Spatial data set similarity calculation method based on convolution and pooling

Abstract

The invention belongs to the field of spatial data set sample search, and discloses a spatial data set similarity calculation method based on convolution and pooling, which comprises the following steps: (1) a data set feature map generation stage: establishing a grid coordinate system for a spatial data set warehouse, and counting the number of two-dimensional coordinates to generate a feature map; (2) a convolution and pooling stage: smoothing the feature map by using a Gaussian smoothing convolution kernel, extracting structural features of the feature map, and selecting a maximum value through a pooling window to generate a pooling feature map; and (3) a similarity calculation stage: flattening the final feature map into a feature vector, carrying out normalization processing, and calculating a relative cosine similarity to obtain the similarity between the data sets. According to the method, the similarity problem of the spatial data sets with the complex structures is solved, the main features of the data sets are effectively extracted, the calculation scale is reduced, accuracy and efficiency are considered, the large-scale spatial data sets with the complex structures are efficiently processed, and the similarity between the data sets is accurately reflected.

Description

Technical Field

[0001] The present invention belongs to the field of spatial data set sample search, and specifically relates to a spatial data set similarity calculation method based on convolution and pooling. Background Art

[0002] Today, as the urbanization process is accelerating, cities, as the main carrier of human activities, have received extensive attention for their development and planning. With the popularization of geographic information systems (GIS), the Internet of Things (IoT) and big data technologies, the ability to collect, process and analyze urban data has been significantly improved. These data not only cover geographic location information, but also incorporate complex characteristics in multiple dimensions such as economy, society and culture, providing a rich information foundation for urban planning, policy making and smart city construction. At the same time, as the core of the GIS system, spatial datasets, which contain geographic location information and complex spatial relationships, play a key role in many fields. Measuring the similarity between urban spatial datasets plays an important role in guiding urban construction planning, helps optimize resource allocation, improve urban management efficiency, and promote the construction and development of smart cities.

[0003] However, existing similarity measurement methods have limitations to varying degrees when facing complex and diverse spatial data sets. Traditional methods such as Euclidean distance and Manhattan distance are effective in processing continuous data, but are incapable of processing complex spatial structures; EMD and Hausdorff distance are widely applicable, but are restricted by high time complexity and sensitivity to outliers; the minimum bounding rectangle (MBR) method focuses on overlapping areas but ignores subtle differences in spatial distribution. Therefore, it is urgent to explore a new method that can efficiently process large-scale spatial data and accurately reflect the similarity between data sets. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a spatial dataset similarity calculation method based on convolution and pooling. The method integrates the multi-dimensional similarity evaluation between cities with the complex geographical location and spatial relationship processing capabilities of spatial datasets. Its innovation will inject new impetus into spatial dataset query technology and promote in-depth development and research and application in the fields of urban planning and smart city construction.

[0005] In order to achieve the above object, the present invention is achieved through the following technical solutions:

[0006] The present invention is a spatial data set similarity calculation method based on convolution and pooling, which includes three stages: the first stage: data set feature map generation stage; the second stage: convolution and pooling stage; the third stage: similarity calculation stage.

[0007] Step 1: Dataset feature map generation stage: Establish a grid coordinate system R for the spatial dataset warehouse D i , count the number of two-dimensional coordinates to generate the original feature map A of the two-dimensional coordinates i ;

[0008] Step 2, convolution and pooling stage: Use Gaussian smoothing convolution kernel to smooth the two-dimensional coordinate original feature map A i , extract the original feature map A of the two-dimensional coordinates i Structural features, and then select the maximum value through the pooling window to generate the sample pooling feature map C i ;

[0009] Step 3: Similarity calculation phase: Pooling feature map C i And according to the sample data set D e The generated sample pooling feature map C is flattened into a feature vector and normalized, and the relative cosine similarity is calculated to obtain the similarity between the data sets.

[0010] A further improvement of the present invention is that step 2 specifically comprises the following steps:

[0011] Step 2.1: Select a Gaussian smoothing convolution kernel with a window size of 3×3 and a standard deviation of α to smooth the original feature map A of the two-dimensional coordinates in subsequent feature extraction. i , retain the structural characteristics of spatial datasets;

[0012] Step 2.2: For the original feature map set A = {A1, A2, ..., A n}, each of the two-dimensional coordinate original feature maps A i ∈A for "zero padding" (Zero Padding), and then place the Gaussian smoothed convolution kernel in the two-dimensional coordinate original feature map A after "zero padding" i The upper left corner, A i The first 3×3 area after “zero padding”, then slide the Gaussian smoothing convolution kernel and perform local weighted summation until the original feature map A of the two-dimensional coordinates is traversed i Generate convolution feature map B i , and obtain the convolution feature map set B corresponding to each spatial dataset in the spatial dataset warehouse D = {B1, B2, …, B n};

[0013] Step 2.3: For the convolutional feature map set B = {B1, B2, ..., B n Each convolution feature map B in i , select one 2 p ×2 p The pooling window is set to 2. p, where 1≤p≤θ-1 places the initial position of the pooling window on the convolutional feature map B i The upper left corner, i.e. B i The first 2 p ×2 p Region, select the maximum value in the region and put it into the corresponding position c of the feature map after pooling i,j,k , where 1≤j≤2 θ-p ,1≤k≤2 θ-p , slide the pooling window until the convolution feature map B is traversed i , get the corresponding pooling feature map C i Finally, we get the pooled feature map set C corresponding to each spatial dataset in the spatial dataset warehouse D = {C1, C2, …, C n}, the formula is as follows:

[0014]

[0015] Among them, 1≤j≤2 θ-p ,1≤k≤2 θ-p , c i,j,k is the pooled feature map C i The value of the jth row and kth column of b i,j,k is the convolution feature map B i The j-th row and k-th column value of the pooling window covers the convolution feature map B in the upper left corner. i The j0th row and k0th column of .

[0016] A further improvement of the present invention is that in step 2.1, a Gaussian smoothing convolution kernel with a window of 3×3 and a standard deviation of α is selected, which specifically includes the following steps:

[0017] Step 2.1.1, calculate each value of the convolution kernel:

[0018]

[0019] Where (x, y) is the coordinate relative to the center of the convolution kernel, σ is the standard deviation, x, y∈{0,±1}, σ=α, and all possible values ​​of (x, y) (-1,-1), (-1,0), (-1,1), (0,-1), (0,0), (0,1), (1,-1), (1,0), (1,1) are calculated to obtain the convolution kernel Kernel, which is expressed as follows:

[0020] Kernel i,j =G(i-2,j-2);

[0021] Step 2.1.2, calculate the sum of all elements of the convolution kernel Sum, and then divide each element by the sum to normalize it to get the normalized convolution kernel NK, which is expressed as follows:

[0022]

[0023] Among them, Kernel i,j is the value of the i-th row and j-th column of the convolution kernel Kernel, 1≤i,j≤3.

[0024] A further improvement of the present invention is that: in step 2.2, the two-dimensional coordinate original feature map A i Perform "zero padding" and sliding convolution kernel and local weighted summation, specifically including the following steps:

[0025] Step 2.2.1: In the two-dimensional coordinate original feature map A i Add a row of 0s above the first row and below the last row, and add a column of 0s to the left of the first column and to the right of the last column, that is, Padding = 1, so that the two-dimensional coordinate original feature map A i The size of 2 θ ×2 θ becomes (2 θ +2)×(2 θ +2) to ensure that the size of the feature map after convolution remains unchanged;

[0026] Step 2.2.2: Add the Gaussian smoothing convolution kernel to the original feature map A in the two-dimensional coordinates i Slide up with a step size of 1. Each position corresponds to a local area, namely the receptive field. In the local area covered by the Gaussian smooth convolution kernel, the Gaussian smooth convolution kernel is combined with the two-dimensional coordinate original feature map A. i The corresponding positions are multiplied element by element, and then all the products are summed to obtain an element value b of the convolution feature map. i,j,k , where 1≤j≤2 θ ,1≤k≤2 θ , this process is repeated until the convolution kernel traverses A i , generate the corresponding convolution feature map B i , and obtain the convolution feature map set B corresponding to each feature map in the two-dimensional coordinate original feature map set A = {B1, B2, ..., B n}, expressed as:

[0027]

[0028] Among them, NK u,v is the value of the u-th row and v-th column of the normalized convolution kernel NK, 1≤u,v≤3.

[0029] A further improvement of the present invention is that the step 3 specifically includes the following steps:

[0030] Step 3.1: For a given query sample dataset D e , according to the query sample data set D e The minimum and maximum values ​​of the x-axis and y-axis of all two-dimensional coordinates in the coordinate system R are established e , count the number of two-dimensional coordinates to generate the sample two-dimensional coordinate original feature map A e , and then use the Gaussian smoothing convolution kernel to smooth the original feature map A of the sample two-dimensional coordinates e Get the sample convolution feature map B e , and then select the maximum value through the pooling window to generate the sample pooling feature map C e ;

[0031] Step 3.2: Pooling feature map C for the sample e And the pooled feature map set C = {C1, C2, ..., C n Each feature map C in i ∈C, flattened into feature vectors v e and v i :

[0032]

[0033] Step 3.3, normalize the eigenvector and calculate v e and v i The similarity of is recorded as the corresponding two-dimensional coordinate space dataset D i And query sample dataset D e The similarity is as follows:

[0034]

[0035] Among them, θ is the partition parameter, v e,t is the vector v e The tth element of i,t is the vector v i The tth element of .

[0036] A further improvement of the present invention is that the step 1 specifically comprises the following steps:

[0037] Step 1.1: Given a spatial dataset warehouse D = {D1, D2, …, D n}, D i ∈D is a spatial data set of two-dimensional coordinates, where D i ={(x1,y1),(x2,y2),…,(x m ,y m )}, find D i The minimum and maximum values ​​of the x-axis and y-axis of all two-dimensional coordinates in are denoted by xi min ,min ,xi max , max ;

[0038] Step 1.2, with (xi min ,yi min ) and (xi max ,yi max ) is used as the diagonal to establish a rectangular area, and then the width of the rectangle is extended to both ends at equal distances to expand the rectangular area into a square, and then the square area is divided into 2 according to the division parameter θ θ ×2 θ grids of equal size, and establish a grid coordinate system R i ;

[0039] Step 1.3: For the spatial dataset warehouse D = {D1, D2, ..., D n Each spatial dataset D in i , statistical spatial dataset D i In the grid coordinate system R i The grid g at row j and column k in j,k The number of two-dimensional coordinate points within is denoted as a i,j,k , where 1≤j≤2 θ ,1≤k≤2 θ , get the spatial dataset D i The two-dimensional coordinate feature map A i , traverse the spatial dataset warehouse D, and obtain the two-dimensional coordinate original feature map set A={A1,A2,…,A n}.

[0040] The beneficial effects of the present invention are:

[0041] The present invention provides an efficient and adaptable method for calculating the similarity of spatial datasets by cleverly combining convolution and pooling with spatial dataset query.

[0042] The present invention not only effectively overcomes the challenge of traditional technology in coping with the diversity and complexity of spatial data through the carefully designed convolution kernel selection and pooling process, but also reduces the calculation scale and improves the search efficiency through the data set feature map generation. At the same time, in the similarity calculation stage, the method of normalizing feature vectors is adopted to ensure the balance and comparability between the features of different spatial data sets, further improving the practicality and accuracy of the method.

[0043] The present invention achieves efficient processing of spatial data sets with complex structures through three carefully designed stages. It not only reduces the computational scale, but also improves the search efficiency, meeting the needs of large-scale spatial data processing. Compared with traditional methods, the present invention can better reflect the complexity and diversity of spatial data sets, and provides strong technical support for urban planning, smart city construction and other fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a schematic diagram of the process of the present invention.

[0045] Figure 2 It is a schematic diagram of the process of generating a two-dimensional coordinate original feature map in the data set feature map generation stage of the present invention.

[0046] Figure 3 It is a schematic diagram of the process of generating a pooling feature map in the convolution and pooling stages of the present invention. DETAILED DESCRIPTION

[0047] The following will disclose the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present application. For the purpose of clear description, many practical details will be described together in the following description. However, it should be understood that these practical details should not be used to limit the present invention. In other words, in some embodiments of the present invention, these practical details are not necessary.

[0048] For the convenience of description, the relevant symbols are defined as follows: Spatial dataset warehouse D = {D1, D2, ..., D n}, contains n spatial datasets, D i ∈D is a spatial data set of two-dimensional coordinates, where D i ={(x1,y1),(x2,y2),…,(x m ,y m )}; the original feature map set of two-dimensional coordinates corresponding to each spatial data set in D is A = {A1, A2, ..., A n}; The convolution feature map set B corresponding to each spatial data set in D = {B1, B2, ..., B n}; The pooled feature map set C corresponding to each spatial data set in D = {C1, C2, ..., C n}; Query sample data set D e And the sample pooling feature map C e .

[0049] like Figure 1 As shown, the present invention is a spatial data set similarity calculation method based on convolution and pooling, and the similarity calculation method includes three stages: the first stage: the data set feature map generation stage; the second stage: the convolution and pooling stage; the third stage: the similarity calculation stage.

[0050] The dataset feature map generation stage includes the following steps:

[0051] (1a) Given a spatial dataset warehouse D = {D1, D2, …, D n}, D i ∈D is a spatial data set of two-dimensional coordinates, where D i ={(x1,y1),(x2,y2),…,(x m ,y m )}, find D i The minimum and maximum values ​​of the x-axis and y-axis of all two-dimensional coordinates in are denoted by xi min , min ,xi max , max .

[0052] (1b) Figure 2 As shown, (xi min ,yi min ) and (xi max ,yi max ) is used as the diagonal to establish a rectangular area, and then the width of the rectangle is extended to both ends at equal distances to expand the rectangular area into a square, and then the square area is divided into 2 according to the division parameter θ θ ×2 θ grids of equal size, and establish a grid coordinate system R i .

[0053] (1c) For a spatial dataset warehouse D = {D1, D2, …, D n Each spatial dataset D in i , count the grid g in the jth row and kth column in the coordinate system R j,k The number of two-dimensional coordinate points within is denoted as a i,j,k , where 1≤j≤2 θ ,1≤k≤2 θ , get D i The two-dimensional coordinate original feature map A i Traverse the spatial data warehouse D and obtain the original feature map set A of two-dimensional coordinates = {A1, A2, ..., A n}.

[0054] The convolution and pooling stage includes the following steps:

[0055] (2a) A Gaussian smoothing convolution kernel with a window size of 3×3 and a standard deviation of α is selected to smooth the feature map in subsequent feature extraction and retain the larger structural features of the spatial data set.

[0056] In this step, a Gaussian smoothing convolution kernel with a window of 3×3 and a standard deviation of α is selected, which specifically includes the following steps:

[0057] (2a1) Calculate each value of the convolution kernel according to the formula, which is as follows:

[0058]

[0059] Where (x, y) is the coordinate relative to the center of the convolution kernel, and σ is the standard deviation, so x, y∈{0,±1}. Take σ=α, and calculate all possible value combinations of (x, y) (-1,-1), (-1,0), (-1,1), (0,-1), (0,0), (0,1), (1,-1), (1,0), (1,1) to obtain the convolution kernel Kernel. The process can be expressed by the following formula:

[0060] Kernel i,j =G(i-2,j-2)

[0061] Among them, Kernel i,j is the value of the i-th row and j-th column of the convolution kernel Kernel, 1≤i,j≤3.

[0062] (2a2) Calculate the sum of all elements of the convolution kernel Sum, and then divide each element by the sum to normalize it to ensure that the total amount of features remains basically unchanged, and obtain the normalized convolution kernel NK. This process can be expressed by the following formula:

[0063]

[0064] (2b) Figure 3 As shown, for the two-dimensional coordinate original feature map set A = {A1, A2, ..., A n}, each of the two-dimensional coordinate original feature maps A i ∈A for "zero padding" and then place the convolution kernel in the original feature map A of the two-dimensional coordinates i The upper left corner, A i The first 3×3 area after “zero padding”, then slide the convolution kernel and perform local weighted summation until all A is traversed i Get the convolution feature map B i Finally, we get the convolution feature map set B corresponding to each spatial data set in D = {B1, B2, ..., B n}.

[0065] In this step, the two-dimensional coordinate original feature map A i Perform "zero padding" and sliding convolution kernel and local weighted summation, specifically including the following steps:

[0066] (2b1) in A i Add a row of 0s above the first row and below the last row, and add a column of 0s to the left of the first column and to the right of the last column, that is, Padding = 1, so that A i The size of 2 θ ×2 θ becomes (2 θ +2)×(2 θ +2) to ensure that the size of the feature map after convolution remains unchanged.

[0067] (2b2) ​​The convolution kernel is placed in A i Slide up, the step size is 1, each position corresponds to a local area, namely the receptive field, in the local area covered by the convolution kernel, the convolution kernel and A i The corresponding positions are multiplied element by element, and then all the products are summed to obtain an element value b of the convolution feature map. i,j,k , where 1≤j≤2 θ ,1≤k≤2 θ This process is repeated until the convolution kernel traverses all A i , get the corresponding convolution feature map B i Finally, we get the convolution feature map set B corresponding to each feature map in A = {B1, B2, ..., B n}. The formula is as follows:

[0068]

[0069] Among them, b i,j,k is the convolution feature map B i The value of the jth row and kth column of a i,j,k A is the original feature map of the two-dimensional coordinates i The value of the jth row and kth column of u,v is the value of the u-th row and v-th column of the normalized convolution kernel NK, 1≤u,v≤3.

[0070] (2c) Figure 3 As shown, for the convolutional feature map set B = {B1, B2, ..., B n Each feature map B in i , select a 2×2 pooling window, set the step size to 2, and place the initial position of the pooling window on the convolution feature map B i The upper left corner, i.e. B i The first 2×2 region of the feature map is selected and the maximum value in the region is placed in the corresponding position c of the feature map after pooling. i,j,k , where 1≤j≤2 θ-1 ,1≤k≤2 θ-1 , slide the pooling window until all B are traversed i, and get the corresponding pooling feature map. The process can be expressed by the following formula:

[0071]

[0072] Among them, 1≤j≤2 θ-1 ,1≤k≤2 θ-1 , c i,j,k is the pooled feature map C i The value of the jth row and kth column of b i,j,k is the convolution feature map B i The j-th row and k-th column value of the pooling window covers the upper left quarter of B i Finally, we get the pooled feature map set C = {C1, C2, ..., C n}.

[0073] The similarity calculation stage comprises the following steps:

[0074] (3a) For a given query sample dataset D e , generate the sample pooling feature map C according to the pooling feature map generation method described in the convolution and pooling stages e .

[0075] (3b) For C e And the pooled feature map set C = {C1, C2, ..., C n Each pooling feature map C in i ∈C, flattened into feature vectors v e and v i , the formula is as follows:

[0076]

[0077]

[0078] (3c) Normalize the eigenvector and calculate v e and v i The similarity of i and D e The similarity is as follows:

[0079]

[0080] Among them, θ is the partition parameter, v e,t is the vector v e The tth element of i,t is the vector v i The tth element of .

[0081] Figure 2The process of generating an original feature map from a spatial dataset is demonstrated. The minimum enclosing rectangle is determined by the four maximum values ​​of the x-axis and y-axis in the dataset and expanded into a minimum enclosing square. A grid coordinate system is then generated based on the division parameters. The number of two-dimensional coordinates falling into each grid is counted to obtain the original feature map.

[0082] Figure 3 The process of generating a convolution feature map and a pooling feature map from an original feature map is demonstrated. The Gaussian smoothing convolution kernel can effectively reduce the impact of outliers in the data set on the data distribution structure. Therefore, the main features are extracted through convolution, and then the data scale is reduced through pooling, which takes into account both accuracy and efficiency for subsequent similarity calculations.

[0083] Therefore, in the present invention, the convolution and pooling algorithm is innovatively applied to the similarity calculation of spatial data sets, and a new method framework is proposed, which effectively overcomes the limitations of traditional methods in processing complex spatial data sets. In the convolution and pooling stage, a Gaussian smoothing convolution kernel is selected to smooth the feature map, retain the main structural features of the spatial data set, and improve the accuracy and robustness of feature extraction. By generating a pooled feature map, the data scale is further reduced, while the key features are retained, providing an efficient data basis for subsequent similarity calculations. In the similarity calculation stage, the relative cosine similarity is used to measure the similarity between data sets, and the balance and comparability between the features of different spatial data sets are ensured by normalizing the feature vector. The present invention realizes efficient processing of spatial data sets with complex structures through three carefully designed stages. Not only does it reduce the calculation scale, but it also improves the search efficiency, meeting the needs of large-scale spatial data processing. Compared with traditional methods, the present invention can better reflect the complexity and diversity of spatial data sets, and provides strong technical support for fields such as urban planning and smart city construction.

[0084] The above description is only an embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of the claims of the present invention.

Claims

1. A method for calculating similarity of spatial data sets based on convolution and pooling, characterized in that: The spatial data set similarity calculation method based on convolution and pooling includes three steps: Step 1: Dataset feature map generation stage: Establish a grid coordinate system R for the spatial dataset warehouse D i , count the number of two-dimensional coordinates to generate the original feature map A of the two-dimensional coordinates i ; Step 2, convolution and pooling stage: Use Gaussian smoothing convolution kernel to smooth the two-dimensional coordinate original feature map A i , extract the original feature map A of the two-dimensional coordinates i Structural features, and then select the maximum value through the pooling window to generate the pooling feature map C i ; Step 3: Similarity calculation phase: Pooling feature map C i And according to the sample data set D e The generated sample pooling feature map C is flattened into a feature vector and normalized, and the relative cosine similarity is calculated to obtain the similarity between the data sets.

2. According to claim 1, a spatial data set similarity calculation method based on convolution and pooling is characterized in that: The step 2 specifically includes the following steps: Step 2.1: Select a Gaussian smoothing convolution kernel with a window size of 3×3 and a standard deviation of α to smooth the original feature map A of the two-dimensional coordinates in subsequent feature extraction. i , retain the structural characteristics of spatial datasets; Step 2.2: For the original feature map set A = {A1, A2, ..., A n }, each of the two-dimensional coordinate original feature maps A i ∈A for "zero padding", and then place the Gaussian smooth convolution kernel in the original feature map A of the two-dimensional coordinates after "zero padding" i The upper left corner, A i The first 3×3 area after "zero padding" is then smoothed with a Gaussian convolution kernel and locally weighted summed until the original feature map A of the two-dimensional coordinates is traversed. i Generate convolution feature map B i , and obtain the convolution feature map set B corresponding to each spatial dataset in the spatial dataset warehouse D = {B1, B2, …, B n }; Step 2.3: For the convolutional feature map set B = {B1, B2, ..., B n Each convolution feature map B in i , select one 2 p ×2 p The pooling window is set to 2. p , where 1≤p≤θ-1 places the initial position of the pooling window on the convolutional feature map B i The upper left corner, i.e. B i The first 2 p ×2 p Region, select the maximum value in the region and put it into the corresponding position c of the feature map after pooling i,j,k , where 1≤j≤2 θ-p ,1≤k≤2 θ-p , slide the pooling window until the convolution feature map B is traversed i , get the corresponding pooling feature map C i Finally, we get the pooled feature map set C corresponding to each spatial dataset in the spatial dataset warehouse D = {C1, C2, …, C n }, the formula is as follows: Among them, 1≤j≤2 θ-p ,1≤k≤2 θ-p , c i,j,k is the pooled feature map C i The value of the jth row and kth column of b i,j,k is the convolution feature map B i The j-th row and k-th column value of the pooling window covers the convolution feature map B in the upper left corner. i The j0th row and k0th column of .

3. The method for calculating similarity of spatial data sets based on convolution and pooling according to claim 1, characterized in that: In step 2.1, a Gaussian smoothing convolution kernel with a window size of 3×3 and a standard deviation of α is selected, which specifically includes the following steps: Step 2.1.1, calculate each value of the convolution kernel: Where (x, y) is the coordinate relative to the center of the convolution kernel, σ is the standard deviation, x, y∈{0,±1}, σ=α, and all possible values ​​of (x, y) (-1,-1), (-1,0), (-1,1), (0,-1), (0,0), (0,1), (1,-1), (1,0), (1,1) are calculated to obtain the convolution kernel Kernel, which is expressed as follows: Kernel i,j =G(i-2,j-2); Step 2.1.2, calculate the sum of all elements of the convolution kernel Sum, and then divide each element by the sum to normalize it to get the normalized convolution kernel NK, which is expressed as follows: Among them, Kernel i,j is the value of the i-th row and j-th column of the convolution kernel Kernel, 1≤i,j≤3.

4. The method for calculating similarity of spatial data sets based on convolution and pooling according to claim 1, characterized in that: In step 2.2, the two-dimensional coordinate original feature map A i Perform "zero padding" and sliding convolution kernel and local weighted summation, including the following steps: Step 2.2.1: In the two-dimensional coordinate original feature map A i Add a row of 0s above the first row and below the last row, and add a column of 0s to the left of the first column and to the right of the last column, that is, Padding = 1, so that the two-dimensional coordinate original feature map A i The size of 2 θ ×2 θ becomes (2 θ +2)×(2 θ +2) to ensure that the size of the feature map after convolution remains unchanged; Step 2.2.2: Add the Gaussian smoothing convolution kernel to the original feature map A in the two-dimensional coordinates i Slide up with a step size of 1. Each position corresponds to a local area, namely the receptive field. In the local area covered by the Gaussian smooth convolution kernel, the Gaussian smooth convolution kernel is combined with the two-dimensional coordinate original feature map A. i The corresponding positions are multiplied element by element, and then all the products are summed to obtain an element value b of the convolution feature map. i,j,k , where 1≤j≤2 θ ,1≤k≤2 θ , this process is repeated until the convolution kernel traverses A i , generate the corresponding convolution feature map B i , and obtain the convolution feature map set B corresponding to each feature map in the two-dimensional coordinate original feature map set A = {B1, B2, ..., B n }, expressed as: Among them, NK u,v is the value of the u-th row and v-th column of the normalized convolution kernel NK, 1≤u,v≤3.

5. The method for calculating similarity of spatial data sets based on convolution and pooling according to claim 1, characterized in that: The step 3 specifically comprises the following steps: Step 3.1: For a given query sample dataset D e , according to the query sample data set D e The minimum and maximum values ​​of the x-axis and y-axis of all two-dimensional coordinates in the coordinate system R are established e , count the number of two-dimensional coordinates to generate the sample two-dimensional coordinate original feature map A e , and then use the Gaussian smoothing convolution kernel to smooth the original feature map A of the sample two-dimensional coordinates e Get the sample convolution feature map B e , and then select the maximum value through the pooling window to generate the sample pooling feature map C e ; Step 3.2: Pooling feature map C for the sample e And the pooled feature map set C = {C1, C2, ..., C n Each feature map C in i ∈C, flattened into feature vectors v e and v i : Step 3.3, normalize the eigenvector and calculate v e and v i The similarity of is recorded as the corresponding two-dimensional coordinate space dataset D i And query sample dataset D e The similarity is as follows: Among them, θ is the partition parameter, v e,t is the vector v e The tth element of i,t is the vector v i The tth element of .

6. The method for calculating similarity of spatial data sets based on convolution and pooling according to claim 1, characterized in that: The step 1 specifically comprises the following steps: Step 1.1: Given a spatial dataset warehouse D = {D1, D2, …, D n }, D i ∈D is a spatial data set of two-dimensional coordinates, where D i ={(x1,y1),(x2,y2),…,(x m ,y m )}, find D i The minimum and maximum values ​​of the x-axis and y-axis of all two-dimensional coordinates in are denoted by xi min , min ,xi max , max ; Step 1.2, with (xi min ,yi min ) and (xi max ,yi max ) is used as the diagonal to establish a rectangular area, and then the width of the rectangle is extended to both ends at equal distances to expand the rectangular area into a square, and then the square area is divided into 2 according to the division parameter θ θ ×2 θ grids of equal size, and establish a grid coordinate system R i ; Step 1.3: For the spatial dataset warehouse D = {D1, D2, ..., D n Each spatial dataset D in i , statistical spatial dataset D i In the grid coordinate system R i The grid g at row j and column k in j,k The number of two-dimensional coordinate points within is denoted as a i,j,k , where 1≤j≤2 θ ,1≤k≤2 θ , get the spatial dataset D i The two-dimensional coordinate feature map A i , traverse the spatial dataset warehouse D, and obtain the two-dimensional coordinate original feature map set A={A1,A2,…,A n }.

Images