Group target grouping method based on algebraic graph theory

Abstract

The invention discloses a group target grouping method based on an algebraic graph theory, which belongs to the technical field of information processing, and comprises the following steps: continuously receiving state update data of all targets; according to the grouping period T, target data of the current period are obtained, the target grouping similarity is calculated, and a grouping matrix A of the current period is constructed; performing row-column exchange operation on the grouping matrix A to form a matrix A '; performing target group identification on the matrix A '; comparing the target group identifier of the last period T-1, judging the merging, splitting and maintaining states of the target group, and recording the target group identifier of the current period; and repeating the processes from S2 to S5 until the interruption exits. The group target grouping method based on the algebraic graph theory has the advantages of being easy to implement and understand and accurate in grouping.

Description

Technical Field

[0001] This invention relates to the field of information processing technology, and in particular to a group target grouping method based on algebraic graph theory. Background Technology

[0002] Target formation identification is a crucial application problem in situation estimation, and target clustering is one of the important techniques for achieving target formation identification. Existing target clustering methods mainly fall into two categories: clustering methods and template methods. Most clustering algorithms introduce uncertainty into the target clustering results due to the random given initial state. For example, the K-means algorithm requires specifying the final number of clusters k in advance, and the Chameleon algorithm requires a given merging threshold. These limitations affect the continuity and real-time performance of the clustering results. Template-based target clustering algorithms have the advantages of high clustering accuracy and strong interpretability of clustering results, but their disadvantage is that formulating the template is very complex and difficult, making this method difficult to generalize and apply.

[0003] Target clustering is essentially a process of periodically and dynamically clustering target data that changes over time, involving operations such as cluster maintenance, splitting, and merging. Most existing target clustering methods are static solutions, suffering from long clustering times and high storage consumption, making them unsuitable for real-time situational target clustering applications. Summary of the Invention

[0004] The purpose of this invention is to provide a grouping method for target grouping based on algebraic graph theory, so as to solve the problems existing in the background technology.

[0005] To achieve the above objectives, this invention provides a group objective grouping method based on algebraic graph theory, comprising the following steps:

[0006] S1. Continuously receive status update data from all targets;

[0007] S2. According to the clustering period T, obtain the target data of the current period, calculate the target cluster similarity, and construct the clustering matrix A of the current period;

[0008] S3. Perform row and column exchange operations on the grouping matrix A to form matrix A';

[0009] S4. Identify the target group for matrix A';

[0010] S5. Compare the target group identifier with the target group identifier of the previous period T-1, determine the merging, splitting and maintenance status of the target group, and record the target group identifier of this period.

[0011] S6. Repeat the process from S2 to S5 until the process is interrupted and exited.

[0012] Preferably, when calculating target group similarity, a hybrid calculation based on continuous and discrete attributes is used, and the similarity between targets is measured by distance. and target The distance between them is defined as:

[0013]

[0014] in, It is the distance of the continuous attribute portion of the target; β is the distance to the discrete attribute portion of the target; β is the weight of the discrete attribute portion. and Representing the target and target The value of the kth consecutive attribute, k∈[1,c]; and Representing the target and target The value of the l-th discrete attribute item, l∈[1,b]; where:

[0015]

[0016] Similarity is defined as follows:

[0017]

[0018] In the formula, Similarity for continuous attributes; The similarity of discrete attributes is defined as follows:

[0019]

[0020] Where, β l Weights for each discrete attribute.

[0021] Preferably, the content of the clustering matrix constructed in S2 is as follows:

[0022] Assume there are N targets Then matrix A is represented as:

[0023]

[0024] in, σ is the number of the Nth target; ij For the goal and target Similarity of groups.

[0025] Preferably, the S3 operation rules are as follows:

[0026] Starting from the element in the 2nd row and 3rd column of matrix A, if σ 12 =1, then the rows and columns of matrix A remain unchanged; if σ 12 =0, then find σ sequentially. 1j Element, if σ 1j If σ = 1, then the elements in the 3rd column and the jth column are swapped, and the elements in the 3rd row and the jth row are also swapped; if σ 1j If the value is 0, then no row or column swapping is performed. Then, starting from the element in the 2nd row and 4th column of matrix A, the above matrix element value judgment and row-to-row and column-to-column swapping are performed, and so on until the (N+1)th column is found.

[0027] Preferably, the process of identifying the target group in S4 is as follows:

[0028] In matrix A', starting from the element in the 2nd row and 3rd column, if σ 1j If the values ​​are consecutively 1, then the targets corresponding to the row and column are a group. That is, if different targets are a group, then the local matrix formed by the elements of their row and column is the identity matrix. This process continues until all elements have been traversed. If the identity matrix contains only one element, it means that the target it corresponds to is a separate group.

[0029] Preferably, the criteria for determining group splitting in S5 are as follows:

[0030] When the number of group members decreases in the clustering matrix of two adjacent periods, the group is in a split state. There are two possible outcomes for the number of group members: splitting into two or more independent groups; or the split group members merging into other groups.

[0031] The criteria for determining whether to merge groups are as follows:

[0032] When the number of group members increases in the clustering matrix of two adjacent periods, the group is in a merging state. There are three cases for the increase of group members: the first case is that two previously independent groups merge into one group; the second case is that a newly emerging target merges into the group; and the third case is that members split off from other groups merge into the group.

[0033] The criteria for determining the maintenance state are as follows:

[0034] If the group members remain unchanged in a clustering matrix for three consecutive periods, then the group has reached a stable state and is in a state of maintenance.

[0035] Therefore, the present invention employs the above-mentioned grouping method for group objectives based on algebraic graph theory, which has the following beneficial effects:

[0036] (1) When forming the target group matrix, it is mainly constructed by the similarity between pairs of targets or pairs of target groups. The selection of similarity calculation parameters is closely related to the actual application. In the actual application process, appropriate target attributes can be selected for calculation according to different target grouping rules. At the same time, different weights can be added to different attribute items to improve the accuracy of grouping.

[0037] (2) It is easy to implement, easy to understand and accurate in grouping, and has good real-time performance. The separation of groups can be directly obtained through the grouping matrix, which is suitable for target grouping calculation when the number of targets is large.

[0038] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0039] Figure 1 This is a flowchart illustrating a group objective grouping method based on algebraic graph theory according to the present invention.

[0040] Figure 2 This is a schematic diagram of the target space distribution in a simulation scene according to an embodiment of the present invention. Detailed Implementation

[0041] The following detailed description of embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0042] Please see Figure 1 A group objective grouping method based on algebraic graph theory includes:

[0043] Assume there are N targets Set the clustering period to time T, and the target The data update cycle is t, and the clustering cycle and the target data update cycle can be the same or different.

[0044] S1. Continuously receive status update data from all targets.

[0045] S2. According to the clustering period T, obtain the target data of the current period, calculate the target cluster similarity, and construct the clustering matrix A of the current period.

[0046] When calculating target group similarity, a hybrid calculation based on continuous and discrete attributes is used. Distance is used to measure the similarity between targets. and target The distance between them is defined as:

[0047]

[0048] in, It is the distance of the continuous attribute portion of the target; β is the distance to the discrete attribute portion of the target; β is the weight of the discrete attribute portion. and Representing the target and target The value of the kth consecutive attribute, k∈[1,c]; and Representing the target and target The value of the l-th discrete attribute item, l∈[1,b]; where:

[0049]

[0050] Similarity is defined as follows:

[0051]

[0052] In the formula, Similarity for continuous attributes; The similarity of discrete attributes is defined as follows:

[0053]

[0054] Where, β l Weights for each discrete attribute.

[0055] The clustering matrix is ​​constructed as follows:

[0056] Assume there are N targets Then matrix A is represented as:

[0057]

[0058] in, σ is the number of the Nth target; ij For the goal and target Similarity of groups.

[0059] When i = j, σ ij =1, all elements on the diagonal of matrix A are 1, which means that the target itself and itself definitely belong to the same group;

[0060] In matrix A, by σ ij The constituent elements are only 0 and 1, where 1 indicates that the targets corresponding to the row and column are in the same group, and 0 indicates that the targets corresponding to the row and column are not in the same group;

[0061] Matrix A is a symmetric matrix.

[0062] S3. Perform row and column exchange operations on the grouping matrix A to form matrix A'.

[0063] The S3 operation rules are as follows:

[0064] Starting from the element in the 2nd row and 3rd column of matrix A, if σ 12 =1, then the rows and columns of matrix A remain unchanged; if σ 12 =0, then find σ sequentially. 1j Element, if σ 1j If σ = 1, then the elements in the 3rd column and the jth column are swapped, and the elements in the 3rd row and the jth row are also swapped; if σ 1j If the value is 0, then no row or column swapping is performed. Then, starting from the element in the 2nd row and 4th column of matrix A, the above matrix element value judgment and row-to-row and column-to-column swapping are performed, and so on until the (N+1)th column is found.

[0065] S4. Identify the target group for matrix A'.

[0066] The process of identifying the target group is as follows:

[0067] In matrix A', starting from the element in the 2nd row and 3rd column, if σ 1j If the values ​​are consecutively 1, then the targets corresponding to the row and column are a group. That is, if different targets are a group, then the local matrix formed by the elements of their row and column is the identity matrix. This process continues until all elements have been traversed. If the identity matrix contains only one element, it means that the target it corresponds to is a separate group.

[0068] S5. Compare the target group identifier with the target group identifier of the previous period T-1, determine the merging, splitting and maintenance status of the target group, and record the target group identifier of this period.

[0069] The criteria for determining group splitting are as follows:

[0070] When the number of group members decreases in the clustering matrix of two adjacent periods, the group is in a split state. There are two possible outcomes for the number of group members: splitting into two or more independent groups; or the split group members merging into other groups.

[0071] The criteria for determining whether to merge groups are as follows:

[0072] When the number of group members increases in the clustering matrix of two adjacent periods, the group is in a merging state. There are three cases for the increase of group members: the first case is that two previously independent groups merge into one group; the second case is that a newly emerging target merges into the group; and the third case is that members split off from other groups merge into the group.

[0073] The criteria for determining the maintenance state are as follows:

[0074] If the group members remain unchanged in a clustering matrix for three consecutive periods, then the group has reached a stable state and is in a state of maintenance.

[0075] S6. Repeat the process from S2 to S5 until the process is interrupted and exited.

[0076] Example:

[0077] The simulation scenario assumes 10 groups of targets, all belonging to the same party in terms of their affiliation. From the scenario setup, the targets... and Set it to a group, and set the target and Set as a group, target and It is a group, with a target They are a separate group. During the simulation design, these 10 groups of targets move at a set speed, and their state data is output according to a time period T. At time T... k The spatial distribution of the target location at any given time is as follows Figure 2 As shown, the trailing trajectory behind the target point represents the target's historical flight path.

[0078] Assuming that the similarity of target clusters is calculated solely based on target location attributes, the aim is to verify the correctness of the target clustering matrix algorithm. The following clustering matrix is ​​constructed:

[0079]

[0080] For matrix and After performing row-by-row and column-by-column swaps, the following matrix is ​​formed:

[0081]

[0082] sequentially on the matrix and and and and After performing row-by-row and column-by-column swaps, the following matrix is ​​formed:

[0083]

[0084] After several row-to-row and column-to-column swaps, following rule S3 until no further swaps are possible, the final matrix shows that, according to S4, the target... and Set it to a group, and set the target and Set as a group, target and It is a group, with a target It is a separate group.

[0085] For target cluster maintenance, the cluster matrix construction and calculation described above need to be performed again in the next cluster calculation cycle. Then, by comparing the cluster matrices from the two cycles, the maintenance, classification, and merging status of the target cluster can be determined. Figure 2 In the simulation scenario design shown, the target After k+n cycles, it gradually approaches the target. and They are combined into a single formation, which allows for verification of target cluster maintenance.

[0086] Therefore, the present invention adopts the above-mentioned group target grouping method based on algebraic graph theory, which has the characteristics of being easy to implement, easy to understand, and accurate in grouping. Moreover, it has good real-time performance, and the separation of groups can be directly obtained through the grouping matrix, making it suitable for target grouping calculation when the number of targets is large.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A grouping method for target groups based on algebraic graph theory, characterized in that, Includes the following steps: S1. Continuously receive status update data from all targets; S2. According to the clustering period T, obtain the target data of the current period, calculate the target cluster similarity, and construct the clustering matrix A of the current period; S3. Perform row and column interchange operations on the grouping matrix A to form matrix A''. ' ; S4. For matrix A ' Target group identification; S5. Compare the target group identifier with the target group identifier of the previous period T-1, determine the merging, splitting and maintenance status of the target group, and record the target group identifier of this period. S6. Repeat the process from S2 to S5 until the process is interrupted and exited.

2. The grouping method for target grouping based on algebraic graph theory according to claim 1, characterized in that: When calculating target group similarity, a hybrid calculation based on continuous and discrete attributes is used. Distance is used to measure the similarity between targets. and target The distance between them is defined as: in, It is the distance of the continuous attribute portion of the target; β is the distance to the discrete attribute portion of the target; β is the weight of the discrete attribute portion. and Representing the target and target The value of the kth consecutive attribute, k∈[1,c]; and Representing the target and target The value of the l-th discrete attribute item, l∈[1,b]; where: Similarity is defined as follows: In the formula, Similarity for continuous attributes; The similarity of discrete attributes is defined as follows: Where, β l Weights for each discrete attribute.

3. The grouping method for target grouping based on algebraic graph theory according to claim 2, characterized in that, The grouping matrix is ​​constructed in S2 as follows: Assume there are N targets Then matrix A is represented as: in, σ is the number of the Nth target; ij For the goal and target Similarity of groups.

4. The grouping method for target grouping based on algebraic graph theory according to claim 3, characterized in that, The S3 operation rules are as follows: Starting from the element in the 2nd row and 3rd column of matrix A, if σ 12 =1, then the rows and columns of matrix A remain unchanged; if σ 12 =0, then find σ sequentially. 1j Element, if σ 1j If σ = 1, then the elements in the 3rd column and the jth column are swapped, and the elements in the 3rd row and the jth row are also swapped; if σ 1j If the value is 0, then no row or column swapping is performed. Then, starting from the element in the 2nd row and 4th column of matrix A, the above matrix element value judgment and row-to-row and column-to-column swapping are performed, and so on until the (N+1)th column is found.

5. The grouping method for target grouping based on algebraic graph theory according to claim 4, characterized in that, The process of identifying the target group in S4 is as follows: In matrix A', starting from the element in the 2nd row and 3rd column, if σ 1j If the values ​​are consecutively 1, then the targets corresponding to the row and column are a group. If different targets are in a group, then the local matrix formed by the elements of the row and column they belong to is the identity matrix. This process continues until all elements have been traversed. If there is an element in the identity matrix, it means that the target it corresponds to is a separate group.

6. The grouping method for target grouping based on algebraic graph theory according to claim 5, characterized in that: The criteria for determining group splitting in S5 are as follows: When the number of group members decreases in the clustering matrix of two adjacent periods, the group is in a split state. The criteria for determining whether to merge groups are as follows: When the number of group members increases in the clustering matrix of two adjacent periods, the group is in a merging state. The criteria for determining the maintenance state are as follows: If the group members remain unchanged in a clustering matrix for three consecutive periods, then the group has reached a stable state and is in a state of maintenance.

Images