A method, device and medium for adaptive segmentation of migratory insect swarms

Abstract

The embodiments of this specification disclose an adaptive segmentation method, device and medium for migrating insect swarms, which relate to the field of data processing technology and are used to solve the problem of noise interference and low accuracy in traditional insect swarm segmentation methods. The method includes: determining the four-dimensional coordinates of each initial data point by high-resolution radar measurement; performing adaptive noise removal based on the average sparsity corresponding to the four-dimensional coordinates to obtain filtered data points; obtaining the nearest neighbor set of each data point to define an adjacency matrix based on the nearest neighbor sets of any two data points; determining the undirected weighted graph corresponding to the data point to determine the Laplace matrix of the current migrating insect swarm based on the pairwise similarity in the adjacency matrix; searching the Laplace matrix based on the preset subgraph optimal segmentation function to obtain the optimal segmentation of the subgraph, so as to segment the data points corresponding to the optimal segmentation of the subgraph as the segmentation result of the current migrating insect swarm.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method, device and medium for adaptively segmenting migratory insect swarms. Background Art

[0002] Insect migration is the main cause of off-site outbreaks of agricultural pests and diseases, which seriously threatens food production and the ecological environment. Radar technology, with its all-day and all-weather observation capabilities, has become an effective means of monitoring insect migration. The distance resolution of super-resolution radar can reach 0.2 meters. It can observe each insect in detail in dense scenes and accurately measure the flight height, weight, wing beat frequency and other characteristics of insects, which provides the possibility for in-depth understanding of the migration behavior of insect groups. Most insect species show the characteristics of aggregated migration, migrating long distances in groups at a specific time and space. If the study can be carried out in groups, it will help to understand the migration patterns and behaviors of insects in more detail. Since the study of insect migration can provide a scientific basis for the prevention and control of agricultural pests and diseases, thereby ensuring food security and ecological security, it is very important to accurately study insect groups.

[0003] Traditional insect swarm segmentation methods mainly rely on the two-dimensional or three-dimensional measurement features of insects. However, there are many types of insects, and the biological parameters vary greatly. The boundaries of some insect swarms are fuzzy and sticky to each other. It is difficult to achieve fine population separation based on two-dimensional or three-dimensional parameters alone. In addition, due to the large density differences and complex spatial distribution of different insect swarms, there may be adjacent or overlapping occlusions between insect swarms, and actual measurements will also be disturbed by noise. Existing clustering methods are highly dependent on parameters and usually require manual adjustment of clustering parameters, which is time-consuming and laborious, and has low accuracy. Therefore, it is urgent to develop an accurate insect swarm segmentation method that does not require manual adjustment of parameters, can utilize higher-dimensional features, and automatically remove noise points. Summary of the invention

[0004] In order to solve the above problems, one or more embodiments of the present specification provide an adaptive segmentation method, device and medium for migrating insect swarms, which are used to finely segment the insect swarms according to the four-dimensional radar measurement characteristics without manual adjustment of parameters, thereby automatically achieving noise removal and fine population segmentation, which helps to distinguish a wide variety of insect swarms that are adhered to each other.

[0005] One or more embodiments of this specification adopt the following technical solutions:

[0006] One or more embodiments of the present specification provide a method for adaptively segmenting a migratory insect swarm, the method comprising:

[0007] Determine the four-dimensional coordinates of each initial data point based on the four-dimensional characteristics of the current migratory insect swarm measured by high-resolution radar;

[0008] Based on the average sparsity corresponding to the four-dimensional coordinates of each initial data point, adaptively remove noise from the initial data point to obtain filtered data points;

[0009] Obtaining the nearest neighbor set of each data point to define an adjacency matrix based on the nearest neighbor sets of any two data points; wherein the adjacency matrix is ​​constructed based on the pairwise similarity of any two data points, and the diagonal value is 0;

[0010] Determine an undirected weighted graph corresponding to the data point to determine the Laplacian matrix of the current migrating insect swarm according to the pairwise similarities in the adjacency matrix;

[0011] The Laplace matrix is ​​searched based on a preset subgraph optimal segmentation function to obtain an optimal subgraph segmentation, so that the data point segmentation corresponding to the subgraph optimal segmentation is used as the segmentation result of the current migrating insect swarm.

[0012] Optionally, in one or more embodiments of the present specification, based on the average sparsity corresponding to the four-dimensional coordinates of each initial data point, adaptively removing noise from the initial data point to obtain filtered data points specifically includes:

[0013] Based on the four-dimensional coordinates of each of the initial data points, obtaining the Euclidean distance between each of the initial data points, so as to obtain a preset number of nearest neighbor data points of each of the initial data points according to the Euclidean distance;

[0014] Taking the average value of the Euclidean distance between each nearest neighbor data point and the initial data point as the average sparsity of the initial data point, and sorting the average sparsity of each initial data point to obtain the distribution characteristics of the average sparsity of the initial data points;

[0015] Performing SG filtering on the curve corresponding to the distribution characteristic to determine the noise threshold value according to the intersection of the filtered curve and the preset slope tangent;

[0016] The average sparsity of each of the initial data points is compared with the noise threshold value to filter out the initial data points that are higher than the noise threshold value to obtain filtered data points.

[0017] Optionally, in one or more embodiments of the present specification, obtaining the nearest neighbor set of each data point to define an adjacency matrix according to the nearest neighbor sets of any two data points specifically includes:

[0018] Obtain the nearest neighbor set of each data point, and determine the pairwise similarity between any two data points based on the intersection of the nearest neighbor sets of any two data points;

[0019] Processing the pairwise similarities according to a preset formula to obtain the pairwise similarities before normalization;

[0020] Performing linear normalization on the pairwise similarities before normalization to obtain normalized pairwise similarities;

[0021] The normalized pairwise similarities are sorted to construct an adjacency matrix of the data points.

[0022] Optionally, in one or more embodiments of this specification, the preset formula is:

[0023] ; where the data points are , Represents data points The nearest neighbor set of Not included , and The set of shared neighbors between , yes No. Neighbor, yes No. Neighbors, Yes The number of all nearest neighbors, Represents the pairwise similarity before normalization.

[0024] Optionally, in one or more embodiments of the present specification, determining the undirected weighted graph corresponding to the data point to determine the Laplacian matrix of the current migrating insect swarm according to the pairwise similarities in the adjacency matrix specifically includes:

[0025] Taking all data points of the current migrating insect swarm as an undirected weighted graph, and obtaining the sum of pairwise similarities of each data point in the undirected weighted graph;

[0026] The sum of the pairwise similarities is taken as the degree of the data point, and the degrees of the data points are summarized to obtain a degree matrix;

[0027] The Laplace matrix of the current migrating insect swarm is determined according to the difference between the degree matrix and the adjacency matrix.

[0028] Optionally, in one or more embodiments of the present specification, searching the Laplacian matrix based on a preset subgraph optimal segmentation function to obtain a subgraph optimal segmentation, so as to segment the data points corresponding to the subgraph optimal segmentation as the segmentation result of the current migrating insect swarm, specifically includes:

[0029] Based on the preset optimization function and the optimization goal of the current migrating insect swarm, a preset subgraph optimal segmentation function is defined; wherein the preset optimization function is:

[0030] , where M is the clustering parameter, Represents the subgraphs corresponding to the clustering parameters, Representative subgraph The complement of For subgraph The modulus of The number of midpoints, ,and , the optimization goal is to find the minimum value of the function;

[0031] Find the M minimum eigenvalues ​​of the Laplacian matrix according to the preset subgraph optimal segmentation function, and determine the optimal segmentation of the M subgraphs;

[0032] Arranging the eigenvectors corresponding to the minimum eigenvalue to obtain a characteristic matrix, so as to cluster the characteristic matrix to obtain a cluster division result, and corresponding the cluster division result to the data point to obtain a clustering result of the current migratory insect swarm;

[0033] According to the eigenvalue distribution of the Laplace matrix and the DB index corresponding to the clustering result, the optimal clustering parameter is determined, so as to determine the segmentation result of the current migratory insect swarm based on the optimal clustering parameter.

[0034] Optionally, in one or more embodiments of the present specification, determining an optimal clustering parameter according to the eigenvalue distribution of the Laplace matrix and the DB index corresponding to the clustering result, so as to determine the segmentation result of the current migratory insect swarm based on the optimal clustering parameter, specifically includes:

[0035] Calculating and sorting the eigenvalues ​​of the Laplacian matrix to extract the clustering eigenvalues ​​whose next eigenvalue in the Laplacian matrix is ​​greater than twice the previous eigenvalue;

[0036] Calculating the grouping characteristic value to use the calculation result as the grouping number, and obtaining the grouping result corresponding to the grouping number;

[0037] The DB index of the corresponding clustering result is obtained to determine the optimal clustering parameter according to the DB index, and the subgraph optimal segmentation of the optimal clustering parameter is used as the segmentation result of the current migratory insect swarm.

[0038] Optionally, in one or more embodiments of the present specification, determining the four-dimensional coordinates of each initial data point according to the four-dimensional features obtained by measuring the current migrating insect swarm using a high-resolution radar specifically includes:

[0039] The four-dimensional features of the current migrating insect swarm are obtained by measuring the four-dimensional features using the four-dimensional features as the initial four-dimensional coordinates of the initial data points;

[0040] The initial four-dimensional coordinates are obtained and normalized based on a minimum-maximum normalization method to determine the four-dimensional coordinates of each initial data point.

[0041] One or more embodiments of the present specification provide an adaptive segmentation device for migrating insect swarms, the device comprising:

[0042] at least one processor; and,

[0043] a memory communicatively connected to the at least one processor; wherein,

[0044] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to: execute any of the above-mentioned methods.

[0045] One or more embodiments of the present specification provide a non-volatile computer storage medium storing computer executable instructions, wherein the computer executable instructions are configured to execute any of the above-described methods.

[0046] At least one of the above technical solutions adopted in the embodiments of this specification can achieve the following beneficial effects:

[0047] After the four-dimensional features of high-resolution radar measurement, the adjacency matrix is ​​defined by using the shared nearest neighbor distance and the optimal subgraph segmentation method, which integrates the spatiotemporal clustering of insect swarms and the clustering of biological parameters to achieve fine segmentation of insect swarms. And according to the average sparsity corresponding to the four-dimensional coordinates, adaptive noise removal is achieved, and different data sets of migrating insect swarms can be automatically removed. Based on the preset subgraph optimal segmentation function, the Laplace matrix is ​​searched, and the optimal clustering parameters are calculated in the process of obtaining the optimal subgraph segmentation, which realizes the accurate extraction of the segmentation results of migrating insect swarms. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of this specification or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art description. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. In the drawings:

[0049] Figure 1 A schematic diagram of a flow chart of an adaptive segmentation method for migrating insect swarms provided in an embodiment of this specification;

[0050] Figure 2 A schematic diagram of the distribution of average sparsity provided in the embodiments of this specification;

[0051] Figure 3 A schematic diagram of a four-dimensional coordinate of a data point provided in an embodiment of the present specification, wherein (a) is a scatter plot of data points displayed in the dimensions of time, height, and weight, (b) is a scatter plot of data points displayed in the dimensions of time, height, and wing-flapping, (c) is a scatter plot of data points displayed in the dimensions of time, weight, and wing-flapping, and (d) is a scatter plot of data points displayed in the dimensions of height, weight, and wing-flapping;

[0052] Figure 4 A schematic diagram of insect swarm segmentation results provided in an embodiment of the present specification, wherein (a) is the segmentation result displayed in the dimensions of time, height, and weight, (b) is the segmentation result displayed in the dimensions of time, height, and wing flapping, (c) is the segmentation result displayed in the dimensions of time, weight, and wing flapping, and (d) is the segmentation result displayed in the dimensions of height, weight, and wing flapping. DETAILED DESCRIPTION

[0053] The embodiments of this specification provide a method, device and medium for adaptively segmenting migratory insect swarms.

[0054] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0055] like Figure 1 As shown, the present specification embodiment provides a schematic flow chart of an adaptive segmentation method for migrating insect swarms. Figure 1 It can be seen that in one or more embodiments of this specification, a method for adaptively segmenting a migratory insect swarm includes the following steps:

[0056] S101: Determine the four-dimensional coordinates of each initial data point based on the four-dimensional features obtained by measuring the current migrating insect swarm using a high-resolution radar.

[0057] After obtaining the four-dimensional features of the current migrating insect swarm using high-resolution radar measurements, the four-dimensional features measured by the high-resolution radar are regarded as coordinates in four-dimensional space in the embodiments of this specification, that is, the four-dimensional coordinates of each initial data point are determined.

[0058] Specifically, since different dimensions are used between different features, the coordinates need to be normalized. Therefore, in one or more embodiments of the present specification, the four-dimensional coordinates of each initial data point are determined according to the four-dimensional features obtained by measuring the current migrating insect swarm by a high-resolution radar, which specifically includes the following process:

[0059] First, the four-dimensional features of the current migratory insect swarms are measured by high-resolution radar, and the four-dimensional features are used as the initial four-dimensional coordinates of the initial data points; the initial four-dimensional coordinates are obtained and normalized based on the minimum and maximum normalization method, and the initial four-dimensional coordinates, that is, the minimum value of the original data, are mapped to 0 and the maximum value is mapped to 1 through linear transformation, so as to determine the four-dimensional coordinates of each initial data point while ensuring the relative proportion of the data. The specific formula in this process is as follows:

[0060] (1)

[0061] in, represents the normalized four-dimensional coordinates, represents the initial four-dimensional coordinates, represents the minimum value of the initial four-dimensional coordinate, Indicates the maximum value of the initial four-dimensional coordinate.

[0062] S102: Based on the average sparsity corresponding to the four-dimensional coordinates of each initial data point, adaptively remove noise from the initial data point to obtain filtered data points.

[0063] After the four-dimensional coordinates of each initial data point are obtained based on the above step S101, in order to automatically complete noise removal and improve the accuracy of insect swarm segmentation, the embodiments of this specification will perform adaptive noise removal on the initial data points based on the average sparsity corresponding to the four-dimensional coordinates of each initial data point to obtain filtered data points. Specifically, in one or more embodiments of this specification, based on the average sparsity corresponding to the four-dimensional coordinates of each initial data point, adaptive noise removal is performed on the initial data points to obtain filtered data points, which specifically includes the following process:

[0064] First, based on the four-dimensional coordinates of each initial data point, the Euclidean distance between each initial data point is obtained, so as to obtain a preset number of nearest neighbor data points of each initial data point according to the Euclidean distance. That is, after obtaining the four-dimensional coordinates of each initial data point, the Euclidean distance of each initial data point to other initial data points is calculated. For a data point, the Euclidean distance between it and its The data point with the smallest distance is called his The average value of the Euclidean distance between each nearest neighbor data point and the initial data point is then used as the average sparsity of the initial data point. By sorting the average sparsity of each initial data point, the distribution characteristics of the average sparsity of the initial data point can be obtained. Figure 2 The figure is a schematic diagram of the distribution characteristics of the average sparsity degree in an application scenario provided by an embodiment of this specification.

[0065] After a large amount of actual measured data verification, it is found that the distribution of the average sparsity always conforms to a similar exponential function form, so in the embodiment of this specification, the curve corresponding to the distribution characteristic is subjected to SG filtering processing to determine the noise threshold value according to the intersection of the filtered curve and the preset slope tangent. Figure 2 As shown, we can find the only point whose tangent slope is 45° and its corresponding tangent, namely Figure 2 The straight line in . Because the average sparsity increases slowly before this point, and increases rapidly after this point. Therefore, the average sparsity of this point is defined as the noise threshold, and the data with an average sparsity higher than this point is noise. Therefore, by comparing the average sparsity of each initial data point with the noise threshold value, the initial data points higher than the noise threshold value can be filtered to obtain the filtered data points. Through this process, the outliers in the space are effectively eliminated, thereby improving the reliability and accuracy of subsequent swarm segmentation.

[0066] S103: Obtain the nearest neighbor set of each data point to define an adjacency matrix according to the nearest neighbor sets of any two data points; wherein the adjacency matrix is ​​constructed based on the pairwise similarity of any two data points, and the diagonal value is 0.

[0067] In the clustering problem, it is necessary to define a similarity measure, that is, how to determine the similarity between points. Therefore, in order to achieve the segmentation of migratory insect swarms, after filtering the initial data points, the nearest neighbor set of each data point is obtained in the embodiment of this specification, so as to define the adjacency matrix based on the nearest neighbor set of any two data points. It should be noted that the adjacency matrix is ​​constructed based on the pairwise similarity of any two data points.

[0068] When determining the similarity between points, in low-dimensional problems (such as two-dimensional clustering), Euclidean distance is usually used as the most intuitive metric. However, when the dimension of data features increases, Euclidean distance becomes unreliable. Close to data point It can be said that they are actually close, but this may not be reliable. However, if they also have many of the same nearest neighbors, it can be more confidently said that they are close, and their shared neighbors can confirm this similarity. Therefore, in one or more embodiments of this specification, the nearest neighbor set of each data point is obtained to define an adjacency matrix based on the nearest neighbor sets of any two data points, which is implemented by the following process:

[0069] First, obtain the nearest neighbor set of each data point. For example, suppose Represents data points The nearest neighbor set of Not included , then the pairwise similarity between any two data points is determined based on the intersection of their nearest neighbor sets. and The set of shared neighbors between , through the collection to measure the pairwise similarity between two data points. Considering The point in The closeness of is not exactly the same, so we need to use weighted pairwise similarity. So at this time, we assume yes No. Neighbor, yes No. neighbors, the pairwise similarity needs to be processed according to the preset formula to obtain the pairwise similarity before normalization. The specific calculation formula is as follows:

[0070] (2)

[0071] in, Yes The number of all nearest neighbors, Represents the pairwise similarity before normalization. Then the pairwise similarity before normalization is linearly normalized to obtain the normalized pairwise similarity. The formula is as follows:

[0072] (3)

[0073] The normalized pairwise similarities are sorted to construct the adjacency matrix of the data points:

[0074] (4)

[0075] In this matrix Line The values ​​in the columns represent data points and Since the nearest neighbors of the same point are exactly the same and meaningless, we set the diagonal values ​​in the matrix to 0.

[0076] S104: Determine an undirected weighted graph corresponding to the data point, so as to determine the Laplacian matrix of the current migrating insect swarm according to the pairwise similarities in the adjacency matrix.

[0077] In order to achieve cluster segmentation of the current migrating insect swarm, we can regard all data as an undirected weighted graph, that is, determine the undirected weighted graph corresponding to the data point, and then determine the Laplacian matrix of the current migrating insect swarm based on the pairwise similarity in the adjacency matrix. Specifically, in one or more embodiments of the present specification, determining the undirected weighted graph corresponding to the data point to determine the Laplacian matrix of the current migrating insect swarm based on the pairwise similarity in the adjacency matrix specifically includes the following process:

[0078] All data points of the current migrating insect swarm are taken as an undirected weighted graph. For example, for such a Plot of data points , we use the set of points and the set of edges To describe, that is . is the set of all data points in the dataset, that is Since we have generated the adjacency matrix using SNN distance , the weight of the edge is described by the SNN distance. At this time, the sum of the pairwise similarities of each data point in the undirected weighted graph is called the degree of the point, that is:

[0079] (5)

[0080] Summarizing the degrees of each data point, the degree matrix is:

[0081] (6)

[0082] According to the difference between the degree matrix and the adjacency matrix, the Laplace matrix of the current migrating insect swarm is determined as:

[0083] (7)

[0084] S105: searching the Laplace matrix based on a preset subgraph optimal segmentation function to obtain an optimal subgraph segmentation, so as to segment the data points corresponding to the optimal subgraph segmentation as a segmentation result of the current migrating insect swarm.

[0085] Spectral clustering is achieved by cutting the entire graph into several subgraphs. Our goal is to cut the graph into pieces that are not connected to each other. subgraphs, and the set of points in each subgraph is , for each cutting method, all subgraphs do not overlap, and each data point is divided into a subgraph. ,and . In order to achieve the purpose of clustering, we need to consider the following three optimization goals at the same time: the sum of weights in each subgraph is maximized; the sum of weights between different subgraphs is minimized; the scale of each subgraph is as large as possible. At this time, the embodiment of this specification will search for the Laplacian matrix according to the preset subgraph optimal segmentation function to obtain the optimal segmentation of the subgraph, so that the data point segmentation corresponding to the optimal segmentation of the subgraph is used as the segmentation result of the current migrating insect swarm.

[0086] Specifically, in one or more embodiments of the present specification, searching for the Laplacian matrix based on a preset subgraph optimal segmentation function to obtain the optimal segmentation of the subgraph, so as to segment the data points corresponding to the optimal segmentation of the subgraph as the segmentation result of the current migratory insect swarm, specifically includes:

[0087] Based on the preset optimization function and the optimization goal of the current migrating insect swarm, the preset subgraph optimal segmentation function is defined; wherein the preset optimization function is:

[0088] (8)

[0089] Among them, M is the clustering parameter, Represents the subgraphs corresponding to the clustering parameters, Representative subgraph The complement of For subgraph The modulus of The number of midpoints, ,and , the optimization goal is to find the minimum value of the function.

[0090] for subgraphs, the indicator vector for each subgraph needs to be defined From the definition of indicator vector, we can know that the number of indicator vectors is related to the number of clusters. Equal and indicating vector dimensions and number of data points Equal, the specific definition is as follows:

[0091] (9)

[0092] Its meaning is that if Data points in the subgraph In The value of is as shown in formula (9), and the rest of the positions are 0. It can be proved that the following relationship exists:

[0093] (10)

[0094] Therefore, we have:

[0095] (11)

[0096] Therefore, the problem becomes finding of minimum value. is an orthonormal basis, of The minimum value is The smallest eigenvalue. Then the eigenvector corresponding to the smallest eigenvalue is Arrange and obtain the feature matrix , in order to cluster the feature matrix to obtain cluster partition results The clustering result is matched to the data point to obtain the clustering result of the current migratory insect swarm. It should be noted that the clustering method of this process can be k-means clustering.

[0097] From the above, it can be seen that the eigenvalue of the Laplace matrix is ​​the optimization indicator for testing the subgraph segmentation effect. Therefore, in order to obtain the best clustering parameter, that is, the number of clusters M, the optimal clustering parameter will be determined according to the eigenvalue distribution of the Laplace matrix and the DB index corresponding to the clustering result, and then the segmentation result of the current migratory insect swarm will be determined according to the optimal clustering parameter.

[0098] Further, in one or more embodiments of the present specification, the above-mentioned method of determining the optimal clustering parameter according to the eigenvalue distribution of the Laplace matrix and the DB index corresponding to the clustering result, so as to determine the segmentation result of the current migratory insect swarm based on the optimal clustering parameter, specifically includes the following process:

[0099] The eigenvalues ​​of the Laplace matrix are calculated and sorted. After verification by simulation data and a large amount of measured data, it is found that the optimal number of clusters is It often corresponds to the position where the eigenvalue suddenly increases sharply, which is manifested as the next eigenvalue being more than twice the previous one, that is:

[0100] (12)

[0101] in Representative Since the number of insect swarms is not 1, we take In order to avoid the insect swarm being too scattered, the maximum number of swarms is 8, so . Based on the above-mentioned embodiment of this specification, the next eigenvalue in the Laplace matrix is ​​extracted to be greater than twice the previous eigenvalue. Then all the extracted grouping eigenvalues, that is, all the grouping eigenvalues ​​that meet the conditions are calculated, and the calculation results are set as the number of groups in turn for grouping, and the grouping results corresponding to the number of groups are obtained. The grouping results are evaluated using the DB value, that is, the DB index of the corresponding grouping results is obtained by the following calculation formula:

[0102] (13)

[0103] in, It is The divergence of a cluster (the compactness within the cluster), usually the average distance from all points in the cluster to the cluster center. It is Clusters and The distance between clusters is usually the Euclidean distance between cluster centers. is the total number of clusters. The smaller the DB value, the less adhesion between different insect swarms, the higher the degree of internal aggregation of the insect swarm, and the more in line with the requirements of this method. Therefore, the clustering result with the smallest DB value is taken as the optimal clustering parameter. Therefore, in the embodiment of this specification, the optimal clustering parameter can be determined according to the DB index, and the optimal segmentation of the subgraph with the optimal clustering parameter is used as the segmentation result of the current migratory insect swarm.

[0104] To further prove that the above method can effectively achieve fine segmentation of insect swarms, a preferred embodiment is selected below to illustrate the effectiveness of the method. The four-dimensional features of the data measured by the high-resolution radar are time, height, weight, and wing-flapping frequency. The data is mapped into a four-dimensional coordinate diagram as shown in Figure 3 As shown, in order to facilitate intuitive understanding and display of four-dimensional coordinates, Figure 3 The data are presented in different combinations of three dimensions. Each figure provides partial information about the relationship between the four dimensions, so technicians in the field can Figure 3 From (a) to (d), we can understand the four-dimensional coordinates of the original data in terms of time, height, weight, and wing-flapping frequency. Then, we first use the noise filtering algorithm mentioned above to remove the noise, and the average sparsity distribution is: Figure 2 As shown in the figure, the straight line in the figure is the noise threshold value. Then, the required insect swarm is segmented according to the subsequent steps, and the final insect swarm is as follows Figure 4 As shown, Figure 4 The four-dimensional segmentation results are displayed from different combinations of three dimensions. Those skilled in the art can Figure 4 (a) to (d) determine the optimal parameters selected as , and obvious insect swarms can be seen, and its DB value is calculated to be 0.5978.

[0105] The embodiment of this specification also provides an adaptive segmentation device for migrating insect swarms, the device comprising:

[0106] at least one processor; and,

[0107] a memory communicatively connected to the at least one processor; wherein,

[0108] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can perform any of the above methods.

[0109] The embodiments of this specification also provide a non-volatile storage medium storing computer executable instructions, wherein the computer executable instructions can execute any of the above-described methods.

[0110] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device, equipment, and non-volatile computer storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0111] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0112] The above description is only one or more embodiments of this specification and is not intended to limit this specification. For those skilled in the art, one or more embodiments of this specification may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of one or more embodiments of this specification shall be included in the scope of the claims of this specification.

Claims

1. An adaptive segmentation method for migratory insect swarms, characterized in that: The method comprises: Determine the four-dimensional coordinates of each initial data point based on the four-dimensional features obtained by measuring the current migrating insect swarm using high-resolution radar; wherein the four-dimensional features are time, height, weight, and wing-flapping frequency; Based on the average sparsity corresponding to the four-dimensional coordinates of each initial data point, adaptively remove noise from the initial data point to obtain filtered data points; Obtaining the nearest neighbor set of each data point to define an adjacency matrix based on the nearest neighbor sets of any two data points; wherein the adjacency matrix is ​​constructed based on the pairwise similarity of any two data points, and the diagonal value is 0; Determine an undirected weighted graph corresponding to the data point to determine the Laplacian matrix of the current migrating insect swarm according to the pairwise similarities in the adjacency matrix; Searching the Laplace matrix based on a preset subgraph optimal segmentation function to obtain an optimal segmentation of the subgraph, so as to segment the data points corresponding to the optimal segmentation of the subgraph as a segmentation result of the current migrating insect swarm; The method of adaptively removing noise from the initial data points based on the average sparsity corresponding to the four-dimensional coordinates of the initial data points to obtain filtered data points specifically includes: Based on the four-dimensional coordinates of each of the initial data points, obtaining the Euclidean distance between each of the initial data points, so as to obtain a preset number of nearest neighbor data points of each of the initial data points according to the Euclidean distance; Taking the average value of the Euclidean distance between each nearest neighbor data point and the initial data point as the average sparsity of the initial data point, and sorting the average sparsity of each initial data point to obtain the distribution characteristics of the average sparsity of the initial data points; Performing SG filtering on the curve corresponding to the distribution characteristic to determine the noise threshold value according to the intersection of the filtered curve and the preset slope tangent; Comparing the average sparsity of each of the initial data points with the noise threshold value to filter out the initial data points that are higher than the noise threshold value to obtain filtered data points; The nearest neighbor set of each data point is obtained to define an adjacency matrix according to the nearest neighbor sets of any two data points, specifically including: Obtain the nearest neighbor set of each data point, and determine the pairwise similarity between any two data points based on the intersection of the nearest neighbor sets of any two data points; Processing the pairwise similarities according to a preset formula to obtain the pairwise similarities before normalization; Performing linear normalization on the pairwise similarities before normalization to obtain normalized pairwise similarities; Sorting the normalized pairwise similarities to construct an adjacency matrix of the data points; The preset formula is: ; The data points are , Represents data points The nearest neighbor set of Not included , and The set of shared neighbors between , yes No. Neighbor, yes No. Neighbors, Yes The number of all nearest neighbors, Represents the pairwise similarity before normalization.

2. The adaptive segmentation method of migratory insect swarms according to claim 1, characterized in that: Determining an undirected weighted graph corresponding to the data point to determine the Laplacian matrix of the current migrating insect swarm according to the pairwise similarities in the adjacency matrix specifically includes: Taking all data points of the current migrating insect swarm as an undirected weighted graph, and obtaining the sum of pairwise similarities of each data point in the undirected weighted graph; The sum of the pairwise similarities is taken as the degree of the data point, and the degrees of the data points are summarized to obtain a degree matrix; The Laplace matrix of the current migrating insect swarm is determined according to the difference between the degree matrix and the adjacency matrix.

3. The adaptive segmentation method of migratory insect swarms according to claim 1, characterized in that: The searching of the Laplacian matrix based on the preset subgraph optimal segmentation function to obtain the subgraph optimal segmentation, so as to segment the data points corresponding to the subgraph optimal segmentation as the segmentation result of the current migrating insect swarm, specifically includes: Based on the preset optimization function and the optimization goal of the current migrating insect swarm, a preset subgraph optimal segmentation function is defined; wherein the preset optimization function is: , where M is the clustering parameter, Represents the subgraphs corresponding to the clustering parameters, Representative subgraph The complement of For subgraph The modulus of The number of midpoints, ,and , P is a set of data points, and the optimization goal is to find the minimum value of the function; Find the M minimum eigenvalues ​​of the Laplacian matrix according to the preset subgraph optimal segmentation function, and determine the optimal segmentation of the M subgraphs; Arranging the eigenvectors corresponding to the minimum eigenvalue to obtain a characteristic matrix, so as to cluster the characteristic matrix to obtain a cluster division result, and corresponding the cluster division result to the data point to obtain a clustering result of the current migratory insect swarm; According to the eigenvalue distribution of the Laplace matrix and the DB index corresponding to the clustering result, the optimal clustering parameter is determined, so as to determine the segmentation result of the current migratory insect swarm based on the optimal clustering parameter.

4. The adaptive segmentation method of migratory insect swarms according to claim 3, characterized in that: Determining an optimal clustering parameter according to the eigenvalue distribution of the Laplace matrix and the DB index corresponding to the clustering result, so as to determine the segmentation result of the current migratory insect swarm based on the optimal clustering parameter, specifically includes: Calculating and sorting the eigenvalues ​​of the Laplacian matrix to extract the clustering eigenvalues ​​whose next eigenvalue in the Laplacian matrix is ​​greater than twice the previous eigenvalue; Calculating the grouping characteristic value to use the calculation result as the grouping number, and obtaining the grouping result corresponding to the grouping number; The DB index of the corresponding clustering result is obtained to determine the optimal clustering parameter according to the DB index, and the subgraph optimal segmentation of the optimal clustering parameter is used as the segmentation result of the current migratory insect swarm.

5. The adaptive segmentation method of migratory insect swarms according to claim 1, characterized in that: The method of determining the four-dimensional coordinates of each initial data point based on the four-dimensional features obtained by measuring the current migrating insect swarm using a high-resolution radar specifically includes: The four-dimensional features of the current migrating insect swarm are obtained by measuring the four-dimensional features using the four-dimensional features as the initial four-dimensional coordinates of the initial data points; The initial four-dimensional coordinates are obtained and normalized based on a minimum-maximum normalization method to determine the four-dimensional coordinates of each initial data point.

6. An adaptive segmentation device for migrating insect swarms, characterized in that: The device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute any one of the methods described in claims 1-5.

7. A non-volatile storage medium storing computer executable instructions, characterized in that: The computer executable instructions can execute the method according to any one of claims 1 to 5.

Images