Clustering hybrid accumulation optimization method for multi-unmanned aerial vehicle radar target signal detection

Abstract

The invention discloses a clustering hybrid accumulation optimization method for multi-unmanned aerial vehicle radar target signal detection, which is applied to the technical field of radar signals, and aims to solve the problems that in the multi-unmanned aerial vehicle radar detection process, part of unmanned aerial vehicles are sparsely distributed in space, full-coherent accumulation cannot be realized, and full-non-coherent accumulation is relatively low in signal-to-noise ratio gain. According to the method, a clustering hybrid accumulation processing architecture (coherent accumulation in clusters and non-coherent accumulation among clusters) is constructed, the calculation amount and hybrid accumulation signal-to-noise ratio gain in the hybrid accumulation process are analyzed and calculated, a clustering hybrid accumulation optimization model is established according to the two indexes, and then a meta-heuristic bidirectional clustering processing method is designed. And thus, an optimized clustering processing result is obtained, effective accumulation of echo signals is realized, and finally, target detection is realized through a CFAR (Constant False Alarm Rate) detector.

Description

Technical Field

[0001] The present invention belongs to the technical field of radar signals, and particularly relates to a radar echo signal hybrid accumulation technology. Background Art

[0002] The multi-UAV-borne radar system is based on the UAV platform and uses multiple UAV-borne radars to simultaneously receive the signals of a certain emission source to realize the accumulation and fusion of multi-channel echo signals. To improve the detection performance of the multi-UAV-borne radar, the key lies in achieving high-gain accumulation of the echoes of the multi-UAV-borne radar; at the same time, it is also necessary to consider reducing the computational amount required for the accumulation process as much as possible to improve the real-time performance of detection. According to the signal accumulation methods, they can be mainly divided into: coherent accumulation and non-coherent accumulation. Coherent accumulation can obtain a higher accumulation gain, but the computational amount is higher than that of non-coherent accumulation, and a certain echo correlation is required. In the UAV-borne radar system, some radar nodes are relatively close, and the target echo correlation is strong, so coherent accumulation can be adopted. While some nodes are relatively far away, and the echo correlation is weak. At this time, non-coherent accumulation is appropriate. Therefore, in order to obtain the best accumulation gain, it is necessary to cluster the radar nodes so that the echoes of the radar nodes within the cluster have a high correlation, which is convenient for coherent accumulation processing. The echo correlation between clusters is low, and non-coherent accumulation processing is appropriate.

[0003] M. Wang et al. proposed a method of inter-channel coherent accumulation, which can be used for in-cluster coherent accumulation. To realize non-coherent accumulation of signals, C. Wang et al. proposed a CS-based inter-channel non-coherent accumulation algorithm, which can be used for inter-cluster non-coherent accumulation. However, the above two methods ignore the problem of how to cluster the nodes to realize the optimized accumulation of signal clustering.

[0004] In terms of clustering the nodes, relevant scholars have studied some algorithms. Fahad designed a clustering method based on the grey wolf algorithm to improve the stability of the vehicle ad hoc network during the communication process. Aadil designed a clustering method based on the K-Means algorithm, which prolongs the life cycle of the UAV network and reduces the routing consumption. However, these methods pay more attention to communication stability or the longest life cycle of the UAV network and do not consider the performance indicators and constraint conditions of radar detection, so they cannot be directly applied to the field of radar signal processing. Therefore, it is of great significance to propose a clustering optimization method for the hybrid accumulation of radar echo signals. Summary of the Invention

[0005] Aiming at the problem of efficient accumulation of the echo signals of the multi-UAV-borne radar, the present invention proposes a clustering hybrid accumulation optimization method for the detection of the target signals of the multi-UAV-borne radar, which has achieved good results in improving the signal-to-noise ratio gain of the hybrid accumulation and reducing the computational amount.

[0006] The technical solution adopted by the present invention is as follows: a clustering hybrid accumulation optimization method for multi-UAV borne radar target signal detection, including:

[0007] S1. Construct a clustering hybrid accumulation architecture, perform coherent accumulation on the received signals of radar nodes within a cluster, and perform non-coherent accumulation on signals between clusters;

[0008] S2. Perform pulse compression processing and multi-pulse accumulation on the received echo signals of multi-UAV borne radar;

[0009] S3. Calculate the computational amount and the signal-to-noise ratio gain of hybrid accumulation in the process of clustering hybrid accumulation of radar echo signals;

[0010] S4. Construct a clustering optimization model with the computational amount and the signal-to-noise ratio gain of hybrid accumulation as indicators;

[0011] S5. Design a clustering processing method based on meta-heuristic bidirection to solve the clustering optimization problem and obtain the optimized clustering processing result;

[0012] S6. Perform clustering hybrid accumulation processing according to the clustering result to obtain the output result after hybrid accumulation of multi-UAV borne radar echoes;

[0013] S7. Use the output result obtained by clustering hybrid accumulation to perform CFAR detection, and finally obtain the detection result.

[0014] The beneficial effects of the present invention: The present invention constructs a clustering hybrid accumulation processing architecture (coherent accumulation within a cluster and non-coherent accumulation between clusters), analyzes the computational amount and the signal-to-noise ratio gain of hybrid accumulation in the calculation process, and establishes a clustering hybrid accumulation optimization model based on these two indicators. Furthermore, a clustering processing method based on meta-heuristic bidirection is designed to obtain the optimized clustering processing result and realize the effective accumulation of echo signals. Finally, target detection is achieved through a CFAR detector. Description of the Drawings

[0015] Figure 1 is the flow chart of the embodiment of the present invention.

[0016] Figure 2 is the hybrid accumulation processing architecture;

[0017] Figure 3 is the flow chart of the meta-heuristic bidirectional clustering algorithm;

[0018] Figure 4 is the initial position of the UAV borne radar;

[0019] Figure 5 is the iteration curve of the meta-heuristic bidirectional clustering algorithm;

[0020] Figure 6 is the clustering result of the UAV borne radar;

[0021] Among them, Figure 6 (a) is the clustering result of the meta - heuristic bidirectional clustering algorithm, Figure 6 (b) is the clustering result when the number of cluster heads is set to 3, Figure 6 (c) is the clustering result when the number of cluster heads is set to 6, Figure 6 (d) is the clustering result when the number of cluster heads is set to 9;

[0022] Figure 7 This is the bar chart comparing the performance of the meta - heuristic bidirectional clustering algorithm and the Kmeans algorithm of the present invention;

[0023] Figure 8 It is the detection performance curve. Detailed implementation method

[0024] The present invention is verified by means of Matlab simulation experiments, and the correctness and effectiveness of the present invention are verified on the scientific computing software Matlab R2021b. The embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0025] As Figure 1 shown, the present invention proposes a clustering hybrid accumulation optimization method for multi - UAV - borne radar target signal detection, which is specifically implemented through the following processes:

[0026] S1. Construct a clustering hybrid accumulation architecture, perform coherent accumulation on the received signals of radar nodes within the cluster, and perform non - coherent accumulation on the signals between clusters;

[0027] S2. Perform pulse compression processing and multi - pulse accumulation on the received echo signals of multi - UAV - borne radar;

[0028] S3. Establish a model to calculate the computational amount and accumulation gain of the signal in the clustering hybrid accumulation;

[0029] S4. Construct a clustering optimization model with the computational amount of the hybrid accumulation process and the signal - to - noise ratio gain of the hybrid accumulation as indicators;

[0030] S5. Design a clustering processing method based on meta - heuristic bidirection to solve the clustering optimization model and obtain the optimized clustering processing result;

[0031] S6. Perform clustering hybrid accumulation according to the clustering result to obtain the signal after hybrid accumulation;

[0032] S7. Use the signal obtained by clustering hybrid accumulation to perform CFAR (Constant False Alarm Rate Detector) detection, and finally obtain the detection result.

[0033] Step S1, in a multi-UAV radar system composed of a transmitting source and N UAV nodes, the UAV nodes act as receivers of radar signals, and cooperate to receive and process signals to improve detection performance. As Figure 2 shown, based on the multi-UAV radar system, the present invention constructs a clustered hybrid accumulation processing architecture. If the clustering method obtained by solving the clustering model in step S4 divides N UAVs into M clusters, where N is greater than or equal to M, and there are n i UAV nodes in the i-th cluster. The echo signal received by the j-th UAV in the i-th cluster is accumulated between pulses to obtain a signal Then, coherent accumulation is performed on the UAVs in the i-th cluster to obtain a signal Non-coherent accumulation is performed between clusters, and finally CFAR detection is performed according to the accumulation result.

[0034] Step S2, in the multi-UAV radar system, the signal transmitted by the external radiation source is

[0035] s(t) = u(t)exp(j2πf c t)

[0036] where u(t) is the baseband signal, t is the fast time, and f c is the carrier frequency of the transmitted signal.

[0037] During the radar observation time T of the multi-UAV-borne radar, the radar receives a total of K pulses. The echo signal received by the j-th UAV in the i-th cluster for the k-th pulse can be expressed as:

[0038]

[0039] where A ij represents the complex amplitude of the echo after the transmitted signal is reflected by the target and reaches the receiving node of the j-th UAV in the i-th cluster, and k = 1, 2, 3,..., Ν represents the k-th pulse. After down-conversion processing, the echo signal is:

[0040]

[0041] The initial radial distance of the external radiation source relative to the target is The initial radial distance of the j-th UAV in the i-th cluster relative to the target is The radial moving speed v of the external radiation source relative to the target e , and the initial radial speed of the j-th UAV in the i-th cluster relative to the target is v ij . Then the distance that the signal transmitted by the external radiation source travels to reach the target is:

[0042]

[0043] where Tp is the pulse repetition time.

[0044] The distance that the external radiation source signal reflects from the target to the UAV is:

[0045]

[0046] where R ij is the radial distance of the target relative to the external radiation source, and v UVA is the radial velocity of the target relative to the external radiation source. The time delay τ ij (k) from signal transmission to reception is:

[0047]

[0048] where c represents the speed of light.

[0049] After performing matched filtering on the echo signal, the signal obtained is:

[0050]

[0051] where ξ represents the fast-time symbol in the integration, and h(t) is the matched filter of the transmitted signal.

[0052] Performing FFT processing on the matched-filtered echo along the slow-time dimension, the multi-pulse accumulation processing result s ij p (f,k) is obtained.

[0053] Step 3, calculate the computational amount and accumulation gain of the signal in the cluster hybrid accumulation. During the hybrid accumulation process, when performing intra-cluster channel alignment, the search amount for range alignment is The search amount for Doppler alignment is The computational amount for phase alignment between the searched range offset and Doppler offset is c p , then the computational amount of coherent accumulation of the i-th cluster with the j-th node as the reference can be expressed as:

[0054]

[0055] where n i is the number of nodes included in the i-th cluster. Among the M clusters, with the cluster head of the m k -th cluster as the reference, when performing inter-cluster non-coherent accumulation, no phase alignment is required, and its computational cost can be expressed as:

[0056]

[0057] In the hybrid accumulation, N nodes are divided into M clusters, and the computational amount of the hybrid accumulation is

[0058]

[0059] Clustering method {n 1 ,n 2 ,…,n m The mixed accumulation gain under} can be expressed as

[0060]

[0061] (1-0.140lg(N ncl )+0.018310(lg(N ncl )) 2 )

[0062] Among them, P D is the required detection probability, P fa is the false alarm rate, N ncl For clustering method {n 1 ,n 2 ,…,nm}, the total number of equivalent nodes is:

[0063]

[0064] Among them, f(n i ,P fa ,P D ) is:

[0065]

[0066] The expression of ζ is:

[0067]

[0068] The expression of ψ is:

[0069]

[0070] Step 4: An optimization model is established based on the two indicators of mixed accumulation calculation amount and signal-to-noise ratio gain. a Weighted summation of mixed accumulation calculation amount CA ω 1 CA+ω 2 G a As the objective function, the maximum distance between two node echo signals that meets the coherence condition is taken as the node distance constraint within the cluster to construct the optimization model. The optimization model is expressed as:

[0071] max W=ω 1 CA+ω 2 G a

[0072]

[0073] k = 1, 2, 3,....., M

[0074] i, j = 1, 2, 3,......, n k

[0075] Among them, ω 1 is the weight of the mixed accumulation calculation amount index, and ω 2 is the weight of the mixed accumulation signal-to-noise ratio gain index, satisfying ω 1 + ω 2 = 1. is the distance between the i-th UAV and the j-th UAV in the k-th cluster, M is the number of divided UAV clusters, and N k is the number of UAVs in the k-th cluster, is the maximum distance for the echo signals of two nodes to meet the coherent condition. The mixed accumulation signal-to-noise ratio gain and calculation amount in the objective function are normalized indexes.

[0076] Step 5. For the optimized model of cluster-based mixed accumulation, design a meta-heuristic two-way clustering algorithm to optimize a better clustering result. This algorithm takes the position, speed of the UAV-borne radar, and the prior information of the target as inputs. As Figure 3 shown, the algorithm mainly includes three parts: initialization operation, fusion processing, and splitting processing. Among them, the initialization operation obtains the number N of UAV-borne radar nodes and generates an initial matrix C of the clustering result:

[0077] C = [1, 2, 3…, N] T

[0078] The superscript T represents transpose;

[0079] The m-th row element in the clustering matrix C represents the number of the UAV-borne radar node assigned to the m-th cluster. At this time, the initialized clustering matrix means that each of the current N UAV nodes forms a cluster by itself.

[0080] The fusion processing selects two clusters from all clusters for fusion and operates in the direction of turning zero into an integer; the splitting processing selects one cluster from all clusters and splits each node in it into independent clusters and operates in the direction of turning an integer into zero. The algorithm randomly performs fusion and splitting processing during the iteration to iterate out a better clustering result. The specific operations of the fusion processing and the splitting processing are as follows:

[0081] Step 5.1 Fusion processing:

[0082] In the fusion processing, in order to improve the optimization ability of the algorithm, two clusters are selected from all clusters for fusion, and the fusion objective value matrix is calculated according to the objective function, and further the clustering objective value change matrix is calculated:

[0083] △Va(i, j) = Va(i, j) - Va′

[0084] Among them, Va(i, j) represents the clustering target value after the fusion of the i-th and j-th clusters, and Va′ represents the clustering target value in the previous generation.

[0085] When there are elements greater than 0 in the clustering target change matrix, the fusion probability matrix is:

[0086]

[0087] Otherwise, set the fusion probability matrix as:

[0088]

[0089] Among them, V + is the positive element matrix of the clustering target value change matrix, expressed as:

[0090]

[0091] M' is the current number of clusters.

[0092] The fusion probability matrix calculates the probability of each fusion method being selected in this round of iteration. In the fusion process, two clusters to be fused in this processing are selected according to the fusion probability matrix. Convert the fusion probability matrix into a fusion probability distribution matrix:

[0093]

[0094] Among them, J is the index of the last element in the second dimension of the fusion probability distribution matrix. Generate a random number rand uniformly distributed between 0 and 1, and search for the index [i r , j r that satisfies:

[0095]

[0096] Then merge the i r -th cluster and the j r -th cluster into one cluster. Compare the clustering result with the clustering result of the previous generation and retain the better clustering result.

[0097] Step 5.2 Splitting process:

[0098] In the splitting process, randomly select several clusters with the number of points greater than 2 for splitting. According to the greedy strategy, let the current best clustering method be retained as the iteration progresses, and set the optimal solution set of the current generation not to participate in splitting. The splitting process will be carried out according to the quality of each cluster, and the poorer clusters have a greater probability of being split. Calculate the index values of all M clusters in the current clustering method through the objective function, and obtain the index value vector V of the M clusters T, the index value vector reflects the quality of each cluster under the evaluation system given by the objective function, and the index value vector is normalized:

[0099]

[0100] Calculate the splitting probability vector according to the index value vector:

[0101]

[0102] Among them, there is Calculate the splitting probability distribution vector

[0103]

[0104] Generate a random number rand uniformly distributed between 0 and 1, and find the index m r Satisfy

[0105]

[0106] Then split each node in the mth r cluster into separate clusters.

[0107] After the above processing, a clustering result can be finally obtained, and this clustering result has achieved good results in improving the mixing accumulation signal-to-noise ratio gain and reducing the computational complexity.

[0108] Step 6, perform signal clustering mixing accumulation according to the final clustering result. Coherently accumulate the echo signal s ij p (f,k) received by the nodes within the cluster to obtain the intra-cluster accumulation signal, and further perform non-coherent accumulation on the intra-cluster accumulation signals of each cluster to obtain the final mixing accumulation result.

[0109] Step 7, perform CFAR detection according to the mixing accumulation result to obtain the final detection result.

[0110] Figure 4 The current position of the airborne radar in the embodiment is given, Figure 5 The iterative curve graph of the clustering result calculated by the meta-heuristic two-way clustering processing method is given. Figure 4 、 Figure 5 The corresponding parameters are shown in Table 1. It can be seen from the figure that the meta-heuristic two-way clustering algorithm can obtain a better clustering result through iteration. Figure 6 For the clustering result of the airborne radar, among them, Figure 6 (a) is the clustering result obtained by the meta-heuristic two-way clustering processing method, Figure 6 (b) is the clustering result when the number of cluster heads is set to 3, Figure 6 (c) is the clustering result when the number of cluster heads is set to 6,Figure 6 (d) shows the clustering result when the number of cluster heads is set to 9. Figure 7 A bar chart comparing the clustering hybrid accumulation performance is given. Figure 8 Detection curves corresponding to different clustering results obtained by the meta-heuristic bidirectional clustering processing method and the Kmeans algorithm are given. Those skilled in the art should know that at a certain signal-to-noise ratio, the higher the detection probability, the higher the signal accumulation gain, indicating better performance of the method.

[0111] Table 1 Parameters

[0112]

[0113] Those of ordinary skill in the art will realize that the embodiments described herein are to assist the reader in understanding the principles of the present invention and should be understood that the scope of protection of the present invention is not limited to such specific statements and embodiments. For those skilled in the art, various changes and modifications can be made to the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the scope of the claims of the present invention.

Claims

1. A clustering hybrid accumulation optimization method for multi-UAV radar target signal detection, characterized in that: include: S1. Construct a clustered hybrid accumulation architecture to perform coherent accumulation of the received signals of the radar nodes within the cluster and non-coherent accumulation of the signals between clusters; S2. Perform pulse compression processing and multi-pulse accumulation on the echo signals received by multiple UAV-mounted radars; S3. Calculate the computational effort and mixed accumulation signal-to-noise ratio gain in the radar echo signal clustering mixed accumulation process; S4. Taking the computational amount of the mixed accumulation process and the mixed accumulation signal-to-noise ratio gain as indicators, a clustering optimization model is constructed; S5. Design a meta-inspired bidirectional clustering processing method to solve the clustering optimization problem and obtain the optimized clustering processing result; S6. Perform clustering mixed accumulation processing according to the clustering results to obtain the output result after mixed accumulation of multiple UAV-mounted radar echoes; S7. Perform CFAR detection using the output results obtained by clustering mixed accumulation to finally obtain a detection result.

2. The clustering hybrid accumulation optimization method for multi-UAV radar target signal detection according to claim 1 is characterized in that: The computational complexity of the mixed accumulation process in step S3 is expressed as: Where N represents the number of nodes, M represents the number of clusters, and ca i represents the computational amount of the coherent accumulation of the ith cluster, ca ncl Represents the computational cost when performing inter-cluster non-coherent accumulation.

3. The clustering hybrid accumulation optimization method for multi-UAV radar target signal detection according to claim 2 is characterized in that: The mixed accumulation signal-to-noise ratio gain in step S3 is expressed as: (1-0.140lg(N ncl )+0.018310(lg(N ncl , 2 , Among them, P D is the required detection probability, P fa is the false alarm rate, N ncl is the total number of equivalent nodes in the current clustering mode.

4. The clustering hybrid accumulation optimization method for multi-UAV radar target signal detection according to claim 3 is characterized in that: The clustering optimization model described in step S4 is expressed as: max W=ω1CA+ω2G a k=1,2,3,.....,M i,j=1,2,3,......,n k Among them, ω1 is the weight of the mixed accumulation calculation amount indicator, ω2 is the weight of the mixed accumulation signal-to-noise ratio gain indicator, ω1+ω2=1, is the distance between the ith UAV and the jth UAV in the kth cluster, n k is the number of drones in the kth cluster, It is the maximum distance at which the echo signals of two nodes meet the coherence condition.

5. The clustering hybrid accumulation optimization method for multi-UAV radar target signal detection according to claim 4 is characterized in that: The implementation process of step S5 includes: S51, obtain the number N of drone-borne radar nodes through initialization operation, and generate a clustering result initialization matrix C: C=[1,2,3…,N] T The m-th row element in the clustering matrix C represents the number of the UAV radar node that is classified into the m-th cluster. At this time, the initialized clustering matrix indicates that the current N UAV nodes each become a cluster; S52, selecting two clusters that need to be fused in this process according to the fusion probability matrix, and the fusion probability determination process is as follows: According to the objective function, the fusion target value matrix is ​​calculated, and the clustering target value change matrix is ​​further calculated: △Va(i,j)=Va(i,j)-Va′ Among them, Va(i,j) represents the clustering target value after the fusion of the i-th and j-th clusters, and Va′ represents the clustering target value in the previous generation; The fusion probability is determined based on the target value change matrix; S53, randomly selecting a number of clusters with a number of points greater than 2, and splitting them according to the splitting probability; The process of determining the split probability is as follows: all the M clusters in the current clustering method are calculated through the objective function to obtain the index value of each cluster, and the index value vector V of the M clusters is obtained. T , according to V T Determine the split probability.

6. The clustering hybrid accumulation optimization method for multi-UAV radar target signal detection according to claim 5 is characterized in that: The fusion probability is determined according to the target value change matrix, specifically: When the clustering target change matrix has elements greater than 0, the fusion probability matrix is: Otherwise, set the fusion probability matrix to: Among them, V + is the positive element matrix of the clustering target value change matrix.

7. The clustering hybrid accumulation optimization method for multi-UAV radar target signal detection according to claim 6 is characterized in that: The split probability calculation formula is: Among them, V T '(m) represents V T The normalized result of .

Images