Multi-base sonar joint detection method based on non-uniform quantized data fusion

Through non-uniform quantized data fusion and differential evolution algorithms to optimize the quantization threshold, the problem of data transmission bandwidth limitation in multi-base sonar detection is solved, efficient underwater target detection is achieved, and the requirements of real-time and detection performance are met.

CN118938228APending Publication Date: 2024-11-12KUNMING SHIP EQUIPMENT RESEARCH & TESTING CENTER (CHINA SHIPBUILDING CORP 750 TEST SITE)
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411028076.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-30
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

Traditional multi-base sonar detection methods cannot meet real-time requirements due to the limited data transmission bandwidth in underwater detection scenarios, and simple downsampling will lead to insufficient information utilization and reduce detection performance.

Method used

The non-uniform quantized data fusion method is adopted to adaptively compress and transmit the data collected by the sonar of each node, and differentiated processing is performed in the fusion center. The global detector is designed based on the generalized likelihood ratio criterion, and the quantization threshold is optimized through the differential evolution algorithm. Meet detection performance and real-time requirements.

Benefits of technology

Without reducing detection performance, the multi-node data transmission volume is significantly reduced, which is suitable for scenarios with limited bandwidth, and solves the problem of inability to utilize information caused by insufficient transmission rate, and improves real-time detection and performance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118938228A_ABST
    Figure CN118938228A_ABST
Patent Text Reader

Abstract

According to the multi-base sonar joint detection method based on non-uniform quantized data fusion, a plurality of sonars and a data quantization and transmission device are used for receiving signals emitted by a target in water and transmitting the signals to a fusion center, the data quantization and transmission device is installed on the sonars, and the data quantization and transmission device transmits the signals to the fusion center. The data quantization and transmission device quantizes and compresses the signal received by the sonar and then transmits the signal to the fusion center; the method comprises the following steps: a plurality of sonars collect signals emitted by the same target together; after the plurality of sonar acquisition signals are subjected to non-uniform quantization compression by the data quantization and transmission device, the data are transmitted to the fusion center; the fusion center obtains a detection statistic based on a Rao criterion; obtaining a judgment threshold by deducing asymptotic distribution of the detection statistics; providing for setting and optimization of a quantization threshold value based on a differential evolution algorithm; and comparing the detection statistics with a judgment threshold to obtain a final judgment result.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the field of multi-base sonar detection, and in particular relates to a multi-base sonar joint detection method based on non-uniform quantization data fusion. Background Art

[0002] With the continuous improvement of ship technology and the continuous expansion of ship combat range, especially in the face of the continuous innovation of underwater cluster weapons, the traditional single sonar can no longer meet the requirements of modern marine warfare due to its limited detection range and single information acquisition. Therefore, in order to break through the performance limitations of single sonar, a large number of sensors with communication transmission capabilities are dispersed in the spatial dimension to realize single-channel signal acquisition, and then the data obtained by each channel is centrally processed according to the fusion criteria, and the multi-base sonar system outputs the final judgment result has been rapidly developed in recent years. The superiority of this system lies in that, on the one hand, by fusing multiple node sonar detection data, the signal gain at the spatial level in the water is obtained, the detection range of the system is expanded, and the detection capability of the system is improved; on the other hand, the system-level interconnection and fusion also reduces the performance requirements for a single node, avoiding the high manufacturing cost due to technical bottlenecks. In short, the multi-base sonar system has broad development space in the field of underwater target detection and tracking applications with its unmatched advantages over single sonar.

[0003] A typical multi-base sonar system realizes multi-source information fusion by centrally processing the sensing data of each node. This process requires combining the prior information of the received signal in the detection environment and designing relevant detectors in the fusion center according to certain fusion criteria. The likelihood ratio detector, as the detector with the best performance, has been widely used in signal detection. In the literature "Kay S M. Fundamentals of statistical signal processing [J]. Technometrics, 1993, 37 (4): 465-466." and "Braca P, Goldhahn R, Ferri G, et al. Distributed Information Fusion in Multistatic Sensor Networks for Underwater Surveillance [J]. IEEE Sensors Journal, 2016, 16 (11): 4003-4014", the authors respectively conducted a detailed study on the theoretical derivation of the likelihood ratio detector and its promotion and application in multi-base sonar systems, and proposed a data fusion strategy suitable for underwater sonar detection. However, in the above research, it is considered that under ideal conditions, each node sonar will transmit the collected signal data completely to the fusion center, and the fusion detector is also designed based on the statistical characteristics of the complete signal. In actual engineering applications, underwater detection scenes are usually complex and changeable, especially when each node sonar is far away from the fusion center. When wireless communication is used, the communication bandwidth between the node and the fusion center is severely limited. The above detector cannot meet the real-time requirements of the system. And if the collected signal is simply downsampled and low-precision data is transmitted, it will lead to insufficient information utilization and reduce the detection performance of the system. Therefore, from a long-term perspective, it is of great significance to research and develop multi-base sonar detection technology with independent intellectual property rights for compressed transmission data. Summary of the invention

[0004] The purpose of the invention of this application is to address the defects mentioned in the background technology, and to study and design a multi-base sonar joint detection method based on non-uniform quantization data fusion. When the multi-base sonar system performs target detection, the data collected by the sonar of each node can be non-uniformly quantized and compressed to reduce the amount of transmitted data, thereby meeting the real-time transmission of the system, and effectively solving the problems of low efficiency in target information acquisition and high communication bandwidth requirements of traditional multi-base sonar detection methods.

[0005] The technical solution adopted by the present invention is as follows: according to the real-time requirements of multi-base sonar system detection and the actual conditions of data transmission, the data collected by the sensors of each node are adaptively non-uniformly quantized, the data size is compressed, and the data is transmitted to the fusion center; the fusion center then performs information association on data of different bits according to the statistical characteristics of the data after differential processing; then, based on the generalized likelihood ratio criterion, a global detector is designed to obtain the calculation expression of the detection statistic; finally, the detection statistic is compared with the detection threshold, and the global detection result is output.

[0006] In order to achieve the invention purpose of this application, the present invention adopts the following technical solutions:

[0007] The present invention discloses a multi-base sonar joint detection method based on non-uniform quantization data fusion. The method uses multiple sonars and data quantization and transmission devices to receive and transmit signals emitted by underwater targets to a fusion center. The data quantization and transmission device is installed on the sonar. The data quantization and transmission device quantizes and compresses the signals received by the sonar and then transmits them to the fusion center. The multi-base sonar joint detection method based on non-uniform quantization data fusion includes the following steps:

[0008] (1) Multiple sonars collect signals emitted by the same target

[0009] At the same time, multiple sonars jointly detect the signals emitted by the same unknown target. The total number of sonars is N, and the data transmission between different sonars is independent of each other. The observation model of each sonar receiving channel is expressed as follows:

[0010] x n =β n s n +w n ---Formula (1)

[0011] Among them, x n =[x n (1),x n (2),...,x n (k)] T represents the set of signals received by n sonars (1) within a period of time; β n represents the signal strength set of target (3), s n =[s n (1),s n (2),...,s n (k)] T represents the set of transmitted signals, w n =[w n (1),w n (2),...,w n (k)] Trepresents the set of noises of each receiving channel during the above period of time, where w n (k) has a mean of 0 and a variance of Gaussian noise distribution, k represents the time order of reception;

[0012] (ii) After multiple sonar acquisition signals are quantized and non-uniformly quantized and compressed by the transmission device, the data is transmitted to the fusion center

[0013] Use the following formula to quantize and compress the signal collected by the sonar

[0014]

[0015] Among them, U n (x n (k)) represents the quantization encoding processing of the signal, represents the quantization threshold, represents a binary code sequence with a word length of q after compression, and k represents the time order of reception;

[0016] The signal obtained from the sonar (1) is transmitted to the fusion center after data quantization and non-uniform quantization compression of the transmission device using formula (2);

[0017] (III) The fusion center obtains the detection statistic based on the Rao criterion

[0018] The detection statistics obtained by N sonars are expressed as:

[0019]

[0020] Where Y = [y 1 ,y 2 ,...,y N ] T and β = [β 1 ,β 2 ,...,β N ] T The compressed data and target signal strength transmitted in all channels are represented by sets, P(YH 1 ;β) and P(Y|H 0 ) respectively represent the conditions H 1 (with target) and condition H 0 The probability mass function of the compressed data under (no target); FI(β) represents the Fisher information matrix, β 0 Indicates that β is in H 0 The true value under the condition;

[0021] Based on the compressed coded signal form, the fusion center is expressed as follows using the following binary hypothesis model:

[0022]

[0023] (iv) obtaining the decision threshold by deriving the asymptotic distribution of the detection statistic;

[0024] In order to meet the constant false alarm characteristics of the detection, the asymptotic distribution of the detection statistic is detected, and the detection statistic is compared with the decision threshold to output the detection result. The specific steps are as follows:

[0025] (A1) According to the expression of the detection statistic and the statistical distribution law of the received data, the modified detection statistic 2Λ is constructed R The asymptotic distribution of :

[0026]

[0027] Where "a" means asymptotic compliance, represents the central chi-square distribution with N degrees of freedom, represents a noncentral chi-square distribution with N degrees of freedom,

[0028] (A2) Calculate the expression of the non-central parameter λ:

[0029]

[0030] Here, the calculation of λ requires the given β in H 1 The true value under the assumption When obtaining the true detection performance of a signal, the signal is certain. When performing false alarm detection of unknown targets, in order to meet the constant false alarm characteristic, the false alarm probability P fa , using H 0 The asymptotic distribution characteristics under the condition are inversely solved to calculate the decision threshold η. According to the decision threshold η and the false alarm probability P fa The relationship can be reversed to get η:

[0031] R N,0 (η)=1-P fa →η=R N,0 (1-P fa )-----Formula (7)

[0032] R N,λ (·) is the noncentral chi-square distribution function, which is used to represent the probability of a random variable having a given value when it obeys a specified distribution. Since its probability density function is not integrable and has no primitive function, its value has been formed into a corresponding index table through computer data fitting. When using it, it can be obtained by looking up the table.

[0033] (V) Provide reference for setting and optimizing quantization thresholds based on differential evolution algorithm;

[0034] (a) Select the number of compression bits

[0035] Select the number of compression bits q according to actual needs;

[0036] (b) Using differential evolution algorithm, we can find the optimal quantization threshold.

[0037] (1) Initialize the parameters and generate N randomly and uniformly in the solution interval. P Individuals, each individual represents a solution to a set of quantitative thresholds, then the mth individual z m No. j τ j The solution is expressed as:

[0038] z m (j) = z min (j)+rand(0,1)×(z max (j)-z min (j))---Formula (8)

[0039] Among them, z m (j) is the jth component of the mth individual, i.e., τ j A solution of min (j) and z max (j) represents the lower and upper bounds of the jth component, rand(0,1) represents generating a random number between 0 and 1. Repeat the above operation to generate N P Group Z m , substitute all solutions into the objective function formula (6),

[0040] st 1 <τ 2 <···<τ 2 , and select the one with the largest target value as the initial target solution z max,0 ;

[0041] (2) After the initial individuals are generated, the above N P Group Z m In the first iteration, each individual z m,p (j) The generated variant solution is expressed as

[0042] g m,p =z r1 +F×(z r3 -z r2 )---Formula (9)

[0043] Among them, z r1 , z r2 , z r3 For the above N P Group Z mThe individual solutions are randomly selected from the set and are different from the target individual. F∈(0,2) is a constant factor that controls the scaling of the mutation variable.

[0044] (3) Set the crossover probability factor CR = 0.5 to generate the crossover solution u m , the jth component generated in step (2) is calculated as follows:

[0045]

[0046] (4) The crossover solutions are grouped into a set of τ, substituted into the following formula (11) for optimality comparison, and a new target solution is iterated:

[0047]

[0048] Iteration 10 5 The resulting set That is the optimal quantization threshold The solution is to replace the optimal quantization threshold in step (2) The optimization of the quantization threshold in step (5) is based on the prior information of the known signal. When the detection scene does not change, the value of the optimal quantization threshold will not change. That is, step (5) only needs to be performed once in the detection cycle.

[0049] (6) Compare the detection statistic with the decision threshold to obtain the final decision result;

[0050] According to the given false alarm probability P fa Obtain the decision threshold and substitute the compressed data received by the fusion center (4) into the following decision expression:

[0051]

[0052] When the corrected detection statistic is greater than the decision threshold, the hypothesis H is determined. 1 If it is established, the target signal exists; otherwise, the hypothesis H 0 Established, the target signal does not exist;

[0053] Repeat the above steps (ii) to (iv) and (vi) for each collected data until the cycle completes the detection of all collected data.

[0054] The multi-base sonar joint detection method based on non-uniform quantization data fusion of the present invention, wherein: in step (b), the iteration parameter N P =8, the upper and lower bounds of the solution vector z min (j) = -100 and z max (j)=100.

[0055] The multi-base sonar joint detection method based on non-uniform quantization data fusion of the present invention, wherein: the false alarm probability P fa =10 -4 .

[0056] The multi-base sonar joint detection method based on non-uniform quantization data fusion of the present invention, wherein: the total number of node sonars is 8, the received signal s n =[1,1,...,1] T , noise power The number of Monte Carlo experiments is 10 5 , target signal to noise power ratio SNR = -8dB.

[0057] The multi-base sonar joint detection method based on non-uniform quantization data fusion of the present invention, wherein: in step (iii), the specific calculation of the detection statistic is as follows:

[0058] (B1), based on the prior information, the background noise is considered as Gaussian noise, and H 1 (with signal) and H 0 (No signal) The probability mass function of compressed data under the assumption:

[0059] In H 1 Assume that;

[0060] In H 0 Assume that P(y n (k) = b i ;H 0 )=Q(τ i-1 / σ)-Q(τ i / σ),-----Formula (15)

[0061] Where Q(·) is the standard normal cumulative distribution function;

[0062] (B2), calculate the Fisher information matrix, FI(β 0 ) is a diagonal matrix, and its n-th diagonal element is simplified as follows:

[0063]

[0064] in, is the standard normal probability density function, I i (u) represents the encoding function, only when u = b i When i (u) = 1; otherwise 0; s n (k) is the transmission signal of the nth channel at time k; y n (k) is the quantized data of the received signal; σ is the noise power;

[0065] (B3) Substitute into Rao detector formula (7) and calculate the detection statistic.

[0066] The multi-base sonar joint detection method based on non-uniform quantization data fusion of the present invention, wherein: the Fisher information matrix FI (β) is expressed as follows:

[0067]

[0068] Where β = [β 1 ,β 2 ,...,β N ] T The set represents the strength of N sonar received echo signals, and P(β) represents the probability density function of the received signal.

[0069] The multi-base sonar joint detection method based on non-uniform quantization data fusion of the present invention, wherein: the standard normal cumulative distribution function Q(·) is

[0070]

[0071] Among them: w represents the integral variable.

[0072] The multi-base sonar joint detection method based on non-uniform quantization data fusion of the present invention, wherein: the non-central parameter λ is the non-central chi-square cumulative distribution function R n,λ (·), which means as follows:

[0073]

[0074] Among them, t represents the independent variable of the cumulative distribution function, that is, it provides the value of the function, w represents the integral variable, I n2-1 represents the Bessel function, σ is the noise power, and λ is the non-central parameter.

[0075] The beneficial effect of the present invention is that in a multi-base sonar system, by non-uniformly quantizing the target signals collected by each node sonar, and designing a matching joint detector in the fusion center according to the Rao criterion, the present invention greatly reduces the transmission data of multiple nodes while basically not affecting the detection performance of the system, and realizes the fusion detection of multi-node data. In addition, by obtaining the asymptotic statistical characteristics of the detector, the present invention also provides a reference for the setting of the quantization threshold, meets the constant false alarm requirement of the detector, and further improves the content of the invention. The advantage of the present invention is that compared with the multi-base sonar detection method of transmitting complete collected data, the method proposed by the present invention can significantly reduce the amount of data transmission, and is suitable for scenarios where the transmission bandwidth between the node and the fusion center is severely limited, especially in the case of high real-time detection requirements, compressing the number of data bits can effectively solve the problem that the system multi-source information cannot be used due to insufficient transmission rate. In addition, further optimization of the quantization threshold based on the differential evolution algorithm can guarantee the detection performance of the multi-base sonar system to the greatest extent. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 A flowchart of the method of the present invention;

[0077] Figure 2 This is a schematic diagram of the multi-base sonar system jointly detecting targets mentioned in the present invention;

[0078] Figure 3 A schematic diagram of obtaining quantization threshold distribution based on a differential evolution algorithm provided in an embodiment of the present invention;

[0079] Figure 4 A comparison diagram of normalized waveforms of detection statistics of different bit symbols provided by an embodiment of the present invention;

[0080] Figure 5 The detection performance curves of the different bit symbol detectors provided by the embodiments of the present invention are shown. DETAILED DESCRIPTION

[0081] The present invention is mainly verified by simulation experiment method, and all steps and conclusions are verified to be correct in Matlab 2022. To facilitate those skilled in the art to understand the technical content of the present invention, the content of the present invention is further explained below in conjunction with the accompanying drawings.

[0082] like Figure 2 As shown, the multi-base sonar joint detection method based on non-uniform quantization data fusion of the present invention uses multiple sonars 1 and data quantization and transmission devices 2 to receive and transmit the signal emitted by the underwater target 3 to the fusion center 4. The data quantization and transmission device 2 is installed on the sonar 1. The data quantization and transmission device 2 quantizes and compresses the signal received by the sonar 1 and then transmits it to the fusion center 4.

[0083] like Figure 1 As shown, the multi-base sonar joint detection method based on non-uniform quantization data fusion of the present invention comprises the following steps:

[0084] (a) Multiple sonars 1 jointly collect signals emitted by the same target 3

[0085] At the same time, multiple sonars 1 jointly detect the signal emitted by the same unknown target 3. The total number of sonars 1 is N, and the data transmission between different sonars 1 is independent of each other. The observation model of each sonar 1 receiving channel is expressed as follows:

[0086] x n =β n s n +w n ---Formula (1)

[0087] Among them, x n =[x n (1),x n (2),...,x n (k)] T represents the set of signals received by n sonars (1) within a period of time; βn represents the set of signal strengths of the target (3), s n =[s n (1),s n (2),...,s n (k)] T represents the set of transmitted signals, w n =[w n (1),w n (2),...,w n (k)] T represents the set of noises of each receiving channel during the above period of time, where w n (k) has a mean of 0 and a variance of Gaussian noise distribution, k represents the time order of reception, for example: the total number of node sonars is 8, and the received signal s n =[1,1,...,1] T , noise power The number of Monte Carlo experiments is 10 5 , target signal to noise power ratio SNR = -8dB;

[0088] (ii) After the signals collected by multiple sonars 1 are quantized and non-uniformly quantized and compressed by the transmission device 2, the data are transmitted to the fusion center 4

[0089] Use the following formula to quantize and compress the signal collected by sonar 1

[0090]

[0091] Among them, U n (x n (k)) represents the quantization encoding processing of the signal, represents the quantization threshold, represents a binary code sequence with a word length of q after compression, and k represents the time sequence of reception, for example: Figure 2 In the multi-base sonar system shown, since each node sonar and the fusion center use wireless communication, the transmission bandwidth is severely limited. Each node sonar will perform non-uniform quantization on the collected data, compress the data volume and transmit it. Here, three types of binary compression codes are mainly set, including: 1-bit code sequence "0", "1", 2-bit code sequence "00", "01", "10", "11", 3-bit code sequence "000", "001", "010", "011", "100", "101", "110", "111", and the setting of non-uniform quantization threshold is shown in S;

[0092] The signal obtained from the sonar 1 is transmitted to the fusion center 4 after data quantization and non-uniform quantization compression by the transmission device 2 using formula (2);

[0093] (III) Fusion center 4 obtains the detection statistic based on Rao criterion

[0094] Using formula (3) to express N sonars 1, we can get the detection statistic:

[0095]

[0096] Where Y = [y 1 ,y 2 ,...,y N ] T and β = [β 1 ,β 2 ,...,β N ] T The compressed data and target signal strength transmitted in all channels are represented by sets, P(YH 1 ;β) and P(Y|H 0 ) respectively represent the conditions H 1 (with target) and condition H 0 The probability mass function of the compressed data under (no target); FI(β) represents the Fisher information matrix, β 0 Indicates that β is in H 0 The true value under the condition;

[0097] Based on the compressed coded signal form, the fusion center 4 uses the following binary hypothesis model to express it as follows:

[0098]

[0099] The specific calculation of the detection statistic is as follows:

[0100] (B1), based on the prior information, the background noise is considered as Gaussian noise, and H 1 (with signal) and H 0 (No signal) The probability mass function of compressed data under the assumption:

[0101] In H 1 Assume that;

[0102] In H 0 Assume that P(y n (k) = b i ;H 0 )=Q(τ i-1 / σ)-Q(τ i / σ),-----Formula (15)

[0103] Where Q(·) is the standard normal cumulative distribution function;

[0104] The standard normal cumulative distribution function Q(·) is

[0105]

[0106] Among them: w represents the integral variable;

[0107] (B2), calculate the Fisher information matrix, FI(β 0 ) is a diagonal matrix, and its n-th diagonal element is simplified as follows:

[0108]

[0109] in, is the standard normal probability density function, I i (u) represents the encoding function. Only when u = bi, I i (u) = 1; otherwise 0; s n (k) is the transmission signal of the nth channel at time k; y n (k) is the quantized data of the received signal; σ is the noise power;

[0110] (B3), Substitute into Rao detector formula (7) and calculate the detection statistic;

[0111] (iv) obtaining the decision threshold by deriving the asymptotic distribution of the detection statistic;

[0112] In order to meet the constant false alarm characteristics of the detection, for example: the false alarm probability P fa=10 -4 , detect the asymptotic distribution of the statistic, compare the detection statistic with the decision threshold, and then output the detection result. The specific steps are as follows:

[0113] (A1) According to the expression of the detection statistic and the statistical distribution law of the received data, the modified detection statistic 2Λ is constructed R The asymptotic distribution of :

[0114]

[0115] Where "a" means asymptotic compliance, represents the central chi-square distribution with N degrees of freedom, represents a noncentral chi-square distribution with N degrees of freedom,

[0116] (A2) Calculate the expression of the non-central parameter λ:

[0117]

[0118] The noncentral parameter λ is the noncentral chi-square cumulative distribution function R n,λ (·), which means as follows:

[0119]

[0120] Among them, t represents the independent variable of the cumulative distribution function, that is, it provides the value of the function, w represents the integral variable, I n2-1 represents the Bessel function, σ is the noise power, and λ is the non-central parameter;

[0121] The Fisher information matrix FI(β) is expressed as follows:

[0122]

[0123] Where β = [β 1 ,β 2 ,...,β N ] T The set represents the strength of N sonar received echo signals, and P(β) represents the probability density function of the received signal;

[0124] Here, the calculation of λ requires the given β in H 1 The true value under the assumption When obtaining the true detection performance of a signal, the signal is certain. When performing false alarm detection of unknown targets, in order to meet the constant false alarm characteristic, the false alarm probability P fa , using H 0 The asymptotic distribution characteristics under the condition are inversely solved to calculate the decision threshold η. According to the decision threshold η and the false alarm probability P fa The relationship can be reversed to get η:

[0125] R N,0 (η)=1-P fa →η=R N,0 (1-P fa )-----Formula (7)

[0126] R N,λ (·) is the noncentral chi-squared cumulative distribution function, which is used to represent the probability of a random variable having a given value when it obeys a specified distribution. Since its probability density function is not integrable and has no primitive function, its value has been formed into a corresponding index table through computer data fitting. When using it, it can be obtained by looking up the table.

[0127] (V) Provide reference for setting and optimizing quantization thresholds based on differential evolution algorithm;

[0128] (a) Select the number of compression bits

[0129] Select the number of compression bits q according to actual needs;

[0130] (b) Using differential evolution algorithm, we can find the optimal quantization threshold.

[0131] (1) Initialize the parameters and generate N randomly and uniformly in the solution interval. P Individuals, each individual represents a solution to a set of quantitative thresholds, then the mth individual z m No. j τ j The solution is expressed as:

[0132] z m (j) = z min (j)+rand(0,1)×(z max (j)-z min (j))---Formula (8)

[0133] Among them, z m (j) is the jth component of the mth individual, i.e., τ j A solution of min (j) and z max (j) represents the lower and upper bounds of the jth component, for example: the iteration parameter N P =8, the upper and lower bounds of the solution vector z min (j) = -100 and z max (j) = 100, which means generating a random number from 0 to 1. Repeat the above operation to generate N P Group Z m , substitute all solutions into the objective function formula (6),

[0134] st 1<τ 2 <···<τ 2 , and select the one with the largest target value as the initial target solution z max,0 ;

[0135] (2) After the initial individuals are generated, the above N P Group zm performs the first iteration, each individual z m,p (j) The generated variant solution is expressed as

[0136] g m,p =z r1 +F×(z r3 -z r2 )---Formula (9)

[0137] Among them, z r1 , z r2 , z r3 For the above N P Group Z m The individual solutions are randomly selected from the set and are different from the target individual. F∈(0,2) is a constant factor that controls the scaling of the mutation variable.

[0138] (3) Set the crossover probability factor CR = 0.5 to generate the crossover solution u m , the jth component generated in step (2) is calculated as follows:

[0139]

[0140] (4) The crossover solutions are grouped into a set of τ, substituted into the following formula (11) for optimality comparison, and a new target solution is iterated:

[0141]

[0142] Iteration 10 5 The resulting set That is the optimal quantization threshold The solution is to replace the optimal quantization threshold in step (2) The optimization of the quantization threshold in step (5) is based on the prior information of the known signal. When the detection scene does not change, the value of the optimal quantization threshold will not change. That is, step (5) only needs to be performed once in the detection cycle.

[0143] (6) Compare the detection statistic with the decision threshold to obtain the final decision result;

[0144] According to the given false alarm probability P fa Obtain the decision threshold and substitute the compressed data received by the fusion center (4) into the following decision expression:

[0145]

[0146] When the corrected detection statistic is greater than the decision threshold, the hypothesis H is determined. 1 If it is established, the target signal exists; otherwise, the hypothesis H 0 Established, the target signal does not exist;

[0147] Repeat the above steps (ii) to (iv) and (vi) for each collected data until the cycle completes the detection of all collected data.

[0148] like Figure 4 As shown in Figure 1, the parameter settings are the same as S1. According to the Monte Carlo method, the signal acquisition samples are simulated and generated, and the algorithm flow is input for calculation. The normalized waveform results of the detection statistics of different quantization bits are intercepted for comparison. According to the simulation results, the lower the number of quantization bits, the more obvious the waveform distortion of the detection statistic, and as the number of quantization bits increases, the waveform of the detection statistic becomes more similar. In particular, the waveform of the 3-bit quantization symbol data has basically fitted the waveform of the uncompressed data, which intuitively illustrates the superiority of the algorithm in extracting target information by optimizing, reducing data redundancy, and retaining the characteristics of the target information.

[0149] like Figure 5 As shown, with 1dB as the interval step, the target signal-to-noise ratio is set from -15dB to 5dB, and the other parameters are set the same as S1. According to the Monte Carlo method, the signal acquisition samples are simulated and generated, and the algorithm flow is input for calculation to generate the detection performance curves of different bit code detectors. It should be noted that under the condition of constant false alarm probability, the higher the detection probability at the same signal-to-noise ratio, the stronger the detection performance of the detector. According to the simulation results, the lower the number of quantization bits of data transmitted, the weaker the detection performance of the detector, and as the number of quantization bits increases, the performance of the detector gradually increases, especially when transmitting 3-bit quantized code data, the detection performance curve of the detector has basically fitted the detection performance curve of the unquantized data transmission, indicating that the 3-bit quantized code detector can exchange a very small performance loss for a large compression of the data volume, which fully verifies the effectiveness of the invention. At the same time, observing the trend of the detection performance curve, it can be seen that the quantized code detector improves with the increase of the signal-to-noise ratio, which is consistent with the law of the unquantized code detector, further illustrating the adaptability and rationality of the invention.

[0150] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the present invention.

Claims

1. A multi-base sonar joint detection method based on non-uniform quantization data fusion, the method using multiple sonars (1) and a data quantization and transmission device (2) to receive and transmit signals emitted by an underwater target (3) to a fusion center (4), the data quantization and transmission device (2) being mounted on the sonar (1), the data quantization and transmission device (2) quantizes and compresses the signals received by the sonar (1) and then transmits them to the fusion center (4); the method is characterized in that: The multi-base sonar joint detection method based on non-uniform quantization data fusion includes the following steps: (a) Multiple sonars (1) collect signals emitted by the same target (3) At the same time, multiple sonars (1) jointly detect the signal emitted by the same unknown target (3). The total number of sonars (1) is N, and the data transmission between different sonars (1) is independent of each other. The observation model of each sonar (1) receiving channel is expressed as follows: x n =β n s n +w n ---Formula (1) Among them, x n =[x n (1),x n (2),...,x n (k)] T represents the set of signals received by n sonars (1) within a period of time; β n represents the signal strength set of target (3), s n =[s n (1),s n (2),...,s n (k)] T represents the set of transmitted signals, w n =[w n (1),w n (2),...,w n (k)] T represents the set of noises of each receiving channel during the above period of time, where w n (k) has a mean of 0 and a variance of Gaussian noise distribution, k represents the time order of reception; (ii) After the signals collected by the multiple sonars (1) are subjected to data quantization and non-uniform quantization compression by the transmission device (2), the data are transmitted to the fusion center (4) The signal collected by sonar (1) is quantized and compressed using the following formula: Among them, U n (x n (k)) represents the quantization encoding processing of the signal, represents the quantization threshold, represents a binary code sequence with a word length of q after compression, and k represents the time order of reception; The signal obtained from the sonar (1) is quantized using formula (2) and non-uniformly quantized and compressed by the transmission device (2), and then transmitted to the fusion center (4); (III) Fusion center (4) Based on Rao criterion, the detection statistic is obtained Using formula (3) to express N sonars (1) we get the detection statistic: Where Y = [y1, y2, ..., y N ] T and β=[β1,β2,...,β N ] T The sets represent the compressed data and target signal strength transmitted in all channels, P(YH1; β) and P(YH0) represent the probability mass function of the compressed data under condition H1 (with target) and condition H0 (without target), respectively; FI(β) represents the Fisher information matrix, and β0 represents the true value of β under condition H0; Based on the compressed coded signal form, the fusion center (4) is expressed as follows using the following binary hypothesis model: (iv) obtaining the decision threshold by deriving the asymptotic distribution of the detection statistic; In order to meet the constant false alarm characteristics of the detection, the asymptotic distribution of the detection statistic is detected, and the detection statistic is compared with the decision threshold to output the detection result. The specific steps are as follows: (A1) According to the expression of the detection statistic and the statistical distribution law of the received data, the modified detection statistic 2Λ is constructed R The asymptotic distribution of : Where "a" means asymptotic compliance, represents the central chi-square distribution with N degrees of freedom, represents a noncentral chi-square distribution with N degrees of freedom, (A2) Calculate the expression of the non-central parameter λ: Here, the calculation of λ requires the true value of β under the H1 hypothesis. When obtaining the true detection performance of a signal, the signal is certain. When performing false alarm detection of unknown targets, in order to meet the constant false alarm characteristic, the false alarm probability P fa , the decision threshold η is calculated by inverse solution using the asymptotic distribution characteristics under the H0 condition. According to the decision threshold η and the false alarm probability P fa The relationship can be reversed to get η: R N,0 (η) = 1 - P fa → η = R N,0 (1 - P fa )-----Formula (7) R N,λ (·) is the noncentral chi-squared cumulative distribution function, which is used to represent the probability of a random variable having a given value when it obeys a specified distribution. Since its probability density function is not integrable and has no primitive function, its value has been formed into a corresponding index table through computer data fitting. When using it, it can be obtained by looking up the table. (V) Provide reference for setting and optimizing quantization thresholds based on differential evolution algorithm; (a) Select the number of compression bits Select the number of compression bits q according to actual needs; (b) Using differential evolution algorithm, we can find the optimal quantization threshold. (1) Initialize the parameters and generate N randomly and uniformly in the solution interval. P Individuals, each individual represents a solution to a set of quantitative thresholds, then the mth individual z m No. j τ j The solution is expressed as: z m f(j) = z min f(j) + rand(0,1) × (z max f(j) - z min f(j)) --- Equation (8) Among them, z m (j) is the jth component of the mth individual, i.e., τ j A solution of min (j) and z max (j) represents the lower and upper bounds of the jth component, rand(0,1) represents generating a random number between 0 and 1. Repeat the above operation to generate N P Group Z m , substitute all solutions into the objective function formula (6), And select the one with the largest target value as the initial target solution z max,0 ; (2) After the initial individuals are generated, the above N P Group Z m In the first iteration, each individual z m,p (j) The generated variant solution is expressed as g m,p = z r1 + F × (z r3 - z r2 )---Formula (9) Among them, z r1 , z r2 , z r3 For the above N P Group Z m The individual solutions are randomly selected from the set and are different from the target individual. F∈(0,2) is a constant factor that controls the scaling of the mutation variable. (3) Set the crossover probability factor CR = 0.5 to generate the crossover solution u m , the jth component generated in step (2) is calculated as follows: (4) The crossover solutions are grouped into a set of τ, substituted into the following formula (11) for optimality comparison, and a new target solution is iterated: Iteration 10 5 The resulting set That is the optimal quantization threshold The solution is to replace the optimal quantization threshold in step (2) The optimization of the quantization threshold in step (5) is based on the prior information of the known signal. When the detection scene does not change, the value of the optimal quantization threshold will not change. That is, step (5) only needs to be performed once in the detection cycle. (6) Compare the detection statistic with the decision threshold to obtain the final decision result; According to the given false alarm probability P fa Obtain the decision threshold and substitute the compressed data received by the fusion center (4) into the following decision expression: When the corrected detection statistic is greater than the decision threshold, the hypothesis H1 is established and the target signal exists; otherwise, the hypothesis H0 is established and the target signal does not exist. Repeat the above steps (ii) to (iv) and (vi) for each collected data until the cycle completes the detection of all collected data.

2. The multi-base sonar joint detection method based on non-uniform quantization data fusion as claimed in claim 1, characterized in that: In step (b), the iteration parameter N P =8, the upper and lower bounds of the solution vector z min (j) = -100 and z max (j)=100.

3. The multi-base sonar joint detection method based on non-uniform quantization data fusion as claimed in claim 2, characterized in that: The false alarm probability P fa =10 -4 .

4. The multi-base sonar joint detection method based on non-uniform quantization data fusion as claimed in claim 3 is characterized in that: The total number of node sonars is 8, and the received signal s n =[1,1,...,1] T , noise power The number of Monte Carlo experiments is 10 5 , target signal to noise power ratio SNR = -8dB.

5. The multi-base sonar joint detection method based on non-uniform quantization data fusion as claimed in claim 4 is characterized in that: In step (iii), the specific calculation of the detection statistic is as follows: (B1) Based on the prior information, the background noise is considered as Gaussian noise, and the probability mass function of the compressed data under the assumptions of H1 (signal) and H0 (no signal) is obtained: Under the H1 hypothesis; Under the H0 hypothesis, P(y n (k) = b i ; H0)=Q(τ i-1 / σ)-Q(τ i / σ),-----Formula (15) Where Q(·) is the standard normal cumulative distribution function; (B2) Calculate the Fisher information matrix, FI(β0) is a diagonal matrix, and its n-th diagonal element is simplified as follows: in, is the standard normal probability density function, I i (u) represents the encoding function, only when u = b i When i (u) = 1; otherwise 0; s n (k) is the transmission signal of the nth channel at time k; y n (k) is the quantized data of the received signal; σ is the noise power; (B3) Substitute into Rao detector formula (7) and calculate the detection statistic.

6. The multi-base sonar joint detection method based on non-uniform quantization data fusion as claimed in claim 5, characterized in that: The Fisher information matrix FI(β) is expressed as follows: Where β=[β1,β2,...,β N ] T The set represents the strength of N sonar received echo signals, and P(β) represents the probability density function of the received signal.

7. The multi-base sonar joint detection method based on non-uniform quantization data fusion as claimed in claim 6, characterized in that: The standard normal cumulative distribution function Q(·) is Among them: w represents the integral variable.

8. The multi-base sonar joint detection method based on non-uniform quantization data fusion as claimed in claim 7, characterized in that: The noncentral parameter λ is the noncentral chi-square cumulative distribution function R n,λ (·), which means as follows: Among them, t represents the independent variable of the cumulative distribution function, that is, it provides the value of the function, w represents the integral variable, I n2-1 represents the Bessel function, σ is the noise power, and λ is the non-central parameter.