dynamic correction technology to obtain the preliminary zoning layout.

[0008] S3. For the initial zoning layout, extract the soil moisture weight and temperature influence factor for each zoning, combine historical data comparison and yield prediction benchmark, calculate the expected yield fluctuation range for each zoning, and determine if the fluctuation range exceeds the preset threshold to trigger boundary smoothing processing and determine the optimized zoning structure.

[0009] S4. Through the optimized partitioning structure, analyze the environmental disturbance variables and growth cycle parameters of each partition, combine the regional crop characteristics, assess the local seed accumulation risk points, and obtain the preliminary density distribution of seed distribution.

[0010] S5. For the initial density distribution, detect the coordinates of the missing areas under the balanced area state of the partition, and perform density correction by superimposing the gradient data affected by drift. If the density distribution is still uneven after correction, adjust the region overlap ratio through boundary dynamic correction to obtain the final density distribution.

[0011] S6. For the final density distribution, obtain soil moisture boundary and temperature distribution difference data, and integrate the data by overlay to obtain the path planning basis;

[0012] Based on the path planning, a drone seeding path is generated. If the path does not cover all zones, the missing areas are identified, and adjustments are determined.

[0013] As can be seen from the above technical solution, the present invention has the following beneficial effects:

[0014] This sowing method, designed to improve the uniformity of light forage seed dispersal, constructs an initial sowing distribution map by acquiring soil moisture weights and temperature influence factors, combined with topographic relief and seed drift characteristics. Through steps such as region division, density assessment, fluctuation judgment, and dynamic boundary correction, the sowing parameters are progressively refined to achieve differentiated sowing control for different sub-regions. Finally, by combining the generated density distribution data and path coverage algorithm, a complete instruction sequence for UAV sowing operations is output. This method effectively reduces seed accumulation or omissions caused by airflow interference and environmental heterogeneity, improves spatial uniformity of sowing and consistency of forage seedling emergence, and significantly enhances the accuracy and efficiency of large-area grassland sowing operations. It is suitable for high-precision forage planting scenarios such as degraded grassland restoration and ecological engineering construction. Attached Figure Description

[0015] Figure 1 This is a flowchart of a sowing method for improving the uniformity of light forage seed sowing according to the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] like Figure 1 As shown, the present invention provides a technical solution: a sowing method for improving the uniformity of light forage seed sowing, comprising:

[0018] S1. Obtain soil moisture weight and temperature influence factor data of the target plot from the preset database, construct an initial distribution map by integrating soil moisture boundary and temperature distribution differences, perform data calibration for regional crop characteristics, and determine the basic influence parameters of seed drift within the plot;

[0019] S2. Based on the basic influence parameters of seed drift, combined with terrain undulation constraints and drift influence gradient, the plots are divided into regions. The region overlap ratio is adjusted through boundary dynamic correction technology to obtain the preliminary zoning layout.

[0020] S3. For the initial zoning layout, extract the soil moisture weight and temperature influence factor for each zoning, combine historical data comparison and yield prediction benchmark, calculate the expected yield fluctuation range for each zoning, and determine if the fluctuation range exceeds the preset threshold to trigger boundary smoothing processing and determine the optimized zoning structure.

[0021] S4. Through the optimized partitioning structure, analyze the environmental disturbance variables and growth cycle parameters of each partition, combine the regional crop characteristics, assess the local seed accumulation risk points, and obtain the preliminary density distribution of seed distribution.

[0022] S5. For the initial density distribution, detect the coordinates of the missing areas under the balanced area state of the partitions, and perform density correction by superimposing the gradient data of the drift influence. If the density distribution is still uneven after correction, adjust the area overlap ratio through boundary dynamic correction to obtain the final density distribution. S6. Based on the final density distribution, integrate the soil moisture boundary and temperature distribution difference data to generate the UAV seeding path plan. Analyze whether the path covers all partitions. If not, adjust the path trajectory to obtain a complete operation instruction sequence.

[0023] In the above implementation, a pre-set database records soil moisture weights and temperature influence factors obtained from long-term monitoring of the target plots. The system first uses this data to construct an initial distribution map, and then further calibrates the data in conjunction with local crop characteristics to make the drift influence parameters more closely match the actual sowing environment. Based on the calibrated parameters, the system divides the region under the constraints of topographic relief and drift gradient, and uses boundary dynamic correction technology to control the region overlap ratio, ensuring that each region maintains spatial independence and continuity. After obtaining the preliminary zoning layout, the system extracts the moisture weights and temperature factors of each region and calculates the expected yield fluctuation range in conjunction with historical yield data. If the fluctuation exceeds a threshold, it indicates that the region division is unstable, and the system automatically performs boundary smoothing to form a more optimal zoning structure. Subsequently, based on the optimized regional structure, the system further analyzes environmental interference factors (including wind speed, airflow turbulence, topographic shading, etc.) and crop growth cycle to locate potential seed accumulation risk points and form a preliminary density distribution map. After obtaining the preliminary density distribution, the system detects the regional area balance and identifies the coordinates of missing areas, and then corrects the density by superimposing the drift influence gradient. If uneven density still exists after correction, the boundary dynamic correction technology is invoked again to adjust the overlap ratio to form a more accurate final density distribution. Finally, the system generates a drone seeding path plan based on this density distribution and humidity and temperature difference data to ensure that the trajectory completely covers all zones; if there is a gap in coverage, the system replans the trajectory and outputs a complete sequence of operation instructions, enabling the drone to deliver highly uniform forage seeds according to the optimized path.

[0024] S1 includes obtaining soil moisture weight and temperature influence factor data of the target plot from a preset database, constructing an initial distribution map by integrating soil moisture boundary and temperature distribution differences; calibrating the data of the initial distribution map according to regional crop characteristics, obtaining plot slope factor data from the preset database and fusing it into the initial distribution map; determining the seed drift path based on the fused distribution map, and adjusting the moisture weight and temperature influence factor if the path deviation exceeds a preset threshold; and determining the basic influence parameters of seed drift within the plot through the adjusted map.

[0025] In this implementation, the target plot is first divided into several grid cells of equal area before implementation. A record is created for each grid cell in a preset database. The record includes the geographical coordinates of the grid cell, soil moisture values ​​measured at multiple time points, surface temperature values ​​measured at multiple time points, and slope factor values ​​calculated from previous elevation measurements. During the execution step, the set of soil moisture and surface temperature values ​​closest to the current sowing time is read from the preset database. The soil moisture values ​​of each grid cell are then normalized by calculating the minimum and maximum soil moisture values ​​among all grid cells. The soil moisture weight is calculated by subtracting the minimum soil moisture value from the soil moisture value of each grid cell, and then dividing by the difference between the maximum and minimum soil moisture values. This yields a soil moisture weight between 0 and 1, where a weight of 1 indicates that the soil in that grid cell is saturated. Simultaneously, based on the known optimal upper and lower moisture limits for the target crop in the region, the lower limit is used as the lower boundary of the soil moisture boundary, and the upper limit as the upper boundary. Soil moisture weights within the boundary range are marked as favorable for seed germination, while those outside the boundary range are marked as unfavorable for seed germination. The determination of temperature influence factors follows a similar method. The average current surface temperature, the optimal upper and lower temperature limits for the target crop are calculated across all grid cells. The surface temperature of each grid cell is compared to the optimal temperature range. When the temperature falls between the optimal upper and lower temperature limits, a temperature influence factor of 1 is assigned to that grid cell. When the temperature is below the optimal lower temperature limit or above the optimal upper temperature limit, the temperature influence factor is decreased proportionally to the deviation from the optimal range until it reaches 0. By setting the temperature influence factor for each grid cell, a temperature index representing the degree of favorable temperature conditions is formed. The distribution of influencing factors is determined, and the difference between the temperature influencing factor values ​​of adjacent grid cells is used as the temperature distribution difference to represent the range of temperature condition changes between different locations. After the soil moisture weight, soil moisture boundary, temperature influencing factor and temperature distribution difference are determined, the geographic coordinates, soil moisture weight, soil moisture boundary, temperature influencing factor and temperature distribution difference of each grid cell are stored in memory according to the grid location order and visualized and mapped to generate an initial distribution map on a two-dimensional plane. Each grid cell in the initial distribution map corresponds to a display cell, and different colors or gray values ​​are used to represent the comprehensive environmental state of the grid cell.

[0026] Then, based on the specific crop type planted in the target plot, the regional crop characteristics corresponding to that crop are read from a preset database. These regional crop characteristics include the optimal range of soil moisture and surface temperature during the sowing period, as well as the requirements for sowing density during the initial growth period. The optimal moisture and temperature ranges in the regional crop characteristics are compared one by one with the soil moisture weights and temperature influence factors of each grid cell in the initial distribution map. When the actual moisture value corresponding to the soil moisture weight of a certain grid cell is higher than the upper limit of the optimal moisture value of the regional crop characteristics, the soil moisture weight of that grid cell is adjusted to the weight corresponding to the upper limit of the optimal moisture value; when it is lower than the lower limit of the optimal moisture value, it is adjusted to the weight corresponding to the lower limit of the optimal moisture value. When the actual temperature value corresponding to the temperature influence factor of a certain grid cell is higher than the upper limit of the optimal temperature value of the regional crop characteristics or lower than the lower limit of the optimal temperature value, the temperature influence factor of that grid cell is adjusted to a value that matches the corresponding boundary temperature value. Through the above calibration process, the values ​​in the initial distribution map are made consistent with the regional crop characteristics, resulting in a calibrated initial distribution map.

[0027] After calibration, the slope factor data for each grid cell is retrieved from the preset database. The slope factor is a dimensionless value calculated based on the elevation change within the grid cell; a larger value indicates a steeper slope. To facilitate calculations in conjunction with soil moisture weighting and temperature influence factors, the slope factor is normalized. This normalization process involves calculating the minimum and maximum slope factors across the entire field, subtracting the minimum slope factor from the slope factor of each grid cell, and then dividing by the difference between the maximum and minimum slope factors to obtain a normalized slope factor between 0 and 1. Then, the normalized slope factor is fused with the calibrated initial distribution map, and a normalized slope factor field is added to the record of each grid cell to form the fused distribution map. In the fused distribution map, to determine the seed drift path, for each target grid cell where seeds are planned to be released, the main direction of seed movement along the ground surface is first determined according to the magnitude of the normalized slope factor of the grid cell and the directional relationship of the normalized slope factors of the surrounding grid cells. The direction of the adjacent grid cell with the largest slope value is taken as the seed slip direction. The normalized slope factor is multiplied by the plot scale to obtain the seed drift path. The sliding distance of the seed is calculated, and then the sliding direction and sliding distance are applied to the initial landing position of the seed to calculate the grid cell where the next possible landing point is located. After obtaining the next grid cell, the soil moisture weight, soil moisture boundary, and temperature influence factor of the grid cell are read. When the actual moisture value corresponding to the soil moisture weight of the grid cell is within the soil moisture boundary range and the temperature influence factor is close to 1, the grid cell is determined as the final landing point of the seed, and the coordinates of the grid cell are recorded as the endpoint on the seed drift path. When the above conditions are not met, the sliding calculation is repeated in the direction indicated by the normalized slope factor until a grid cell that meets the conditions is encountered or the plot boundary is reached. By recording the coordinates of all grid cells that the seed passes through from the release position to the final landing point in chronological order, a seed drift path composed of multiple landing points is obtained. The above calculation process is repeated for all target grid cells to obtain multiple seed drift paths. The distance between the coordinates of all landing points on each seed drift path and the center coordinates of the corresponding target grid cell is calculated to obtain the average offset distance from all landing points in each path to the center of the grid cell. This average offset distance is used as the path deviation value.

[0028] During the method deployment phase, multiple sowing experiments were conducted on similar plots to record seed drift paths formed under different sowing heights and operating conditions. The maximum average offset distance of each seed drift path was calculated when actual germination was nearly uniform. This maximum value was determined as a preset threshold and stored in a preset database. During actual sowing, the average offset distance of each seed drift path was compared with the preset threshold. If the average offset distance exceeded the preset threshold, the path was considered to have a path deviation. When a path deviation occurred, the soil moisture weight and temperature influence factor of all grid cells related to that path were adjusted. For grid cells with excessive seed concentration near the path endpoint and an actual seeding density significantly exceeding the target sowing density, the soil moisture weight and temperature influence factor values ​​were reduced by a certain percentage according to the degree of concentration. For grid cells with significantly fewer seeds in the middle section of the path… According to the number of missing seeds, a certain proportion of soil moisture weight and temperature influence factor values ​​are increased. Through the above increase and decrease operations, the attraction of different grid cells to the seed landing point is redistributed. After completing one adjustment of moisture weight and temperature influence factor, the fused distribution map is regenerated and the seed drift path is re-evaluated. The path deviation is recalculated and compared with the preset threshold. If there is still a path deviation, the adjustment process continues until the average offset distance of all seed drift paths is no greater than the preset threshold. At this time, in the adjusted map, the soil moisture weight, temperature influence factor and normalized slope factor of each grid cell together constitute the basic influence parameters of seed drift within the plot. The basic influence parameters include the drift direction parameter reflecting the orientation of the seed landing point, the drift distance parameter reflecting the horizontal distance of the seed from the release position to the final landing point, and the drift concentration parameter reflecting the concentration of seeds in adjacent grid cells.

[0029] The optimal humidity range refers to the soil moisture content interval that ensures normal seed water absorption and swelling, and the initiation of physiological metabolic processes, during the sowing and germination stages of the target crop. This range is determined by long-term planting experience of the crop in the target area, agronomic parameters, or existing experimental data, and is pre-stored in a database as an optimal lower and upper humidity limit. Below the optimal lower humidity limit, the soil moisture is insufficient, making it difficult for seeds to complete the normal water absorption process, resulting in a reduced germination rate. Above the optimal upper humidity limit, the soil pore water content is too high, easily leading to an oxygen-deficient environment, which is also unfavorable for seed germination. Therefore, this range is used as the soil moisture boundary to determine whether soil moisture conditions in different grid cells are conducive to seed germination, and to mark or calibrate the soil moisture weight accordingly, which helps to spatially distinguish between areas favorable and unfavorable to seed germination.

[0030] The optimal temperature range refers to the range of surface temperatures within which the target crop can maintain seed physiological activity and promote normal growth of the radicle and plumule during the sowing and initial growth stages. This range is also predetermined and stored in the form of an optimal lower limit and an optimal upper limit. When the surface temperature is within this range, the enzymatic reactions and metabolic activities inside the seed are in a relatively stable state, which is conducive to uniform emergence. When the temperature is below the optimal lower limit, the metabolic rate slows down, and germination is delayed. When the temperature is above the optimal upper limit, physiological stress is easily triggered, and even seed inactivation may occur. Based on the above mechanism, by comparing the actual surface temperature of each grid cell with the optimal temperature range and assigning different temperature influence factors accordingly, the degree of benefit of temperature conditions at different spatial locations to seed germination and early growth can be quantitatively represented, providing a basis for subsequent seed drift path judgment and sowing regulation.

[0031] S2 includes: obtaining drift influence gradient data from a pre-set database based on the basic influence parameters of seed drift and the constraints of terrain undulation; integrating the parameter values ​​by weighted averaging the basic influence parameters and terrain undulation constraints to obtain a gradient constraint matrix; dividing the land parcels into regions using the gradient constraint matrix; determining the range of the division boundary lines by integrating the drift influence gradients and using a boundary expansion method; using boundary dynamic correction technology to extract the regional overlap ratio data from the division boundary lines; adjusting the overlap ratio to correct the boundary by scaling to obtain a corrected boundary set; integrating zoning parameter calibration data from the pre-set database based on the corrected boundary set; determining that if the overlap ratio is lower than a pre-set threshold, integrating the layout boundary optimization by boundary smoothing to determine the optimized zoning map; and using the optimized zoning map and constraint gradient fusion to process the data by gradient superposition to obtain preliminary zoning layout data, resulting in the preliminary zoning layout of the land parcels.

[0032] In this embodiment, after obtaining the seed drift basic influence parameters, the target plot is first divided into multiple grid cells arranged in rows and columns along the east-west and north-south directions according to fixed grid side lengths. A record is created for each grid cell in a preset database, storing the center geographic coordinates of the grid cell, the seed drift basic influence parameters determined in step S1, and the terrain relief constraint values ​​obtained during the terrain survey stage. Drift influence gradient data between each grid cell and its adjacent grid cells is also pre-stored in the preset database. The seed drift basic influence parameters include drift direction parameters, drift distance parameters, and drift concentration parameters. The drift direction parameter is the dominant drift direction angle value obtained by statistically analyzing multiple simulated drift paths in step S1. The drift distance parameter is the average distance from the seed release position to the final landing point. The drift concentration parameter is the average number of landing points per unit area near the grid cell and the target seeding. The density ratio and terrain undulation constraint value are obtained by calculating the difference between the highest and lowest elevations within the grid cell and normalizing it with the maximum elevation difference of the entire plot. The normalization process involves subtracting the minimum elevation difference across the entire field from the elevation difference of each grid cell and then dividing by the difference between the maximum and minimum elevation differences across the entire field, ensuring that the terrain undulation constraint value falls within the range of 0 to 1. The drift influence gradient data are the difference data obtained when comparing the seed drift base influence parameters of each pair of adjacent grid cells during the method deployment phase. During the deployment phase, the drift direction parameter difference, drift distance parameter difference, and drift concentration parameter difference are calculated for each pair of adjacent grid cells. The three differences are weighted, summed, and normalized to obtain a single drift influence gradient value. The larger this value, the more significant the difference in seed drift characteristics between adjacent grid cells. This drift influence gradient value is stored in a preset database and read by grid index when executing step S2.In step S2, the drift direction parameter, drift distance parameter, drift concentration parameter, terrain undulation constraint value, and drift influence gradient value adjacent to the grid cell are read sequentially from the preset database. The drift direction parameter, drift distance parameter, and drift concentration parameter are normalized according to their minimum and maximum values ​​in the entire field. The processing method is to calculate the minimum and maximum values ​​of each type of parameter in the entire field, subtract the minimum value from the actual value of the parameter of that type in the grid cell, and then divide by the difference between the maximum and minimum values, so that the normalized drift direction influence value, drift distance influence value, and drift concentration influence value all fall between 0 and 1. Then, fixed weight values ​​are set for the seed drift basic influence parameter and terrain undulation constraint. During the deployment phase, multiple tests are conducted on different plots, and after comparing the evaluation index of seeding uniformity under different weight combinations, a set of weight values ​​is selected to determine the drift direction. The weight ratios corresponding to the influence values, drift distance influence values, drift concentration influence values, and terrain undulation constraint values ​​are assigned to the respective values ​​and written into a preset database, where they remain unchanged during execution. When calculating the comprehensive constraint value of each grid cell, the drift direction influence value, drift distance influence value, drift concentration influence value, and terrain undulation constraint value of the grid cell are multiplied by their respective weights. These four products are then added together to obtain the comprehensive constraint value of the grid cell. This comprehensive constraint value reflects the combined constraint strength of seed drift characteristics and terrain undulation on the seeding effect at that location. The comprehensive constraint values ​​of all grid cells are arranged in order of their row and column numbers in the plot to form a gradient constraint matrix. Each row in the matrix corresponds to a grid band in the east-west direction of the plot, and each column corresponds to a grid band in the north-south direction of the plot.

[0033] After constructing the gradient constraint matrix, to divide the land parcels into regions based on the gradient constraint matrix and drift influence gradient data, the first step in the deployment phase is to select a region division difference upper limit parameter based on historical test records. This parameter is determined by using different comprehensive constraint difference upper limits for region division on multiple sample land parcels, quantitatively evaluating the uniformity of seeding within each region corresponding to the difference upper limit, and selecting a fixed value for the difference upper limit parameter that ensures minimal variation in the comprehensive constraint values ​​within most land parcels without resulting in an excessive number of regions. This value is then written into a preset database. During execution, a grid cell without assigned region identifiers is selected from the gradient constraint matrix as a seed cell, and its comprehensive constraint value is recorded as the current region seed value. Then, the grid cells above, below, to the left, and to the right of the seed cell are checked, and calculations are performed for each adjacent cell. The difference between the comprehensive constraint value and the seed value is calculated. When the difference does not exceed the upper limit parameter of the region division difference, the adjacent cell is assigned to the current region and the adjacent cell is used as the new expansion cell. The above comparison process is repeated until the difference between all adjacent cells is greater than the upper limit parameter of the region division difference. During this expansion process, the drift influence gradient value between the new cell and the outer adjacent cell is read synchronously each time a new cell is added to the current region. The grid boundary with the larger drift influence gradient value is marked as a region boundary candidate. When the expansion stops, the region is spatially separated from the adjacent undivided regions along all grid boundaries marked as region boundary candidates, thereby generating the division boundary line corresponding to the region. The above expansion and separation method is repeated for all grid cells that have not yet been assigned a region identifier until all grid cells are assigned to a certain region and the corresponding division boundary line is generated, thus completing the initial region division.After obtaining all the boundary lines, to control the overlap ratio between adjacent areas and correct the boundaries, firstly, the sets of grid cells belonging to the left and right regions are determined on both sides of each boundary line. The area ratios of the boundary zones corresponding to the left and right regions are obtained by calculating the ratio of the sum of the areas of the grid cells adjacent to the boundary in the left region to the total area of ​​the left region, and the sum of the areas of the grid cells adjacent to the boundary in the right region to the total area of ​​the right region. Then, the ratio of the overlapping area of ​​the grid cells on both sides of the boundary line that overlap spatially to the total area of ​​the smaller region is determined as the area overlap ratio data for that pair of adjacent regions. This area overlap ratio data is recorded in memory for later use. During the deployment phase, statistics from multiple test plots show that when the area overlap ratio is below a certain fixed ratio, the boundaries are easily too sharp, and the spreading path frequently starts and stops near the boundary, leading to decreased operational efficiency. When the area overlap ratio is above a certain fixed ratio, the division between regions becomes unclear, and the spreading control strategy is difficult to distinguish. Based on this, during the deployment phase... The minimum allowable overlap ratio that simultaneously balances boundary smoothness and regional independence is selected as the preset threshold. This preset threshold is written into the preset database and remains unchanged during execution. During actual execution, the regional overlap ratio data of each pair of adjacent regions is compared with the preset threshold. When the regional overlap ratio is significantly higher than the preset threshold, the boundary is expanded outward according to the scaling method. That is, based on the current boundary, the boundary control points are moved at equal distances along the internal direction of the region on both sides of the boundary line, so that the proportion of the overlapping part of the two regions in the grid cells near the boundary is reduced to close to the preset threshold. When the regional overlap ratio is lower than the preset threshold, the boundary control points are moved in the direction of the overlapping region between the two regions according to a certain ratio, based on the current boundary, so that the area proportion of the overlapping region on both sides of the boundary is increased to no less than the preset threshold. During the above movement, the total area of ​​each region is kept within a certain proportion of the total area to prevent the region size from changing too much. After multiple iterations of movement, a set of corrected boundaries that meet the preset overlap ratio requirements is obtained.

[0034] Subsequently, based on the corrected boundary set, partition parameter calibration data is read from the preset database. This data is obtained during the deployment phase by statistically analyzing historical operation records and includes at least the target area range, target seeding density range, and target path complexity level for each type of region. During execution, the actual area and region overlap ratio of each region are compared with the target range in the partition parameter calibration data. If the region overlap ratio of a certain region is not lower than the preset threshold, but the boundary exhibits multiple sharp folds in space, resulting in a long and narrow or strip-shaped region, it is determined that this region will have high path complexity in subsequent path planning, and boundary smoothing is initiated. The method involves sequentially selecting multiple boundary points in space for the corresponding boundary lines in the modified boundary set at a fixed step size. The coordinates of each adjacent set of boundary points are averaged to obtain the coordinates of new smooth boundary points. The original boundary points are replaced with the new smooth boundary points, and the area is recalculated after smoothing. If the area deviation exceeds the preset allowable area deviation range, the boundary position is fine-tuned by adding or removing a small number of boundary points on the outside or inside of the boundary to restore the area to the allowable area deviation range. By performing the above smoothing and fine-tuning operations on all boundary lines in sequence, an optimized partition map with continuous boundary shape, fewer corners, and meeting the area and overlap ratio requirements is finally obtained. Finally, based on the optimized zoning map, a constraint gradient fusion process is performed. The comprehensive constraint values ​​in the gradient constraint matrix are correlated with the optimized zoning results. All grid cells in each region are traversed in row and column order, and the comprehensive constraint values ​​of all grid cells in the region are accumulated one by one to obtain the comprehensive constraint sum of the region. At the same time, the number of grid cells in the region is recorded, and the average constraint intensity of the region is obtained by dividing the comprehensive constraint sum by the number of grid cells. Then, the comprehensive constraint values ​​of each row and column in the region are accumulated in the east-west and north-south directions respectively to obtain the gradient distribution curves in the two directions within the region. The average constraint intensity, the characteristic values ​​of the gradient distribution curves in the two directions, as well as the geometric boundary coordinates, area, and one-dimensional index number of the region are combined to form the zoning layout data of the region. The zoning layout data of all regions are merged to form the preliminary zoning layout data. The preliminary zoning layout data fully records the location range and internal constraint gradient characteristics of each region of the plot, serving as the preliminary zoning layout of the plot.

[0035] S3 includes: for the initial zoning layout, obtaining the soil moisture weight and temperature influence factor of each zone from a preset database; calculating a comprehensive environmental index by fusing the soil moisture weight and the temperature influence factor in a weighted manner to obtain a zoning environmental matrix; based on the zoning environmental matrix, extracting the yield prediction benchmark from the preset database by comparing with historical data; calculating the expected yield fluctuation range by superimposing the comprehensive environmental index and the yield prediction benchmark to determine the fluctuation range set; for the fluctuation range set, determining if the fluctuation range exceeds a preset threshold, triggering a crop adaptability assessment; obtaining adaptability parameters from the fluctuation range set; processing the data by integrating the adaptability parameters in a way that matches crop type to obtain adaptability assessment data; based on the adaptability assessment data, integrating irrigation adjustment plans using boundary smoothing processing; correcting the boundaries in a smooth manner by expanding the adaptability assessment data; and fusing the soil moisture weight to adjust the boundary curve to determine the corrected boundary group; for the corrected boundary group, fusing zoning parameter calibration data from the preset database; and processing the layout by superimposing the corrected boundary group and the zoning parameter calibration data to determine the optimized zoning structure.

[0036] In this embodiment, firstly, based on the preliminary zoning layout obtained in step S2, it is determined which grid units constitute each zoning. The corresponding soil moisture weight and temperature influence factor are stored for each grid unit in a preset database, and the zoning number to which the grid unit belongs is recorded. During execution, all grid units are grouped according to the zoning number. The soil moisture weight and temperature influence factor of each grid unit within each zoning are read one by one. The zoning soil moisture weight of the zoning is obtained by calculating the arithmetic mean of the soil moisture weights of all grid units within the zoning, and the zoning temperature influence factor of the zoning is obtained by calculating the arithmetic mean of the temperature influence factors of all grid units within the zoning. During the deployment phase, the influence coefficients of the soil moisture weight and temperature influence factor in the overall environment are determined in advance based on multiple yield statistics of similar plots. These two influence coefficients are written into the preset database as fixed weighting coefficients and remain unchanged during execution. When calculating the overall environmental index of each zoning, the zoning soil moisture weight of the zoning is first multiplied by the fixed moisture weighting coefficient, and then the zoning temperature influence factor of the zoning is multiplied by... A fixed temperature weighting coefficient is used, and the two products are added together to obtain the comprehensive environmental index of the zone. The larger the comprehensive environmental index value, the closer the humidity and temperature conditions of the zone are to the optimal growth range of the target crop. The comprehensive environmental indices of all zones are filled into a row-column arranged data table according to the spatial location of the zones in the plot. Each row in the data table corresponds to an east-west zone, and each column corresponds to a north-south zone. Each table position stores the comprehensive environmental index of the corresponding zone. This data table constitutes the zone environmental matrix and is stored in memory for subsequent calculations. After the zone environmental matrix is ​​obtained, the historical yield data and environmental record data corresponding to each zone position are read from the preset database according to the zone number. In the deployment phase, the historical comprehensive environmental index has been obtained by performing the same weighting calculation on the soil moisture weight and temperature influence factor in the historical environmental record data. The average yield value and yield fluctuation range in each segment are statistically calculated according to the comprehensive environmental index. The average yield value and yield fluctuation width corresponding to each comprehensive environmental index segment are stored in the preset database as the yield prediction benchmark.During execution, based on the comprehensive environmental index of each partition, the yield prediction benchmark falling within the same segment as that comprehensive environmental index is found in the preset database. The corresponding benchmark yield value and benchmark fluctuation width are obtained. The deviation between the benchmark yield value and the current comprehensive environmental index of the partition is superimposed. The ratio of the difference between the current comprehensive environmental index and the comprehensive environmental index at the center of the segment to the width of that segment is multiplied by the benchmark fluctuation width to obtain the additional fluctuation amount for that partition. The benchmark fluctuation width and the additional fluctuation amount are added to obtain the expected yield fluctuation width for that partition. The expected minimum yield for that partition is determined by subtracting the expected yield fluctuation width from the benchmark yield value, and the expected maximum yield is determined by adding the expected yield fluctuation width to the benchmark yield value. The expected minimum yield, expected maximum yield, and expected yield fluctuation width for each partition are recorded together. The above data set for all partitions constitutes the fluctuation range set. During the deployment phase, statistical analysis of a large number of historical plots is conducted to find the maximum allowable value of the expected yield fluctuation width for each partition, provided that the germination rate and final yield balance meet agronomic requirements. This maximum value is determined as a fixed preset threshold and written into the preset database, and is not adjusted during execution.

[0037] For the set of fluctuation ranges, the expected yield fluctuation width of each partition is compared with a preset threshold one by one. When the expected yield fluctuation width of a partition is greater than the preset threshold, the partition is marked as a partition requiring crop adaptability assessment. Adaptability parameters are extracted from the fluctuation range data corresponding to that partition. Adaptability parameters include at least the ratio of the expected yield fluctuation width to the preset threshold, the difference between the expected minimum yield and the target standard yield, and the difference between the expected maximum yield and the target standard yield. The target standard yield is determined as a fixed value based on the multi-year average high-stability yield of similar crops in the local area during the deployment phase. After obtaining the above adaptability parameters, the crop type information of the currently planted crop and the corresponding adaptability requirement parameters are read from the preset database. Adaptability requirement parameters include the upper limit of the tolerance ratio for yield fluctuation, the upper limit of the tolerance difference for yield reduction direction deviation, and the upper limit of the tolerance difference for yield increase direction deviation. The adaptability parameters of each partition are compared with the crop's adaptability requirement parameters one by one. When the yield fluctuation ratio exceeds the upper limit of the tolerance ratio or the expected minimum yield is lower than the lower limit of the allowable target standard yield, the partition is determined to be adapted. The adaptability is categorized as follows: poor adaptability (when the yield fluctuation ratio is close to but does not exceed the tolerance limit), medium adaptability (when the yield fluctuation ratio is far below the tolerance limit and the expected minimum yield is higher than the lower limit of the target standard yield), and high adaptability (when the yield fluctuation ratio is far below the tolerance limit and the expected minimum yield is higher than the lower limit of the target standard yield). Adaptability assessment data for each zone is compiled together with adaptability parameters and stored corresponding to the zone number. When boundary and plan adjustments are made based on the adaptability assessment data, adjacent zones in the initial zoning layout are first checked in pairs. When the adaptability levels of two adjacent zones differ significantly and their yield fluctuation directions are opposite, the boundary line is marked as the boundary requiring adjustment. These boundaries are then smoothed in conjunction with the irrigation adjustment plan. During the deployment phase, the irrigation adjustment plan pre-determines the range of irrigation water volume that should be increased or decreased for each zone under different adaptability levels based on the local irrigation facility layout and maximum water supply capacity, and stores this information in a pre-set database. During execution, the specific irrigation adjustment direction and magnitude are determined for each zone based on its adaptability assessment data. When two adjacent zones show significantly opposite trends in irrigation adjustment direction and magnitude, this boundary is prioritized for smoothing adjustment to reduce the complexity of irrigation control.

[0038] In the specific smoothing process, a polyline composed of several boundary control points is selected along both sides of the marked boundary. The comprehensive environmental index and soil moisture weight of the grid cells on both sides of the boundary are compared. Grid cells with soil moisture weights close to the target moisture range and higher adaptability levels are prioritized and retained in the same area. Grid cells with lower adaptability levels and soil moisture weights deviating from the target moisture range are merged into adjacent areas with higher adaptability through boundary smoothing. During each merging operation, the coordinates of the boundary control points are adjusted to positions where the comprehensive environmental index changes more gradually and the difference in soil moisture weights is smaller at the boundary between the two areas. By repeating the above moving and merging operations, the original jagged boundary is gradually adjusted into a smooth boundary with continuous curvature changes, while ensuring that the total area change of each area is controlled within the area deviation allowable range pre-set during the deployment phase. During the boundary adjustment process, the boundary influence range is expanded by superimposing adaptability assessment data, that is, by extending a fixed number of grid cells inward on both sides of the adjusted boundary. The adaptability level and soil moisture weight of these grid cells are included in the smoothing calculation, so that the boundary adjustment is not limited to a single grid line, but covers a strip area. The overall process is smoothed, ultimately resulting in a set of corrected boundary groups that take into account both irrigation needs and adaptability. After the corrected boundary groups are determined, for each partition enclosed by the corrected boundaries, partition parameter calibration data is integrated from a preset database. The partition parameter calibration data is obtained during the deployment phase by statistically analyzing existing high-yield plots, including the optimal area range, optimal average comprehensive environmental index range, and optimal average soil moisture weight range for each partition. During execution, the actual area, average comprehensive environmental index, and average soil moisture weight of each corrected partition are compared with the corresponding ranges in the partition parameter calibration data. When a certain indicator exceeds the corresponding range, several boundary grid cells are selected along the boundary line of the corrected boundary group that borders the partition, and the partition is transferred to an adjacent partition with environmental characteristics closer to the target range. The area and environmental indicators of the two partitions are updated simultaneously, and the results are compared with the partition parameter calibration data again. If the results are still not satisfactory, the boundaries are fine-tuned in the same way until the area and main environmental indicators of all partitions fall within the corresponding calibration range. At this point, all partitions and their boundaries are fixed as the optimized partition structure and stored in the preset database.

[0039] S4 includes obtaining environmental disturbance variables and growth cycle parameters for each partition through an optimized partitioning structure, and calculating the disturbance impact index using a weighted average method combined with regional crop characteristics. The environmental disturbance variables are multiplied by crop characteristic weights and then summed to obtain a set of disturbance parameters. For this set of disturbance parameters, a local risk assessment method is used to integrate seed accumulation risk points through grid partitioning. If the number of risk points exceeds a preset threshold, a density adjustment mechanism is triggered to handle them according to the principle of uniform dispersion, thus determining a risk point distribution map. Based on the risk point distribution map, crop adaptation integration data is extracted from a preset database. The distribution is then processed by overlaying the disturbance parameter set and crop adaptation integration data in a point-by-point addition manner to obtain a preliminary density distribution. For the preliminary density distribution, the density distribution mapping and risk point location are fused and uniformity calibration is performed through curve fitting to determine the seed distribution uniformity index.

[0040] In this embodiment, based on the optimized partitioning structure obtained in the previous step, the set of grid cells contained in each partition is first defined. In a preset database, environmental disturbance variables and growth cycle parameters are pre-stored for each grid cell. The environmental disturbance variables are a set of quantitative indicators describing the degree of deviation between the actual seed landing and the target location. These include at least the average wind speed variation during the sowing period, the frequency of wind direction changes, the concentration of historical rainfall in a short period of time, the degree of surface compaction, and the degree of surface roughness. During the system deployment phase, these environmental disturbance variables are discretized into several levels after statistical analysis of historical operation records and on-site monitoring records, and each level is assigned a fixed value and recorded in the preset database. The growth cycle parameters are a set of quantitative indicators describing the sensitivity of the target crop to the above-mentioned environmental disturbances at different growth stages. These include at least the number of days from sowing to emergence, the number of days from emergence to turf establishment, and the allowable range of sowing density fluctuations at each stage. During the deployment phase, these growth cycle parameters are determined as fixed values ​​based on local experimental data of the crop over many years and written into the preset database. During execution, the system processes each partition sequentially according to the partition number. For each grid cell within the partition, it reads the corresponding environmental disturbance variable values ​​and growth cycle parameter values ​​from the preset database. At the same time, it reads the weight coefficients of the crop characteristics in that area for different disturbance factors from the preset database. The crop characteristic weights are determined as a set of fixed weighting coefficients during the deployment phase based on the yield sensitivity analysis results of the crop to factors such as wind speed, rainfall, and surface conditions, and are not changed during runtime.

[0041] When calculating the disturbance impact index, each environmental disturbance variable in each grid cell is first multiplied by the corresponding crop characteristic weight coefficient. Then, all products are summed by item within the grid cell to obtain the total disturbance impact value of the grid cell. This total value is then divided by the sum of all crop characteristic weight coefficients involved in the calculation to obtain the disturbance impact index of the grid cell. The disturbance impact index value is made to fall within a pre-defined limited range. The larger the value, the stronger the adverse impact of the grid cell on seed accumulation under the current crop and current growth stage. Subsequently, at the partition level, the arithmetic mean of the disturbance impact indices of all grid cells in each partition is calculated to obtain the partition disturbance impact index of that partition. The disturbance impact index of each grid cell in the same partition is then weighted and averaged with the growth cycle parameter of that partition to make the disturbance impact from sowing to emergence stage account for a higher proportion in the index. This results in a partition disturbance impact index that comprehensively considers environmental disturbance and growth cycle. The index of all partitions is filled into a partition environment matrix stored in a row and column structure according to the partition position. Each position in the partition environment matrix corresponds to the partition disturbance impact index of a partition.

[0042] After obtaining the zoning environment matrix, the production prediction benchmark corresponding to each zoning location is extracted from the preset database according to the zoning number. The production prediction benchmark is obtained by processing a large amount of historical land data during the deployment phase. It includes at least the benchmark unit area production value corresponding to different interference impact index intervals and the production fluctuation width observed historically within the interval. During the deployment phase, the intervals are divided using a fixed interference impact index, and the benchmark data for each interval are written into the preset database. During execution, for each zoning, the interference impact index interval into which it falls is found according to its zoning interference impact index, and the benchmark unit area production value and benchmark production fluctuation width corresponding to the zoning are obtained. Then, according to the relative position of the interference impact index of the zoning in the interval, the benchmark production fluctuation width is linearly amplified or reduced by calculating the distance ratio between the interference impact index of the zoning and the lower and upper limits of the interval, so as to obtain the expected production fluctuation width of the zoning. The expected minimum production is obtained by subtracting the expected production fluctuation width from the benchmark unit area production value, and the expected maximum production is obtained by adding the expected production fluctuation width to the benchmark unit area production value. The expected minimum production, expected maximum production, and expected production fluctuation width of all zoning are combined to form a fluctuation range set. During the deployment phase, the system uses long-term statistical results of the same crop on different plots to select the upper limit of the expected yield fluctuation width of each partition as a preset threshold when the yield spatial distribution is considered to meet the agronomic evenness requirements. This threshold is written into a preset database as a fixed value and remains unchanged during operation. During execution, for each partition in the fluctuation range set, its expected yield fluctuation width is compared with the preset threshold. When the expected yield fluctuation width of a partition exceeds the preset threshold, the partition is determined to have a high risk of seed accumulation or sparseness. Adaptive parameters are extracted from the fluctuation range data of that partition. These adaptive parameters include at least the expected yield of that partition. The ratio of the yield fluctuation width to the preset threshold, the deviation between the expected minimum yield and the regional target yield, and the deviation between the expected maximum yield and the regional target yield are used. Here, the regional target yield is determined as a fixed value based on the multi-year average of high-stability fields in the region during the deployment phase and written into the preset database. The above adaptation parameters are compared with the crop adaptation standards recorded in the preset database. The crop adaptation standards include the allowable range for different fluctuation ratios and yield deviations. When the fluctuation ratio and negative deviation exceed the allowable range, the partition is marked as a high-risk partition, and the adaptation level and location coordinates of the partition are recorded in the adaptation assessment data.

[0043] Subsequently, a local risk assessment method was employed within each partition. This involved further subdividing the partition into multiple assessment grid cells with fixed side lengths. For each grid cell, the local disturbance risk value was calculated using the same weighted summation method as described above, based on the partition interference impact index of its respective partition and the corresponding environmental disturbance variable. This local disturbance risk value was then amplified by incorporating the weights of the sowing-to-emergence stage from the growth cycle parameters. Finally, the local disturbance risk value of the grid cell was compared with a local risk threshold recorded in a pre-defined database. The local risk threshold, during the deployment phase, was based on the threshold corresponding to a significant increase in seed accumulation observed in the small-plot experiment. The risk value is set as a fixed value. When the local interference risk value is greater than the local risk threshold, the grid cell is marked as a seed accumulation risk point. For each partition, the ratio of the number of risk points within it to the total number of evaluation grid cells is calculated. This ratio is compared with the risk point ratio threshold stored in the preset database. The risk point ratio threshold is set as a fixed value during the deployment phase based on the upper limit of the risk point ratio when there is obvious unevenness in the seedling growth survey after sowing. When the risk point ratio of a certain partition is greater than the risk point ratio threshold, the density adjustment mechanism is triggered. In the density adjustment mechanism, all evaluation grid cells marked as risk points in the partition are processed according to the principle of uniform dispersion.

[0044] Specifically, the difference between the uniform seeding density of the target in the partition and the current density prediction value is first calculated. Then, seed quantity is allocated to several non-risk grid cells around each risk point as the center. The predicted density of the risk point grid cell is reduced by a fixed ratio, while the predicted density of the adjacent non-risk grid cells is increased by a corresponding ratio. This makes the arithmetic mean of the density of all grid cells in a local area remain unchanged while the local density difference decreases. By repeating the above operation on all partitions that trigger the density adjustment mechanism, a risk point distribution map containing whether each evaluated grid cell is a risk point and the adjusted density prediction value is obtained. Based on the risk point distribution map, for each assessment grid unit, crop adaptation integration data is read from a pre-set database. The crop adaptation integration data is statistically determined during the deployment phase based on different crop varieties and different growth cycle parameters. It includes at least the target plant spacing, target row spacing, and allowable density correction coefficient corresponding to each level of environmental disturbance. The target density value corresponding to the level of environmental disturbance and growth stage of the grid unit in the crop adaptation integration data is added point by point to the local disturbance risk value of the grid unit in the aforementioned disturbance parameter set. When the local disturbance risk value is high, it is reduced based on the target density; when the local disturbance risk value is low, it is moderately increased based on the target density. The preliminary density distribution value of each assessment grid unit is obtained through this point-by-point superposition method. The preliminary density distribution values ​​of all grid units are plotted as a density distribution map according to their spatial location in the plot.

[0045] For the initial density distribution mapping, to perform uniformity calibration by integrating the density distribution mapping and risk point location information, the system selects several profile lines traversing all grid cells in the east-west and north-south directions. The initial density distribution values ​​of each evaluation grid cell on each profile line are arranged spatially into a density sequence. This density sequence is used as discrete data points. A continuously varying density fitting curve is constructed along the profile line using a piecewise smooth curve fitting method. Simultaneously, risk point location information is introduced during the fitting process, giving higher weight to discrete data points located at risk points, making the fitting curve more closely resemble the actual distribution near the risk points. Subsequently, the density value of the fitted curve at each point on each profile line is calculated. The absolute value of the difference in average density values ​​of all points on the profile line is summed and divided by the product of the number of points and the average density value on the profile line to obtain the density uniformity missing ratio on the profile line. Then, the missing ratio is subtracted from the fixed value to obtain the density uniformity index of the profile line. The arithmetic mean of the density uniformity indices of all profile lines is calculated in the east-west and north-south directions to obtain the seed distribution uniformity index of the entire plot. This index is defined as a dimensionless value that varies between 0 and 1 during the deployment phase, where a value close to 1 indicates that the seeding density distribution is close to an ideal uniform state, and a value close to 0 indicates that there are obvious accumulation or sparse areas. The final seed distribution uniformity index is used as the evaluation output of this step.

[0046] S5 includes obtaining the coordinates of missing areas under the balanced area state of the initial density distribution, integrating drift influence gradient data by superimposing soil permeability data extracted from a preset database, where soil permeability data represents the regional water flow rate and drift influence gradient data reflects the degree of seed position offset, and processing the two in a weighted summation manner to obtain a corrected density distribution; if the corrected density distribution is uneven, the deviation of the regional overlap ratio is judged, and the deviation is adjusted by boundary dynamic correction, where boundary dynamic correction moves the partition edge according to the deviation magnitude to determine the balanced distribution parameters; extracting partition coordinate positioning information from the balanced distribution parameters, and performing density calibration by integrating influence data, where influence data integration is derived from the superposition result of drift influence gradient data and soil permeability data, and calibrating by point-by-point comparison to obtain an optimized density map; and triggering a dynamic mechanism to obtain the final density distribution based on the area balanced state of the optimized density map, where the dynamic mechanism iteratively and evenly disperses the seed positions through the balanced distribution parameters.

[0047] In this embodiment, based on the preliminary density distribution and optimized zoning structure obtained in step S4, the entire plot is further subdivided into several evaluation grid units according to the aforementioned grid division method. A preset database stores for each evaluation grid unit its zoning number, the number of seeds per unit area under the preliminary density distribution, and the soil permeability and drift gradient values. The soil permeability value is obtained by measuring the depth or flow rate of water infiltration per unit time through on-site infiltration tests during the system deployment phase. The minimum and maximum values ​​of the measured infiltration data for all grids are then statistically analyzed. The minimum value is subtracted from the measured value of each grid, and the result is divided by the difference between the maximum and minimum values, yielding a dimensionless soil permeability value between 0 and 1. The closer this value is to 1, the faster the water flow rate of that grid. The drift gradient value is a dimensionless value obtained by normalizing the difference in seed drift baseline parameters between adjacent grids in the aforementioned steps. The closer this value is to 1, the greater the degree of seed drop point shift near that location.

[0048] When executing step S5, the system first calculates the "coordinates of the empty areas under the balanced area state of the partition". Specifically, for each partition, using the partition area data and target sowing density recorded in the preset database, the target seed quantity of the partition is calculated as the product of the partition area and the target sowing density. Then, the preliminary density distribution values ​​of all evaluation grid units in the partition are accumulated and multiplied by their respective grid areas to obtain the actual predicted seed quantity. The actual predicted seed quantity is compared with the target seed quantity. When the actual predicted seed quantity is significantly lower than the target seed quantity and there are several evaluation grid units in the partition whose preliminary density values ​​are lower than the density lower limit threshold determined in the deployment stage, these evaluation grid units whose density is lower than the density lower limit threshold and whose surrounding adjacent grid densities are significantly higher than the threshold are marked as empty areas. The center coordinates of these evaluation grid units are used as the coordinates of the empty areas. The density lower limit threshold is determined in the deployment stage by statistically analyzing the lowest number of seeds per unit area that can ensure normal seedling emergence and no obvious sparseness in similar plots, and the corresponding value is written into the preset database as a fixed threshold.

[0049] After determining the coordinates of the missing areas, the soil permeability and drift gradient values ​​are read for each evaluation grid cell. Pre-calibrated soil permeability weights and drift gradient weights are then retrieved from a pre-defined database. These two weights are fixed values ​​selected during the deployment phase by comparing seed uniformity indices across multiple experimental combinations on sample plots. They are used to balance the impact of soil moisture redistribution on density correction with the impact of seed drift on density correction. Within each evaluation grid cell, the soil permeability value is multiplied by the soil permeability weight, and the drift gradient value is multiplied by the drift gradient weight. Finally, the two multipliers are multiplied... The environmental adjustment factor for the grid is formed by summing the products. Then, the number of seeds per unit area of ​​the grid in the initial density distribution is superimposed on the environmental adjustment factor. When the environmental adjustment factor is too large and the initial density of the grid is lower than the target sowing density, the number of seeds in the grid is increased proportionally. When the environmental adjustment factor is too large and the initial density of the grid is higher than the target sowing density, the number of seeds in the grid is decreased proportionally. When the environmental adjustment factor is too small, the density change is kept small. By performing the above weighted summation and superposition adjustment operations on all evaluation grids in the field, the corrected density distribution considering the combined effects of soil permeability and drift gradient is obtained.

[0050] After correcting the density distribution, the system calculates the ratio of the difference between the maximum and minimum density values ​​to the target sowing density for all grids within each partition, based on the density uniformity judgment criteria pre-set during the deployment phase. If this ratio exceeds a preset density uniformity threshold, the system determines that the corrected density distribution for that partition is still uneven. The density uniformity threshold is the maximum density fluctuation ratio observed in the test plot that does not affect the uniformity of forage emergence, and it is written into a preset database as a fixed value. For partitions determined to be uneven, the system further judges the regional overlap ratio deviation. The regional overlap ratio is the ratio of the area of ​​the overlapping zone used for buffering sowing transition on both sides of the common boundary between the partition and adjacent partitions to the entire area of ​​the partition. This target overlap ratio is determined as a fixed value during the deployment phase based on the combined effect of path coverage and density transition. When the difference between the actual overlap ratio and the target overlap ratio exceeds the allowable range for overlap ratio deviation set during the deployment phase, it is determined that there is a regional overlap ratio deviation. Initiate dynamic boundary correction. During dynamic boundary correction, the system sequentially checks each boundary segment along the partition edge, sums the evaluation grid cell faces adjacent to each side of the boundary segment, and calculates the actual overlapping area around the boundary. When the actual overlapping area is less than the area corresponding to the target overlapping ratio, the partition edge is moved in the overlapping direction by a fixed step size of several grid widths, and some boundary grids are drawn from adjacent partitions to increase the overlapping area. When the actual overlapping area is greater than the area corresponding to the target overlapping ratio, the boundary is shrunk back inwards into the partition by a fixed step size of several grid widths. Through multiple iterations, the difference between the overlapping area of ​​all boundary segments and the target overlapping area does not exceed the allowable range. During this process, the system records the target overlapping ratio, adjusted boundary position, updated partition area, and target seeding density for each partition. Based on the above data, a balanced distribution parameter is generated. The balanced distribution parameter reflects the comprehensive constraints of each partition in terms of area balance, overlap balance, and density target.

[0051] The system then extracts the latest boundary coordinates, area, target unit area seed quantity, and overlapping area coordinates of each partition from the equilibrium distribution parameters. This partition coordinate information is then overlaid with the environmental adjustment factor obtained by weighted summation of soil permeability and drift gradient values ​​to form an integrated impact data. For each evaluation grid cell, its current corrected density value, target unit area seed quantity of its partition, and integrated impact data value are read. Density calibration is performed through point-by-point comparison: when the corrected density value of a grid cell is significantly higher than the target unit area seed quantity of its partition and the integrated impact data value is not obtained, the system will perform a point-by-point comparison. When the combined value is large, the grid density is reduced proportionally, and the reduced number of seeds is distributed among several grids with significantly lower density around the grid according to the combined value of the data. When the corrected density value of a certain grid is significantly lower than the number of seeds per unit area of ​​the target area in the partition and the combined value of the data is large, the number of predicted seeds is equally deducted from the surrounding grids with significantly higher density and added to the grid. This ensures that the total number of seeds in each partition is consistent with the target number of seeds determined by the equilibrium distribution parameters. By performing the above point-by-point comparison and adjustment operations on all grids, an optimized density map that has been adjusted within each partition and at the boundaries of each partition is obtained. Finally, based on the optimized density map, the system checks the area balance of the entire field again, comparing the actual area of ​​each partition with the area balance target set during the deployment phase. When the area deviation of all partitions is within the allowable range, a dynamic mechanism is triggered to obtain the final density distribution. The specific process of the dynamic mechanism is as follows: within each partition, the grid density values ​​in the optimized density map are adjusted in multiple rounds. In each round, the highest density and lowest density grids within the partition are identified. The portion of the highest density grid that exceeds the target unit area seed number is transferred to the lowest density grid according to a certain proportion, so that the difference between the maximum and minimum density within the partition decreases after each iteration. At the same time, the transfer range is controlled on the partition boundary band according to the overlapping area width recorded in the balance distribution parameters to avoid crossing too many partitions and causing damage to the overall structure. The iteration is repeated until the density difference between adjacent grids within each partition is lower than the density difference allowable threshold set during the deployment phase. At this point, the density values ​​of all evaluated grid cells in the entire field constitute the final density distribution, and the final density distribution, along with the corresponding grid coordinates and partition number, is stored in the preset database.

[0052] S6 includes acquiring soil moisture boundary and temperature distribution difference data for the final density distribution, integrating the data through overlay to obtain the path planning basis; generating a drone seeding path based on the path planning basis; if the path does not cover all zones, identifying the missing coverage areas and determining adjustment needs; dynamically correcting the path trajectory based on the adjustment needs, processing the trajectory by integrating wind speed influence data and regional slope information to obtain an optimized path; extracting operation instructions from the optimized path, integrating zone coordinate information and seeding density thresholds to output a complete sequence.

[0053] In this embodiment, after obtaining the final density distribution in step S5, the final density distribution is mapped to the soil moisture boundary data and temperature distribution difference data already stored in the preset database in the previous steps using the same grid coordinate system. The grid coordinate system is the plot grid division result used in steps S1 to S5. Each grid cell has a unique planar coordinate and the corresponding final number of seeds per unit area, soil moisture boundary, and temperature distribution difference value. The soil moisture boundary is the lower and upper limit of humidity determined in step S1 based on the optimal humidity range of the target crop. The temperature distribution difference is the normalized value of the difference between the current temperature influence factor of each grid and the average temperature influence factor of the plot during the deployment stage. The larger the value, the more obvious the deviation of the temperature condition of the grid from the overall optimal state. Before executing path planning, the system pre-determines the target unit area sowing density in the deployment stage according to the sowing requirements of the target crop, and calculates two fixed values: the upper limit and the lower limit of sowing density. These two values ​​are written into the preset database as the sowing density threshold and are uniformly called during subsequent path planning and operation instruction generation.

[0054] For each grid cell, the system reads the final density value, humidity boundary, temperature distribution difference value, and the influence weight of the crop on humidity and temperature during the sowing period from a preset database. These weights are fixed to specific values ​​during the deployment phase after statistical analysis of historical yields and environmental conditions of similar plots, and remain unchanged during operation. When calculating the basis for path planning, the system first determines whether the actual soil moisture corresponding to the grid is within the humidity boundary range. When the actual moisture is between the lower and upper humidity limits, the optimal humidity level of the grid is recorded as high-level; when it is below the lower limit or above the upper limit, it is recorded as low-level. Two preset values ​​are used to represent the humidity weights of high-level and low-level, respectively. Subsequently, the system divides the temperature distribution difference range across the entire field into several continuous level intervals based on the magnitude of the temperature distribution difference value. During the deployment phase, the optimal temperature coefficient corresponding to each level interval is fixed through statistical analysis. The smaller the temperature distribution difference, the larger the temperature optimal coefficient. For each grid cell, the system reads the corresponding temperature optimal coefficient based on the temperature distribution difference interval, multiplies the humidity weight by the humidity influence weight, multiplies the temperature optimal coefficient by the temperature influence weight, and adds the two to obtain the environmental optimal index of the grid. At the same time, the density deviation value is calculated based on the difference between the final density value of the grid and the target sowing density. When the final density is lower than the target sowing density, the density deviation value is positive; when the final density is higher than the target sowing density, the density deviation value is negative. The density deviation value is then scaled according to the density sensitivity coefficient determined in the deployment stage to make it the same order of magnitude as the environmental optimal index. Finally, the environmental optimal index and the density deviation scaling value are added to form the path planning weight of the grid. The larger the weight value, the more important the coverage or the denser the flight should be during path planning. The path planning weights of all grid cells are arranged into a two-dimensional data table in spatial order, and this data table is stored in memory as the basis for path planning.

[0055] Next, the system generates the UAV seeding path based on the path planning. During the deployment phase, the width of the ground strip that can be covered by each straight flight has been determined based on the effective seeding width of the UAV seeding device, and a fixed safety factor for the flight strip spacing has been determined. The effective seeding width is multiplied by a safety factor less than 1 to obtain the fixed value of the flight strip design spacing. During the actual path generation, the system reads the azimuth of the long side of the plot and the prevailing wind direction during the seeding period from the preset database based on the direction of the circumscribed long side of the plot outline and the prevailing wind direction. The two main flight path schemes, one perpendicular to the prevailing wind direction and the other along the prevailing wind direction, are substituted into the flight strip spacing calculation to estimate the full field coverage. The system determines the required number of flight strips and the total length of flight segments. Based on the path planning data, it counts the number of times the flight strip passes through areas with high path planning weights under each scheme. The scheme with a moderate total flight segment length and a high number of coverages of high-weight areas is selected as the main heading for this operation. After determining the main heading, the system divides the land into several parallel strip areas along this direction with the designed spacing of the flight strips. A flight line is generated on the center line of each strip area from one boundary of the land to the other. The start and end points of adjacent flight lines are alternately connected to form a round-trip zigzag path. At the same time, the coordinates of the inflection point of the flight track are recorded at each turning point of the zigzag path to form the initial UAV seeding path.

[0056] After generating the initial path, the system checks the path coverage. The check method is to calculate the vertical distance from the center point of each grid cell to the nearest flight line. If the distance is less than or equal to half of the designed flight line spacing, the grid is determined to be covered by the path. If the distance is greater than half of the designed flight line spacing, the grid is determined to be a grid with missing coverage. All grids with missing coverage are clustered according to partition number and spatial location. Sets of consecutive or adjacent grids with missing coverage are marked as missing coverage areas. If there is a missing coverage area in any partition, it is considered that the path does not cover all partitions. Based on this, the system determines the adjustment requirements, which may include adding new flight lines in the direction of the missing coverage area or making minor adjustments to the positions of adjacent flight lines. To adjust requirements and dynamically correct path trajectories, the system reads wind speed impact data and regional slope information for each grid during the sowing period from a pre-set database. The wind speed impact data is derived from wind speed monitoring records during deployment, converting different wind speed ranges into definite values ​​representing the strength of seed horizontal drift. The regional slope information is calculated from topographic measurements, determining the slope magnitude and direction, and then normalizing the values. Within the strip area corresponding to each flight line, the system calculates the average and maximum values ​​of the wind speed impact data for that area. When the maximum wind speed impact value exceeds the wind speed impact benchmark value determined based on experimental results during deployment, the strip area containing that flight line is marked as a high-wind-speed impact zone. Within the high-wind-speed impact zone, the system compensates for seed drift by shifting the entire flight line upwind by a fixed distance. This shift distance is calculated during deployment based on measured results of different wind speed levels and actual seed drift distances, using the wind speed impact data range as a reference. The data is stored in a pre-set database in a fixed table format. Simultaneously, the system calculates the average and maximum slope based on the slope information of each flight path through the grid. When the average or maximum slope exceeds the slope benchmark value determined during deployment, to prevent the drone from flying too fast in steep areas and causing uneven seeding, the operating speed of the flight path in steep sections is set to a low speed. The low speed value is determined experimentally during deployment to be the minimum safe speed to ensure stable seeding by the seeding device, and this value is written into the pre-set database. While adjusting the flight path position and speed, for areas with missing coverage, the system inserts one or more new auxiliary flight paths in the direction with higher weights based on the distribution of path planning weights within that area. The spacing of the auxiliary flight paths is set according to the design spacing of the flight strips, and the start and end point coordinates are determined based on the outer rectangular boundary of the area with missing coverage. These auxiliary flight paths also participate in wind speed and slope correction. After the above dynamic correction based on wind speed impact data and regional slope information, an optimized path is obtained.

[0057] Finally, the system extracts operational instructions from the optimized path, breaking down each flight line into ordered operational segments according to the requirements of the UAV control system. Each operational segment includes the coordinates of the starting and ending track points, flight altitude, operational speed, seeding start position, and seeding end position. The flight altitude is determined during the deployment phase based on the UAV model and seeding characteristics to ensure the seed distribution range matches the designed spacing of the flight strip, and is stored in a preset database for direct retrieval in this step. The operational speed has already been differentiated into normal speed and low speed in the aforementioned slope correction. When generating the seeding control parameters for each operational segment, the system reads the final density distribution and path planning weights of all grid cells covered by the segment, first accumulating the target seeding amount for each grid cell within the coverage area, and then dividing the accumulated value by the area of ​​the region to obtain the seeding amount. The system calculates the target seed output per unit time for each segment and compares it with the upper and lower limits of the sowing density threshold. If the calculated result exceeds the upper limit, the seed output is limited to the level corresponding to the upper limit. If it is lower than the lower limit, the seed output is increased to the level corresponding to the lower limit. This ensures that the local sowing density during actual operation is not lower than the minimum requirement to guarantee germination and not higher than the upper limit to avoid excessive accumulation. Subsequently, the system binds each operation segment to its corresponding partition coordinate information, including the partition number covered by the segment, the coordinates of the boundary vertex of the partition, and the segment number sequence. The target seed output calibrated by the density threshold is also written in. The instructions of all operation segments are arranged into a complete operation instruction sequence according to the time sequence of takeoff preparation, operation flight, and return landing, which is the final output of this method.

[0058] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A sowing method for improving the uniformity of light forage seed dispersal, characterized in that, include: S1. Obtain soil moisture weight and temperature influence factor data of the target plot from the preset database, construct an initial distribution map by integrating soil moisture boundary and temperature distribution differences, perform data calibration for regional crop characteristics, and determine the basic influence parameters of seed drift within the plot; S2. Based on the basic influence parameters of seed drift, combined with terrain undulation constraints and drift influence gradient, the plots are divided into regions. The region overlap ratio is adjusted through boundary dynamic correction technology to obtain the preliminary zoning layout. S2 includes: Based on the fundamental influence parameters of seed drift and combined with terrain undulation constraints, drift influence gradient data are obtained from a pre-set database. The gradient constraint matrix is ​​obtained by integrating the parameter values ​​in a weighted average manner by fusing the fundamental influence parameters and terrain undulation constraints. The land parcels are divided into regions using the gradient constraint matrix, and the drift influence gradient is integrated to determine the range and boundary line of the division using the boundary expansion method. To address the boundary line division, a dynamic boundary correction technique is employed. This involves extracting regional overlap ratio data from the boundary line division, adjusting the overlap ratio, and correcting the boundary using a scaling method to obtain a corrected boundary set. Based on the corrected boundary set, the partition parameter calibration data is fused from the preset database. If the overlap ratio is lower than the preset threshold, the layout boundary is integrated and optimized through the boundary smoothing method to determine the optimized partition map. By optimizing the zoning map and combining constrained gradient fusion to process the data in a gradient superposition manner, preliminary zoning layout data is obtained, resulting in the preliminary zoning layout of the land parcels. S3. For the initial zoning layout, extract the soil moisture weight and temperature influence factor for each zoning, combine historical data comparison and yield prediction benchmark, calculate the expected yield fluctuation range for each zoning, and determine if the fluctuation range exceeds the preset threshold to trigger boundary smoothing processing and determine the optimized zoning structure. S4. Through the optimized partitioning structure, analyze the environmental disturbance variables and growth cycle parameters of each partition, combine the regional crop characteristics, assess the local seed accumulation risk points, and obtain the preliminary density distribution of seed distribution. S5. For the initial density distribution, detect the coordinates of the missing areas under the balanced area state of the partition, and perform density correction by superimposing the gradient data affected by drift. If the density distribution is still uneven after correction, adjust the region overlap ratio through boundary dynamic correction to obtain the final density distribution. S6. For the final density distribution, obtain soil moisture boundary and temperature distribution difference data, and integrate the data by overlay to obtain the path planning basis; Based on the path planning, a drone seeding path is generated. If the path does not cover all zones, the missing areas are identified, and adjustments are determined.

2. The sowing method for improving the uniformity of light forage seed sowing according to claim 1, characterized in that: S1 includes: Soil moisture weights and temperature influence factor data for the target plots are obtained from a pre-defined database, and an initial distribution map is constructed by integrating soil moisture boundaries and temperature distribution differences. The initial distribution map is calibrated based on the characteristics of regional crops, and the plot slope factor data is obtained from a preset database and fused into the initial distribution map. The seed drift path is determined based on the fused distribution map. If the path deviation exceeds the preset threshold, the humidity weight and temperature influence factor are adjusted. The basic influence parameters of seed drift within the plot were determined by adjusting the atlas.

3. The sowing method for improving the uniformity of light forage seed sowing according to claim 1, characterized in that: S3 includes: For the initial zoning layout, the soil moisture weight and temperature influence factor of each zone are obtained from the preset database. The comprehensive environmental index is calculated by integrating the soil moisture weight and the temperature influence factor in a weighted manner to obtain the zoning environmental matrix. Based on the zoning environment matrix and historical data comparison, the production forecast benchmark is extracted from the preset database. The expected production fluctuation range is calculated by superimposing the comprehensive environment index and the production forecast benchmark, and the fluctuation range set is determined. For the set of fluctuation ranges, if the fluctuation range exceeds a preset threshold, a crop adaptability assessment is triggered. Adaptability parameters are obtained from the set of fluctuation ranges, and the data is processed by integrating the adaptability parameters in a way that matches the crop type to obtain the adaptability assessment data. Based on the adaptability assessment data, the irrigation adjustment plan was integrated using boundary smoothing. The boundary was corrected in a smooth manner by expanding the adaptability assessment data, and the boundary curve was adjusted by incorporating soil moisture weights to determine the corrected boundary group. For the correction boundary group, partition parameter calibration data is fused from a preset database, and the optimized partition structure is determined by overlaying the correction boundary group and the partition parameter calibration data processing layout.

4. The sowing method for improving the uniformity of light forage seed sowing according to claim 1, characterized in that: S4 includes: By using the optimized partitioning structure, the environmental disturbance variables and growth cycle parameters of each partition are obtained. The disturbance impact index is calculated by weighted averaging in combination with the regional crop characteristics. The environmental disturbance variables are multiplied by the crop characteristic weights and then summed to obtain the set of disturbance parameters. For the set of interference parameters, a local risk assessment method is used to integrate seed accumulation risk points through grid division. If the number of risk points exceeds the preset threshold, a density adjustment mechanism is triggered to handle them according to the principle of uniform dispersion, and the distribution map of risk points is determined.

5. A sowing method for improving the uniformity of light forage seed sowing according to claim 4, characterized in that: S4 further includes: Based on the risk point distribution map, crop adaptation integration data is extracted from the preset database. The distribution is processed by overlaying the set of interference parameters and the crop adaptation integration data in a point-by-point addition manner to obtain a preliminary density distribution. For the initial density distribution, the density distribution mapping and risk point location are integrated and uniformity is calibrated by curve fitting to determine the seed distribution uniformity index.

6. A sowing method for improving the uniformity of light forage seed sowing according to claim 1, characterized in that: S5 includes: For the initial density distribution, the coordinates of the missing areas under the balanced area state of the partition are obtained. The drift influence gradient data are integrated by superimposing soil permeability data extracted from the preset database. The soil permeability data represents the regional water flow rate, and the drift influence gradient data reflects the degree of seed position offset. The two are processed by weighted summation to obtain the corrected density distribution. If the density distribution is uneven, the deviation of the regional overlap ratio is determined, and the deviation is adjusted by boundary dynamic correction. The boundary dynamic correction moves the partition edge according to the deviation magnitude to determine the balanced distribution parameters.

7. A sowing method for improving the uniformity of light forage seed sowing according to claim 6, characterized in that: The S5 also includes: The partition coordinate positioning information is extracted from the equilibrium distribution parameters, and density calibration is performed by integrating the influence data. The influence data integration is derived from the superposition of drift influence gradient data and soil permeability data. The calibration is performed by point-by-point comparison to obtain the optimized density map. Based on the optimized density map fusion area equilibrium state, a dynamic mechanism is triggered to obtain the final density distribution, in which the dynamic mechanism iteratively and uniformly disperses the seed positions through the equilibrium distribution parameters.

8. A sowing method for improving the uniformity of light forage seed dispersal according to claim 1, characterized in that, S6 further includes: By adjusting the demand and dynamically correcting the path trajectory, and by integrating wind speed impact data and regional slope information to process the trajectory, an optimized path is obtained. The operation instructions are extracted from the optimized path, and the complete sequence is output by integrating the partition coordinate information and the sowing density threshold.

Images