Unmanned aerial vehicle partition variable fertilization method based on prescription map dynamic update

Abstract

This invention discloses a method and system for variable-rate fertilization by unmanned aerial vehicles (UAVs) based on dynamic updates of prescription maps, belonging to the field of precision agriculture variable-rate fertilization technology. The method addresses the problem of lateral shift of the actual deposition zone relative to the center of the existing nitrogen fertilizer prescription grid during UAV granular fertilizer application. It acquires multi-source operational sampling values ​​and the existing nitrogen fertilizer prescription map, constructs a time-aligned original observation sequence, inverts the lateral deposition shift sequence of granular fertilizer, performs reverse compensation on the prescription grid center, and generates a control sequence containing lateral tracking correction and fertilizer application rate. This invention improves the alignment accuracy between the granular deposition center and the target prescription grid center without regenerating the nitrogen fertilizer prescription map, reducing missed and incorrect applications.

Description

Technical Field

[0001] This invention relates to the field of precision agriculture variable fertilization technology using big data processing, and in particular to a drone-based zoned variable fertilization method based on dynamic updating of prescription maps. Background Technology

[0002] With the development of smart agriculture and precision agriculture technologies, using UAV remote sensing imagery to acquire crop growth information and combining it with vegetation indices, nitrogen nutrient diagnostic models, and spatial grid management methods to generate nitrogen fertilizer prescription maps has become an important technical approach for variable-rate fertilization in farmland. This type of method can determine the target fertilizer application rate for each operational grid based on the differences in crop nitrogen requirements in different areas of the field, thereby achieving on-demand fertilization, reducing fertilizer waste, lowering agricultural non-point source pollution, and improving crop yield and quality stability.

[0003] In hilly areas, scattered plots, and small- to medium-sized farmlands, drones offer advantages such as operational flexibility, rapid response, and strong terrain adaptability. Utilizing drones equipped with granular fertilizer spreaders to perform prescription map-based variable-rate fertilization is an important direction for promoting intelligent and refined management of agricultural production. During actual granular fertilizer variable-rate spread, the fertilizer particles dispensing from the centrifugal spreader are affected by factors such as drone attitude changes, crosswind disturbances, rotor downwash airflow, initial particle velocity, particle size differences, and the spreader's installation position, causing the actual deposition zone of the granular fertilizer to shift laterally relative to the flight path center or prescription grid center.

[0004] Existing methods typically assume that the granular fertilizer deposition center coincides with the UAV flight path center or the target prescription grid center, and primarily rely on the UAV's current position and the fertilizer application amount in the target grid for variable control. They lack an online identification and compensation mechanism for lateral offsets of the actual deposition center. When the actual deposition center deviates from the target prescription grid center, even if the prescription map itself is accurately generated and the fertilizer application amount is correctly controlled, high-fertilizer-requirement grids may not receive sufficient fertilizer, while adjacent low-fertilizer-requirement grids may be mistakenly or excessively fertilized. This weakens the accuracy of variable fertilization and the effectiveness of prescription map execution.

[0005] Therefore, a method is needed to identify and compensate for lateral deposition offsets in UAV granular fertilizer application online without regenerating nitrogen fertilizer prescription maps, so that the actual granular deposition center is aligned with the center of the target prescription grid, thereby improving the operational accuracy of UAV variable fertilization and reducing fertilizer mismatch. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for drone-based zonal variable fertilization based on prescription map dynamic updates. This method is used to identify the lateral deposition offset of drone-applied granular fertilizer when an existing nitrogen fertilizer prescription map has been generated, and to align the actual deposition center of the granular fertilizer with the target prescription grid center through prescription grid center reverse compensation and control sequence updates.

[0007] To achieve the above objectives, this invention provides a method for drone-based zoned variable fertilization based on prescription map dynamic updates, comprising: S1, Obtain the raw sampled values ​​from the equipment and existing nitrogen fertilizer prescription maps, and construct the raw sampled values ​​from the equipment and existing nitrogen fertilizer prescription maps into a time-aligned raw observation sequence of drone particle application; S2, Based on the original observation sequence of UAV particle application with time alignment, generate the particle fertilizer lateral deposition migration sequence that has been completed and inverted online; S3, based on the lateral depositional migration sequence of the granular fertilizer that has been inverted online, form the prescription grid center sequence that has been compensated for lateral migration. S4, generated from the prescription grid center sequence that has completed lateral offset reverse compensation, produces the UAV partitioned fertilization control sequence that has completed lateral deposition alignment.

[0008] Further, step S1 includes: S11. Extract the UAV RTK position coordinate sampling sequence, heading angle sampling sequence, roll angle sampling sequence, pitch angle sampling sequence, centrifuge disk rotation speed sampling sequence, fertilizer discharge port opening sampling sequence, crosswind speed sampling sequence, particle motion image pixel frame sequence, and particle motion image pixel frame timestamp to generate a control cycle timestamp sequence. The control cycle timestamp sequence contains multiple consecutive control cycle timestamps. Perform linear interpolation on the sampled values ​​in the sequence to obtain synchronous sampled values ​​at different control cycle timestamps, forming control cycle synchronous sampled values. S12, match the control cycle synchronous sampling value with the prescription grid boundary coordinates in the existing nitrogen fertilizer prescription map, and write the prescription grid center coordinates and the prescription grid target fertilizer amount of the successfully matched prescription grid into the control cycle synchronous sampling value to obtain the prescription grid matching synchronous sampling value of different prescription grids at different control cycle timestamps, where the existing nitrogen fertilizer prescription map contains multiple prescription grids; S13: Read the camera intrinsic parameters, camera extrinsic parameters, fertilizer spreader installation position offset, and particle landing point projection height from the UAV calibration file. Extract the prescription grid matching synchronous sampling values ​​with the same prescription grid identifier and the control cycle timestamp. Sort the extracted content according to the control cycle timestamp order and stitch together the camera intrinsic parameters, camera extrinsic parameters, fertilizer spreader installation position offset, and particle landing point projection height to obtain the original observation sequence of UAV particle application with time-aligned prescription grids. The prescription grid identifier is a unique identifier for the prescription grid and is used to distinguish different prescription grids.

[0009] Further, step S12 includes: S121, convert the UAV RTK position coordinates in the control cycle synchronous sampling values ​​and the prescription grid boundary coordinates in the existing nitrogen fertilizer prescription map to a unified geographic coordinate system and a unified length unit; S122, establish a two-dimensional prescription grid spatial index based on the minimum horizontal coordinate, maximum horizontal coordinate, minimum vertical coordinate, and maximum vertical coordinate of each prescription grid boundary coordinate; S123, verify the inclusion relationship between the UAV RTK position coordinates and the candidate prescription grid boundary coordinates according to the boundary determination rules of horizontal left closed and right open and vertical bottom closed and top open, and include the outermost horizontal maximum boundary and vertical maximum boundary of the existing nitrogen fertilizer prescription map into the adjacent prescription grid. S124: Read the center coordinates of the prescription grid and the target fertilizer application amount of the prescription grid from the uniquely corresponding prescription grid, and write the corresponding control cycle synchronous sampling value.

[0010] Further, step S2 includes: S21. Extract particle motion image pixel frames from the original observation sequence of drone particle dispersal with time alignment, filter the bright spot pixels of particles by connected components and sort them by control cycle timestamp to form a set of particle image trajectory points. S22, converts the pixel centroid coordinates of the particle image trajectory point set into the observation ray of the camera coordinate system, the observation ray of the body coordinate system and the observation ray of the geographic coordinate system, and obtains the lateral center coordinates of the particle deposition in the control period based on the projection height of the particle landing point. S23, calculate the lateral depositional offset values ​​with direction signs in the vertical heading direction between the lateral center coordinates of the control period particle deposition and the center coordinates of the corresponding prescription grid. Arrange the lateral depositional offset values ​​in ascending order according to the control period timestamp to generate the particle fertilizer lateral depositional offset sequence that has been completed in online inversion.

[0011] Further, step S21 includes: S211, convert the pixel frames of the particle motion image into a grayscale image matrix, and perform brightness threshold segmentation on the grayscale image matrix to obtain the binary pixel region of the bright spot of the particle. S212, based on the area range of the connected region and the aspect ratio of the connected region, remove the noisy bright spot region from the binary pixel region of the bright spot; S213, calculate the centroid coordinates of pixels that retain connected components and the control period timestamp, and arrange the centroid coordinates of pixels in ascending order according to the control period timestamp to form a set of particle image trajectory points.

[0012] Further, step S22 includes: S221, using camera intrinsic parameters, the pixel centroid coordinates in the particle image trajectory point set are back-projected into the observation ray in the camera coordinate system; S222, using camera extrinsic parameters from the camera coordinate system to the body coordinate system to convert the observation ray in the camera coordinate system into the observation ray in the body coordinate system; S223, combining heading angle, roll angle and pitch angle to convert the observation ray in the body coordinate system into the observation ray in the geographic coordinate system; S224: The RTK position coordinates of the UAV are superimposed with the offset of the fertilizer spreader installation position to form the coordinates of the application reference point. The intersection of the observation ray in the geographic coordinate system and the projection height plane of the particle landing point is calculated to obtain the lateral center coordinates of the particle deposition in the control period.

[0013] Further, step S3 includes: S31, Horizontal sedimentary offset values ​​whose absolute values ​​of horizontal sedimentary offset values ​​in the online inversion granular fertilizer horizontal sedimentary offset sequence are less than the pre-configured maximum allowable compensation distance threshold are selected as restricted horizontal sedimentary offset values. S32, calculate the reverse translation of the restricted lateral deposition offset value in the vertical heading direction, move the center coordinates of the prescription grid along the reverse translation while keeping the target fertilizer application amount of the prescription grid unchanged, and form the center coordinates of the prescription grid after compensation. S33, after encoding the coordinates of the prescription grid center, the target fertilization amount of the prescription grid, and the corresponding control cycle timestamp, generate the prescription grid center sequence after completing the lateral offset reverse compensation.

[0014] Further, step S32 includes: S321, Calculate the unit vector of the vertical heading direction in the geographic coordinate system based on the corresponding heading angle; S322, take the opposite of the restricted lateral deposition offset value to obtain the reverse compensation distance, and multiply the reverse compensation distance by the unit vector of the vertical heading direction in the geographic coordinate system to obtain the reverse translation amount in the geographic coordinate system. S323, the coordinates of the center of the prescription grid are shifted along the reverse translation amount in the geographic coordinate system to obtain the compensated coordinates of the center of the prescription grid; S324 establishes a correspondence between the target fertilizer application amount of the prescription grid and the center coordinates of the compensated prescription grid.

[0015] Further, step S4 includes: S41, Match the coordinates of the compensated prescription grid center in the prescription grid center sequence after the lateral offset reverse compensation is completed with the coordinates of the current RTK position of the UAV, and calculate the lateral tracking correction amount in the vertical heading direction; S42, calculate the product of the target fertilizer application amount of the prescription grid in the prescription grid center sequence after the lateral offset reverse compensation is completed, and the UAV ground speed, effective spreading width and control cycle duration, to obtain the fertilizer application amount of the fertilizer spreader control cycle. S43, generate a UAV partitioned fertilization control sequence that has completed lateral deposition alignment, including lateral tracking correction amount, fertilizer discharge amount of fertilizer spreader control cycle, and corresponding control cycle timestamp.

[0016] This invention also provides a dynamic prescription map update system for UAV-based zoned variable fertilization, comprising: a sampling input unit, a prescription map matching unit, a particle deposition migration inversion unit, a prescription grid center compensation unit, a fertilization control sequence generation unit, a storage unit, and a controller; the sampling input unit receives raw sampled values ​​obtained from the device and existing nitrogen fertilizer prescription maps; the prescription map matching unit forms a time-aligned original observation sequence of UAV particle application; the particle deposition migration inversion unit generates a lateral deposition migration sequence of particle fertilizer that has been inverted online; the prescription grid center compensation unit generates a prescription grid center sequence that has been compensated for lateral migration; the fertilization control sequence generation unit generates a UAV zoned fertilization control sequence that has been aligned laterally; the storage unit stores existing nitrogen fertilizer prescription maps, camera intrinsic parameters, camera extrinsic parameters, fertilizer applicator installation position offset, particle landing point projection height, maximum allowable compensation distance threshold, and control cycle duration; the controller is connected to the sampling input unit, prescription map matching unit, particle deposition migration inversion unit, prescription grid center compensation unit, fertilization control sequence generation unit, and storage unit, and is configured to execute the aforementioned UAV zoned variable fertilization method based on prescription map dynamic update.

[0017] The beneficial effects of this invention are as follows: By unifying the UAV RTK position coordinates, attitude, fertilizer spreader status, crosswind speed, particle motion image pixel frames, and existing nitrogen fertilizer prescription maps into the same control cycle and coordinate system, a complete, time-ordered, and traceable original observation basis can be formed; through particle bright spot recognition, camera back projection, attitude transformation, and ground projection, the lateral offset of the actual deposition center of the granular fertilizer relative to the target prescription grid center can be obtained online; through threshold constraints and reverse translation of the prescription grid center, the control target can be dynamically updated while keeping the target fertilizer application amount of the existing nitrogen fertilizer prescription map unchanged; by simultaneously generating the lateral tracking correction amount and the fertilizer discharge amount of the fertilizer spreader control cycle, the synchronization between the lateral correction of the flight path and the fertilizer discharge execution can be improved, so that the actual deposition center of the granular fertilizer falls more accurately into the corresponding prescription grid center, reducing the omission of high-fertilizer-demand grids and the misapplication of adjacent low-fertilizer-demand grids. Attached Figure Description

[0018] Figure 1 is a comparison of the execution effects of existing nitrogen fertilizer prescription diagrams in the process of drone granular fertilizer application.

[0019] Figure 2 is a comparison of the transverse deposition profiles of granular fertilizer before and after compensation.

[0020] Figure 3 is a flowchart of the technical solution. Detailed Implementation

[0021] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. These specific embodiments are only for explaining the present invention and do not limit the scope of protection of the present invention.

[0022] like Figure 3 As shown, an embodiment of the present invention provides a method for drone-based zonal variable fertilization based on prescription map dynamic updates, comprising: S1, Obtain the raw sampled values ​​from the equipment and existing nitrogen fertilizer prescription maps, and construct the raw sampled values ​​from the equipment and existing nitrogen fertilizer prescription maps into a time-aligned raw observation sequence of drone particle application; S2, Based on the original observation sequence of UAV particle application with time alignment, generate the particle fertilizer lateral deposition migration sequence that has been completed and inverted online; S3, Based on the lateral depositional migration sequence of the granular fertilizer that has been inverted online, a prescription grid center sequence with lateral migration reverse compensation is formed; S4, generated from the prescription grid center sequence that has completed lateral offset reverse compensation, produces the UAV partitioned fertilization control sequence that has completed lateral deposition alignment.

[0023] In one embodiment, the input objects in step S1 include: UAV RTK position coordinates, heading angle, roll angle, pitch angle, and timestamp read from the UAV flight control recorder; centrifuge disc rotation speed, fertilizer outlet opening, and timestamp read from the fertilizer spreader controller; crosswind speed sampling value and timestamp collected from the airborne crosswind sensor; pixel frames and exposure timestamps of particle motion images collected from the airborne particle motion camera; prescription grid center coordinates, prescription grid boundary coordinates, and prescription grid target fertilizer application amount read from an existing nitrogen fertilizer prescription map; camera intrinsic parameters, camera extrinsic parameters, fertilizer spreader installation position offset, and particle landing point projection height read from the calibration file. The fertilizer spreader installation position offset represents the lateral offset of the fertilizer spreader outlet center relative to the origin of the UAV body coordinate system. The particle landing point projection height represents the target ground height used in the particle motion trajectory calculation process. Optionally, the particle landing point projection height can be calculated by subtracting the UAV flight altitude from laser altimetry or RTK elevation. The output object in step S1 is the UAV zonal fertilization control sequence that completes lateral deposition alignment.

[0024] Specifically, firstly, time synchronization of multi-source sampled values ​​is performed on the input object in step S1, including: reading the UAV RTK position coordinate sampling sequence, heading angle sampling sequence, roll angle sampling sequence, pitch angle sampling sequence, centrifuge disk rotation speed sampling sequence, fertilizer discharge port opening sampling sequence, crosswind speed sampling sequence, particle motion image pixel frame sequence, particle motion image pixel frame timestamp, and control cycle duration. As an embodiment of the present invention, the UAV RTK position coordinate sampling record includes at least a position sampling timestamp, lateral position coordinates, longitudinal position coordinates, and altitude position coordinates; the heading angle sampling record includes at least a heading angle sampling timestamp and a heading angle value; the particle motion image pixel frame can be a two-dimensional grayscale pixel matrix or a three-channel color pixel matrix. The unified time reference can be an integer microsecond value corresponding to Coordinated Universal Time, or a floating-point second value with the UAV operation start time as zero. The lateral position coordinates, longitudinal position coordinates, and altitude position coordinates are uniformly organized into meters; the heading angle, roll angle, and pitch angle are uniformly organized into radians; the centrifuge disc rotation speed is uniformly organized into revolutions per minute; the fertilizer discharge port opening is uniformly organized into percentages; and the crosswind speed is uniformly organized into meters per second.

[0025] Furthermore, the timestamps of the UAV RTK position coordinate sampling sequence, heading angle sampling sequence, roll angle sampling sequence, pitch angle sampling sequence, centrifuge disc rotation speed sampling sequence, fertilizer outlet opening sampling sequence, crosswind speed sampling sequence, and particle motion image pixel frames are cleaned of duplicate records and sorted in ascending order. The timestamps of the first and last valid sampling records of each sequence are read, and the maximum value among the timestamps of the first valid sampling records is determined as the common valid start timestamp, and the minimum value among the timestamps of the last valid sampling records is determined as the common valid end timestamp. The control cycle duration can be 0.10 seconds. A control cycle timestamp sequence is generated according to the common valid start timestamp, control cycle duration, and common valid end timestamp. The control cycle timestamp sequence contains multiple consecutive control cycle timestamps, and the time interval between adjacent control cycle timestamps is equal to the control cycle duration.

[0026] For each control cycle timestamp in the control cycle timestamp sequence, the preceding and following adjacent sampling records are retrieved from the UAV RTK position coordinate sampling sequence, roll angle sampling sequence, pitch angle sampling sequence, centrifuge disc rotation speed sampling sequence, fertilizer discharge port opening sampling sequence, and crosswind speed sampling sequence. When the control cycle timestamp is located between the preceding and following adjacent sampling timestamps, linear interpolation is performed on the corresponding physical quantities according to the time interval ratio.

[0027] in, This represents the synchronous sampled value at the timestamp of the control cycle. Indicates the timestamp of the control cycle The adjacent sampled values ​​on the front side, Indicates the timestamp of the control cycle The adjacent sampled data values ​​on the back side, Indicates the control period timestamp. Indicates control cycle timestamp The timestamp of the preceding adjacent sampling record. Indicates control cycle timestamp The timestamp of the adjacent sampling record behind it.

[0028] For the heading angle sampling sequence, the heading angle sampling records are first read in ascending order according to the heading angle sampling timestamp, and the heading angle value of the first heading angle sampling record is written into the continuous heading angle sequence. Then, subsequent heading angle sampling records are read one by one, the shortest signed angle difference between the current heading angle value and the previous heading angle value is calculated, and the shortest signed angle difference is added to the previous continuous heading angle value to eliminate the heading angle variation. and Periodic jumps occurring at the boundary:

[0029] in, Indicates the first The consecutive heading angle values ​​corresponding to the heading angle sampling records. Indicates the first The consecutive heading angle values ​​corresponding to the heading angle sampling records. Indicates the first The original heading angle values ​​in the heading angle sampling record. Indicates the first The original heading angle values ​​in the heading angle sampling record. This means normalizing the angle difference to... Angle normalization calculation within the range. After linear interpolation of the continuous heading angle sequence, the interpolation results of the continuous heading angles during the control cycle are normalized to... Within the range:

[0030] in, This represents the normalized heading angle synchronization sample value at the timestamp of the control cycle. This represents the continuous heading angle interpolation result at the control cycle timestamp. The particle motion image pixel frames and their timestamps are extracted from the particle motion image pixel frame sequence. The particle motion image pixel frames are then assigned to their corresponding control cycle timestamps according to their timestamps, forming time-aligned control cycle synchronization sampling values.

[0031] Specifically, the control cycle synchronous sampling value includes the original sampling value at the control cycle timestamp obtained from the equipment and the existing nitrogen fertilizer prescription map; The existing nitrogen fertilizer prescription map is variable fertilizer prescription data generated and stored before the start of the UAV fertilization operation. This existing nitrogen fertilizer prescription map contains multiple prescription grids. Matching the control cycle synchronous sampling values ​​with the existing nitrogen fertilizer prescription map involves: organizing the control cycle synchronous sampling values ​​into a set of control cycle synchronous sampling records arranged in ascending order of control cycle timestamps; organizing the existing nitrogen fertilizer prescription map into a set of prescription grid records, where each prescription grid record contains multiple prescription grid records. Each prescription grid record represents the grid attribute of the prescription grid, and each prescription grid record contains at least the position of the prescription grid. The system identifies the grid, the grid boundary coordinates, the grid center coordinates, and the target fertilizer application rate for each grid. It then converts the UAV RTK position coordinates from the synchronous sampling values ​​of the control cycle and the grid boundary coordinates in the existing nitrogen fertilizer prescription map to a unified geographic coordinate system and a unified length unit (default setting is meters). A two-dimensional grid index is generated for the grids associated with the grid boundary coordinates, resulting in a candidate grid set. This two-dimensional grid index includes row and column indices. The row index represents the horizontal order of the grids in the existing nitrogen fertilizer prescription map, and the column index represents the vertical order of the grids in the existing nitrogen fertilizer prescription map. As an embodiment of the present invention, for prescription grids in the form of regular grids, a two-dimensional prescription grid index in the form of row and column mapping index can be directly established according to the row and column numbers of the prescription grid; for prescription grids with different grid sizes, the rectangular bounding box of the prescription grid can be generated first in the form of a rectangular bounding box, and a two-dimensional prescription grid index in the form of a rectangular bounding box in the form of row and column mapping index can be established according to the row and column numbers of the prescription grid.

[0032] Optionally, the prescription grid boundary coordinates can be represented by the coordinates of four vertices, or by the minimum horizontal coordinate, the maximum horizontal coordinate, the minimum vertical coordinate, and the maximum vertical coordinate. As an embodiment of the present invention, the minimum horizontal coordinate, the maximum horizontal coordinate, the minimum vertical coordinate, and the maximum vertical coordinate are used by default to represent the prescription grid boundary coordinates. Based on the extracted and transformed UAV RTK position coordinates, boundary inclusion checks are performed on each prescription grid in the candidate prescription grid set. The boundary determination rules of left-closed and right-open horizontally and bottom-closed and top-open vertically are used to determine whether the current UAV RTK position coordinates fall within the prescription grid in the candidate prescription grid set.

[0033] in, This indicates the result of determining whether the UAV's RTK position coordinates (x, y) fall within the prescription grid. A value of 1 indicates that the UAV's RTK position coordinates (x, y) fall within the prescription grid. A value of 0 indicates that the UAV's RTK position coordinates (x, y) do not fall within the prescription grid. This represents the lateral position coordinates (x, y) of the UAV's RTK position coordinates. This represents the longitudinal position coordinates (x, y) in the RTK position coordinate system of the UAV. This represents the minimum horizontal coordinate of the prescription grid. This represents the maximum horizontal coordinate of the prescription grid. This represents the minimum vertical coordinate of the prescription grid. This represents the maximum vertical coordinate of the prescription grid. For the outermost boundary of an existing nitrogen fertilizer prescription map, the global maximum horizontal boundary and the global maximum vertical boundary are incorporated into the adjacent prescription grid. The center coordinates of the prescription grid and the target fertilizer application rate of the prescription grid are read from the uniquely corresponding prescription grid and written into the synchronous sampling value of the corresponding control period to form the prescription grid matching synchronous sampling value.

[0034] In one embodiment of the present invention, the camera intrinsic parameters include at least the lateral focal length, longitudinal focal length, lateral coordinates of the principal point, longitudinal coordinates of the principal point, and imaging distortion parameters. When the pixel skew value is not configured in the input data in step S1, the pixel skew value is written as zero. A 3x3 matrix form of camera intrinsic parameters is constructed based on the lateral focal length, longitudinal focal length, pixel skew value, lateral coordinates of the principal point, and longitudinal coordinates of the principal point.

[0035] in, Indicates camera intrinsic parameters. Indicates the lateral focal length. Indicates the vertical focal length. Indicates the pixel skew amount. Indicates the horizontal coordinate of the principal point. This represents the longitudinal coordinate of the principal point. The camera extrinsic parameters are the 4x4 homogeneous transformation matrix from the camera coordinate system to the body coordinate system.

[0036] in, This represents the 4x4 homogeneous transformation matrix from the camera coordinate system to the body coordinate system. This represents the rotation matrix from the body coordinate system to the camera coordinate system. Let T denote the translation vector from the body coordinate system to the camera coordinate system, and let T denote the transpose. This represents a 1x3 vector filled with zeros. The fertilizer applicator installation position offset is uniformly organized into a 3D position vector in the body coordinate system, and the particle landing point projection height is uniformly organized into a height scalar in a unified geographic coordinate system. The prescription grid matching synchronous sampling values, camera intrinsic parameters, the 4x4 homogeneous transformation matrix from the camera coordinate system to the body coordinate system (camera extrinsic parameters), fertilizer applicator installation position offset, particle landing point projection height, prescription grid center coordinates, and prescription grid target fertilizer application amount are encapsulated in a fixed field order to obtain the original observation sequence of UAV particle application with completion time alignment.

[0037] In the online inversion process of lateral deposition migration of granular fertilizer, the original observation records of each control cycle timestamp are read from the time-aligned original observation sequence of UAV particle application, and the granular motion image pixel frames and their control cycle timestamps are analyzed. For a three-channel granular motion image pixel frame with a resolution of 1920×1080 pixels, the red, green, and blue channels are first converted into an 8-bit grayscale image matrix according to a fixed brightness mapping relationship. The grayscale image matrix is ​​then segmented by a brightness threshold to obtain the binary pixel regions of bright granular spots. Noisy bright spot regions are removed based on the area range and aspect ratio of the connected regions. The centroid coordinates and control cycle timestamps of the pixels in the retained connected regions are calculated, and the centroid coordinates are arranged in ascending order according to the control cycle timestamps to form a set of granular image trajectory points.

[0038] Specifically, the area range of the retained connected region must be located in... Between, and the aspect ratio must be within between, For the preset area threshold, The preset ratio threshold is set to the default value. It is 6 pixels. For 180 pixels, =1, It is 3; For each particle image trajectory point, the pixel centroid's lateral coordinates, vertical coordinates, and control cycle timestamp are read. The corresponding camera intrinsic parameters, camera extrinsic parameters, UAV RTK position coordinates, heading angle, roll angle, pitch angle, fertilizer spreader installation position offset, and particle landing point projection height are also read. Further, the pixel centroid coordinates (including the pixel centroid's lateral and vertical coordinates) are back-projected into a camera coordinate system observation ray using the camera intrinsic parameters; this ray is then converted into a body coordinate system observation ray using the camera extrinsic parameters; finally, the body coordinate system observation ray is combined with the heading angle, roll angle, and pitch angle to convert it into a geographic coordinate system observation ray. Specifically, an attitude rotation matrix is ​​constructed based on the heading angle, roll angle, and pitch angle. The product between the attitude rotation matrix and the observation ray in the body coordinate system is calculated and used as the observation ray in the geographic coordinate system. The attitude rotation matrix includes a rotation matrix around the roll angle, a rotation matrix around the pitch angle, and a rotation matrix around the heading angle. Furthermore, the RTK position coordinates of the UAV and the offset of the fertilizer spreader installation position are superimposed to form the coordinates of the application reference point. The intersection of the observation ray of the geographic coordinate system and the projection height plane of the particle landing point is calculated as the particle ground projection point. The center position of multiple particle ground projection points in the vertical heading direction within the same control cycle time stamp is taken as the control cycle particle deposition lateral center coordinate of that control cycle time stamp. The control cycle particle deposition lateral center coordinates under the same prescription grid are sorted in ascending order according to the control cycle time stamp to obtain the control cycle particle deposition lateral center coordinate sequence of that prescription grid.

[0039] The sequence of lateral center coordinates of grain deposition during the control period of the prescription grid is read, along with the prescription grid center coordinates, heading angle, and control period timestamp from the original observation sequence of UAV grain application completed with time alignment. A unit vector in the vertical heading direction is calculated based on the heading angle. The coordinate difference vector between the lateral center coordinates of the grain deposition during the control period and the center coordinates of the prescription grid is calculated. This coordinate difference vector is projected onto the unit vector in the vertical heading direction to obtain the lateral depositional offset values ​​with direction signs for the prescription grid at different control period timestamps. These lateral depositional offset values ​​are then arranged in ascending order according to the control period timestamps to generate the lateral depositional offset sequence of the grains corresponding to the prescription grid that has undergone online inversion.

[0040] During the lateral migration reverse compensation process, the lateral depositional migration sequence of the granular fertilizer and the maximum allowable compensation distance threshold, which can be set to 0.60 meters, are read after the online inversion is completed. This maximum allowable compensation distance threshold is used to limit the range of data that can enter subsequent compensation in a single control cycle. The lateral depositional migration values ​​are parsed item by item according to the control cycle timestamp order. The absolute value of the lateral depositional migration value is calculated. The lateral depositional migration values ​​with an absolute value less than the maximum allowable compensation distance threshold are retained as restricted lateral depositional migration values, and the control cycle timestamp corresponding to the restricted lateral depositional migration values ​​is retained simultaneously.

[0041] Read the restricted lateral deposition offset value carrying the control cycle timestamp, the center coordinates of the prescription grid, the target fertilization amount of the prescription grid, and the corresponding heading angle; further, calculate the unit vector of the vertical heading direction in the geographic coordinate system based on the corresponding heading angle, take the negative of the restricted lateral deposition offset value to obtain the reverse compensation distance, so that positive lateral deposition offset corresponds to negative compensation, and negative lateral deposition offset corresponds to positive compensation, and multiply the reverse compensation distance by the unit vector of the vertical heading direction to obtain the reverse translation amount in the geographic coordinate system:

[0042] in, This represents the reverse translation in the geographic coordinate system. Indicates the restricted lateral deposition offset value. This represents the unit vector of the vertical heading direction in the geographic coordinate system. The prescription grid center coordinates are shifted along the reverse translation amount to obtain the compensated prescription grid center coordinates:

[0043] in, Indicates the coordinates of the center of the prescription grid after compensation. This indicates the coordinates of the center of the prescription grid before compensation. This represents the reverse translation amount in the geographic coordinate system. In another embodiment of the invention, for the center coordinates of a three-dimensional prescription grid, only the lateral planar coordinate component is reverse translated, while the height component remains consistent with the original prescription grid center coordinates; for the center coordinates of a two-dimensional prescription grid, only the lateral and vertical coordinate components in the same geographic coordinate system are reverse translated. The compensated prescription grid center coordinates, the target fertilization amount of the prescription grid, and the corresponding control cycle timestamp are encapsulated into a control cycle compensation record, generating a prescription grid center sequence that has completed the lateral offset reverse compensation for the corresponding prescription grid.

[0044] During the online compensation application process, the coordinates of the compensated prescription grid center, the target fertilizer application amount, and the corresponding control cycle timestamp are read from the prescription grid center sequence after lateral offset reverse compensation. Simultaneously, the current RTK position coordinates and current heading angle of the UAV are read. A time index relationship between the compensated prescription grid center coordinates and the UAV's current RTK position coordinates is established according to the corresponding control cycle timestamp. A unit vector in the vertical heading direction is calculated based on the current heading angle. The coordinate difference vector between the compensated prescription grid center coordinates and the UAV's current RTK position coordinates is calculated. This coordinate difference vector is projected onto the unit vector in the vertical heading direction to obtain the lateral tracking correction amount with a direction sign. This lateral tracking correction amount is then added to the control cycle compensation record to form the lateral tracking correction record.

[0045] The system reads the target fertilizer application rate, lateral tracking correction amount, and corresponding control cycle compensation record from the lateral tracking correction record, and simultaneously reads the UAV ground speed, effective spreading width, and control cycle duration. The target fertilizer application rate is standardized to fertilizer mass per unit area, the UAV ground speed to meters per second, the effective spreading width to meters, and the control cycle duration to seconds. When the target fertilizer application rate is expressed in kilograms per square meter, it is directly used in the calculation; when it is expressed in kilograms per acre, it is converted to kilograms per square meter based on the equation 1 acre equals 666.67 square meters. The product of the target fertilizer application rate, UAV ground speed, effective spreading width, and control cycle duration is calculated to obtain the fertilizer application rate per control cycle of the fertilizer spreader. The lateral tracking correction amount, fertilizer discharge amount in the fertilizer spreader control cycle, and the corresponding control cycle timestamp are encoded as track control field, fertilizer discharge execution field, and control cycle trigger field, respectively. They are then encapsulated in ascending order of control cycle timestamps into UAV partitioned fertilization control records, which serve as UAV partitioned fertilization control sequences for prescription grids that have completed lateral deposition alignment. The UAV partitioned fertilization control sequences reflect the fertilizer discharge amounts of different prescription grids at consecutive control cycle timestamps.

[0046] Figure 1 shows a comparison of the execution effects of existing nitrogen fertilizer prescription maps during UAV granular fertilizer application. From left to right, the figure represents three states: target prescription, uncompensated, and compensated. The target prescription represents the target fertilizer application rate distribution of each prescription grid in the existing prescription map; the uncompensated state indicates that the actual granular deposition zone has shifted laterally relative to the target prescription area, resulting in the omission of some high-fertilizer-requirement grids and misapplication in adjacent areas; the compensated state represents the deposition effect after reverse compensation based on the lateral depositional shift inversion results, with the actual deposition area more consistent with the target prescription grid.

[0047] Figure 2 shows the changes in the lateral depositional profile of the granular fertilizer before and after compensation. The lateral position in the figure is referenced to the center of the target prescription grid, and the vertical position represents the size of the granular deposition. The peak position of the uncompensated curve deviates from the target center, indicating that the actual granular deposition center does not fall at the center of the target prescription grid. The peak position of the compensated curve returns to the vicinity of the target center, indicating that after online inversion and reverse compensation through lateral offset, the granular deposition center is basically aligned with the center of the target prescription grid.

[0048] In another embodiment, the pixel frames of the particle motion image can be acquired using a grayscale camera, and the brightness threshold segmentation can be replaced by adaptive threshold segmentation based on local background estimation; the two-dimensional prescription grid spatial index can be replaced by a regular row and column index or a rectangular bounding box retrieval index; the current RTK position coordinates of the UAV can be local planar coordinates or geographic coordinates after projection transformation. None of the above replacements change the overall inventive concept of using the online inversion results of lateral depositional migration of granular fertilizer to inversely compensate the prescription grid center and generate a fertilization control sequence.

Claims

1. A method for applying regional variable fertilization by unmanned aerial vehicles (UAVs) based on dynamically updated prescription maps, characterized in that, include: S1, Obtain the raw sampled values ​​from the equipment and existing nitrogen fertilizer prescription maps, and construct the raw sampled values ​​from the equipment and existing nitrogen fertilizer prescription maps into a time-aligned raw observation sequence of drone particle application; S2, Based on the original observation sequence of UAV particle application with time alignment, generate the particle fertilizer lateral deposition migration sequence that has been completed and inverted online; S3, Based on the lateral depositional migration sequence of the granular fertilizer that has been inverted online, a prescription grid center sequence with lateral migration reverse compensation is formed; S4, generated from the prescription grid center sequence that has completed lateral offset reverse compensation, produces the UAV partitioned fertilization control sequence that has completed lateral deposition alignment.

2. The UAV-based regional variable fertilization method based on prescription map dynamic updating according to claim 1, characterized in that, Step S1 includes: S11. Extract the UAV RTK position coordinate sampling sequence, heading angle sampling sequence, roll angle sampling sequence, pitch angle sampling sequence, centrifuge disk rotation speed sampling sequence, fertilizer discharge port opening sampling sequence, crosswind speed sampling sequence, particle motion image pixel frame sequence, and particle motion image pixel frame timestamp to generate a control cycle timestamp sequence. The control cycle timestamp sequence contains multiple consecutive control cycle timestamps. Perform linear interpolation on the sampled values ​​in the sequence to obtain synchronous sampled values ​​at different control cycle timestamps, forming control cycle synchronous sampled values. S12, match the control cycle synchronous sampling value with the prescription grid boundary coordinates in the existing nitrogen fertilizer prescription map, and write the prescription grid center coordinates and the prescription grid target fertilizer amount of the successfully matched prescription grid into the control cycle synchronous sampling value to obtain the prescription grid matching synchronous sampling value of different prescription grids at different control cycle timestamps, where the existing nitrogen fertilizer prescription map contains multiple prescription grids; S13: Read the camera intrinsic parameters, camera extrinsic parameters, fertilizer spreader installation position offset, and particle landing point projection height from the UAV calibration file. Extract the prescription grid matching synchronous sampling values ​​with the same prescription grid identifier and the control cycle timestamp. Sort the extracted content according to the control cycle timestamp order and stitch together the camera intrinsic parameters, camera extrinsic parameters, fertilizer spreader installation position offset, and particle landing point projection height to obtain the original observation sequence of UAV particle application with time-aligned prescription grids. The prescription grid identifier is a unique identifier for the prescription grid and is used to distinguish different prescription grids.

3. The UAV-based regional variable fertilization method based on prescription map dynamic updating according to claim 2, characterized in that, Step S12 includes: S121, convert the UAV RTK position coordinates in the control cycle synchronous sampling values ​​and the prescription grid boundary coordinates in the existing nitrogen fertilizer prescription map to a unified geographic coordinate system and a unified length unit; S122, establish a two-dimensional prescription grid spatial index based on the minimum horizontal coordinate, maximum horizontal coordinate, minimum vertical coordinate, and maximum vertical coordinate of each prescription grid boundary coordinate; S123, verify the inclusion relationship between the UAV RTK position coordinates and the candidate prescription grid boundary coordinates according to the boundary determination rules of horizontal left closed and right open and vertical bottom closed and top open, and include the outermost horizontal maximum boundary and vertical maximum boundary of the existing nitrogen fertilizer prescription map into the adjacent prescription grid. S124: Read the center coordinates of the prescription grid and the target fertilizer application amount of the prescription grid from the uniquely corresponding prescription grid, and write the corresponding control cycle synchronous sampling value.

4. The UAV-based regional variable fertilization method based on prescription map dynamic updating according to claim 1, characterized in that, Step S2 includes: S21. Extract particle motion image pixel frames from the original observation sequence of drone particle dispersal with time alignment, filter the bright spot pixels of particles by connected components and sort them by control cycle timestamp to form a set of particle image trajectory points. S22, converts the pixel centroid coordinates of the particle image trajectory point set into the observation ray of the camera coordinate system, the observation ray of the body coordinate system and the observation ray of the geographic coordinate system, and obtains the lateral center coordinates of the particle deposition in the control period based on the projection height of the particle landing point. S23, calculate the lateral depositional offset values ​​with direction signs in the vertical heading direction between the lateral center coordinates of the control period particle deposition and the center coordinates of the corresponding prescription grid. Arrange the lateral depositional offset values ​​in ascending order according to the control period timestamp to generate the particle fertilizer lateral depositional offset sequence that has been completed in online inversion.

5. The UAV zonal variable fertilization method based on prescription map dynamic updating according to claim 4, characterized in that, Step S21 includes: S211, convert the pixel frames of the particle motion image into a grayscale image matrix, and perform brightness threshold segmentation on the grayscale image matrix to obtain the binary pixel region of the bright spot of the particle. S212, based on the area range of the connected region and the aspect ratio of the connected region, remove the noisy bright spot region from the binary pixel region of the bright spot; S213, calculate the centroid coordinates of pixels that retain connected components and the control period timestamp, and arrange the centroid coordinates of pixels in ascending order according to the control period timestamp to form a set of particle image trajectory points.

6. The UAV zonal variable fertilization method based on prescription map dynamic updating according to claim 4, characterized in that, Step S22 includes: S221, using camera intrinsic parameters, the pixel centroid coordinates in the particle image trajectory point set are back-projected into the observation ray in the camera coordinate system; S222, using camera extrinsic parameters from the camera coordinate system to the body coordinate system to convert the observation ray in the camera coordinate system into the observation ray in the body coordinate system; S223, combining heading angle, roll angle and pitch angle to convert the observation ray in the body coordinate system into the observation ray in the geographic coordinate system; S224: The RTK position coordinates of the UAV are superimposed with the offset of the fertilizer spreader installation position to form the coordinates of the application reference point. The intersection of the observation ray in the geographic coordinate system and the projection height plane of the particle landing point is calculated to obtain the lateral center coordinates of the particle deposition in the control period.

7. The UAV-based regional variable fertilization method based on prescription map dynamic updating according to claim 1, characterized in that, Step S3 includes: S31, Horizontal sedimentary offset values ​​whose absolute values ​​of horizontal sedimentary offset values ​​in the online inversion granular fertilizer horizontal sedimentary offset sequence are less than the pre-configured maximum allowable compensation distance threshold are selected as restricted horizontal sedimentary offset values. S32, calculate the reverse translation of the restricted lateral deposition offset value in the vertical heading direction, move the center coordinates of the prescription grid along the reverse translation while keeping the target fertilizer application amount of the prescription grid unchanged, and form the center coordinates of the prescription grid after compensation. S33, after encoding the coordinates of the prescription grid center, the target fertilization amount of the prescription grid, and the corresponding control cycle timestamp, generate the prescription grid center sequence after completing the lateral offset reverse compensation.

8. The UAV zonal variable fertilization method based on prescription map dynamic updating according to claim 7, characterized in that, Step S32 includes: S321, Calculate the unit vector of the vertical heading direction in the geographic coordinate system based on the corresponding heading angle; S322, take the opposite of the restricted lateral deposition offset value to obtain the reverse compensation distance, and multiply the reverse compensation distance by the unit vector of the vertical heading direction in the geographic coordinate system to obtain the reverse translation amount in the geographic coordinate system. S323, the coordinates of the center of the prescription grid are shifted along the reverse translation amount in the geographic coordinate system to obtain the compensated coordinates of the center of the prescription grid; S324 establishes a correspondence between the target fertilizer application amount of the prescription grid and the center coordinates of the compensated prescription grid.

9. The UAV-based regional variable fertilization method based on prescription map dynamic updating according to claim 1, characterized in that, Step S4 includes: S41, Match the coordinates of the compensated prescription grid center in the prescription grid center sequence after the lateral offset reverse compensation is completed with the coordinates of the current RTK position of the UAV, and calculate the lateral tracking correction amount in the vertical heading direction; S42, calculate the product of the target fertilizer application amount of the prescription grid in the prescription grid center sequence after the lateral offset reverse compensation is completed, and the UAV ground speed, effective spreading width and control cycle duration, to obtain the fertilizer application amount of the fertilizer spreader control cycle. S43, generate a UAV partitioned fertilization control sequence that has completed lateral deposition alignment, including lateral tracking correction amount, fertilizer discharge amount of fertilizer spreader control cycle, and corresponding control cycle timestamp.

10. A dynamic prescription map update system for unmanned aerial vehicle (UAV) zoned variable fertilization, characterized in that, include: The system comprises a sampling input unit, a prescription map matching unit, a particle deposition migration inversion unit, a prescription grid center compensation unit, a fertilization control sequence generation unit, a storage unit, and a controller. The sampling input unit receives raw sampled values ​​from the device and existing nitrogen fertilizer prescription maps. The prescription map matching unit generates a time-aligned raw observation sequence of UAV particle application. The particle deposition migration inversion unit generates a lateral deposition migration sequence of the granular fertilizer after online inversion. The prescription grid center compensation unit generates a prescription grid center sequence after lateral migration inverse compensation. The fertilization control sequence generation unit generates a UAV zonal fertilization control sequence after lateral deposition alignment. The storage unit stores existing nitrogen fertilizer prescription maps, camera intrinsic parameters, camera extrinsic parameters, fertilizer applicator installation position offset, particle landing point projection height, maximum allowable compensation distance threshold, and control cycle duration. The controller is connected to the sampling input unit, prescription map matching unit, particle deposition migration inversion unit, prescription grid center compensation unit, fertilization control sequence generation unit, and storage unit, and is configured to execute the UAV zonal variable fertilization method based on dynamic prescription map updates as described in any one of claims 1 to 9.

Images