A method for constructing a high spatiotemporal resolution emission inventory of primary microplastics

By identifying the range of spatial layer changes and using B-spline registration technology, the data consistency problem of the primary microplastic emission inventory when the objects change was solved, and the accurate construction of the high spatiotemporal resolution emission inventory and the reproducibility of historical data were achieved.

CN123115285APending Publication Date: 2026-11-17GUANGDONG UNIV OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611064669.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-07-17
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

The existing primary microplastic emission inventory cannot effectively maintain the total provincial emission volume when roads are split, land parcels are merged, or wastewater zones are adjusted, while also maintaining consistency between unchanged grid records and historical versions, resulting in repeated rewriting and data unreproducibility.

Method used

By identifying the actual range of change between the old and new spatial layers, and combining factor map matching with B-spline free deformation registration, the emission inheritance relationship is determined, and spatial redistribution and monthly splitting are performed only on non-inherited emissions to generate a high spatiotemporal resolution emission inventory.

Benefits of technology

It enables the transfer of emission contributions only within the actual range of change when spatial objects change, reduces repeated rewriting of unchanged grids, maintains consistency between provincial total emissions and historical versions, and improves the accuracy and reproducibility of emission inventories.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123115285A_ABST
    Figure CN123115285A_ABST
Patent Text Reader

Abstract

This invention discloses a method for constructing a primary microplastic high spatiotemporal resolution emission inventory, specifically relating to the fields of microplastic emission inventory construction and geographic information data processing. The method includes acquiring a spatial layer update task initiated by an industrial cloud computing platform; generating a source year number by combining the emission source number with the accounting year; reading old objects, old monthly values, new objects, and annual quantities from the geographic information database according to the source year number; generating CPU capacity based on the number of changed tiles and the number of idle cores in the central processing unit to obtain a fragmented set; for old objects within the fragmented set, finding new objects whose bounding boxes intersect using the spatial index of the geographic information database; constructing a factor map using the old and new objects; determining the emission inheritance relationship by identifying the actual change range of the old and new spatial layers, combining factor map matching and B-spline free deformation registration; and performing spatial redistribution, monthly splitting, and grid aggregation only on non-inherited emissions.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of microplastic emission inventory construction and geographic information data processing technology, and more specifically, to a primary method for constructing a high spatiotemporal resolution microplastic emission inventory. Background Technology

[0002] Existing primary microplastic emission inventories typically first calculate the annual emissions from each source based on provincial activity level data and emission factors. Then, using spatial data such as road density, population density, industrial land, built-up areas, and wastewater collection zones, the provincial emissions are allocated to a 1km grid and updated in the cloud according to administrative regions or spatial tiles. In practical applications, road data may be split from one record into multiple records due to surveying revisions, industrial plots may be merged due to boundary adjustments, and sewage collection zones may be re-divided due to changes in the service area of ​​treatment facilities. After such changes occur, it is necessary to keep the provincial annual total emissions unchanged, avoid rewriting grid records that have not undergone spatial changes without cause, and at the same time retain the reproducibility of the existing versions. Current processing typically involves recalculating spatial weights based on the updated object numbers. This can easily result in the retention of both old and new object emissions within the same geographic area, or in the re-normalization of grid values ​​across the entire province due to changes in local objects, causing emissions in unchanged areas to change synchronously and making it impossible to restore historical inventory data from the original data. The technical problem this application aims to solve is: when spatial objects change due to road splitting, land parcel merging, or sewage zoning adjustments, how to transfer only the emission contribution corresponding to the actual change range, while maintaining consistency between the provincial total emission, unchanged grid records, and historical version results. Summary of the Invention

[0003] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a method for constructing a primary microplastic high spatiotemporal resolution emission inventory. This method identifies the actual range of change between old and new spatial layers, combines factor map matching with B-spline free deformation registration to determine emission inheritance relationships, and performs spatial reallocation, monthly splitting, and grid aggregation only on non-inherited emissions, thereby solving the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a method for constructing a high spatiotemporal resolution emission inventory of primary microplastics, comprising: S1. Obtain the spatial layer update task initiated by the industrial cloud computing platform, generate the source year number by combining the emission source number with the accounting year, read the old object, old month value, new object and annual quantity in the geographic information database according to the source year number, and generate the CPU quota based on the number of changed tiles and the number of idle cores of the central processing unit to obtain the fragment set. S2. For old objects within the fragment set, use the spatial index of the geographic information database to find new objects that intersect the bounding boxes, and construct a factor graph with the old and new objects. Generate edge costs by taking the inverse value of the intersection ratio, the difference of the surrogate value, the difference of the centroid grid number, and the difference of the source year. Subtract the remaining geometric quantities of the old and new objects in ascending order of edge costs to obtain the inherited edges. S3. Based on the inherited edge, execute the B-spline free deformation registration algorithm to generate a control grid within the outer rectangle of the old object, pair the sampling points of the old object boundary to the boundary points of the new object, generate the control grid displacement field according to the point-pair displacement, and use the cubic B-spline basis function to map the old object to the coordinate domain of the new object to obtain the registered object. S4. Calculate the inheritance rate based on the intersection geometry of the registered object and the new object, write the old month value into the new object according to the inheritance rate, classify the unwritten month value into the recycling amount, and generate the receiving object quota according to the proxy value of the new object to obtain the migration table. S5. Summarize the migration table and subtract the inherited value from the annual total to obtain the allocation quantity. Write the quota into the receiving object. Split the annual allocation value of the receiving object according to the monthly proxy sequence. Then correct the monthly value of the hit object according to the verification record and return the same amount of allocation value in the unhit receiving object. Summarize the monthly value of the object according to the geometric value of the intersection between the new object and the target grid. Output the monthly allocation list of the target grid with the layer version number.

[0005] In a preferred embodiment, S1 includes: S11. After receiving the spatial layer update task on the industrial cloud computing platform, the emission source number is written into the high digit and the accounting year is written into the low digit to form the source year number. The version spatial index of the geographic information database is retrieved using the source year number. First, the old objects are read along the old version address and the old month value is retrieved back. Then, the new objects are read along the new version address and the year quantity is read from the accounting table to form a version object set. The emission source number refers to the fixed-length binary code assigned to each type of emission source according to the primary microplastic emission source classification table. Each type of emission source in the same classification table corresponds to only one emission source number. Spatial layer update tasks, annual records and object records all establish source correspondence through emission source numbers. Version spatial index refers to a three-level index structure in a geographic information database that uses the source year number to locate the emission source year partition, the layer version number to locate the old version partition or the new version partition, and the object bounding box to locate the object record page. It is used to limit the retrieval scope of old objects and new objects. The version address refers to the partition number and record page start offset obtained by mapping the layer version number in the geographic information database. The old version address points to the old object record area, and the new version address points to the new object record area. The reading process starts from the start offset of the corresponding record page and stops at the last page of the partition. The accounting table refers to a database table that stores the annual amount of primary microplastics with the source year as the primary key and the administrative region code as the secondary key. When querying, the source year is used to locate the annual record of the emission source, and the corresponding annual amount and unit of measurement are read by the administrative region code. S12. Cut the version object set according to the target grid boundary, calculate the geometric union of old objects falling into the same target grid, and perform a symmetric difference operation on the geometric union of the obtained geometric union and the geometric union of new objects in the same target grid. Write the target grid with non-empty symmetric difference result as a change tile, and then generate Hilbert code according to the coordinates of the center point of the change tile to form a sequence of change tiles arranged in ascending order of Hilbert code. The process of finding the geometric union of old objects falling into the same target grid is as follows: First, each old object is cut off by the boundary of the target grid to obtain old pieces located inside the target grid; for line objects, duplicate line segments are deleted and line segments with overlapping endpoints are joined; for face objects, intersecting face regions are merged and the common boundary inside the merged area is deleted. The resulting set of lines or face sets is written as the union of old objects. The process of performing symmetric difference operation is as follows: within the same target grid, the old subtracted piece is obtained by subtracting the new object union from the old object union, and the new piece is obtained by subtracting the old object union from the new object union. Then, the old subtracted piece and the new piece are merged. When the length or area of ​​the merged result is non-zero, the corresponding target grid is written as a changed tile. When the length or area of ​​the merged result is zero, no changed tile is written. The process of generating Hilbert codes based on the coordinates of the center points of the changing tiles is as follows: the coordinates of the center points are converted into horizontal and vertical grid numbers using a unified geographic origin and the side length of the target grid. The two grid numbers are written as binary numbers with the same bit width. Then, the horizontal and vertical bits are read bit by bit from the high bit to the low bit. After each reading, the horizontal bit is multiplied by three and then XORed with the vertical bit to obtain a quaternary code. When the vertical bit is zero, the remaining low bit sequence is swapped. When the horizontal bit is one, the remaining low bit sequence is inverted bit by bit. Then, the quaternary codes are concatenated according to the reading order to obtain the Hilbert code. S13. Divide the number of variable tiles by the number of idle cores of the central processing unit to obtain the quotient and remainder. Write the CPU amount obtained by adding one to the quotient to the idle core corresponding to the remainder in order of idle cores, and write the CPU amount obtained by the quotient to the remaining idle cores. Then, along the variable tile sequence, extract the same number of variable tiles as each CPU amount to form a fragment set. Idle cores refer to logical processing cores whose status word is schedulable and whose local run queue is empty after the industrial cloud computing platform reads the central processing unit's scheduling status when generating shard sets; the idle cores are generated in ascending order according to the logical processing core numbers.

[0006] In a preferred embodiment, S2 includes: S21. For each old object in the fragment set, call the R-tree index of the geographic information database to read the new object that intersects the bounding box. For each pair of objects, calculate the exact geometric intersection. Line objects are measured by the intersection length, and polygon objects are measured by the intersection area. Write the geometric union of the intersection of the same old object as an inheritable quantity. Starting with the value obtained by subtracting the intersection ratio, successively add the surrogate value difference normalized by the sum of the surrogate values ​​of the two objects, the centroid grid difference normalized by the fragment edge length, and the source year difference and same position to obtain the edge cost of the candidate edge and form the candidate graph. R-tree index refers to a hierarchical index built in a geographic information database according to spatial bounding relationships. Non-leaf nodes store the coverage area and page address of child nodes, while leaf nodes store the bounding box of the new object and the address of the object record. During retrieval, it starts from the root node and only enters the child nodes that intersect with the bounding box of the old object until the leaf node returns the corresponding new object. A bounding box is an axis-aligned rectangle formed by the lower bound of the x-coordinate, the lower bound of the y-coordinate, the upper bound of the x-coordinate, and the upper bound of the y-coordinate of all coordinate points of an object. It is used to exclude new objects whose spatial extents do not intersect before performing precise geometric operations. Geometric union quantity refers to the geometric measure obtained by performing a union operation on the exact geometric intersection formed by the same old object and each new object. The overlapping part is counted only once. When the old object is a line object, the union line length is taken; when the old object is a face object, the union area is taken. The proxy value is a numerical value corresponding to the source year and used to characterize the emission distribution contribution of a spatial object. During the calculation, the spatial object intersects with the proxy grid cell by cell. For line objects, the value is calculated by multiplying the corresponding grid value by the ratio of the overlap length to the length of the object within the grid, and then summing the values. For area objects, the value is calculated by multiplying the corresponding grid value by the ratio of the overlap area to the area of ​​the grid, and then summing the values. The centroid grid number refers to the horizontal and vertical grid numbers obtained by converting the centroid coordinates of the object using the origin of the target grid and the grid side length; the difference between the centroid grid numbers of two objects is the sum of the absolute values ​​of the differences in the horizontal and vertical grid numbers. The process of finding the exact geometric intersection for each pair of objects is as follows: First, transform the old object and the new object to the same projected coordinate system and round them to the same coordinate precision. Then, split the object boundary into line segments, calculate the intersection point of the two sets of line segments and insert the intersection point into the corresponding boundary. Then, retain the line segments that are simultaneously located inside or on the boundary of the two objects. For line objects, the retained line segments will be connected by their endpoints to form an intersection line. For surface objects, the retained line segments will be connected end to end by direction to form an intersection surface. S22. Establish a factor graph matching target with geometric conservation constraints on the candidate graph. The sum of the product of the candidate edge transit amount and the edge cost is the value to be deducted. Limit the total outgoing edge transit amount of each old object to be equal to the inheritable amount. Ensure that the total incoming edge transit amount of each new object does not exceed the geometric amount of the new object. Ensure that the candidate edge transit amount does not exceed the geometric amount of the corresponding intersection. Fix the transit amount of the candidate edge with the source year number being one to zero. Then write the transit amount and constraint multiplier to zero and write the reciprocal of the number of candidate edges into the initial step size to form the initial matching state. The geometric conservation constraint means that the total amount of transmission allocated to each candidate edge by each old object is equal to the corresponding inheritable amount, the total amount of transmission received by each new object from the candidate edges does not exceed the geometric amount of the new object, the amount of transmission carried by each candidate edge does not exceed the geometric amount of the corresponding intersection, and the amount of transmission of candidate edges with different source year numbers remains zero. The candidate edge transitivity is the geometric amount of the intersection of a candidate edge from the old object to the new object; the outgoing edge transitivity is the sum of the transitivity of all candidate edges connected to the same old object; the incoming edge transitivity is the sum of the transitivity of all candidate edges connected to the same new object. Geometric quantities refer to the numerical values ​​obtained by measuring spatial objects in a unified projected coordinate system. Line objects use length, and surface objects use area. Old objects, new objects, and candidate edges within the same candidate image use the same measurement type. Constraint multipliers are iterative variables attached to the constraints of object nodes. The old object multiplier corresponds to the difference between the outgoing edge propagation and the inheritable quantity, while the new object multiplier corresponds to the difference between the incoming edge propagation and the new object geometry. After each round of calculation, the old object multiplier is added to the product of the current step size and the outgoing edge difference, and the new object multiplier is added to the product of the current step size and the incoming edge difference. When the resulting new object multiplier is less than zero, it is written as zero.

[0007] In a preferred embodiment, S2 further includes: S23. Based on the initial matching state, the reduced cost is obtained by superimposing the constraint multipliers at both ends on the edge cost in each round. The transit amount is updated according to the reduced cost and the current step size. The update value below zero is written as zero and the update value above the intersection geometry is written as the intersection geometry. The constraint multipliers are then rewritten according to the deviation of the outgoing edge transit amount and the deviation of the incoming edge transit amount. When the value to be reduced in this round is higher than the value to be reduced in the previous round, the current step size is halved and the update in this round is canceled until the transit amounts of the two adjacent rounds are consistent bit by bit and all geometric conservation constraints are met, thus obtaining the transit edge set. The update value refers to the transmission amount obtained by subtracting the product of the current step size and the reduced cost from the transmission amount of the candidate edge before this round of update. The reduced cost is generated by superimposing the candidate edge cost and the constraint multipliers of the objects at both ends of the candidate edge. The update value is then limited to zero and the corresponding intersection geometry. S24. Based on the transit edge set, in the intersection geometry corresponding to each candidate edge, extract the inherited pieces whose cumulative geometric quantity is equal to the transit quantity in ascending order of Hilbert code. Find the geometric intersection of the inherited pieces that connect the same old object or the same new object. Keep the overlapping pieces in the candidate edges that are ranked first in the edge cost ascending order and deduct them from the other candidate edges. Then, rewrite the upper limit of the transit quantity of the corresponding candidate edge with the geometric quantity of the inherited piece after deduction and return to S23 to recalculate until there is no geometric overlap between the inherited pieces and the transit quantities of two adjacent times are consistent bit by bit. Write the candidate edges with non-zero transit quantities as inherited edges. Inherited slices refer to the spatial portions extracted from the corresponding intersection geometry according to the candidate edge transit quantities. When extracting, the intersection geometry is first divided along the boundary of the target grid, and then the geometric quantities are accumulated in ascending order according to the Hilbert codes of the center points of each divided portion. When the accumulated quantity reaches the divided portion where the transit quantity is located, the last part is cut off according to the missing geometric quantity. The resulting spatial range is written as inherited slices. Overlapping pieces refer to the non-zero geometric parts formed after performing precise geometric intersection on two candidate edges that connect the same old object or the same new object and their corresponding inherited pieces; line objects are confirmed by a non-zero intersection length, and face objects are confirmed by a non-zero intersection area.

[0008] In a preferred embodiment, S3 includes: S31. Based on the inherited edge, read the old object boundary and the new object boundary from the geographic information database and convert them to the same projection coordinate system. Write the boundary vertices and the intersections of the boundary and the target grid lines as sampling points. Round up the square root of the number of old sampling points to obtain the control grid order. Then set control points according to the equal distribution of the old object's outer rectangle span to form the initial control grid. The control points refer to the grid points in the initial control grid used to bear two-dimensional displacements. Each control point stores the lateral displacement value and the longitudinal displacement value. The deformation of any coordinate within the old object is obtained by weighting the displacements of the sixteen control points surrounding that coordinate according to the cubic B-spline basis function. The outer rectangle refers to the axis-aligned rectangle formed by the lower bound of the horizontal coordinate, the lower bound of the vertical coordinate, the upper bound of the horizontal coordinate, and the upper bound of the vertical coordinate in all the boundary coordinates of the old object. It is used to limit the layout range of the control grid and to calculate the diagonal length of the outer rectangle of the old object. S32. Accumulate the distance between adjacent sampling points along the boundary of the old object and the boundary of the new object respectively, and divide by the total length of the corresponding boundary to obtain the arc length order. For each old sampling point, first determine the retrieval order of the new sampling point by the absolute value of the difference in arc length order, then rewrite the retrieval order by the distance of the old sampling point projected along the normal to the boundary of the new object, and retain the sampling points that point to each other from the retrieval results from the old object to the new object and from the new object to the old object, forming a registration point pair.

[0009] In a preferred embodiment, S3 further includes: S33. Based on the registration point pairs, execute the B-spline free deformation registration algorithm. Divide the sum of squares of the deformation residuals of the registration point pairs by the square of the diagonal of the outer rectangle of the old object to obtain the point pair term. Divide the sum of squares of the second-order difference of the displacements of adjacent control points by the number of control points to obtain the deformation term. Then, divide the square of the difference between the intersection geometry of the mapped old object and the new object and the inherited edge transfer amount by the square of the geometry of the new object to obtain the inheritance term. With the goal of decreasing the sum of the three terms and limiting the deformation Jacobian determinant of the control unit corners and center points to be positive, update the control point displacements using the Gauss-Newton direction. If the target value does not decrease or the deformation Jacobian determinant is not positive, halve the update step size and recalculate until the displacement update of each control point is zero after being rounded according to the coordinate quantization unit of the geographic information database, and obtain the control grid displacement field. The deformation residual refers to the two-dimensional coordinate difference obtained by subtracting the coordinates of the corresponding new sampling point from the coordinates of the old sampling point obtained by mapping the displacement field of the current control grid; the square of the deformation residual is the sum of the square of the horizontal coordinate difference and the square of the vertical coordinate difference. The control unit refers to the rectangular area enclosed by two horizontally adjacent control points and two vertically adjacent control points in the initial control grid or the refined control grid. After the coordinates of the old object fall into the control unit, the four rows and four columns of control points around the control unit participate in the cubic B-spline displacement calculation. The deformation Jacobian determinant is the second-order matrix obtained by taking the partial derivatives of the horizontal and vertical coordinates of the old object after mapping with respect to the horizontal and vertical coordinates before mapping. A positive deformation Jacobian determinant indicates that the local mapping direction within the corresponding control unit has not been flipped. The target value is the sum of the point-to-point item, deformation item, and inheritance item in the current round. The point-to-point item reflects the coordinate deviation of the registration point pair, the deformation item reflects the bending degree of the control grid displacement, and the inheritance item reflects the difference between the intersection geometry and the inherited edge transfer quantity after mapping. Displacement update refers to the control point displacement increment obtained by multiplying the Gauss-Newton direction of the current cycle by the current step size. Each control point corresponds to a lateral increment and a longitudinal increment. The new displacement of the control point is obtained by adding the corresponding displacement update to the current displacement of the control point. The process of updating the control point displacement using the Gauss-Newton direction is as follows: First, the lateral and longitudinal displacement values ​​of each control point are concatenated in order of control point number to form a displacement vector. Then, the lateral and longitudinal deformation residuals of the registration point pairs, the second-order differences of the displacements of adjacent control points, and the difference between the mapped intersection geometry and the inherited edge transfer quantity are sequentially written into the residual vector. For the registration point pair residuals and the second-order difference residuals, the partial derivatives of the residual vector with respect to the displacement vector are calculated according to the cubic B-spline basis function and its corresponding derivative. For the intersection geometry residuals, a coordinate quantization unit is added to each displacement component of each control point sequentially. The intersection geometry is recalculated, and the corresponding partial derivative is obtained by dividing the increment of the intersection geometry by the coordinate quantization unit. The residual Jacobian matrix is ​​formed by all partial derivatives. Then, the least squares solution that makes the product of the residual Jacobian matrix and the direction vector approximate the negative value of the residual vector is obtained by QR decomposition. This least squares solution is used as the Gauss-Newton direction, and the displacement update is generated with 1 as the current step size. If the updated target value is not lower than the target value before the update, or if the deformation Jacobian determinant of any control unit corner point and center point is not positive, the current displacement update is canceled, the current step size is halved, and the displacement update is regenerated.

[0010] In a preferred embodiment, S3 further includes: S34. Apply the control grid displacement field to the old object using the cubic B-spline basis function to form candidate registration objects. Then, use the reverse displacement field obtained after exchanging the old and new objects to map the candidate registration objects back to the coordinate domain of the old object. Take the first and second error points in descending order of the distance between the mapped sampling points and the old sampling points. Add a control point at the center of the control unit where the error point is located and return to S33. Continue until all mapped distances are rounded to zero after being quantized by coordinate units, and the candidate registration objects do not have boundary self-intersections or cross the boundaries of unchanged adjacent objects. Write the candidate registration objects into the geographic information database and output the registration objects. The process of applying the control grid displacement field to the old object to form candidate registration objects using cubic B-spline basis functions is as follows: For each boundary coordinate of the old object, the corresponding control unit is first determined based on the horizontal and vertical positions of the coordinate in the control grid. Then, the displacement values ​​of the four rows and four columns of control points around the control unit are read, and the horizontal and vertical decimal positions of the boundary coordinates relative to the starting point of the control grid are calculated respectively. Based on this, four horizontal cubic B-spline basis function values ​​and four vertical cubic B-spline basis function values ​​are obtained. The horizontal displacement of each control point is multiplied by the corresponding horizontal and vertical basis function values, and the sixteen products are summed to obtain the horizontal deformation of the boundary coordinates. The vertical deformation is obtained in the same way. The two deformations are added to the original boundary coordinates to form the mapped coordinates. All mapped coordinates are connected in the original boundary order of the old object to form candidate registration objects. The retrospective distance refers to the Euclidean distance between the mapped sampling point in the candidate registration object and the corresponding old sampling point after being mapped back to the coordinate domain of the old object by the reverse displacement field. It is the square root of the sum of the square of the retrospective horizontal coordinate difference and the square of the retrospective vertical coordinate difference.

[0011] In a preferred embodiment, S4 includes: S41. For each inherited edge, the registration object and the new object are overlaid through the geographic information database, and the intersection geometry is cut along the target grid boundary. The repeated intersection pieces are retained in the first inherited edge and deducted from the subsequent inherited edges in ascending order of edge cost. The inheritance rate is obtained by dividing the amount of intersection geometry after deduction by the amount of geometry of the old object, and an inheritance rate table is formed. S42. Multiply the old month value by each inheritance rate according to the inheritance rate table to obtain the inheritance month value. Write the inheritance month value into the corresponding new object by month. Then subtract the total inheritance month value of the same month from the old month value to obtain the recovery amount. Write the source old object number and intersection piece number of the inheritance month value into the geographic information database to form an inheritance record. S43. Using the intersection of the recycled amount as the supply point, and the new objects with non-zero proxy values ​​within the range of change as the receiving objects, a spatial transfer graph is established according to the shared edges of the target grid. The path cost is obtained by adding the number of grid edges traversed by the path to the reciprocal of the proxy value of the receiving object. The paths are read one by one in ascending order of path cost and path code. For each path, the sum of the recycled amount and the remaining proxy share of the corresponding receiving object, minus the absolute value of the difference between the two, is divided by two and written as the transfer amount. The transfer amount is written to the receiving object along the corresponding path and deducted from the recycled amount and the remaining proxy share respectively. After the recycled amount is zeroed, the quota is generated by dividing the cumulative received amount of each receiving object by the original value of the recycled amount, and the migration table is output. Path coding refers to reading the target grid numbers sequentially along the path from the supply point to the receiving object, and then concatenating the target grid numbers in the order of passage to form a fixed-length code; when the path costs of two paths are the same, the reading order is determined by the ascending order of the corresponding path code values. The agency share refers to the acceptable amount obtained by dividing the agency value of the receiving object by the sum of the agency values ​​of all receiving objects, and then multiplying it by the original value of the recovery amount; the remaining agency share is the value obtained by subtracting the accumulated amount received by the receiving object from the acceptable amount.

[0012] In a preferred embodiment, S5 includes: S51. Summarize the inherited monthly values ​​in the migration table according to the source year number and add them up to obtain the total inherited value for the year. Subtract the total inherited value for the year from the annual quantity to obtain the quantity to be allocated. Then multiply the quantity to be allocated by the quota of each receiving object to obtain the annual ranking value of the receiving object, and form the receiving table. S52. Read the monthly agency sequence corresponding to each receiving object from the geographic information database, divide the monthly agency value by the total annual agency value to obtain the monthly share, and multiply the annual ranking value of the receiving object by the monthly share to obtain the monthly value of the object; when the total annual agency value is zero, divide the number of days in each month by the total number of days in the year to obtain the monthly share, and form the monthly table of objects. The monthly proxy sequence refers to an ordered sequence of months established according to the source year and the receiving object. Each month corresponds to a monthly proxy value. The monthly proxy value is obtained by finding the intersection of the receiving object and the time proxy grid of the current month, multiplying each intersecting grid value by the proportion of the corresponding intersection geometry to the geometry of the receiving object, and then summing them up. The months are written in the order of the natural months.

[0013] In a preferred embodiment, S5 further includes: S53. Intersect the geographical range of the verification record with the execution space of the new object in the object monthly table, summarize the monthly value of the hit object according to the verification month, and generate a signed verification difference by subtracting the summary value from the verification value. Then, rewrite the monthly value of the hit object according to the proportion of the intersection geometry to the total number of hit intersection geometry. At the same time, write the callback value with the opposite sign according to the proportion of the original monthly value of the missed receiving object to the total number of missed monthly values, thus forming the verification monthly table. The verification record refers to a data record formed by locating the emission source and accounting year in the geographic information database by the source year number, limiting the verification area by the geographical range, and binding the corresponding emission period with the verification month and writing the verification value, which is used to verify the emission amount of the target monthly table in the corresponding area and month. The verification value refers to the emission amount written in the verification record for the corresponding source year, geographical range, and verification month, which is used as the target amount in the monthly value summary result of the target object; when the result of the verification value minus the summary value is positive, it means that the monthly value of the target object needs to be increased, and when the result is negative, it means that the monthly value of the target object needs to be deducted. The total value refers to the emission amount obtained by multiplying the monthly value of each hit object in the verification month by the ratio of the geometric quantity of the intersection between the hit object and the verification range to the geometric quantity of the hit object; S54. By overlaying and verifying the new objects and target grids in the monthly table through the geographic information database, the grid share is obtained by dividing the intersection geometry by the new object geometry. The monthly value of the object is multiplied by the grid share and then accumulated according to the target grid and the month. When the annual total of the accumulated value is consistent with the annual quantity, the layer version number is written, and the monthly emission list of the target grid is output.

[0014] The technical effects and advantages of this invention are as follows: This approach reduces the repeated rewriting of unchanged grids by identifying changed tiles and migrating only the emission contributions within the actual range of change, thus maintaining consistency between provincial annual data and historical versions. By using factor graph matching to constrain the geometric transfer relationship between new and old objects, it is possible to distinguish between object splitting and merging, and reduce the situation of repeated inheritance of emission values ​​in the same spatial range; By using B-spline free deformation registration to correct boundary translation and local deformation, the inheritance rate can be made to correspond to the actual spatial overlap range, thus improving the accuracy of emission migration after object update. By generating fragment sets based on the number of changing tiles and the number of idle cores in the central processing unit, the amount of redundant processing in industrial cloud computing can be reduced by reducing the participation of unchanged areas in the calculation. By using layer version numbers and version association records to save update results, the source of spatial layers, proxy data, and verification data can be traced, enhancing the reproducibility of historical lists. Attached Figure Description

[0015] Figure 1 This is a flowchart outlining the method steps of 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] Refer to the instruction manual appendix Figure 1 The present invention provides a method for constructing a high spatiotemporal resolution emission inventory of primary microplastics, comprising: S1. Obtain the spatial layer update task initiated by the industrial cloud computing platform, generate the source year number by combining the emission source number with the accounting year, read the old object, old month value, new object and annual quantity in the geographic information database according to the source year number, and generate the CPU quota based on the number of changed tiles and the number of idle cores of the central processing unit to obtain the fragment set. After the spatial layer is updated and entered into the industrial cloud computing platform, the consistency of version, source and administrative scope is first completed, then the target grid that has undergone geometric changes is identified, and finally the fragment set is generated based on the status of the central processing unit. The entire process uses the source year number to run through object reading and year query, uses the target grid to limit the range of geometric changes, and uses the task number to constrain the piecewise writing, so that subsequent object matching only accesses the spatial data that has changed. After the spatial layer update task writes the task number, emission source number, accounting year, administrative region code, old layer version number and new layer version number, the platform forms a version object set according to S11. The emission source number width is taken from the number field width in the primary microplastic emission source classification table, and the accounting year width is taken from the year field width in the accounting table. If the field length is insufficient, zeros are added to the left. If the field length exceeds the corresponding width, an error record is written and the task is terminated. After coding is completed, the emission source number is written to the high digit and the accounting year is written to the low digit to form the source year number. The old layer version number points to the current release version, and the new layer version number points to the release version. The predecessor version number in the new layer version record should be consistent with the old layer version number. If they are inconsistent, a version conflict record should be written. The geographic information database first locates the emission source annual zone by the source year number, then reads the old objects along the old version address, and looks up the old monthly value by the old object number; then it reads the new objects along the new version address, locates the annual record of the accounting table by the source year number, and then reads the annual quantity and unit of measurement by the administrative region code; Old objects, new objects, old month values, and year quantities are uniquely written using task number, source year number, administrative region number, and layer version number to form a version object set; when old month values ​​are missing, object geometry fields are missing, or projection identifiers are missing, the corresponding records are not included in subsequent spatial calculations. After the version object set is formed, S12 is responsible for converging object-level changes to the target grid. Old objects and new objects are first converted to the same planar projection coordinate system, and the horizontal and vertical coordinates are divided by the coordinate quantization unit and then rounded. Line objects delete adjacent duplicate points and zero-length line segments, and surface objects close boundary loops, eliminate boundary self-intersections, and unify the reading direction of the outer and inner loops. Then, the object is clipped along the boundary of the target grid. The coordinates of the object located on the left or lower boundary are assigned to the current target grid, and the coordinates of the object located on the right or upper boundary are assigned to the adjacent target grid. Each clipping result is assigned an object piece number with the object number, the target grid number, and the order within the piece. Old object pieces falling into the same target grid are geometrically joined, and new object pieces are geometrically joined in the same way. The old object union minus the new object union is the old subtracted piece, and the new object union minus the old object union is the new added piece. The two are merged to form a symmetrical difference result. When the length of the symmetry difference of a line object reaches one coordinate quantization unit, the corresponding target grid is written as a variable tile; when the area of ​​the symmetry difference of a surface object reaches one coordinate quantization unit square, the corresponding target grid is written as a variable tile; the remaining results are written as zero. The center point of the changing tile is converted into a horizontal grid number and a vertical grid number relative to the unified geographical origin. The two grid numbers are then padded to the same width and read from the high bit to the low bit. When the vertical bit is zero and the horizontal bit is one, the remaining low bits of the two groups are inverted bit by bit and then swapped. When the vertical bit is zero and the horizontal bit is zero, they are swapped directly. In each round, the horizontal bit is multiplied by three and then XORed with the vertical bit to generate a quaternary code. The Hilbert code is obtained by splicing the code according to the reading order. Finally, the Hilbert code and the target grid number are arranged in sequence to form a sequence of changing tiles. Central processing unit resource allocation is completed by S13; within the same scheduling transaction, the industrial cloud computing platform reads the status of logical processing cores, writes logical processing cores with a status word of schedulable and an empty local run queue as idle cores, and generates an idle core sequence according to the logical processing core number; the idle cores participating in this task are then written with the task occupancy number; When the number of idle cores is zero, the spatial layer update task enters the waiting queue and the scheduling status is reread after the logic processing core is released. When the number of idle cores is non-zero, the number of variable tiles is divided by the number of idle cores to obtain the quotient and remainder. The logic processing cores whose idle cores fall within the remainder range get the quotient plus one corresponding CPU credit, and the remaining logic processing cores get the CPU credit corresponding to the quotient. When the number of variable tiles is less than the number of idle cores, the first few logic processing cores are each allocated one variable tile, and the CPU credit of the remaining logic processing cores is written to zero. The platform sequentially extracts the same number of changing tiles as the CPU along the changing tile sequence, forms a fragment number with the task number, logical processing core number and fragment order, and writes it into the fragment table to obtain the fragment set; When retrying a task, first retrieve the shard table corresponding to the task number. If a shard set already exists, return the original shard set directly. If any logic processing core involved in the allocation fails before submission, cancel the current round of shard writing and release the task occupancy number. The new layer version will not enter the release state. Taking the road layer update corresponding to tire wear emission sources as the actual operation scenario, when an old road is split into multiple new road records, the platform first reads the corresponding old object, old month value, new object and year quantity based on the source year and administrative region code, and then cuts the old and new roads to the same target grid. Target grids whose object numbers change but whose geometric range remains consistent will not be written into the changed tiles; only target grids that still have geometric differences after quantization will be included in the changed tile sequence. After the changed tiles are arranged according to Hilbert codes and assigned to the free kernel, subsequent factor graph matching only processes object tiles within the tile set, thus keeping the unchanged space records from being overwritten and providing version-consistent, administratively defined, and non-duplicate object tiles as input for the migration of old month values ​​to new objects.

[0018] S2. For old objects within the fragment set, use the spatial index of the geographic information database to find new objects that intersect the bounding boxes, and construct a factor graph with the old and new objects. Generate edge costs by taking the inverse value of the intersection ratio, the difference of the surrogate value, the difference of the centroid grid number, and the difference of the source year. Subtract the remaining geometric quantities of the old and new objects in ascending order of edge costs to obtain the inherited edges. After the fragment set enters the object inheritance relationship calculation, the geographic information database first limits the candidate range with spatial index, then constructs candidate maps with geometric intersection, proxy contribution and location difference, and then obtains the transfer amount between objects through factor graph matching with geometric conservation constraints, and realizes the transfer amount as spatially non-overlapping inheritance fragments. All geometric operations use the projection coordinate system, coordinate quantization unit, source year, administrative region code, and object piece number determined by S1; candidate graphs are created for line objects and surface objects respectively to avoid length and area entering the same conservation relationship; The generation of candidate relations starts from the old object pieces in the shard set; after the geographic information database reads the bounding box of the old object piece, S21 calls the R tree index to search for the nodes that intersect with the bounding box of the old object piece layer by layer from the root node, and only returns the corresponding new object piece at the leaf node; Old object slices and new object slices must also have the same source year, administrative region code and geometric type. Records with inconsistent fields are written to the slice cross-write exception table and are not included in the candidate map. After candidate retrieval is completed, the boundaries of the two objects are transformed to the same projected coordinate system and rounded according to the coordinate quantization unit; after the face object is inserted at the boundary intersection point, the line segments that are simultaneously located inside or on the boundary of the two objects are retained, and then connected according to the direction to form the intersection surface, and the intersection area is used as the candidate edge intersection geometry; The line object is first cut according to the target grid. Within the same target grid, the endpoints of the old line segment are projected perpendicularly to the new line segment. The projection interval that falls within the range of the endpoints of the new line segment is written as the line support intersection. Then, the length that the old line segment and the new line segment can support each other within this interval is written as the intersection geometry. When two line segments coincide precisely, the length of the coinciding line is read directly. When the parallel offset causes the exact intersection to be zero, the line support intersection is still retained for S3 to correct the coordinate offset. The union of all intersections connected by the same old object piece is geometrically calculated. The overlapping part is counted only once to obtain the inheritable quantity. Edge costs are calculated synchronously when candidate edges are generated; the intersection ratio is calculated by dividing the intersection geometry of candidate edges by the geometry of the old object piece; the surrogate raster has a numeric type field; When the total amount of the grid is represented numerically, the proxy value for face objects is accumulated by multiplying the grid value by the ratio of the intersection area to the grid area. When the area density is represented numerically, the proxy value is accumulated by multiplying the grid value by the intersection area. For line objects, the proxy value is accumulated by multiplying the grid value by the ratio of the overlapping line length to the line length of the objects within the grid. When the sum of the proxy values ​​of two objects is zero, the normalized proxy value difference is written as zero; otherwise, the absolute value of the proxy value difference is divided by the sum of the proxy values ​​of the two objects. The centroid coordinates of the object are converted into horizontal and vertical grid numbers according to the origin of the target grid and the grid side length. The difference in centroid grid number is the sum of the absolute values ​​of the differences in grid numbers in the two directions, and then divided by the sum of the number of horizontal and vertical grids in the current segment. S21 uses the value obtained by subtracting the intersection ratio from one as the starting value, and superimposes the normalized surrogate value difference and the normalized centroid grid number difference to form the entity candidate edge cost; records with source year numbers that are different from each other do not generate entity candidate edges, and are written into the source year number conflict record. Each old object piece is connected to a recycling node. The cost of a recycling edge is the sum of the costs of all candidate edges for entities in that old object piece plus one. When there are no candidate edges for entities, only recycling edges are retained, thus forming a candidate graph. Before entering the matching solution, S22 converts the candidate edge transit amount into integer geometric quantities. Line objects are represented by a length word with a coordinate quantization unit, and face objects are represented by an area word with a coordinate quantization unit squared. The sum of the outgoing edge propagation and recycling edge propagation of each old object piece is equal to the inheritable amount. The sum of the incoming edge propagation of each new object piece does not exceed the geometric amount of the new object piece. The propagation amount of each candidate edge does not exceed the geometric amount of the corresponding intersection. The propagation amount of the recycling edge does not exceed the inheritable amount of the corresponding old object piece. The initial value of candidate edges for entities is written to zero, and the initial value of recycled edges is written to the inheritable value of the old object slice. The multiplier of the old object corresponds to the equality constraint, which allows writing positive or negative values. The multiplier of the new object corresponds to the capacity constraint of the incoming edges, which is initially written to zero and cannot be lower than zero after the update. The candidate edge propagation amount is multiplied by the edge cost and then accumulated to form the value to be deducted. The recycled edges participate in the same accumulation. The initial step size is the reciprocal of the number of candidate edges. When the candidate graph contains only recycled edges, the initial step size is set to one, forming an initial matching state that can directly enter the iteration. The iterative update is performed by S23 according to the candidate edge number order; the reduction cost of the entity candidate edge is obtained by adding the old object multiplier to the edge cost and then adding the new object multiplier, and the reduction cost of the recycled edge is obtained by adding the old object multiplier to the recycled edge cost; the current transmission amount of the candidate edge is subtracted from the product of the current step size and the reduction cost to obtain the update value of this round; Write zero when the update value is below zero; write the intersection geometry when the update value of the entity candidate edge exceeds the intersection geometry; write the inheritable geometry when the update value of the recycled edge exceeds the inheritable value; and then round down to the integer geometry value. The outgoing edge deviation of each old object piece is the difference between the outgoing edge propagation amount of the entity, the reclamation edge propagation amount, and the inheritable amount, plus the old object multiplier and the product of the current step size and the outgoing edge deviation; the incoming edge deviation of each new object piece is the entity incoming edge propagation amount minus the new object piece geometry, plus the new object multiplier and the product of the current step size and the incoming edge deviation, and zero is written when the result is less than zero. If the value to be reduced in this round is higher than the value to be reduced in the previous round, or if the number of geometric conservation violations has not decreased, cancel the transit quantity and constraint multiplier of this round, halve the current step size and recalculate; when the transit quantities of two adjacent rounds are consistent bitwise according to integer geometric quantities and all geometric conservation relations are valid, output the transit edge set. If the current step size is multiplied by all the reduced costs and the result is rounded to zero, and there are still entity edges that have not met the propagation amount, the remaining propagation amount is written to the corresponding recycled edge, and then the iteration ends to avoid the sharding task from being stuck in a state without effective updates. When the transmission quantity is implemented to a specific spatial range, S24 first divides the intersection geometry corresponding to each candidate edge of the entity along the boundary of the target grid, and generates Hilbert codes based on the center points of the divided parts; when the Hilbert codes are the same, they are arranged in ascending order according to the intersection piece numbers. The line object is divided into segments along the original line direction. When the cumulative value reaches the position of the transmission quantity, the last segment is cut off according to the remaining length. The intersection area of ​​the face objects is accumulated sequentially according to Hilbert code. When the last intersection piece is accumulated, it is divided along the horizontal axis with the coordinate quantization unit as the movement step size. The area difference is entered into an area word and then divided along the vertical axis. The tail difference that does not reach an area word is written into the recycling edge. The resulting space range forms the inherited piece. The inherited pieces that connect the same old object piece or the same new object piece are intersected pairwise. When the intersection length or intersection area is non-zero, an overlapping piece is formed. The overlapping piece retains the candidate edges that are ranked first in the edge cost ascending order. When the edge costs are the same, the edges to be retained are determined by the candidate edge number ascending order. The remaining candidate edges are deducted from the overlapping piece and the upper limit of the transit quantity is rewritten with the geometry of the inherited piece after deduction. Then, the calculation is returned to S23. Each return reduces the upper limit of the transit amount of at least one candidate edge by an integer geometric amount until there is no geometric overlap between inherited slices and the transit amounts of two adjacent slices are consistent. Entity candidate edges with non-zero transit amounts are written as inherited edges, and the geometric amount corresponding to the recycled edges is written into the recycled field for S4 to read. In the scenario of updating the road proxy layer, when an old road is split into two new roads in the new layer, the R-tree index first returns the new road patch where the bounding boxes intersect, the line support intersection retains the connection interval formed by the survey offset, and the edge cost is combined with the road proxy value and the centroid grid number difference to distinguish the two candidate edges. Factor graph matching distributes the inheritable length of the old road to two new road pieces, and lengths that cannot be inherited by the new roads enter the recycling edge; after spatial overlap resolution, the inherited pieces are bound to the corresponding new road pieces respectively. Therefore, the inherited edges output by S2 simultaneously possess a definite correspondence between the old and new objects, an integer transitivity, and a non-overlapping spatial range, which can be directly read by the B-spline free deformation registration algorithm.

[0019] S3. Based on the inherited edge, execute the B-spline free deformation registration algorithm to generate a control grid within the outer rectangle of the old object, pair the sampling points of the old object boundary to the boundary points of the new object, generate the control grid displacement field according to the point-pair displacement, and use the cubic B-spline basis function to map the old object to the coordinate domain of the new object to obtain the registered object. After inheriting the spatial connection between the old and new objects, it is also necessary to eliminate the coordinate deviation caused by surveying revision, boundary translation and local shape changes, and keep the geometric range corresponding to the inherited edge transmission from flipping, going out of bounds or being repeatedly covered. To this end, the geographic information database reads the object boundary according to the inheritance edge, establishes the correspondence of sampling points, obtains the control grid displacement field through the B-spline free deformation registration algorithm, and then refines the control grid by combining the reverse mapping results, finally forming a registration object that can be read for inheritance rate calculation; Before entering the deformation solution, the geographic information database reads the old object boundary and the new object boundary according to the inherited edge number, and converts them to the same plane projection coordinate system used by S1. The horizontal and vertical coordinates are rounded according to the coordinate quantization unit. When the old object is a closed surface object, the first and second points of the boundary coordinate Hilbert code in ascending order are used as the starting point for reading. The outer ring is read in a clockwise direction, and the inner ring is read in a counterclockwise direction. When the old object is a line object, retain the forward boundary sequence and the reverse boundary sequence respectively for selection in subsequent correspondence calculations; Boundary vertices and the intersections of the boundary with the target grid lines are written as sampling points in sequence. For repeated coordinates, only the first record of the ascending number is retained. In S31, the control grid order is obtained by taking the square root of the number of old sampling points upwards. If the order is less than four, it is written as four. Then, the outer rectangle of the old object is divided equally according to the control grid order, and control points are laid out in the horizontal and vertical directions using a clamping node sequence that repeats three times with the end nodes, so that the boundary position of the old object can still read four rows and four columns of control points. The initial lateral and longitudinal displacements of each control point are written to zero. The control point numbers are generated according to the lateral and longitudinal order, thus forming the initial control grid. When the boundary of the old object is empty or the lateral or longitudinal span of the outer rectangle is zero, the corresponding inherited edge transfer amount is written to the recycling amount and will not be included in the subsequent registration. The correspondence of sampling points is determined by the boundary's own order and normal projection; the distance between adjacent old sampling points is accumulated along the old object boundary, and the old arc length order is obtained by dividing the accumulated distance by the total length of the old object boundary; the new arc length order is generated for the new object boundary in the same way. For closed surface objects, the starting point and direction determined by S31 are used directly; for line objects, the total difference in arc length position between the forward and reverse sequences of the old object and the boundary of the new object is calculated, and the first sequence is read in ascending order according to the total value. If the total values ​​are the same, they are read in ascending order according to the boundary sequence number. For each old sampling point, S32 first arranges the new sampling points according to the absolute value of the arc length position difference and the new sampling point number, and then projects them from the old sampling points along the positive and negative normal directions to the boundary of the new object respectively; when there are multiple intersections in the same normal direction, they are arranged according to the projection distance and the order of the new boundary where the intersections are located; when there are no intersections in the positive and negative normal directions, the arc length position arrangement result is retained. Reverse retrieval repeats the same process starting from the new sampling point, only writing the records where the old object points to the new object and the new object points back to the same old object as registration point pairs; When the registration point pair is empty, the inherited edge transfer amount is transferred to the recovery amount; when the registration point pair is not empty but does not cover part of the control unit, the uncovered control points do not participate in the point pair item, and their displacement is only constrained by the deformation item. The control grid displacement field is obtained in S33 through iterative calculation of the uniformly normalized residual vector. First, the coordinates of the old sampling points after being mapped by the current control grid displacement field are subtracted from the coordinates of the corresponding new sampling points. The resulting horizontal and vertical differences are divided by the length of the diagonal of the outer rectangle of the old object and written into the point-pair residual. The second-order difference of the lateral and longitudinal displacements of adjacent control points are divided by the square root of the number of control points and written into the deformation residual; the geometric quantity of the intersection of the old object and the new object after mapping is reduced by the inherited edge transfer quantity and then divided by the geometric quantity of the new object and written into the inheritance residual. The target value is formed by adding the sum of squares of the point-to-point residuals, the sum of squares of the deformation residuals, and the sum of squares of the inherited residuals, ensuring that the residual vector and the target value use the same dimension. The displacements of each control point are spliced ​​together into a displacement vector according to the control point number. The partial derivatives of the point residuals and deformation residuals with respect to the displacement vectors are obtained by the cubic B-spline basis functions and their derivatives. The partial derivatives of the inherited residuals are obtained by successively adding a displacement component and a coordinate quantization unit of a control point, recalculating the intersection geometry, and then dividing the increment of the intersection geometry by the coordinate quantization unit. All partial derivatives form the residual Jacobian matrix. The Gauss-Newton directions are obtained using QR decomposition with column pivoting. After the residual Jacobian matrix is ​​rearranged with column pivoting, the direction components corresponding to the columns whose diagonal elements are rounded to zero according to the coordinate quantization unit are written as zero, and the remaining direction components are used to generate the Gauss-Newton directions according to the back substitution results. The current step size is initially written as 1, and the control point displacement update is obtained by multiplying the Gauss-Newton directions by the current step size. The updated displacement vector recalculates the target value and calculates the deformation Jacobian determinant at each control unit corner point, center point, old sampling point, target grid line intersection, and grid point generated by coordinate quantization unit within the control unit; If the target value is not lower than the target value before the update, or if the deformation Jacobian determinant of any check position is not positive, cancel the current displacement update, halve the current step size, and recalculate; stop iterating when all control point displacement updates are zero after rounding to the coordinate quantization unit. If the geometric conservation relationship holds at the time of stopping, the current displacement vector is written as the control grid displacement field; when the inherited residual still corresponds to a non-zero integer geometric quantity, the difference is written into the recovery quantity, and the inherited edge transmission quantity is written back with the actual intersection geometric quantity. After the control grid displacement field is formed, S34 generates candidate registration objects point by point; for each boundary coordinate of the old object, the corresponding control unit is located first, and then the four rows and four columns of control points around the control unit are read. The four horizontal cubic B-spline basis function values ​​and the four vertical cubic B-spline basis function values ​​are obtained based on the horizontal and vertical decimal positions of the boundary coordinates in the control grid. The lateral displacements of the sixteen control points are multiplied by the corresponding lateral basis function values ​​and longitudinal basis function values, and then summed to obtain the lateral deformation. The longitudinal deformation is obtained in the same way. The two deformations are added to the old boundary coordinates, and then connected according to the original boundary order of the old object to form a candidate registration object. The reverse displacement field regenerates the control grid, registration point pairs, and control grid displacement field by exchanging the old and new objects and then following steps S31 to S33, without directly taking the negative of the forward displacement field; the mapped sampling points in the candidate registration objects are reflected back by the reverse displacement field, and the Euclidean distance between them and the corresponding old sampling points is calculated and written as the reflected distance. The reflected distances are arranged in descending order. When the distances are the same, they are arranged according to the old sampling point number. The control unit where the first and second error points are located simultaneously inserts a horizontal node line and a vertical node line, and the original control unit is thus divided into four sub-units. The displacement of the newly added control point is converted from the displacement of the original control grid according to the B-spline node insertion rule, so that the displacement field at the moment of node insertion remains unchanged, and then return to S33 to continue the solution; When the horizontal or vertical span of the control unit is equal to the coordinate quantization unit, the insertion of node lines stops, and the inherited edge transfer amount with non-zero reflection distance is transferred to the recycling amount. After all the retracing distances are rounded to zero according to the coordinate quantization unit, we continue to check whether the boundaries of the candidate registration objects intersect with each other, and retrieve new objects that intersect with the bounding box of the old object and whose corresponding target grid has not written changed tiles. These objects are regarded as unchanged neighboring objects. When a candidate registration object crosses the boundary of any unchanged neighboring object, the current candidate registration object is cancelled and transferred to the recovery quantity. The candidate registration objects that pass all checks are written into the geographic information database with a unique key consisting of task number, layer version number and inherited edge number, thus forming registration objects. Taking the local translation of the road layer accompanied by road splitting as an example, the old road sampling points first establish mutual pointing registration point pairs with the new road sampling points according to the boundary order. The control grid displacement field is responsible for absorbing the mapping coordinate offset. The inheritance item restricts the road intersection length after mapping to maintain the correspondence with the inheritance edge transfer amount. When a feedback error occurs at a local bend, the control unit refines the information synchronously through the lateral node lines and the longitudinal node lines, while the unchanged road boundary continues to serve as a spatial constraint. The final registered object retains the inheritance range of the old object's monthly emission values ​​and provides a geometric basis for S4 to calculate the inheritance rate without flipping, self-intersection, or intrusion into the unchanged space.

[0020] S4. Calculate the inheritance rate based on the intersection geometry of the registered object and the new object, write the old month value into the new object according to the inheritance rate, classify the unwritten month value into the recycling amount, and generate the receiving object quota according to the proxy value of the new object to obtain the migration table. After the registration object completes geometric correction, the emission value migration needs to simultaneously satisfy the following requirements: no repetition in spatial range, no loss of old month values, and conservation of annual emissions within the same administrative region. Based on this, the geographic information database first determines the actual spatial range that can be inherited from the registration results, then divides the old monthly value into the inherited monthly value and the recovery amount, and then generates the receiving object quota based on the grid adjacency relationship within the range of change. The inherited calculation, recycling records and trial allocation records all use the source year, administrative region code, object piece number, month and layer version number. The emission values ​​are uniformly converted into integer emission characters, and the geometric quantities continue to use the integer length characters or integer area characters determined by S2. After the registered object and the new object are overlaid, the geographic information database divides the intersection geometry corresponding to each inherited edge along the boundary of the target grid, and writes the inherited edge number, old object piece number, new object piece number and intersection piece number to each intersection piece; When intersection pieces connecting the same old object piece or the same new object piece have geometric overlap, the overlapping parts are retained in ascending order of edge cost. If the edge costs are the same, they are retained in ascending order of inherited edge number. The corresponding overlapping parts of the remaining inherited edges are deducted. This process is completed by S41. The intersection geometry after deduction is used as the actual inherited amount. When the actual inherited amount is inconsistent with the amount of transmission written by S2, the inherited edges are written back with the actual inherited amount, and the difference between the two is transferred to the recycling field. The inheritance rate is calculated by dividing the actual amount of inheritance by the geometric quantity of the old object piece. The inheritance rate of the same old object piece is accumulated in ascending order according to the edge cost, and writing stops when the accumulated value reaches one. When the last inheritance rate causes the cumulative value to exceed one, the last inheritance rate is rewritten as one minus the sum of the previous inheritance rates; the resulting inheritance rate table stores the actual inheritance amount, inheritance rate and intersection piece number for old month value splitting and reading. The old month value is split into individual months; after entering S42, the old month value of each old object piece is first converted into an integer word, then multiplied by the corresponding inheritance rate and rounded down to obtain the initial inheritance month value of each inheritance edge. The remaining emission words after rounding are added one by one in ascending order of edge cost and inherited edge number, until the total monthly inherited value of the same month is equal to the emission words that the old object piece can inherit this month. The inherited monthly value is written to the corresponding new object fragment according to the month, and the source old object number and intersection fragment number are saved at the same time; the difference between the old monthly value and the total inherited monthly value of the same month is written as the recycling amount; Spatial location is obtained by subtracting the geometric union of all inherited pieces from the old object piece to obtain the recycled piece. Each connected line segment of a line object forms a recycled piece, and each connected surface region of a surface object forms a recycled piece. When the same old object piece contains multiple recycled pieces, the recycling quantity is split according to the proportion of the geometric quantity of each recycled piece to the total geometric quantity of the recycled pieces, and the rounded-off remainder is added in ascending order of the recycled piece number. Each recycling record is bound to the source year, administrative region code, old object number, month, recycling piece number, and recycling quantity. The sum of the inherited monthly value and the recycling quantity is always equal to the corresponding old monthly value, thus forming an inheritance record. Uninherited emission values ​​generate quotas based on the spatial connectivity between the changing grids; before entering S43, each recycling piece reads the midpoint of the line object as the supply point according to the cumulative line length, and reads the centroid of the area as the supply point for the surface object; when the supply point falls on the grid boundary, it is assigned to the first target grid in ascending order of the target grid number. Each target grid within the range of change serves as a node in the spatial transfer graph. When two target grids share a complete grid edge, a bidirectional transfer edge is established. When there is only corner contact, no transfer edge is established. The transfer edge must not cross the administrative region boundary. A new object with a non-zero proxy value is written as the inheriting object. Its proxy value is first divided by the sum of the proxy values ​​of all inheriting objects under the current source year and administrative region to obtain the dimensionless proxy value. The reciprocal of the proxy value in the path cost is the reciprocal of this dimensionless proxy value. The platform performs a breadth-first search from the target grid where the supply point is located, with the number of grid edges increasing. Only one path is retained for the same receiving object. When the number of grid edges on the path is the same, it is retained in ascending order of path code. The high-order bits of the path code are written with the number of target grids traversed, and the subsequent bits are written with the target grid number in the order of travel. Unused positions are padded with zeros. After the path is formed, the path cost is first obtained by adding the reciprocal of the dimensionless proxy value of the receiving object to the number of grid edges traversed by the path. Then, the path is read in ascending order of path cost and path code. The agency share of each receiving object is obtained by multiplying the original value of the recovered amount by the proportion of the agency value of the receiving object to the total agency value of all receiving objects. The remaining agency share is the agency share minus the accumulated received amount. For the current path, subtract the absolute value of the difference between the sum of the recovered amount and the remaining agency share, divide by two, and write the resulting integer emission amount as the trial transfer amount. The trial transfer amount is deducted from the recovered amount and the remaining agency share respectively, and accumulated to the corresponding receiving object. The trial transfer amount is only written to the trial allocation field of the migration table, without changing the annual or monthly value of the receiving object; after the recovery amount is zeroed, the annual quota is obtained by dividing the cumulative received amount of each receiving object by the original value of the recovery amount. When multiple recycling plates belong to the same source year and administrative region code, the cumulative amount received by each receiving object is first summarized, and then a unified annual quota is generated; If there is no receiving object in the spatial transfer map, or if the supply point is not connected to all receiving objects, the corresponding recovery amount is written into the unallocated record and a missing agent mark is attached. No subsequent grid list of the corresponding source year is generated. Taking the adjustment of industrial land boundaries as the actual operational scenario, after the old industrial land is registered, part of it is inherited by the new industrial land, and the rest forms a recycling area. S41 First, eliminate the overlapping intersections between multiple inherited edges. S42 Then, accurately decompose the old monthly value of each month into the inherited monthly value and the recovery amount. S43 Based on the shared edge relationship of the changed grid, allocate the recovery amount to the new industrial plots with non-zero proxy values. The migration table ultimately saves the inherited monthly value, recovered pieces, trial transfer volume, and annual quota. Subsequently, S5 only writes the annual emission value of the receiving object based on the annual quota, thereby avoiding duplicate measurement of the same emission value in the trial allocation stage and the formal writing stage.

[0021] S5. Summarize the migration table and subtract the inherited value from the annual total to obtain the allocation quantity. Write the quota into the receiving object. Split the annual allocation value of the receiving object according to the monthly agency sequence. Then correct the monthly value of the hit object according to the verification record and return the same amount of allocation value in the unhit receiving object. Summarize the monthly value of the object according to the geometric value of the intersection between the new object and the target grid. Output the monthly allocation list of the target grid with the layer version number. After the migration table is completed, the emission value processing changes from object inheritance to annual supplementation, monthly splitting, spatial verification, and grid aggregation; The geographic information database uses the source year and administrative region code to limit the same accounting scope. The inherited monthly value remains unchanged from the original month. The allocation quantity is only distributed among the receiving objects. The verification difference only applies to the object patches actually covered by the verification scope. Finally, a target grid monthly emission list that can trace back the input version is formed. First, check the total annual value and annual quantity of the old month under the same source year and administrative region code. If the two are inconsistent, write a version error record and stop the current update. After verification, the inherited monthly values ​​in the S51 summary migration table are accumulated by month and then the annual total is calculated to obtain the annual total of the inherited values. The amount to be allocated is obtained by subtracting the annual total of the inherited values ​​from the annual amount. The quotas of each receiving entity are uniformly converted into fixed-point quota words. The remaining amount generated by rounding down the quota word is added one by one in ascending order according to the receiving entity number until the total quota equals one. After multiplying the amount to be allocated by the quota of each receiving entity and rounding down, the remaining emission characters are added one by one in ascending order according to the receiving entity number, so that the total annual emission value of the receiving entities is equal to the amount to be allocated. The annual emission values ​​of the receiving objects are only formally written in the current processing stage. The trial transfer amount stored in S43 does not participate in the accumulation of the emission values ​​of the objects. After completion, a receiving table is formed. Monthly splitting reads time proxy data one by one according to the receiving object; when executing S52, the geographic information database first locates the monthly proxy sequence by the source year, administrative region code and receiving object number, and then intersects the receiving object with the time proxy raster execution space of the corresponding month; When storing the total number of individual cells in a time-proximity raster, the intersection contribution is obtained by multiplying the raster value by the proportion of the intersection geometry to the total raster geometry; when storing the density in a time-proximity raster, the intersection contribution is obtained by multiplying the raster value by the intersection geometry, and all intersection contributions are summed to form the proxy value for that month. When monthly data is missing, a missing month marker is written instead of zero. When all twelve months have valid proxy values ​​and the total proxy value for the year is not zero, the monthly share is obtained by dividing the monthly proxy value by the total proxy value for the year. When the total proxy value for the year is zero or all twelve months are missing, the monthly share is obtained by dividing the number of days in each month by the number of days in the year. February in a leap year is counted as the 29th day. The annual ranking value of the inherited object is multiplied by the monthly share and rounded down to form the monthly value to be assigned. The rounding remainder is added month by month in order. The inherited monthly value and the monthly value to be assigned for the same new object in the same month are added to form the object monthly value. The source fields of the inherited monthly value and the monthly value to be assigned are saved respectively. The summary result is written to the object monthly table. After the verification data enters the geographic information database, the new objects in the monthly table are first segmented along the boundary of the verification range to obtain hit and miss slices, and then the emission values ​​of the corresponding months are processed by S53. The monthly value of an object is divided into each object piece according to the proportion of the geometric quantity of the object piece to the geometric quantity of the new object. The monthly values ​​of the hit pieces are accumulated to form a summary value. The signed verification difference is obtained by subtracting the summary value from the verification value. When the verification difference is positive, the monthly value of the hit pieces is increased according to the proportion of the geometric quantity of the intersection of the hit pieces to the total geometric quantity of the intersection of the hit pieces, and an equal amount of emission words are deducted from the monthly value of the unhit pieces according to the proportion of the original monthly value to the total monthly value of the unhit pieces. When the verification difference is negative, the monthly value of the hit segment is deducted according to the proportion of the original monthly value of the hit segment to the total monthly value of the hit segment, and the equivalent emission words are made up according to the proportion of the original monthly value of the non-hit segment. If the value of any object piece is lower than zero after deduction, the monthly value of the corresponding object piece will be written to zero, and the unfinished deduction amount will be redistributed to object pieces that still have monthly values. When the missing chip is empty and the verification difference requires an increase in the monthly value of the hit chip, write the verification conflict record and retain the monthly value of the object before verification. When multiple verification records cover the same month and the same spatial range, first divide the new object into non-overlapping verification pieces along all verification boundaries, then establish the correspondence between verification records and verification pieces, and uniformly allocate the piece monthly value; when the total piece monthly value of each verification record cannot be equal to the corresponding verification value at the same time, write it into the verification conflict record, and do not submit the corresponding verification monthly table. During the grid aggregation stage, the new objects and target grids in the monthly table are overlaid and verified using a geographic information database; S54 obtains the grid share by dividing the intersection geometry by the geometry of the new object. When a new object spans multiple target grids, each grid share is converted into a fixed-point share word, and the rounded remainder is written into the target grid where the centroid of the new object is located; when the centroid is located at the grid boundary, it is written into the first target grid in ascending order of the target grid number. After multiplying the monthly value of the target grid by the grid share and rounding down, the emission margin is added in ascending order according to the target grid number, so that the total monthly value of all target grids for the same new target is equal to the monthly value of the corresponding target; then, the monthly emission value of the target grid is obtained by summing the target grids and months. When there is a difference between the annual total of the monthly emissions of the target grid and the annual amount, the difference is written back in the order of the month and the target grid number. During the deduction process, the monthly emissions of the grid are kept not lower than zero. After the difference is returned to zero, the layer version number is written back. The version association table simultaneously stores the annual version, spatial proxy layer version, monthly proxy version, verification record version, and task number. Each grid monthly layout record is uniquely keyed by task number, source year, administrative region code, target grid number, and month. After all shards are written, a new version is submitted. If any shard fails, the old version's release address is retained. Taking the annual update of synthetic fiber washing emission sources as an application scenario, the old object inherits the monthly value, first retains the original monthly distribution, and the remaining annual amount is written into the sewage collection zone according to the quota of the receiving object. Then, it is combined with the monthly domestic sewage proxy data to form the monthly value to be allocated. When the verification record covers part of the sewage collection zone, the discharge value is only adjusted by the same amount between the hit zone and the miss zone, and the provincial annual amount remains unchanged; after the new object is overlaid with the target grid, the discharge value is written into the grid record of each month according to the actual intersection range, and the corresponding accounting input is restored through the version association table; The resulting target grid monthly emissions inventory maintains consistency in annual totals, monthly structure, spatial scope, and historical versions.

[0022] Working principle: This solution first uses an industrial cloud computing platform to read the old and new spatial layers and annual emissions for the same emission source and accounting year, identify the target grid that has actually changed, and allocate computing resources. Then, it uses the spatial index of the geographic information database to find the correspondence between the old and new objects, determines the geometric inheritance relationship between one-to-many or many-to-one through the factor graph matching algorithm, and then corrects the boundary translation and local deformation with the B-spline free deformation registration algorithm to obtain the actual spatial range that can inherit the old monthly emission values. The inheritable emission values ​​are written into the new objects according to the inheritance rate, and the uninherited parts are redistributed according to the spatial connectivity and proxy values. Then, it is combined with the monthly proxy sequence to split into monthly emissions, and the local results are corrected according to the verification data. Finally, it is summarized to form a monthly emission list of the target grid with version information. Taking road layer updates as an example, when an old road is split into multiple new roads due to surveying revisions, the system first locates the target grid involved in the road changes, then uses factor maps to determine the connection relationship between each new road and the old road, and uses B-spline free deformation registration to eliminate road boundary offsets; monthly emission values ​​in the old roads that can correspond to the new roads are inherited according to the actual overlapping range, while emission values ​​that cannot correspond enter the recycling process, and are then allocated to the receiving roads according to the road proxy value and grid connectivity within the change range; after completing the annual supplementary allocation, the system combines monthly traffic change data to form monthly emission volumes, and uses verification records to adjust local deviations, and finally summarizes the object emission volumes into the target grid to obtain a microplastic emission inventory that is traceable in terms of spatial location, monthly changes, and historical versions.

[0023] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for constructing a high spatiotemporal resolution emission inventory of primary microplastics, characterized in that, include: S1. Obtain the spatial layer update task initiated by the industrial cloud computing platform, generate the source year number by combining the emission source number with the accounting year, read the old object, old month value, new object and annual quantity in the geographic information database according to the source year number, and generate the CPU quota based on the number of changed tiles and the number of idle cores of the central processing unit to obtain the fragment set. S2. For old objects within the fragment set, use the spatial index of the geographic information database to find new objects that intersect the bounding boxes, and construct a factor graph with the old and new objects. Generate edge costs by taking the inverse value of the intersection ratio, the difference of the surrogate value, the difference of the centroid grid number, and the difference of the source year. Subtract the remaining geometric quantities of the old and new objects in ascending order of edge costs to obtain the inherited edges. S3. Based on the inherited edge, execute the B-spline free deformation registration algorithm to generate a control grid within the outer rectangle of the old object, pair the sampling points of the old object boundary to the boundary points of the new object, generate the control grid displacement field according to the point-pair displacement, and use the cubic B-spline basis function to map the old object to the coordinate domain of the new object to obtain the registered object. S4. Calculate the inheritance rate based on the intersection geometry of the registered object and the new object, write the old month value into the new object according to the inheritance rate, classify the unwritten month value into the recycling amount, and generate the receiving object quota according to the proxy value of the new object to obtain the migration table. S5. Summarize the migration table and subtract the inherited value from the annual total to obtain the allocation quantity. Write the quota into the receiving object. Split the annual allocation value of the receiving object according to the monthly proxy sequence. Then, correct the monthly value of the hit object according to the verification record and return the same amount of allocation value in the unhit receiving objects. Summarize the monthly value of the object according to the geometric value of the intersection between the new object and the target grid. Output the monthly allocation list of the target grid with the layer version number.

2. The method for constructing a high spatiotemporal resolution emission inventory of primary microplastics according to claim 1, characterized in that: S1 includes: S11. After receiving the spatial layer update task on the industrial cloud computing platform, the emission source number is written into the high digit and the accounting year is written into the low digit to form the source year number. The version spatial index of the geographic information database is retrieved using the source year number. First, the old objects are read along the old version address and the old month value is retrieved back. Then, the new objects are read along the new version address and the year quantity is read from the accounting table to form a version object set. S12. Cut the version object set according to the target grid boundary, calculate the geometric union of old objects falling into the same target grid, and perform a symmetric difference operation on the geometric union of the obtained geometric union and the geometric union of new objects in the same target grid. Write the target grid with non-empty symmetric difference result as a change tile, and then generate Hilbert code according to the coordinates of the center point of the change tile to form a sequence of change tiles arranged in ascending order of Hilbert code. S13. Divide the number of variable tiles by the number of idle cores of the central processing unit to obtain the quotient and remainder. Write the CPU amount obtained by adding one to the quotient to the idle core corresponding to the remainder in order of idle core position. Write the CPU amount obtained by the quotient to the remaining idle cores. Then, extract the same number of variable tiles as each CPU amount along the variable tile sequence to form a fragment set.

3. The method for constructing a high spatiotemporal resolution emission inventory of primary microplastics according to claim 2, characterized in that: S2 includes: S21. For each old object in the fragment set, call the R-tree index of the geographic information database to read the new object that intersects the bounding box. For each pair of objects, calculate the exact geometric intersection. Line objects are measured by the intersection length, and polygon objects are measured by the intersection area. Write the geometric union of the intersection of the same old object as an inheritable quantity. Starting with the value obtained by subtracting the intersection ratio, successively add the surrogate value difference normalized by the sum of the surrogate values ​​of the two objects, the centroid grid difference normalized by the fragment edge length, and the source year difference and same position to obtain the edge cost of the candidate edge and form the candidate graph. S22. Establish a factor graph matching target with geometric conservation constraints on the candidate graph. The sum of the product of the candidate edge transit amount and the edge cost is the value to be deducted. The total outgoing edge transit amount of each old object is limited to be equal to the inheritable amount. The total incoming edge transit amount of each new object does not exceed the geometric amount of the new object. The candidate edge transit amount does not exceed the geometric amount of the corresponding intersection. The transit amount of the candidate edge with the source year number being one is fixed to zero. Then, the transit amount and the constraint multiplier are written to zero and the inverse of the number of candidate edges is written to the initial step size to form the initial matching state.

4. The method for constructing a high spatiotemporal resolution emission inventory of primary microplastics according to claim 3, characterized in that: S2 also includes: S23. Based on the initial matching state, the reduced cost is obtained by superimposing the constraint multipliers at both ends on the edge cost in each round. The transit amount is updated according to the reduced cost and the current step size. The update value below zero is written as zero and the update value above the intersection geometry is written as the intersection geometry. The constraint multipliers are then rewritten according to the deviation of the outgoing edge transit amount and the deviation of the incoming edge transit amount. When the value to be reduced in this round is higher than the value to be reduced in the previous round, the current step size is halved and the update in this round is canceled until the transit amounts of the two adjacent rounds are consistent bit by bit and all geometric conservation constraints are met, thus obtaining the transit edge set. S24. Based on the transit edge set, extract the inherited pieces whose cumulative geometric quantity equals the transit quantity in ascending order of Hilbert code within the intersection geometry corresponding to each candidate edge. Calculate the geometric intersection of the inherited pieces that connect the same old object or the same new object. Keep the overlapping pieces in the candidate edges that are ranked first in the edge cost ascending order and deduct them from the remaining candidate edges. Then, rewrite the upper limit of the transit quantity of the corresponding candidate edge with the geometric quantity of the deducted inherited piece and return to S23 to recalculate until there is no geometric overlap between the inherited pieces and the transit quantities of two adjacent times are consistent bit by bit. Write the candidate edges with non-zero transit quantities as inherited edges.

5. The method for constructing a high spatiotemporal resolution emission inventory of primary microplastics according to claim 4, characterized in that: S3 includes: S31. Based on the inherited edge, read the old object boundary and the new object boundary from the geographic information database and convert them to the same projection coordinate system. Write the boundary vertices and the intersections of the boundary and the target grid lines as sampling points. Round up the square root of the number of old sampling points to obtain the control grid order. Then set control points according to the equal distribution of the old object's outer rectangle span to form the initial control grid. S32. Accumulate the distance between adjacent sampling points along the boundary of the old object and the boundary of the new object respectively, and divide by the total length of the corresponding boundary to obtain the arc length order. For each old sampling point, first determine the retrieval order of the new sampling point by the absolute value of the difference in arc length order, then rewrite the retrieval order by the distance of the old sampling point projected along the normal to the boundary of the new object, and retain the sampling points that point to each other from the retrieval results from the old object to the new object and from the new object to the old object, forming a registration point pair.

6. The method for constructing a high spatiotemporal resolution emission inventory of primary microplastics according to claim 5, characterized in that: S3 also includes: S33. Based on the registration point pairs, execute the B-spline free deformation registration algorithm. Divide the sum of squares of the deformation residuals of the registration point pairs by the square of the diagonal of the outer rectangle of the old object to obtain the point pair term. Divide the sum of squares of the second-order difference of the displacements of adjacent control points by the number of control points to obtain the deformation term. Then, divide the square of the difference between the intersection geometry of the mapped old object and the new object and the inherited edge transfer amount by the square of the geometry of the new object to obtain the inheritance term. With the goal of decreasing the sum of the three terms, and limiting the deformation Jacobian determinant of the control unit corners and center points to be positive, update the control point displacements using the Gauss-Newton direction. If the target value does not decrease or the deformation Jacobian determinant is not positive, halve the update step size and recalculate until the displacement update of each control point is zero after being rounded according to the coordinate quantization unit of the geographic information database, thus obtaining the control grid displacement field.

7. The method for constructing a high spatiotemporal resolution emission inventory of primary microplastics according to claim 6, characterized in that: S3 also includes: S34. Using cubic B-spline basis functions, the control grid displacement field is applied to the old object to form candidate registration objects. Then, the candidate registration objects are mapped back to the coordinate domain of the old object using the reverse displacement field obtained after exchanging the old and new objects. The first and second error points are selected in descending order of the distance between the mapped sampling points and the old sampling points. A control point is added at the center of the control unit where the error point is located, and the process is repeated until all mapped distances are rounded to zero according to the coordinate quantization unit. The candidate registration objects do not have boundary self-intersections or cross the boundaries of unchanged adjacent objects. The candidate registration objects are written into the geographic information database and the registration objects are output.

8. The method for constructing a high spatiotemporal resolution emission inventory of primary microplastics according to claim 7, characterized in that: S4 includes: S41. For each inherited edge, the registration object and the new object are overlaid through the geographic information database, and the intersection geometry is cut along the target grid boundary. The repeated intersection pieces are retained in the first inherited edge and deducted from the subsequent inherited edges in ascending order of edge cost. The inheritance rate is obtained by dividing the amount of intersection geometry after deduction by the amount of geometry of the old object, and an inheritance rate table is formed. S42. Multiply the old month value by each inheritance rate according to the inheritance rate table to obtain the inheritance month value. Write the inheritance month value into the corresponding new object by month. Then subtract the total inheritance month value of the same month from the old month value to obtain the recovery amount. Write the source old object number and intersection piece number of the inheritance month value into the geographic information database to form an inheritance record. S43. Using the intersection of the recovered amount as the supply point, and the new objects with non-zero proxy values ​​within the range of change as the receiving objects, a spatial transfer graph is established according to the shared edges of the target grid. The path cost is obtained by adding the number of grid edges traversed by the path to the reciprocal of the proxy value of the receiving object. The paths are read one by one in ascending order of path cost and path code. For each path, the sum of the recovered amount and the remaining proxy share of the corresponding receiving object, minus the absolute value of the difference between the two, is divided by two and written as the transfer amount. The transfer amount is written to the receiving object along the corresponding path and deducted from the recovered amount and the remaining proxy share respectively. After the recovered amount is zeroed, the quota is generated by dividing the cumulative received amount of each receiving object by the original value of the recovered amount, and the migration table is output.

9. The method for constructing a high spatiotemporal resolution emission inventory of primary microplastics according to claim 8, characterized in that: S5 includes: S51. Summarize the inherited monthly values ​​in the migration table according to the source year number and add them up to obtain the total inherited value for the year. Subtract the total inherited value for the year from the annual quantity to obtain the quantity to be allocated. Then multiply the quantity to be allocated by the quota of each receiving object to obtain the annual ranking value of the receiving object, and form the receiving table. S52. Read the monthly agent sequence corresponding to each receiving object from the geographic information database, divide the monthly agent value by the total annual agent value to obtain the monthly share, and multiply the annual ranking value of the receiving object by the monthly share to obtain the monthly value of the object; when the total annual agent value is zero, divide the number of days in each month by the total number of days in the year to obtain the monthly share, and form the monthly table of objects.

10. The method for constructing a high spatiotemporal resolution emission inventory of primary microplastics according to claim 9, characterized in that: S5 also includes: S53. Intersect the geographical range of the verification record with the execution space of the new object in the object monthly table, summarize the monthly value of the hit object according to the verification month, and generate a signed verification difference by subtracting the summary value from the verification value. Then, rewrite the monthly value of the hit object according to the proportion of the intersection geometry to the total number of hit intersection geometry. At the same time, write the callback value with the opposite sign according to the proportion of the original monthly value of the missed receiving object to the total number of missed monthly values, thus forming the verification monthly table. S54. By overlaying and verifying the new objects and target grids in the monthly table through the geographic information database, the grid share is obtained by dividing the intersection geometry by the new object geometry. The monthly value of the object is multiplied by the grid share and then accumulated according to the target grid and the month. When the annual total of the accumulated value is consistent with the annual quantity, the layer version number is written, and the monthly emission list of the target grid is output.