Spectrum map construction method based on dual-scale fusion

Abstract

The invention discloses a spectrum map construction method based on dual-scale fusion, relates to the technical field of spectrum map construction, and aims to improve the adaptability and characterization accuracy of a spectrum map to a complex environment by using agglomerated hierarchical clustering on sampling points; according to the method, a Voronoi diagram of random sampling points is constructed by taking a global region to be interpolated as a boundary, the Voronoi diagram is divided according to an optimal clustering result, space division and the clustering result are combined, the advantages of the Voronoi diagram in space division are fully utilized, and improvement of local precision and overall spatial resolution of a spectrum map is facilitated; the Kriging interpolation method combining the region and the global is adopted, and the local characteristics of different regions and the overall trend of the global can be fully considered, so that the interpolation error in a complex environment is reduced.

Description

Technical Field

[0001] The present invention relates to the field of spectrum map construction, and particularly to a spectrum map construction method based on dual-scale fusion. Background Art

[0002] In modern communication, electronics and other fields, spectrum resources are extremely valuable resources that support the operation of various wireless communication technologies, such as mobile phone communication, satellite communication, Internet of Things communication, etc. Reasonable allocation and effective utilization of spectrum resources are crucial for improving communication quality and promoting industrial development. In order to achieve the rational use of spectrum resources, it is necessary to monitor and analyze the usage of the spectrum in real time. As a tool for intuitively displaying the distribution and usage of spectrum resources, the spectrum map can help spectrum management departments, communication operators, etc. better understand information such as spectrum occupancy and signal strength distribution, so as to provide a basis for spectrum planning, frequency allocation and other decisions.

[0003] Existing technical solutions:

[0004] CN119154978A proposes a multi-band spectrum map construction method, which includes: obtaining map data of a target area, rasterizing the map data to obtain a target raster map; randomly selecting a plurality of acquisition raster grids in the target raster map, and obtaining the radiation signal strength of each acquisition raster grid at different frequency bands; using the position information of each acquisition raster grid and the radiation signal strength of each acquisition raster grid at different frequency bands as heterogeneous graph node information, and modeling each acquisition raster grid at each frequency band as a heterogeneous graph node; constructing a spatial domain graph and a frequency domain graph of the target raster map according to the heterogeneous graph node information of each heterogeneous graph node; obtaining a pre-trained heterogeneous graph neural network, and inputting the spatial domain graph, the frequency domain graph and the heterogeneous graph node information of each heterogeneous graph node into the heterogeneous graph neural network to obtain a multi-band spectrum map corresponding to the target area. This method can construct a high-quality multi-band spectrum map.

[0005] CN119107424A discloses a three-dimensional electromagnetic spectrum map construction method based on tensor completion, including: S1, obtaining a measurement area in a simulation scenario, collecting spectrum data in the measurement area to obtain a missing data set; S2, based on the tensor singular value decomposition method and the missing data set, obtaining a completed tensor; S3, constructing a three-dimensional electromagnetic spectrum map based on the Poisson surface reconstruction algorithm and the completed tensor. This method can construct and reconstruct a three-dimensional electromagnetic spectrum map.

[0006] The prior art does not consider the actual distribution and spatial correlation of sampling point data when dividing regions, resulting in large data differences within the divided regions, being unable to accurately reflect the local change characteristics of the spectrum, and affecting the smoothness and accuracy of the spectrum map. At the same time, the prior art generally uses global interpolation methods to construct spectrum maps, uniformly using the same parameters and models globally without considering the variation differences of the spectrum in different regions. Therefore, it is unable to effectively capture local spectrum characteristics and abnormal situations, resulting in large interpolation errors in complex environments and low accuracy of the spectrum map. Summary of the Invention

[0007] The purpose of the present invention is to: in view of the above problems, the present invention provides a spectrum map construction method based on dual-scale fusion, divides the global region according to uniformity, and obtains the final result through a multi-scale fusion strategy, which can take into account both global and local characteristics and improve the interpolation accuracy.

[0008] The technical solution adopted by the present invention is as follows:

[0009] A spectrum map construction method based on dual-scale fusion, the method includes:

[0010] Obtain a number of sampling points;

[0011] Agglomerative clustering, clustering the obtained sampling points through the agglomerative hierarchical clustering algorithm, respectively specifying different numbers of clusters and clustering, and then respectively calculating the silhouette coefficient and determining the best clustering result;

[0012] Construct a Voronoi diagram, use the global region to be interpolated as the boundary, construct the Voronoi diagram of the sampling points, and divide the Voronoi diagram through the best clustering result to obtain a number of divided regions;

[0013] Calculation of regional and global weights, perform Kriging interpolation on both the region and the global and calculate the uniformity and energy weights respectively;

[0014] Spectrum map construction, calculate the fusion value of all points to be interpolated through the uniformity and energy weights of the region and the global, assign the calculated fusion value to the corresponding points to be interpolated, and obtain a complete spectrum map based on dual-scale fusion after data processing, map drawing, and visualization adjustment.

[0015] Furthermore, the agglomerative clustering includes:

[0016] Specify the number of clusters as k = 2, cluster the obtained sampling points according to the position distance, first calculate the distance matrix D between the sampling points, and use the Euclidean distance:

[0017]

[0018] Then each data point is regarded as a separate cluster, that is, there are initially n clusters. Iteratively solve the similarity between all current clusters, and use average linkage as the merging criterion, that is, calculate the average distance between all pairs of points in two clusters:

[0019]

[0020] After merging the clusters, update the distance matrix, and iterate until all data points are merged into one cluster. Finally, construct a dendrogram and cut the dendrogram according to the specified number of clusters to obtain the clustering results with the specified number of clusters;

[0021] Repeat the above steps until k = k max ;

[0022] Calculate the silhouette coefficient under different clustering clusters. For each sampling point i, calculate its global silhouette coefficient S through the following formula:

[0023]

[0024] In the formula, s(i) is the silhouette coefficient of each point. Compare the global silhouette coefficients under different numbers of clustering clusters, and select the number of clusters with the smallest global silhouette coefficient for clustering to obtain the best clustering results.

[0025] Furthermore, the silhouette coefficient of each point is specifically obtained through the following formula:

[0026]

[0027] In the formula, a(i) is the average distance from sampling point i to other sampling points in its cluster, b(i) is the average distance from sampling point i to all sampling points in the nearest neighbor cluster, C i represents the cluster where point i is located, and C l represents the cluster that point i does not belong to.

[0028] Furthermore, the Kriging interpolation process is specifically as follows:

[0029] For the set of sampling points in region t and the point to be interpolated First, construct a semivariogram function:

[0030]

[0031] In the formula, h i0 represents the distance between each pair of sampling points and the point to be interpolated;

[0032] Select a semivariogram model and fit the model to adapt to the calculated semivariogram data, and then construct a Kriging equation:

[0033]

[0034] where μ is the Lagrange multiplier, and the parameters λ i and μ are obtained by solving;

[0035] Finally, the interpolation of the target point is obtained as:

[0036]

[0037] Repeat the above process until the interpolation within the region is completed.

[0038] Furthermore, the uniformity and energy weight calculation processes of the region and the global are the same, and the calculation of the region uniformity is as follows:

[0039] The region uniformity u is the reciprocal of the standard deviation σ of the areas of all Voronoi diagrams within a region. Suppose there are m Voronoi diagrams within the region, and their areas are S j (j = 1, 2,..., m), then the average area is:

[0040]

[0041] The standard deviation σ of the area is:

[0042]

[0043] Furthermore, the region energy weight is specifically as follows:

[0044] The energy weight v is the sum of the powers of the field strengths within a region, as shown in the following formula:

[0045]

[0046] Furthermore, the fusion value is obtained by weighting the interpolation points of the global and the region.

[0047] Furthermore, the calculation of the fusion value is specifically as follows:

[0048] w range = ((u range / u range + u global ) + (v range / v range + v global )) / 2 (14)

[0049] w global = ((u global / u range + u global ) + (v global / v range + v global )) / 2 (15)

[0050] P = P g ·w global +P r ·w range (16)

[0051] In the formula, P g is the global interpolation point field strength, and w global is its weight; P r is the regional interpolation point field strength, and w range is its weight.

[0052] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are as follows:

[0053] A method for constructing a spectrum map based on dual-scale fusion of the present invention uses agglomerative hierarchical clustering for sampling points, tries different numbers of clusters respectively, and determines the best clustering result according to the silhouette coefficient. This clustering method fully considers the internal distribution structure of the sampling point data, avoids the irrationality brought by artificial arbitrary region division or the use of a fixed clustering mode, can adaptively discover the natural grouping in the data, makes the subsequent clustering-based analysis and processing more conform to the actual spectrum distribution, and improves the adaptability and characterization accuracy of the spectrum map to complex environments; constructs a Voronoi diagram of random sampling points with the global area to be interpolated as the boundary, and divides the Voronoi diagram according to the best clustering result, combines the space division with the clustering result, makes full use of the advantages of the Voronoi diagram in space division, and helps to improve the local accuracy and overall spatial resolution of the spectrum map; adopts a Kriging interpolation method combining region and global, can fully consider the local characteristics of different regions and the overall global trend, so as to reduce the interpolation error in complex environments. Description of the Drawings

[0054] Figure 1 is a flowchart of a method for constructing a spectrum map based on dual-scale fusion of the present invention;

[0055] Figure 2 is a flowchart for determining the best clustering result in a method for constructing a spectrum map based on dual-scale fusion of the present invention;

[0056] Figure 3 is a flowchart of Kriging interpolation in a method for constructing a spectrum map based on dual-scale fusion of the present invention. Detailed Embodiments

[0057] The present invention will be described in detail below with reference to the drawings.

[0058] To make the objectives, technical solutions and advantages of the present invention more comprehensible, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0059] Embodiment

[0060] This embodiment provides a method for constructing a spectrum map based on dual-scale fusion, which uses multi-scale interpolation and fusion to obtain the result. The core of this embodiment lies in dividing the global region according to uniformity, performing Kriging interpolation on the multi-scale regions of the global and local respectively to obtain the interpolation results, calculating the multi-scale weights of the global and local, and fusing the global and local interpolation results according to the weights to obtain the final spectrum map, as Figure 1 shown. The specific method is as follows:

[0061] Obtain sampling points. In the spectrum area to be surveyed, use the sensing node device to obtain the signal field strength value and add the position information. Suppose there are n sampling points in total, then the sampling point set is S = {[x 1 , y 1 , P 1 ,..., [x n , y n , P n};

[0062] Agglomerative clustering. Cluster the obtained sampling points through the agglomerative hierarchical clustering algorithm, specify different numbers of clusters and perform clustering respectively, and then calculate the silhouette coefficient and determine the best clustering result respectively, as Figure 2 shown. Specifically, it includes:

[0063] Specify the number of clusters as k = 2. Cluster the obtained sampling points according to the position distance. First, calculate the distance matrix D between the sampling points, using the Euclidean distance:

[0064]

[0065] Then regard each data point as a separate cluster, that is, there are n clusters at the beginning. Iteratively solve the similarity between all current clusters, and use average linkage as the merging criterion, that is, calculate the average distance between all pairs of points in two clusters:

[0066]

[0067] After merging the clusters, update the distance matrix, and iterate until all data points are merged into one cluster. Finally, construct a dendrogram and cut the dendrogram according to the specified number of clusters to obtain the clustering result of the specified number of clusters;

[0068] Repeat the above steps until k = k max ;

[0069] Calculate the silhouette coefficient for different clusters. For each sampling point i, calculate its global silhouette coefficient S through the following formula:

[0070]

[0071] In the formula, s(i) is the silhouette coefficient of each point. Compare the global silhouette coefficients under different numbers of clusters, and select the number of clusters with the smallest global silhouette coefficient for clustering to obtain the best clustering result;

[0072] The silhouette coefficient of each point is specifically obtained through the following formula:

[0073]

[0074] In the formula, a(i) is the average distance from sampling point i to other sampling points in its cluster, b(i) is the average distance from sampling point i to all sampling points in the nearest neighbor cluster, C i represents the cluster where point i is located, C l represents the cluster that point i does not belong to.

[0075] Region division: Using the global area to be interpolated as the boundary, construct the Voronoi diagram of the sampling points, and divide the Voronoi diagram through the best clustering result to obtain several divided regions;

[0076] Weight calculation for regions and the global area: Perform Kriging interpolation on both the regions and the global area, and calculate the uniformity and energy weights respectively;

[0077] The process of performing Kriging interpolation on a region is specifically as follows, as Figure 3 shown:

[0078] For the set of sampling points in region t and the point to be interpolated First, construct the semivariogram function:

[0079]

[0080] In the formula, h i0 represents the distance between each pair of sampling points and the point to be interpolated;

[0081] Select a suitable semivariogram model, fit the model to adapt to the calculated semivariogram data, and then construct the Kriging equation:

[0082]

[0083] In the formula, μ is the Lagrange multiplier, and solve to obtain the parameters λ i and μ;

[0084] Finally, the interpolation of the target point is:

[0085]

[0086] Repeat the above process until the interpolation within the region is completed.

[0087] Calculation of regional uniformity and energy weight. The regional uniformity u is the reciprocal of the standard deviation σ of the areas of all Voronoi diagrams within a region. Suppose there are m Voronoi diagrams in the region, and their areas are S j (j = 1, 2, …, m), then the average area is:

[0088]

[0089] The standard deviation σ of the area is:

[0090]

[0091] The specific regional energy weight is as follows:

[0092] The energy weight v is the power sum of the field strengths in a region, as shown in the following formula:

[0093]

[0094] Perform Kriging interpolation within the entire global region to be surveyed, and calculate the global uniformity and energy weight using the same method as the calculation of regional uniformity and energy weight;

[0095] Spectrum map construction. Calculate the fusion value of all points to be interpolated through the uniformity and energy weight of the region and the global region, assign the calculated fusion value to the corresponding points to be interpolated, and obtain a complete spectrum map based on dual-scale fusion after data processing, map drawing, and visualization adjustment. The specific process is as follows:

[0096] w range = ((u range / u range + u global ) + (v range / v range + v global )) / 2 (14)

[0097] w global = ((u global / u range + u global ) + (v global / v range + v global )) / 2 (15)

[0098] P = P g ·w global + Pr ·w range (16)

[0099] wherein, P g is the global interpolation point field strength, and w global is its weight; P r is the regional interpolation point field strength, and w range is its weight.

[0100] In summary, in this embodiment, through agglomerative hierarchical clustering and determining the optimal clustering result according to the silhouette coefficient, the natural grouping of data is adaptively discovered, unreasonable division is avoided, and the adaptability and accuracy of the spectrum map to complex environments are improved; the present invention constructs a Voronoi diagram of randomly sampled points based on the global area to be interpolated, divides the Voronoi diagram according to the clustering result, integrates the advantages of spatial division, and enhances the local and overall accuracy of the spectrum map; the present invention uses the Kriging interpolation method combining region and global, taking into account local characteristics and global trends, reducing the interpolation error in complex environments, and the RMSE error has decreased by 3% compared with the original global Kriging algorithm.

[0101] Specific embodiments are used in this article to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and modifications can be made to the present invention, and these improvements and modifications also fall within the protection scope of the claims of the present invention.

Claims

1. A spectrum map construction method based on dual-scale fusion, characterized in that: The method comprises: Get a number of adoption points; Agglomerative hierarchical clustering, clustering the acquired sampling points through the agglomerative hierarchical clustering algorithm, specifying different numbers of clusters and clustering them, and then calculating the silhouette coefficient and determining the best clustering result; Region division: Taking the global interpolation area as the boundary, construct the Voronoi diagram of the sampling points, and divide the Voronoi diagram by the best clustering result to obtain several divided regions; Regional and global weight calculation, Kriging interpolation is performed on both the regional and global data and the uniformity and energy weights are calculated respectively; The spectrum map is constructed by calculating the fusion values ​​of all the points to be interpolated through regional and global uniformity and energy weights, and the calculated fusion values ​​are assigned to the corresponding points to be interpolated. After data processing, map drawing and visualization adjustment, a complete spectrum map based on dual-scale fusion is obtained.

2. The method for constructing a spectrum map based on dual-scale fusion according to claim 1, characterized in that: The cohesive layer clustering includes: Specify the number of clusters as k = 2, cluster the acquired sampling points according to the position distance, and first calculate the distance matrix D between the sampling points, using the Euclidean distance: Then each data point is regarded as a separate cluster, that is, there are n clusters at the beginning, and the similarity between all current clusters is iteratively solved. The average link is used as the merging criterion, that is, the average distance between all point pairs in the two clusters is calculated: After merging the clusters, the distance matrix is ​​updated and iterated until all data points are merged into one cluster. Finally, a dendrogram is constructed and cut according to the specified number of clusters to obtain the clustering results of the specified number of clusters. Repeat the above steps until k=k max ; Calculate the silhouette coefficients under different clusters. For each sampling point i, calculate its global silhouette coefficient S by the following formula: In the formula, s(i) is the silhouette coefficient of each point. The global silhouette coefficients under different cluster numbers are compared, and the number of clusters with the smallest global silhouette coefficient is selected for clustering to obtain the best clustering result.

3. The method for constructing a spectrum map based on dual-scale fusion according to claim 2, characterized in that: The contour coefficient of each point is obtained by the following formula: Where a(i) is the average distance from sampling point i to other sampling points in its cluster, b(i) is the average distance from sampling point i to all sampling points in the nearest neighbor cluster, and C i Indicates the cluster where point i belongs, C l Indicates the cluster to which point i does not belong.

4. The method for constructing a spectrum map based on dual-scale fusion according to claim 1, characterized in that: The Kriging interpolation process is as follows: For the sampling point set in region t and the point to be interpolated First construct the semivariogram function: In the formula, h i0 Indicates the distance between each pair of sampling points and the point to be interpolated; Select the semivariance model and fit the model to the calculated semivariance data, then construct the kriging equations: Where μ is the Lagrange multiplier, and the parameter λ is obtained by solving i and μ; Finally, the interpolation value of the target point is: Repeat the above process until the regional interpolation is completed.

5. The method for constructing a spectrum map based on dual-scale fusion according to claim 4, characterized in that: The calculation process of the regional and global uniformity and energy weight is the same, and the calculation of the regional uniformity is as follows: The regional uniformity u is the inverse of the standard deviation σ of the areas of all Voronoi diagrams in a region. Suppose there are m Voronoi diagrams in the region, and their areas are S j (j=1,2,…,m), then the average area for: The standard deviation σ of the area is:

6. The method for constructing a spectrum map based on dual-scale fusion according to claim 5, characterized in that: The regional energy weights are as follows: The energy weight v is the power sum of the field strength in a region, as follows:

7. The method for constructing a spectrum map based on dual-scale fusion according to claim 1, characterized in that: The fusion value is obtained by weighting the global and regional interpolation points.

8. The method for constructing a spectrum map based on dual-scale fusion according to claim 7, characterized in that: The fusion value is calculated as follows: Where P g is the field strength at the global interpolation point, w global Its weight; P r is the field strength at the regional interpolation point, w range For its weight.

Images