A multi-array bearing-only positioning method and system based on direction-finding line classification combination

Abstract

The application provides a multi-array pure bearing positioning method and system based on a direction finding line classification combination, the method comprising: obtaining multi-array direction finding data, extracting effective measurements exceeding a threshold value after beam forming; constructing a direction finding combination probability model based on maximum likelihood estimation, combining measurement values and target existence probability, introducing a combination cost function containing angle deviation and correlation coefficient constraints, scoring and optimizing all relevant combinations, and outputting target direction estimation and existence probability; forming a complete candidate set by delimiting an initial candidate set, iteratively retaining low-cost combinations and eliminating supersets thereof; selecting a minimum cost combination through classification combination, and realizing multi-target positioning by using a maximum likelihood method. The application has the advantages that: the method significantly reduces the exponential complexity of the original combination problem through reasonable classification and screening strategies.

Description

Technical Field

[0001] This application belongs to the field of underwater acoustic signal processing and passive positioning technology, specifically relating to a multi-array pure azimuth positioning method and system based on direction finding line classification combination, which is suitable for underwater target positioning in low detection rate and multi-target environments. Background Technology

[0002] In underwater passive acoustic positioning, passive sonar systems do not actively emit acoustic signals but rely solely on receiving target radiated noise for detection. Therefore, they typically only obtain the target's azimuth information and cannot directly measure its distance. Consequently, multi-array pure azimuth positioning has become an important technical approach for underwater passive positioning.

[0003] Existing multi-array pure azimuth localization methods typically require correlating direction-finding observations from multiple arrays to determine whether the direction-finding results from different arrays originate from the same target, and then estimating the target position based on this correlation. However, in practical applications, this type of method generally suffers from the following problems: 1. Low detection rate and prominent missed detection problems. In complex marine environments, due to propagation loss, environmental noise and array performance limitations, the detection results of different arrays for the same target are often incomplete, resulting in some arrays having missed detections in direction finding, which makes the traditional data association model that assumes "every target is observed in every array" invalid.

[0004] 2. Multidimensional allocation problems have high computational complexity. Existing data association methods based on maximum likelihood or multidimensional allocation typically model the problem as a constrained combinatorial optimization problem. This type of problem is NP-hard, and the computational complexity increases rapidly with the number of arrays and targets, making it difficult to meet the real-time requirements of engineering projects.

[0005] 3. Insufficient robustness under model mismatch conditions. When the actual detection rate and direction finding error are inconsistent with the theoretical model, traditional association methods based on joint likelihood are prone to association errors, leading to a significant decrease in positioning accuracy.

[0006] Therefore, there is an urgent need for a method that can achieve efficient and robust data association and pure orientation localization under conditions of low detection rate and multiple targets, so as to improve the engineering practicality of multi-array passive localization systems. Summary of the Invention

[0007] To overcome the above-mentioned shortcomings, this application proposes a multi-array pure azimuth positioning method based on direction-finding line classification combination, including: Step S1: Acquire multi-array direction finding data, perform beamforming on the direction finding data of each node, and acquire measurement data exceeding the set threshold; Step S2: By constructing a direction finding combination probability model based on maximum likelihood estimation, combining node measurement values ​​and target existence probability, introducing a combination cost function to score all possible measurement association combinations, and using angle deviation and correlation coefficient constraints to optimize combination selection, the optimized target direction estimate and its corresponding existence probability are obtained. Step S3: Define the initial candidate set and perform iterative optimization based on the combined cost function, retain the combination with lower cost and remove its superset to form a complete candidate set containing the optimal association hypothesis; Step S4: Classify and combine the complete candidate set to obtain the minimum cost combination; Step S5: Use the maximum likelihood method to locate the combination of different targets and obtain the final multi-target localization result.

[0008] As an improvement to the above-mentioned method, step S2 includes: A joint likelihood model for a multi-node direction finding system is constructed, and constraints on angle deviation and inter-measurement correlation are introduced. A combined cost function is defined to evaluate all feasible measurement correlation hypotheses. The correlation constraint filters the correlation coefficients of measurements within the combined system using a preset threshold and applies penalties accordingly. The cost function used is as follows: ; in, Indicates the number of nodes; Indicates the first The measurement number selected for each node. This indicates that no target was detected. Indicates an indicator function, when The value is 1 if it is true, and 0 otherwise. Indicates the first Standard deviation of direction finding error for each node; This represents the difference between the observed azimuth of the maximum likelihood estimated location and the observed data at that node; The penalty factor is determined based on the minimum correlation coefficient within the combination; Represents a node The observation range; For the first Node detection rate; Iterate through all feasible combinations, select the combination that minimizes the cost function as the optimal association result, and output the corresponding target direction estimate and existence confidence.

[0009] As an improvement to the above-mentioned method, step S3 includes: The initial candidate set is formed by combining the measured values ​​of all nodes. That is, satisfying All combinations , Indicates the first The measurement number selected by each node; let the maximum cost of this set be: ; in, Indicate combination The cost; The set is improved through an iterative process, and then... The potential is processed in sequence. The combination of , for each potential is combination Calculate its cost ,if ,Will Add to candidate set Otherwise, in Find all containing combination If it exists Make Then from Delete these and will join in Last updated .

[0010] As an improvement to the above-mentioned method, step S4 includes: The node with the most measured values ​​is selected as the classification node, and the candidate set is... The combinations in the data are classified according to the measurement values ​​used by the classification nodes, and are divided into a total of 10 categories. kind, Number of measurements for the classification node: ; in, Indicates the category node number; Indicates that the classification node uses the first One measurement value or all combinations of not using any measurement values; From each category Select at most one combination from the given list, ensuring that these combinations do not conflict, and calculate the total cost: ; in, This indicates that no combination is selected for this class, and the corresponding cost is... The final result is obtained by solving the following optimization problem: ; Solve this optimization problem using dynamic programming or a greedy algorithm, and output the set of combinations with the minimum cost as the objective association result.

[0011] As an improvement to the above-mentioned method, step S5 includes: For the set of measurement results obtained in step S4, the maximum likelihood method for single-target localization is used to calculate the nonlinear least squares problem expressed by the following formula: ; in, This represents the single-objective estimation result calculated using the maximum likelihood method; , Indicates the first Standard deviation of direction finding error for each node; For the first The observation angle of each node, express Position relative to the first The angle of each node; The Gauss-Newton iterative method is used to solve this nonlinear least squares problem, and the direction-finding error matrix is ​​set as follows: ; Let the Jacobian matrix be: ; Where T represents the matrix transpose; Represents the gradient; Based on the Gauss-Newton approximation, the weight matrix is ​​set as follows: , Indicates the first The weight of the nth node; then the weight of the nth node. t Increment of the next iteration Satisfies the linear equation: ; The update equation is: ; The estimated location of each target can be obtained by using the above formula.

[0012] This application also provides a multi-array pure azimuth positioning system based on direction-finding line classification and combination, implemented based on the above method, the system comprising: The data acquisition module is used to acquire multi-array direction finding data, perform beamforming on the direction finding data of each node, and acquire measurement data exceeding a set threshold. The optimization module is used to construct a direction finding combination probability model based on maximum likelihood estimation, combine node measurement values ​​with target existence probability, introduce a combination cost function to score all possible measurement association combinations, and use angle deviation and correlation coefficient constraints to optimize combination selection, so as to obtain the optimized target direction estimate and its corresponding existence probability. The candidate set optimization module is used to define the initial candidate set and perform iterative optimization based on the combination cost function, retaining combinations with lower costs and eliminating their supersets to form a complete candidate set containing the optimal association hypothesis; The combination classification module is used to classify and combine the complete candidate set to obtain the minimum cost combination; The localization estimation module is used to locate combinations of different targets using the maximum likelihood method to obtain the final multi-target localization result.

[0013] Compared with existing technologies, the advantages of this application are: This method significantly reduces the exponential complexity of the original combinatorial problem through a reasonable classification and selection strategy. Let the number of sensors be... The average measurement value of each sensor is The complexity of the original problem is then... This results in exponential growth; however, the complexity of this method mainly depends on the iterative process in step three, with a complexity of only [missing information]. . Attached Figure Description

[0014] Figure 1 The diagram shows a flowchart of a multi-array pure azimuth localization method based on the classification and combination of direction-finding lines. Figure 2 The diagram shown is a schematic of multiple array nodes; Figure 3 The image shown is a comparison chart of single-shot localization results; Figure 4 The diagram shows the relative positions of each array under experimental conditions and the direction finding results. Figure 5 The image shows a comparison of the spectra of two targets on three arrays; Figure 6 The image shown is a schematic diagram of the positioning results. Detailed Implementation

[0015] The technical solution of this application will be described in detail below with reference to the accompanying drawings.

[0016] This application provides a multi-array pure azimuth positioning method and system based on direction finding line classification and combination. By performing hierarchical screening, classification and combination and optimal subset selection on multi-array direction finding data, the complexity of data association is reduced and the association accuracy and positioning accuracy are improved in low detection rate environments.

[0017] Multi-array pure azimuth localization methods based on direction-finding line classification combinations include: Step 1) Acquisition and preprocessing of multi-array direction finding data.

[0018] Assuming the number of nodes (i.e., the number of arrays) is , No. Two-dimensional coordinates of each node for: (1) The number of sound sources is , No. Two-dimensional coordinates of a sound source for: (2) All sound sources are represented as: (3) No. The sound source is related to the first The DOA of each node is: (4) The observation results are as follows: (5) in For the first The node is the first DOA observation noise for each target Assuming it follows a zero-mean Gaussian distribution with variance of . . No. There are a total of nodes The total number of measurement data points is expressed as follows: (6) In actual array signal processing, the first... Beamforming is performed at each node, and the measurement data of that node is extracted by limiting a threshold.

[0019] Step 2) Construct the likelihood probability model of the direction finding combination.

[0020] Assumption It is a false measurement. Since there is no prior information, we assume that it follows a uniform distribution. ,in Represents a node Observation range ( ).use Indicates spurious measurements, nodes The measured value is expressed as All measured values ​​are expressed as A set of association results is represented as A set of feasible partitioning results is represented as follows: ,in This indicates that at least two arrays have observed the target. This indicates that at most one array has observed the target. (Definition) For the first The detection rate of nodes, under normal circumstances, is the observation combination. For the same goal The joint likelihood function of the measurements is: (7) in Functions for binary indicator variables. hour In other cases, the value is 1.

[0021] To standardize the likelihood function so that it is independent of the number of measurements from each node and the assumed number of targets, the partition is defined as follows under all possible efficient partitions: Division Corresponding to the assumptions of zero target number and all measurements being spurious targets, the maximum joint likelihood ratio can be given in the following form: (8) In the formula Indicates all possible partitions. The definition is as follows: (9) in Indicates division Next node The number of targets associated with it.

[0022] The formula also contains an unknown term for the location of the target sound source. This term can be replaced with the maximum likelihood estimation result under this assignment. ,So Change to : (10) Take the negative logarithm of the joint likelihood ratio: (11) In the formula: (12) in This represents the difference between the observed azimuth of the maximum likelihood estimated location under this partition and the observed data of that node. Indicates the first The standard deviation of the direction finding error of each array.

[0023] For ease of understanding and operation, the following basic concepts are defined: Combination refers to the combination of at most one measurement value from each sensor to form all measurements of a target, denoted as . ,in , . Indicates the first The sensor does not use any measurement values. Indicates the first The sensor uses the first One measurement value. It is the first The number of measurements from each sensor. Each combination has a unique number, and there is a one-to-one correspondence between the number and the combination.

[0024] combination The potential is defined as the number of non-zero measurements used in the combination: (13) in It is an indicator function: (14) Two combinations and The intersection operation is defined as: (15) in: (16) combination Includes combinations , recorded as If and only if for all ,satisfy and ,Right now The index of all non-zero measurements is in They have the same value.

[0025] Based on the above definitions, equation (12) represents the combination cost under a target combination condition: (17) To simplify computation and improve algorithm robustness, utilize In principle, the angle deviation term in the formula is... Replace with This yields a simplified cost threshold: (18) Only retain those with a cost less than All combinations form the initial set of feasible combinations. All subsequent combinations are selected from this set of feasible combinations. It should be noted that when only two actual observations are selected in a combination, the bias term in equation (17) will... At this point, the calculated combined cost becomes a constant: (19) In this situation, conventional data association methods treat all combinations of two observations as identical, making it impossible to distinguish between them. To address this issue, a correlation coefficient threshold for the two observation combinations is introduced. Only combinations exceeding this threshold are retained. During simulation, the correlation coefficient of two observation combinations from the same target is set to be greater than [a certain threshold]. The value, from different targets, is set to less than The value of is used to simulate the correlation effect of this method. In practical data applications, the correlation coefficient of spectral features can be used as the input value, and the threshold value can be used as the input value. Adjust the values ​​according to the actual application environment. When calculating the combined cost of three or more real measurements, calculate the correlation coefficients between each pair of these measurements, and then take the average of the results as the correlation coefficient for the current combination. (Assume a combination.) The calculated correlation coefficient is Therefore, the cost of modifying the current combination is: (20) Step 3) Define the initial candidate set and refine the result candidate set.

[0026] Initial candidate set It consists of a combination of measured values ​​from all sensors, which satisfies All combinations Let the maximum cost in this set be: (twenty one) in Indicate combination The cost.

[0027] The set is improved through an iterative process, making The potential is processed in sequence. The combination of , for each potential is combination Calculate its cost ,if , directly Add to candidate set Otherwise, in Find all containing combination If it exists Make Then from Delete these and will join in Last updated .

[0028] Step 4) Classify and combine the candidate result sets to obtain the minimum cost combination.

[0029] The sensor with the most measured values ​​is selected as the classification sensor. Let's assume this sensor is the [number]th [sensor / sensor]. There are 1 sensor, and the number of its measurements is 1. Candidate set The combination in the middle is according to the first The measurements used by each sensor are categorized into several groups. Classes (including cases where measurements are not used): (twenty two) in Indicates the first The sensor uses the first One measurement value ( ) or not using any measurements ( All combinations of ).

[0030] From each category Select at most one combination such that these combinations do not conflict (i.e., the same sensor does not use the same measurement value in different combinations), and calculate the total cost: (twenty three) in This indicates that no combination is selected for this class, and the corresponding cost is... The final result is obtained by solving the following optimization problem: (twenty four) This optimization problem can be solved efficiently using dynamic programming or a greedy algorithm, and the final output is the set of combinations with the minimum cost as the target association result.

[0031] Step 5) Use the maximum likelihood method to locate the combination of different targets and obtain the final multi-target localization result.

[0032] The above steps yield a combination of measurement results for each target. This combination is then used to calculate the maximum likelihood of single-target localization, thus solving the nonlinear least squares problem expressed in the following equation: (25) in This represents the single-objective estimation result calculated using the maximum likelihood method. , It is the first The observation angle of each sensor node. express The angle of the position relative to that node. Note the angle difference in the above formula. To meet the requirements .

[0033] This nonlinear least squares problem can be solved using the Gauss-Newton iteration method. Let the direction-finding error matrix be: (26) The Jacobian matrix is: (27) Based on the Gauss-Newton approximation, let the weight matrix be... The increment of one iteration Satisfies the linear equation: The update equation is: (28) The estimated location of each target can be obtained by using the above formula.

[0034] Example 1 Multi-array pure azimuth localization methods based on direction-finding line classification combinations include: Step 1) Acquisition and preprocessing of multi-array direction finding data.

[0035] Assuming the number of nodes (i.e., the number of arrays) is , with the first Taking a node as an example, its two-dimensional coordinates for: The number of sound sources is , No. Two-dimensional coordinates of a sound source for: All sound sources are represented as: No. The sound source is related to the first The DOA of each node is: The observation results are as follows: in For the first The node is the first DOA observation noise for each target Assuming it follows a zero-mean Gaussian distribution with variance of . . No. There are a total of nodes The total number of measurement data points is expressed as follows: For the first Beamforming is performed at each node, and the measurement data of that node is extracted by limiting a threshold.

[0036] Step 2) Construct a likelihood probability model for the direction finding combination; by constructing a direction finding combination probability model based on maximum likelihood estimation, combining the node measurement values ​​and the target existence probability, a combination cost function is introduced to score all possible measurement association combinations, and the combination selection is optimized using angle deviation and correlation coefficient constraints to obtain the optimized target direction estimate and its corresponding existence probability.

[0037] Assumption It is a false measurement. Since there is no prior information, we assume that it follows a uniform distribution. ,in Represents a node Observation range ( ).use Indicates spurious measurements, nodes The measured value is expressed as All measured values ​​are expressed as A set of association results is represented as A set of feasible partitioning results is represented as follows: ,in This indicates that at least two arrays have observed the target. This indicates that at most one array has observed the target. (Definition) For the first The detection rate of nodes, under normal circumstances, is the observation combination. For the same goal The joint likelihood function of the measurements is: (29) in Functions for binary indicator variables. hour In other cases, the value is 1.

[0038] To standardize the likelihood function so that it is independent of the number of measurements from each node and the assumed number of targets, the partition is defined as follows under all possible efficient partitions: Division Corresponding to the assumptions of zero target number and all measurements being spurious targets, the maximum joint likelihood ratio can be given in the following form: (30) In the formula Indicates all possible partitions. The definition is as follows: (31) in Indicates division Next node The number of targets associated with it.

[0039] The formula also contains an unknown term for the location of the target sound source. This term can be replaced with the maximum likelihood estimation result under this assignment. ,So Change to : (32) Take the negative logarithm of the joint likelihood ratio: (33) In the formula: (34) in This represents the difference between the observed azimuth of the maximum likelihood estimated location under this division and the observed data of that node.

[0040] Equation (34) represents the combination cost under a target combination condition: (35) To simplify computation and improve algorithm robustness, utilize In principle, the angle deviation term in the formula is... Replace with This yields a simplified cost threshold: (36) Only retain those with a cost less than All combinations form the initial set of feasible combinations. All subsequent combinations are selected from this set of feasible combinations. It should be noted that when only two actual observations are selected in a combination, the bias term in equation (36) will... At this point, the calculated combined cost becomes a constant: (37) In this situation, conventional data association methods treat all combinations of two observations as identical, making it impossible to distinguish between them. To address this issue, a correlation coefficient threshold for the two observation combinations is introduced. Only combinations exceeding this threshold are retained. During simulation, the correlation coefficient of two observation combinations from the same target is set to be greater than [a certain threshold]. The value, from different targets, is set to less than The value of is used to simulate the correlation effect of this method. In practical data applications, the correlation coefficient of spectral features can be used as the input value, and the threshold value can be used as the input value. Adjust the values ​​according to the actual application environment. When calculating the combined cost of three or more real measurements, calculate the correlation coefficients between each pair of these measurements, and then take the average of the results as the correlation coefficient for the current combination. (Assume a combination.) The calculated correlation coefficient is Therefore, the cost of modifying the current combination is: (38) Step 3) Define the initial candidate set and improve the result candidate set; Define the initial candidate set and perform iterative optimization based on the combination cost function, retain the combination with lower cost and remove its superset, forming a complete candidate set containing the optimal association hypothesis.

[0041] Initial candidate set It consists of a combination of measured values ​​from all sensors, which satisfies All combinations Let the maximum cost in this set be: (39) in Indicate combination The cost.

[0042] The set is improved through an iterative process, making The potential is processed in sequence. The combination of , for each potential is combination Calculate its cost ,if , directly Add to candidate set Otherwise, in Find all containing combination If it exists Make Then from Delete these and will join in Last updated .

[0043] Step 4) Classify and combine the candidate result sets to obtain the minimum cost combination.

[0044] The sensor with the most measured values ​​is selected as the classification sensor. Let's assume this sensor is the [number]th [sensor / sensor]. There are 1 sensor, and the number of its measurements is 1. Candidate set The combination in the middle is according to the first The measurements used by each sensor are categorized into several groups. Classes (including cases where measurements are not used): (40) in Indicates the first The sensor uses the first One measurement value ( ) or not using any measurements ( All combinations of ).

[0045] From each category Select at most one combination such that these combinations do not conflict (i.e., the same sensor does not use the same measurement value in different combinations), and calculate the total cost: (41) in This indicates that no combination is selected for this class, and the corresponding cost is... The final result is obtained by solving the following optimization problem: (42) This optimization problem can be solved efficiently using dynamic programming or a greedy algorithm, and the final output is the set of combinations with the minimum cost as the target association result.

[0046] Step 5) Use the maximum likelihood method to locate the combination of different targets and obtain the final multi-target localization result.

[0047] The above steps yield a combination of measurement results for each target. This combination is then used to calculate the maximum likelihood of single-target localization, thus solving the nonlinear least squares problem expressed in the following equation: (43) in , It is the first The observation angle of each sensor node. express The angle of the position relative to that node. Note the angle difference in the above formula. To meet the requirements .

[0048] This nonlinear least squares problem can be solved using the Gauss-Newton iteration method. Let the direction-finding error matrix be: (44) The Jacobian matrix is: (45) Based on the Gauss-Newton approximation, let the weight matrix be... The increment of one iteration Satisfies the linear equation: The updated equation is: (46) The estimated location of each target can be obtained by calculating using the above formula. The overall process of this method is as follows: Figure 1 As shown.

[0049] To verify the applicability of the algorithm, a simulation experiment was conducted. The observation area in the simulation was square, as shown below. Figure 2 As shown, the blue circles represent sensor nodes, distributed at the four vertices of the square area, with the area's side length being... The observation area is the square region enclosed by the nodes. In the figure, the green circle represents the target location, and the dashed line represents the observation azimuth line of the node to the target, i.e., the DOA measurement result. Ideally, the azimuth lines of each node intersect at the target location, consistent with the results shown in the figure. To simplify the simulation, the nodes are fixed at the indicated positions, and the number is fixed at four. To clearly show the difference between the proposed method and traditional methods, we compare and analyze the results of a single simulation. For example... Figure 3 As shown in the figure, the light blue triangles at the four vertices of the square area represent the locations of the sensor array, the gray dashed lines represent the direction finding lines of each sensor node, the red circles represent the true locations of the targets, the purple pentagrams represent the localization results of the proposed method, the blue asterisks represent the localization results of the divide-and-conquer greedy method, the black triangles represent the results based on intersection point clustering, and the green boxes represent the localization results based on Lagrange relaxation multidimensional allocation. It is clear from the figure that the proposed method located all four targets, while other methods showed varying degrees of performance degradation. The FZTX method located three targets, but with some deviation from the true target locations; the IPC method located two targets with deviations and one false target location; the LSD method located only one target with deviations and two false target locations, performing the worst. It can be observed that the false localization results (one black triangle and two green boxes) are all located by intersections of the direction finding lines of three nodes, indicating a high correlation probability and representing a locally optimal correlation combination. Traditional correlation methods can easily solve for this localization result, while the proposed method avoids getting trapped in this local solution.

[0050] To verify the localization effect of the proposed method, experimental data were used for comparison. For example... Figure 4 As shown in the diagram, triangles represent the array positions, numbers represent array numbers, dashed lines represent the current azimuth measurement results of each array, and boxes represent the intersections of each azimuth line. Each array is a horizontal array placed on the seabed. The diagram shows that array 1 has two measurement results, array 2 has three measurement results, and array 3 has two measurement results, indicating that both targets were observed. Array 2 has one interfering azimuth measurement value. To verify that the azimuth measurement results of each array point to the same target, the spectrum of each array's measurement azimuth is compared. Figure 5 The spectral results at that moment are shown in the figure. As can be seen from the figure, the three directions mentioned in the legend all have the same line spectrum characteristics. Figure 5 The measurement results of target 1 on the three arrays shown on the left are as follows: , Figure 5 The measurement results of target 2 on the three arrays shown on the right are as follows: Array 2 has an interference direction. Since the target is non-cooperative, to verify the algorithm's localization results, the result of the single-target maximum likelihood method is used as the true target location, and the standard deviation of the direction-finding error is set to... The localization results of each method are as follows: Figure 6 As shown in the figure, the true location was obtained using the maximum likelihood single-target localization method, and the direction finding results of each array were used to determine which target each array belonged to. It can be seen from the figure that each method successfully located the target. Both targets have observations on each array, and the number of targets in the region is relatively small; therefore, under these relatively ideal conditions, various data association methods all perform well.

[0051] Example 2 This application also provides a multi-array pure azimuth positioning system based on direction-finding line classification and combination, implemented based on the above method, the system comprising: The data acquisition module is used to acquire multi-array direction finding data, perform beamforming on the direction finding data of each node, and acquire measurement data exceeding a set threshold. The optimization module is used to construct a direction finding combination probability model based on maximum likelihood estimation, combine node measurement values ​​with target existence probability, introduce a combination cost function to score all possible measurement association combinations, and use angle deviation and correlation coefficient constraints to optimize combination selection, so as to obtain the optimized target direction estimate and its corresponding existence probability. The candidate set optimization module is used to define the initial candidate set and perform iterative optimization based on the combination cost function, retaining combinations with lower costs and eliminating their supersets to form a complete candidate set containing the optimal association hypothesis; The combination classification module is used to classify and combine the complete candidate set to obtain the minimum cost combination; The localization estimation module is used to locate combinations of different targets using the maximum likelihood method to obtain the final multi-target localization result.

[0052] This application may also provide a computer device, including: at least one processor, memory, at least one network interface, and a user interface. The various components in this device are coupled together via a bus system. It is understood that the bus system is used to implement communication between these components. In addition to a data bus, the bus system also includes a power bus, a control bus, and a status signal bus.

[0053] The user interface can include a display, keyboard, or clicking device. Examples include a mouse, trackball, touchpad, or touchscreen.

[0054] It is understood that the memory in the embodiments disclosed in this application may be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory may be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory may be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced Synchronous DRAM (ESDRAM), Synchlink DRAM (SLDRAM), and Direct Rambus RAM (DRRAM). The memories described herein are intended to include, but are not limited to, these and any other suitable types of memory.

[0055] In some implementations, the memory stores elements such as executable modules or data structures, or subsets thereof, or extended sets thereof: operating systems and applications.

[0056] The operating system includes various system programs, such as the framework layer, core library layer, and driver layer, used to implement various basic business functions and handle hardware-based tasks. The application programs include various applications, such as media players and browsers, used to implement various application functions. Programs implementing the methods of the embodiments of this disclosure can be included in the application programs.

[0057] In the above embodiments, the processor can also invoke programs or instructions stored in memory, specifically programs or instructions stored in an application program, for the following purposes: Follow the steps described above.

[0058] The above methods can be applied to or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above methods can be completed by integrated logic circuits in the processor's hardware or by software instructions. The processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic diagrams disclosed above. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the disclosed methods can be directly implemented by a hardware decoding processor, or by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory; the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above methods.

[0059] It is understood that the embodiments described in this application can be implemented using hardware, software, firmware, middleware, microcode, or a combination thereof. For hardware implementation, the processing unit can be implemented in one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), general-purpose processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described in this application, or combinations thereof.

[0060] For software implementation, the technology of this application can be implemented by executing the functional modules (e.g., procedures, functions, etc.) of this application. The software code can be stored in memory and executed by a processor. The memory can be implemented in the processor or outside the processor.

[0061] This application may also provide a non-volatile storage medium for storing a computer program. When the computer program is executed by a processor, it can implement the steps in the above method embodiments.

[0062] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application and are not intended to limit it. Although this application has been described in detail with reference to the embodiments, those skilled in the art should understand that modifications or equivalent substitutions to the technical solutions of this application do not depart from the spirit and scope of the technical solutions of this application, and should all be covered within the scope of the claims of this application.

Claims

1. A multi-array pure azimuth localization method based on direction-finding line classification and combination, comprising: Step S1: Acquire multi-array direction finding data, perform beamforming on the direction finding data of each node, and acquire measurement data exceeding the set threshold; Step S2: By constructing a direction finding combination probability model based on maximum likelihood estimation, combining node measurement values ​​and target existence probability, introducing a combination cost function to score all possible measurement association combinations, and using angle deviation and correlation coefficient constraints to optimize combination selection, the optimized target direction estimate and its corresponding existence probability are obtained. Step S3: Define the initial candidate set and perform iterative optimization based on the combined cost function, retain the combination with lower cost and remove its superset to form a complete candidate set containing the optimal association hypothesis; Step S4: Classify and combine the complete candidate set to obtain the minimum cost combination; Step S5: Use the maximum likelihood method to locate the combination of different targets and obtain the final multi-target localization result.

2. The multi-array pure azimuth positioning method based on direction-finding line classification combination according to claim 1, characterized in that, Step S2 includes: A joint likelihood model for a multi-node direction finding system is constructed, and constraints on angle deviation and inter-measurement correlation are introduced. A combined cost function is defined to evaluate all feasible measurement correlation hypotheses. The correlation constraint filters the correlation coefficients of measurements within the combined system using a preset threshold and applies penalties accordingly. The cost function used is as follows: ； in, Indicates the number of nodes; Indicates the first The measurement number selected for each node. This indicates that no target was detected. Indicates an indicator function, when The value is 1 if it is true, and 0 otherwise. Indicates the first Standard deviation of direction finding error for each node; This represents the difference between the observed azimuth of the maximum likelihood estimated location and the observed data at that node; The penalty factor is determined based on the minimum correlation coefficient within the combination; Represents a node The observation range; For the first Node detection rate; Iterate through all feasible combinations, select the combination that minimizes the cost function as the optimal association result, and output the corresponding target direction estimate and existence confidence.

3. The multi-array pure azimuth positioning method based on direction-finding line classification combination according to claim 1, characterized in that, Step S3 includes: The initial candidate set is formed by combining the measured values ​​of all nodes. That is, satisfying All combinations , Indicates the first The measurement number selected by each node; let the maximum cost of this set be: ； in, Indicate combination The cost; The set is improved through an iterative process, and then... The potential is processed in sequence. The combination of , for each potential is combination Calculate its cost ,if ,Will Add to candidate set Otherwise, in Find all containing combination If it exists Make Then from Delete these and will join in Last updated .

4. The multi-array pure azimuth positioning method based on direction-finding line classification combination according to claim 3, characterized in that, Step S4 includes: The node with the most measured values ​​is selected as the classification node, and the candidate set is... The combinations in the data are classified according to the measurement values ​​used by the classification nodes, and are divided into a total of 10 categories. kind, Number of measurements for the classification node: ； in, Indicates the category node number; Indicates that the classification node uses the first One measurement value or all combinations of not using any measurement values; From each category Select at most one combination from the given list, ensuring that these combinations do not conflict, and calculate the total cost: ； in, This indicates that no combination is selected for this class, and the corresponding cost is... The final result is obtained by solving the following optimization problem: ； Solve this optimization problem using dynamic programming or a greedy algorithm, and output the set of combinations with the minimum cost as the objective association result.

5. The multi-array pure azimuth positioning method based on direction-finding line classification combination according to claim 4, characterized in that, Step S5 includes: For the set of measurement results obtained in step S4, the maximum likelihood method for single-target localization is used to calculate the nonlinear least squares problem expressed by the following formula: ； in, This represents the single-objective estimation result calculated using the maximum likelihood method; , Indicates the first Standard deviation of direction finding error for each node; For the first The observation angle of each node, express Position relative to the first The angle of each node; The Gauss-Newton iterative method is used to solve this nonlinear least squares problem, and the direction-finding error matrix is ​​set as follows: ； Let the Jacobian matrix be: ； Where T represents the matrix transpose; Represents the gradient; Based on the Gauss-Newton approximation, the weight matrix is ​​set as follows: , Indicates the first The weight of the nth node; then the weight of the nth node. t Increment of the next iteration Satisfies the linear equation: ； The update equation is: ； The estimated location of each target can be obtained by using the above formula.

6. A multi-array pure azimuth positioning system based on direction-finding line classification and combination, implemented according to the method of any one of claims 1-5, characterized in that, The system includes: The data acquisition module is used to acquire multi-array direction finding data, perform beamforming on the direction finding data of each node, and acquire measurement data exceeding a set threshold. The optimization module is used to construct a direction finding combination probability model based on maximum likelihood estimation, combine node measurement values ​​with target existence probability, introduce a combination cost function to score all possible measurement association combinations, and use angle deviation and correlation coefficient constraints to optimize combination selection, so as to obtain the optimized target direction estimate and its corresponding existence probability. The candidate set optimization module is used to define the initial candidate set and perform iterative optimization based on the combination cost function, retaining combinations with lower costs and eliminating their supersets to form a complete candidate set containing the optimal association hypothesis; The combined classification module is used to classify and combine a complete candidate set to obtain the minimum cost combination; and The localization estimation module is used to locate combinations of different targets using the maximum likelihood method to obtain the final multi-target localization result.

Images