A sample query method for geospatial datasets based on density distribution

Abstract

The present invention belongs to the field of data retrieval and discloses a density distribution-based sample query method for geographic spatial datasets. The method comprises the following steps: data processing: processing a spatial dataset, dividing the spatial dataset into a grid, statistically analyzing the density distribution of location points within the grid, obtaining a density distribution element set, and then generating a minimum covering rectangle; query execution: generating a density distribution element set and a minimum covering rectangle based on a query sample dataset; first, during a query, screening a portion of candidate datasets based on the size of the overlap between the sample dataset and the minimum covering rectangle of each dataset; then, calculating the similarity between the candidate datasets and the query sample dataset based on the density distribution element set, and selecting the k datasets with the highest similarity as the query results. The present invention converts the spatial dataset into a density distribution representation, uses the distribution ratio of location points within the overlapping grids between spatiotemporal datasets as a measure of similarity, and thus improves query efficiency.

Description

Technical Field

[0001] The present invention belongs to the field of data retrieval, and in particular relates to an efficient density distribution-based spatial data set sample query method. Background Art

[0002] As an important part of real-world datasets, spatial datasets are increasingly used worldwide. There is an urgent need to effectively obtain the required spatial datasets, namely spatial dataset search.

[0003] Sample queries on spatial datasets require similarity measurements between datasets. Existing literature has proposed various methods for similarity measurement. For example, EMD measures the minimum cost of transforming one distribution into another by representing spatial datasets in a distributed manner and calculating the EMD. MBR calculates similarity based on the size of the overlapping regions of the MBRs of spatial datasets. Among the proposed spatial similarity measures, EMD has a high time complexity, significantly impacting the results. MBR cannot effectively measure situations where two datasets are nearly identical in spatial range and have little similarity in the distribution of their locations; in other words, it fails to consider the spatial distribution of the data.

[0004] Chinese patent application ZL 2023115515813 discloses an efficient spatial dataset query method based on density grayscale images. However, this method converts the spatial dataset into a grayscale image representation and uses a similarity calculation method in the image field. This method has high storage cost and computational overhead and cannot adapt to large-scale spatiotemporal dataset query scenarios. Summary of the Invention

[0005] In order to address the shortcomings of the existing technology, the present invention provides an efficient spatial dataset sample query method based on density distribution. This query method converts the spatial dataset into a density distribution representation, and uses the distribution ratio of location points between spatiotemporal datasets in overlapping grids as the basis for measuring similarity, thereby improving query efficiency.

[0006] In order to achieve the above objectives, the present invention is implemented through the following technical solutions:

[0007] The present invention is an efficient spatial data set query method based on density grayscale map, which includes two stages: the first stage: data processing stage; the second stage: query execution stage.

[0008] The data preprocessing stage includes the following steps:

[0009] 1a) According to the set parameter threshold θ, the global geographic space is divided into θ*θ grids of equal size. Each grid has a unique ID. The set of grid IDs is denoted as G = {g1, g2, ..., g m}, where m = θ*θ.

[0010] 1b) For each geospatial dataset D in the spatial database D i , according to D i The density and distribution of each location point in each grid in G are calculated. i The density distribution element set T i The set of density distribution element sets corresponding to each data set in D is recorded as T = {T1, T2, ..., T n}.

[0011] 1c) For each geospatial dataset D in D i , the calculation can cover D i The minimum covering rectangle R of each point in i =(P LB ,P RU ), R i .P LB and R i .P RU Rectangle R i The position coordinates of the lower left corner and upper right corner of the dataset. The set of minimum covering rectangles corresponding to each dataset in D is denoted as R = {R1, R2, ..., R n}.

[0012] The spatial dataset sample query phase includes the following steps:

[0013] 2a) For the query sample dataset D e , according to D e The density and distribution of each location point in each grid in G are calculated. e The density distribution element set T e , the calculation method is similar to step 1b).

[0014] 2b) Calculate the coverage of D e The minimum covering rectangle R of each point in e =(P LB ,P RU ), R e .P LB and R e .P RU Rectangle R e The position coordinates of the lower left corner and the upper right corner are calculated in a similar way as in step 1c).

[0015] 2c) For each dataset D in D i , calculate D i The minimum covering rectangle R i With D e The minimum covering rectangle R e The relationship between the spatial overlap of and the given candidate filtering threshold τ (0 < τ < 1) determines D i Is it a candidate query result? If so, D i Add candidate query result set D'.

[0016] 2d) Using each dataset D in D' i The corresponding density distribution element set T i and D e The corresponding density distribution element set T e Calculate the candidate dataset D i With D e , and then obtain the k data sets with the highest similarity. These k data sets are the query results.

[0017] Furthermore, in step (1b), according to D i The density and distribution of each location point in each grid in G generate the data set D i The density distribution element set T i , specifically including the following steps:

[0018] 1b1) Scan D i All the location points in the grid are recorded, and the ID of the grid where each location point is located is counted, and the number of location points contained in these grids is counted. i The set of grid IDs of the position points in the grid is recorded as G i = The number of location points contained in these grids corresponds to where N i,j Indicates D i In the ID g i,j The number of location points in the grid.

[0019] 1b2) Calculation Each grid g i,j Contains the density value ρ of the location point i,j ,

[0020]

[0021] Among them, |D i | represents the dataset D i The number of locations contained in .

[0022] 1b3) Generate D i The density distribution element set Ti ={(g i,j ,ρ i,j )|g i,j ∈G i}, where (g i,j ,ρ i,j ) is a grid ID-grid density pair.

[0023] The calculation in step (1c) can cover the data set D i The minimum rectangle R of all points in i =(P LB ,P RU ), specifically including the following steps:

[0024] 1c1) Let’s assume that the dataset D i Contains t position points, denoted as D i ={l1,l2,…,l t}, l j .x and l j .y represents the location point l j The horizontal and vertical coordinates of R i The lower left corner P LB and the upper right corner P RU The position coordinates are calculated as follows:

[0025] R i .P LB .x=min{l j .x|l j ∈D i}

[0026] R i .P LB .y=min{l j .y|l j ∈D i}

[0027] R i .P RU .x=max{l j .x|l j ∈D i}

[0028] R i .P RU .y=max{l j .y|l j ∈D i}

[0029] Among them, (R i .P LB .x,R i .P LB.y) indicates R i The lower left corner P LB The position coordinates, (R i .P RU .x,R i .P RU .y) indicates R i The upper right corner P RU The position coordinates of , min{*} and max{*} respectively represent the minimum and maximum values ​​of the collection.

[0030] Furthermore, in step (2c), according to the data set D i The minimum covering rectangle and query example D e The spatial overlap of the minimum covering rectangle and the relationship with the given candidate filtering threshold τ determine D i Whether it is a candidate query result, specifically including the following steps:

[0031] 2c1) According to D i and D e The minimum covering rectangle R i =(P LB ,P RU ) and R e =(P LB ,P RU ), calculate R i and R e The spatial overlap (R i ,R e ), the calculation formula is as follows:

[0032]

[0033] Where area(X) represents the area of ​​rectangle X, R i ∩R e is a rectangle R i and rectangle R e The overlapping area, if any, is also a rectangle.

[0034] 2c2) If overlap(R i ,R e )>τ, then D i The candidate query results.

[0035] In step (2d), we use the dataset D i The corresponding density distribution element set T i and D e The corresponding density distribution element set T e Calculate the candidate dataset D i With D eThe calculation of the similarity specifically includes the following steps:

[0036] 2d1) Let T i ={(g i,a ,ρ i,a )|g i,a ∈G i}、T e ={(g e,b ,ρ e,b )|g e,b ∈G e}, where G i and G e Represents the dataset D i and D e The set of IDs of the grids where each location point is located, then D i and D e Similarity(D i ,D e ) is calculated as follows:

[0037]

[0038] The beneficial effects of the present invention are:

[0039] The present invention can effectively reduce the amount of calculation and complexity of the query by calculating the similarity between spatiotemporal data sets using density distribution.

[0040] The present invention improves upon existing methods such as MBR and EMD by converting spatiotemporal datasets into density distribution representations. The similarity between spatiotemporal datasets is measured based on the distribution ratio of location points within overlapping grids, including the distance between location points and the similarity of location point distribution characteristics between datasets. This allows for rapid data location and filtering, providing more accurate and comprehensive query results.

[0041] The present invention can adjust the granularity and efficiency of the balanced query through the parameter θ, and has flexibility and scalability. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is an example of a flow chart of the data processing and query phase of the present invention.

[0043] Figure 2 This is an example of the present invention generating a density distribution based on a data set.

[0044] Figure 3 This is an example of generating a minimum bounding rectangle based on a data set in the present invention. DETAILED DESCRIPTION

[0045] The following diagrams illustrate embodiments of the present invention. For clarity, many practical details are included in the following description. However, it should be understood that these practical details are not intended to limit the present invention. In other words, in some embodiments of the present invention, these practical details are not essential.

[0046] For the convenience of description, the relevant symbols are defined as follows: Spatial database D = {D1, D2, ···, D n}, contains n data sets, D i ∈D is a spatial dataset containing several two-dimensional location points, D i ={(x1,y1),(x2,y2),…,(x mi ,y mi )}, the density distribution element set T corresponding to each data set in D = {T1, T2, ..., T n}, the minimum covering rectangle set R={R1,R2,…,R n}; Query sample dataset D e And the corresponding density distribution set T e and the minimum covering rectangle R e .

[0047] The present invention is an efficient spatial dataset query method based on density grayscale map. The query method includes two stages: the first stage: data processing stage; the second stage: query execution stage. Specifically, it includes the following process:

[0048] The data preprocessing stage includes the following steps:

[0049] 1a) According to the set parameter threshold θ, the global geographic space is divided into θ*θ grids of equal size. Each grid has a unique ID. The set of grid IDs is denoted as G = {g1, g2, ..., g m}, where m = θ*θ.

[0050] 1b) For each geospatial dataset D in the spatial database D i , according to D i The density and distribution of each location point in each grid in G are calculated. i The density distribution element set T i The set of density distribution element sets corresponding to each data set in D is recorded as T = {T1, T2, ..., T n}.

[0051] In this step, according to D i The density and distribution of each location point in each grid in G generate the data set D i The density distribution element set T i, specifically including the following steps:

[0052] 1b1) Scan D i All the location points in the grid are recorded, and the ID of the grid where each location point is located is counted, and the number of location points contained in these grids is counted. i The set of grid IDs of the position points in the grid is recorded as G i = The number of location points contained in these grids corresponds to where N i,j Indicates D i In the ID g i,j The number of location points in the grid.

[0053] 1b2) Calculation Each grid g i,j Contains the density value ρ of the location point i,j ,

[0054]

[0055] Among them, |D i | represents the dataset D i The number of locations contained in .

[0056] 1b3) Generate D i The density distribution element set T i ={(g i,j ,ρ i,j )|g i,j ∈G i}, where (g i,j ,ρ i,j ) is a grid ID-grid density tuple, such as Figure 2 Example shown.

[0057] 1c) For each geospatial dataset D in D i , the calculation can cover D i The minimum covering rectangle R of each point in i =(P LB ,P RU ), R i .P LB and R i .P RU Rectangle R i The position coordinates of the lower left corner and upper right corner of the dataset. The set of minimum covering rectangles corresponding to each dataset in D is denoted as R = {R1, R2, ..., R n}.

[0058] In this step, the calculation can cover the data set D i The minimum rectangle R of all points ini =(P LB ,P RU ), specifically including the following steps:

[0059] 1c1) Let’s assume that the dataset D i Contains t position points, denoted as D i ={l1,l2,…,l t}, l j .x and l j .y represents the location point l j The horizontal and vertical coordinates of R i The lower left corner P LB and the upper right corner P RU The position coordinates are calculated as follows:

[0060] R i .P LB .x=min{l j .x|l j ∈D i}

[0061] R i .P LB .y=min{l j .y|l j ∈D i}

[0062] R i .P RU .x=max{l j .x|l j ∈D i}

[0063] R i .P RU .y=max{l j .y|l j ∈D i}

[0064] Among them, (R i .P LB .x,R i .P LB .y) indicates R i The lower left corner P LB The position coordinates, (R i .P RU .x,R i .P RU .y) indicates R i The upper right corner P RU The position coordinates of the set, min{*} and max{*} respectively represent the minimum and maximum values ​​of the set, such as Figure 3Example shown.

[0065] The spatial dataset sample query phase includes the following steps:

[0066] 2a) For the query sample dataset D e , according to D e The density and distribution of each location point in each grid in G are calculated. e The density distribution element set T e , the calculation method is similar to step 1b).

[0067] 2b) Calculate the coverage of D e The minimum covering rectangle R of each point in e =(P LB ,P RU ), R e .P LB and R e .P RU Rectangle R e The position coordinates of the lower left corner and the upper right corner are calculated in a similar way as in step 1c).

[0068] 2c) For each dataset D in D i , calculate D i The minimum covering rectangle R i With D e The minimum covering rectangle R e The relationship between the spatial overlap of D and the given candidate filtering threshold τ, 0<τ<1, determines i Is it a candidate query result? If so, D i Add candidate query result set D'.

[0069] In this step, according to the dataset D i The minimum covering rectangle and query example D e The spatial overlap of the minimum covering rectangle and the relationship with the given candidate filtering threshold τ determine D i Whether it is a candidate query result, specifically including the following steps:

[0070] 2c1) According to D i and D e The minimum covering rectangle R i =(P LB ,P RU ) and R e =(P LB ,P RU ), calculate R i and R e The spatial overlap (R i ,R e), the calculation formula is as follows:

[0071]

[0072] Where area(X) represents the area of ​​rectangle X, R i ∩R e is a rectangle R i and rectangle R e The overlapping area, if any, is also a rectangle.

[0073] 2c2) If overlap(R i ,R e )>τ, then D i The candidate query results.

[0074] 2d) Using each dataset D in D' i The corresponding density distribution element set T i and D e The corresponding density distribution element set T e Calculate the candidate dataset D i With D e , and then obtain the k data sets with the highest similarity. These k data sets are the query results.

[0075] In this step, each dataset D in D' is used i The corresponding density distribution element set T i and D e The corresponding density distribution element set T e Calculate the candidate dataset D i With D e The calculation of the similarity specifically includes the following steps:

[0076] 2d1) Let T i ={(g i,a ,ρ i,a )|g i,a ∈G i}、T e ={(g e,b ,ρ e,b )|g e,b ∈G e}, where G i and G e Represents the dataset D i and D e The set of IDs of the grids where each location point is located, then D i and D e Similarity(D i ,D e ) is calculated as follows:

[0077]

[0078] Figure 2 The geospatial dataset D is presented in i 、D j And the sample dataset D e The larger the overlapping area, the higher the proportion of geospatial datasets distributed in the same area. Therefore, the spatial coincidence between datasets can be calculated based on the overlapping area to filter out geospatial datasets that are obviously dissimilar to the sample dataset.

[0079] Figure 3 The geospatial dataset D is presented in i And the sample dataset D e Transformed into density distribution element set T i and T e In the process of spatial overlap, the distribution characteristics of the internal location points of the dataset with high spatial overlap may not be similar, such as Figure 2 Middle D i and D e The spatial overlap is high but obvious D j and D e More similar, so the element set T is distributed by density i and T e Measuring the similarity of data sets can fully consider the distribution characteristics of location points in the data set and avoid this problem.

[0080] The foregoing is merely an embodiment of the present invention and is not intended to limit the present invention. It will be apparent to those skilled in the art that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention are intended to be included within the scope of the claims of the present invention.

Claims

1. A density distribution-based geospatial dataset sample query method, characterized by: A geospatial dataset sample query is to obtain and query sample geospatial dataset D in geospatial dataset D. e The query operation of the most similar k spatial datasets, the geographic spatial dataset D = {D1, D2, ..., D n }, where D i It is a geospatial dataset containing several two-dimensional spatial location points. Specifically, it includes the data preprocessing stage and the dataset sample query stage: The data preprocessing stage specifically includes the following steps: Step 11: According to the set parameter threshold θ, the geographic spatial dataset D is divided into θ*θ grids of equal size. The IDs of each grid constitute a set G = {g1, g2, ..., g m }, where m = θ * θ; Step 12: For each two-dimensional spatial location point in the geospatial dataset D, i , according to the geospatial dataset D of each two-dimensional spatial location point i The density and distribution of the two-dimensional spatial location points contained in each grid in the ID set G of each grid are used to calculate the geographic spatial dataset D of each two-dimensional spatial location point. i The density distribution element set T i , a geospatial dataset D of two-dimensional spatial location points in a geospatial dataset D i The corresponding density distribution element set T={T1,T2,…,T n }; Step 13: For each two-dimensional spatial location point in the geospatial dataset D, i , calculate the geospatial dataset D that can cover the two-dimensional spatial location points i The minimum covering rectangle R of each two-dimensional space position point i =(P LB ,P RU ), R i .P LB and R i .P RU Rectangle R i The position coordinates of the lower left corner and the upper right corner of the geospatial dataset D, the geospatial dataset D of the two-dimensional spatial position point i The corresponding minimum covering rectangle set R={R1,R2,…,R n }; The dataset sample query phase specifically includes the following steps: Step 21: Query the sample geospatial dataset D e , according to the query sample geospatial dataset D e The density and distribution of each two-dimensional spatial location point in each grid in the geographic spatial dataset G are calculated to query the sample geographic spatial dataset D e The density distribution element set T e ; Step 22: Calculate the geospatial dataset D that can cover the query sample e The minimum covering rectangle R of the two-dimensional space position e =(P LB ,P RU ), R e .P LB and R e .P RU Rectangle R e The position coordinates of the lower left corner and upper right corner; Step 23: For each two-dimensional spatial location point in the geospatial dataset D, i , calculate the geospatial dataset D of the two-dimensional spatial location points i The minimum covering rectangle R i and query the sample geospatial dataset D e The minimum covering rectangle R e The relationship between the spatial coincidence of the given candidate filter threshold τ, 0<τ<1, determines the geographic spatial dataset D of the two-dimensional spatial location point i Is it a candidate query result? If so, the geographic spatial dataset D of the two-dimensional spatial location point is i Add candidate query result set D'; Step 24: Use the geospatial dataset D of each two-dimensional spatial location point in the candidate query result set D' i The corresponding density distribution element set T i and query the sample geospatial dataset D e The corresponding density distribution element set T e Calculate D in the candidate dataset i and query the sample geospatial dataset D e The k data sets with the highest similarity are obtained, and the k data sets are the query results.

2. The density distribution-based geospatial dataset sample query method according to claim 1, characterized in that: In step 12, the geographic spatial dataset D of each two-dimensional spatial location point is used. i The density and distribution of each two-dimensional spatial location point in each grid in the ID set G of each grid are calculated, and the geographic spatial dataset D of each two-dimensional spatial location point is calculated. i The density distribution element set T i The process includes the following steps: Step 121: Scan the geospatial dataset D of the two-dimensional spatial location point i All the two-dimensional spatial location points in the grid, record the ID of the grid where each two-dimensional spatial location point is located, and count the number of location points contained in the grid: Let the geographic spatial dataset D containing two-dimensional spatial location points be i The ID set of the grid of two-dimensional space position points in g i,j is a grid, and the number of two-dimensional space position points contained in the grid corresponds to Step 122: Calculate ID set Each grid g i,j Contains the density value ρ of the location point i,j : Among them, |D i |A geospatial dataset D representing two-dimensional spatial location points i The number of location points contained in N i,j A geospatial dataset D representing two-dimensional spatial locations i In the ID g i,j The number of location points in the grid; Step 123: Generate a geospatial dataset D of two-dimensional spatial location points i The density distribution element set T i ={(g i,j ,ρ i,j )|g i,j ∈G i }, where (g i,j ,ρ i,j ) is a grid ID-grid density pair.

3. The density distribution-based geospatial dataset sample query method according to claim 1, characterized in that: In step 13, the geospatial dataset D that can cover the two-dimensional spatial location points is calculated. i The minimum covering rectangle R of each two-dimensional space position point i =(P LB ,P RU ) is as follows: Let the geographic spatial dataset D of the two-dimensional spatial location points be i Contains t two-dimensional space position points, denoted as D i ={l1,l2,…,l t }, l j .x and l j .y represent the two-dimensional space position point l j The horizontal and vertical coordinates of the minimum covering rectangle R i The lower left corner P LB and the upper right corner P RU The position coordinates are calculated as follows: R i .P LB .x=min{l j .x|l j ∈D i } R i .P LB .y=min{l j .y|l j ∈D i } R i .P RU .x=max{l j .x|l j ∈D i } R i .P RU .y=max{l j .y|l j ∈D i } Among them, (R i .P LB .x,R i .P LB .y) represents the minimum covering rectangle R i The lower left corner P LB The position coordinates, (R i .P RU .x,R i .P RU .y) represents the minimum covering rectangle R i The upper right corner P RU The position coordinates of , min{*} and max{*} respectively represent the minimum and maximum values ​​of the collection.

4. The density distribution-based geospatial dataset sample query method according to claim 1, characterized in that: In step 23, the geographic space dataset D of the two-dimensional space location point is used. i The minimum covering rectangle and query sample geospatial dataset D e The spatial coincidence of the minimum covering rectangle and the relationship with the given candidate filtering threshold τ determine the geospatial dataset D of the two-dimensional spatial location point i The process of determining whether a result is a candidate is as follows: Step 231: According to the minimum covering rectangle R i =(P LB ,P RU ) and R e =(P LB ,P RU ), calculate R i and R e The spatial overlap (R i ,R e ), the calculation formula is as follows: Among them, area(X) represents the area of ​​rectangle X, R i ∩R e is a rectangle R i and rectangle R e The overlapping area, if any, is also a rectangle; Step 232: If the spatial overlap (R i ,R e )>τ, then D i Add candidate query results.

5. The density distribution-based geospatial dataset sample query method according to claim 1, characterized in that: In step 24, the dataset D is used i The corresponding density distribution element set T i and D e The corresponding density distribution element set T e Calculate the geospatial dataset D of the two-dimensional spatial location points in the candidate query result set D' i With D e The similarity calculation process is as follows: Let T i = {(g i,a , ρ i,a ) | g i,a ∈ G i}, T e = {(g e,b , ρ e,b ) | g e,b ∈ G e} Among them G i and G e The geospatial dataset D represents the two-dimensional spatial location points. i and query the sample geospatial dataset D e The set of IDs of the grids where each two-dimensional space position point is located is D i and D e Similarity(D i ,D e ) is calculated as follows:

6. The density distribution-based geospatial dataset sample query method according to claim 1, characterized in that: In step 21, the query sample geospatial dataset D e , according to the query sample geospatial dataset D e The density and distribution of each two-dimensional spatial location point in each grid in the geographic spatial dataset G are calculated to query the sample geographic spatial dataset D e The density distribution element set T e The specific process: Step 211: Scan the geographic spatial dataset D of the two-dimensional spatial location point i All the two-dimensional spatial location points in the grid, record the ID of the grid where each two-dimensional spatial location point is located, and count the number of location points contained in the grid: Let the geographic spatial dataset D containing two-dimensional spatial location points be i The ID set of the grid of two-dimensional space position points in g i,j is a grid, and the number of two-dimensional space position points contained in the grid corresponds to Step 212: Calculate ID set Each grid g i,j Contains the density value ρ of the location point i,j : Among them, |D i |A geospatial dataset D representing two-dimensional spatial location points i The number of location points contained in N i,j A geospatial dataset D representing two-dimensional spatial locations i In the ID g i,j The number of location points in the grid; Step 213: Generate a geospatial dataset D of two-dimensional spatial location points i The density distribution element set T i ={(g i,j ,ρ i,j )|g i,j ∈G i }, where (g i,j ,ρ i,j ) is a grid ID-grid density pair.

7. The density distribution-based geospatial dataset sample query method according to claim 1, characterized in that: The step 22 calculates the geospatial dataset D that can cover the query sample. e The minimum covering rectangle R of the two-dimensional space position e =(P LB ,P RU ), R e .P LB and R e .P RU Rectangle R e The position coordinates of the lower left corner and upper right corner are calculated as follows: R i .P LB .x=min{l j .x|l j ∈D i } R i .P LB .y=min{l j .y|l j ∈D i } R i .P RU .x=max{l j .x|l j ∈D i } R i .P RU .y=max{l j .y|l j ∈D i } Among them, (R i .P LB .x,R i .P LB .y) represents the minimum covering rectangle R i The lower left corner P LB The position coordinates, (R i .P RU .x,R i .P RU .y) represents the minimum covering rectangle R i The upper right corner P RU The position coordinates of , min{*} and max{*} respectively represent the minimum and maximum values ​​of the collection.

Images