Active sonar target dynamic and static attribute recognition feature extraction method based on tracking trajectory

Abstract

The invention belongs to the technical field of underwater acoustic signal processing, and particularly relates to an active sonar target dynamic and static attribute recognition feature extraction method based on a tracking trajectory, which comprises the following steps of: S1, respectively performing quadratic polynomial least square fitting on azimuth and distance information on the tracking trajectory; s2, eliminating gross observation errors, and evaluating azimuth and distance information change indexes on a tracking trajectory; and S3, finally, extracting a movement index of the target by synthesizing the azimuth and distance change indexes as a target dynamic and static attribute identification feature. In the active sonar target dynamic and static attribute identification application, the target dynamic and static attributes can be identified according to the target orientation and distance information time sequence obtained on the tracking trajectory.

Description

Technical Field

[0001] The present invention belongs to the technical field of underwater acoustic signal processing, and in particular relates to a method for extracting dynamic and static attribute recognition features of active sonar targets based on tracking trajectories. The method is mainly used for moving target / stationary clutter recognition in active sonar applications, and aims to identify the static and static attributes of echoes by analyzing tracking trajectory information, thereby providing support for highlighting moving targets and suppressing interference from stationary clutter. Background Art

[0002] The judgment of dynamic and static attributes is the basis for active sonar to identify moving targets and suppress static clutter interference. Trackable moving targets differ from trackable static interference in terms of azimuth-range trajectory information. In-depth analysis of the trajectory information of echoes and extraction of their features are important technical approaches to identify the static and static attributes of echoes.

[0003] There are some related studies on the determination of target dynamic and static attributes based on tracking trajectories in the radar field, but they are not suitable for active sonar applications. The main reasons are the following two aspects: First, the waveguide speed of radar is much faster than that of sonar, it is close to a stationary random signal, the time interval of multi-cycle echoes is short, the correlation is strong, which is conducive to multi-pulse accumulation and stationary clutter interference cancellation, while the sonar signal is a non-stationary random signal, the time interval of multi-cycle echoes is long, the correlation is low, which is not conducive to multi-pulse accumulation, and stationary clutter is difficult to cancel; second, the radar carrier frequency is high, the signal bandwidth is large, and the ranging is accurate, while the sonar carrier frequency is low, the available signal bandwidth is much smaller than that of radar, and the ranging accuracy is significantly lower than that of radar. In reality, the radar can accurately image the target of interest, while the sonar can only obtain a rough echo of multiple bright spots.

[0004] Active sonar underwater moving target identification is a long-standing international problem. Most of the available active sonar target dynamic and static attribute identification methods are based on radial velocity measurement, such as measuring the radial velocity of echoes through Doppler sensitive signals such as single-frequency pulse / frequency modulated pulse train / sine frequency modulation signal / up-down frequency modulation combination signal. This type of method is based on the premise that the target should have a radial velocity that is significantly different from stationary clutter. In actual engineering applications, this premise is often not met, mainly due to three factors. First, the target of interest is often a low-speed target, and the difference between the speed and stationary clutter itself is not very significant; second, when the target can be tracked, it is often in the vicinity of the positive horizontal posture with a large target intensity. In this situation, the radial velocity of the target, which is not high, is even less different from the stationary clutter from the sonar perspective; third, the uncertainty of sonar observation information causes the radial velocity measurement error to be higher than the radial velocity difference information between the moving target and the clutter, making it difficult to identify the dynamic and static attributes of the moving target with the help of radial velocity measurement. For the vast majority of common stable trackable echoes, it is difficult to identify the dynamic and static attributes by measuring the radial velocity, and new technical approaches must be developed to identify the dynamic and static attributes of the target.

[0005] Active sonar usually directly obtains the azimuth and distance information of the echo. Due to the limitation of the aperture size of the acoustic array and the bandwidth of the transmitting waveform, the azimuth / distance measurement of the sonar contains significant uncertainty errors. In addition, the accuracy of azimuth and distance measurement is also related to the stability of the multi-bright spot dispersion of the echo. Underwater low-speed moving targets of interest, such as manned submarines and unmanned underwater vehicles, are usually artificial targets with smooth surfaces. When they are tracked continuously for multiple cycles, the distribution of strong scattering points is relatively stable; the surface of naturally formed stationary clutter, such as reefs and seamounts, is irregular and not smooth. When they are tracked continuously for multiple cycles, the distribution of strong scattering points is relatively unstable. The difference in the stability of the distribution of strong scattering points between the moving target of interest and the stationary clutter is inevitably reflected in the stability of their respective echo multi-bright spot dispersion. Therefore, the multi-bright spot dispersion stability of the moving target is stronger than that of the stationary clutter. On the one hand, the moving target must have a more significant change in at least one quantity in azimuth or distance compared to the stationary target; on the other hand, the multi-bright spot distribution of the echo of the artificial moving target is more stable than that of the stationary clutter. Due to the combined effect of the above two reasons, although there is uncertainty in the echo azimuth and distance measurement from the perspective of active sonar, examining whether there is a significant and reasonable change in the echo azimuth or distance information is still a feasible technical approach to identify the dynamic and static attributes of the target. Summary of the invention

[0006] The purpose of the present invention is to provide a method for extracting dynamic and static attributes of active sonar targets based on tracking trajectory azimuth and distance information. In the application of active sonar target dynamic and static attribute identification, the dynamic and static attributes of the target can be identified according to the time series of target azimuth and distance information obtained on the tracking trajectory.

[0007] In order to solve the above technical problems, the present invention provides a method for extracting dynamic and static attribute recognition features of active sonar targets based on tracking trajectory azimuth distance information, comprising the following steps:

[0008] Step S1: firstly perform quadratic polynomial least squares fitting on the azimuth and distance information on the tracking trajectory respectively;

[0009] Step S2: Eliminate the gross observation error and then evaluate the change index of the azimuth and distance information on the tracking trajectory;

[0010] Step S3: Finally, the target movement index is extracted by integrating the direction and distance change index as the target dynamic and static attribute identification feature.

[0011] Preferably, the step S1 specifically includes: applying Matlab software to perform adaptive quadratic polynomial least squares fitting on the observation sequence, and the Matlab library functions called are polyfit and polyval.

[0012] Preferably, the step S1 further comprises: defining an observation distance sequence vector R=[r1, r2, ..., r N ], observation position sequence A = [a1, a2, ..., a N ] and the corresponding observation time sequence T = [t1, t2, ..., t N ], where N is the number of observation sequence points; the observation time sequence refers to its own starting time, that is, t1=0; the active sonar tracking target refers to the position of the ship at the time of sonar emission with reference to the same point.

[0013] Preferably, eliminating gross observation errors in step S2 specifically includes: when the number of trajectory points is greater than or equal to 15 points, taking a maximum error observation value for every 15 points, replacing it with a fitting value and refitting, and rounding it when the remaining trajectory points are less than 15 points; when the number of trajectory points is less than 15, no gross error correction processing is performed, and no refitting is performed; the Matlab library function called is the sorting function sort.

[0014] Preferably, evaluating the change index of the azimuth and distance information on the tracking trajectory in step S2 specifically includes: calculating the standard deviation between the distance / azimuth sequence fitting value and the observed value after gross error correction, respectively denoted as σ R ,σ A , the Matlab library function called is the standard deviation solution function std;

[0015] The peak change value on the fitting curve is calculated as the ratio between the steady-state variable and the standard deviation, which are recorded as the distance change index C R , Azimuth change index C A The Matlab library functions called are the maximum value function max and the minimum value function min.

[0016] Preferably, the step S3 specifically includes: using the distance change index C R , Azimuth change index C A The larger value of is taken as the target movement index M, that is, M = max(C R ,C A ).

[0017] Preferably, when performing quadratic polynomial least squares curve fitting, the upper and lower error limits corresponding to each fitting value are indicated respectively according to 3 times the mean square deviation value during the fitting.

[0018] Preferably, a numerical value is set as a decision threshold of the movement index, and when the movement index of a target is higher than the decision threshold, it is judged as a moving target, and conversely, when the movement index of the target is lower than the decision threshold, it is judged as a stationary target.

[0019] Preferably, when the decision threshold of the mobile index changes, it also includes missed alarm rate, false alarm rate, and confidence parameters; wherein missed alarm means that it is actually a moving target but is judged as stationary clutter; false alarm means that it is actually stationary clutter but is judged as a moving target; the missed alarm rate is defined as the number of missed alarms divided by the total number of real moving target samples; the false alarm rate is defined as the number of false alarms divided by the total number of reported moving targets; the confidence is defined as the number of correct reports in the reported moving targets divided by the total number of reported moving targets; the false alarm rate and the confidence are complementary, and the sum of the two is always 100%.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] 1) The present invention creatively proposes a target dynamic and static attribute evaluation mechanism based on the movement index. This mechanism gets rid of the embarrassing situation that the radial velocity difference information of the low-speed target / stationary clutter of the active sonar is too small, and the absolute velocity difference information is unreliable and unstable. Compared with the traditional technical approach of judging the dynamic and static attributes of the target by speed measurement, the present invention only uses the objective physical fact that the moving target must have obvious regular changes in at least one of the azimuth and distance observation sequences to extract the dynamic and static attribute features, and directly evaluates whether the target moves regularly based on the azimuth and distance information measured by the sonar. It has a solid theoretical foundation, few intermediate links in data processing, simple and clear cause-and-effect relationship, and has strong feasibility verified by measured data.

[0022] 2) The present invention proposes a method based on quadratic polynomial fitting to evaluate whether the target azimuth / distance observation sequence has regular changes, which is reliable and easy to implement. For artificial underwater low-speed inertial motion targets, quadratic polynomials are sufficient to describe the true value changes of their azimuth and distance observation sequences, including but not limited to common situations such as uniform linear motion of the target, accelerated / decelerated linear motion, uniform speed turning, and variable speed turning. Even when the underwater target performs complex variable speed serpentine maneuvers, it always takes a long time, which can also be accurately described by the time-varying quadratic polynomial fitting with adaptive parameter adjustment.

[0023] 3) The present invention describes the azimuth / range variation index of the tracked target echo by measuring the ratio between the steady-state variation (peak-to-peak variation on the fitting curve) and the transient uncertainty (error value between the observed sampling value and the fitting curve) in the azimuth / range observation time series. This method conforms to the physical reality that the steady-state variation of the moving target is larger than the transient uncertainty, while the steady-state variation of the stationary clutter is smaller than the transient uncertainty, and can highlight the different characteristics of the moving target and the stationary clutter.

[0024] 4) The present invention adopts the larger value of the distance change index and the azimuth change index as the quantitative measurement value of the target movement index, so as to quantitatively characterize the dynamic and static properties of the target. The steady-state change of at least one dimension of the azimuth or distance information of the non-cooperative moving target should significantly exceed the observation error, while the steady-state change of any dimension of the stationary interference information should not significantly exceed the observation error. The mobile index measurement method proposed in the present invention has a solid physical foundation, strong model robustness, and simple calculation.

[0025] 5) The present invention adopts an outlier elimination method in the process of fitting the curve. The outliers are selected in proportion to the number of trajectory points, which can effectively reduce gross errors and reduce the impact of mis-associated trajectory points on the results during the tracking process. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 The present invention is a flow chart of a method for extracting dynamic and static attribute recognition features of an active sonar target based on tracking trajectory azimuth distance information.

[0027] Figure 2 It is a schematic diagram of the distance observation sequence, fitting curve and error limit of a moving target of the present invention.

[0028] Figure 3 It is a schematic diagram of the azimuth observation sequence, fitting curve and error limit of a moving target of the present invention.

[0029] Figure 4 It is a schematic diagram of the distance observation sequence, fitting curve and error limit of a certain stationary clutter of the present invention.

[0030] Figure 5 It is a schematic diagram of the azimuth observation sequence, fitting curve and error limit of a certain stationary clutter of the present invention.

[0031] Figure 6 It is a distribution curve diagram of the moving index-sample number of the moving target VS the stationary clutter of the present invention.

[0032] Figure 7 It is a curve diagram of the estimated signal-to-noise ratio of a target-free beam in a certain period B of the present invention. DETAILED DESCRIPTION

[0033] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. The advantages and features of the present invention will become more apparent from the following description. It should be noted that the accompanying drawings are in very simplified form and in non-precise proportions, and are only used to conveniently and clearly assist in explaining the purpose of the embodiments of the present invention.

[0034] like Figure 1 As shown, an embodiment of the present invention provides a method for extracting dynamic and static attribute recognition features of active sonar targets based on tracking trajectory azimuth distance information, which specifically includes:

[0035] It is known that the observation distance sequence vector R = [r1, r2, ..., r N ], observation position sequence A = [a1, a2, ..., a N ] and the corresponding observation time sequence T = [t1, t2, ..., t N ], where N is the number of observation sequence points. Without loss of generality, the observation time sequence refers to its own starting time, that is, t1=0.

[0036] 1) Matlab software (version MATLAB R2018b) is used to perform adaptive quadratic polynomial least squares fitting on the observation sequence. The Matlab library functions called are polyfit and polyval. The Matlab code of the specific implementation method is as follows:

[0037] [P R ] = polyfit(T,R,2);

[0038] R1=polyval(P R ,T);

[0039] [P A ] = polyfit(T,A,2);

[0040] A1=polyval(P A ,T);

[0041] 2) To avoid the influence of gross errors on the observed data, when the number of trajectory points exceeds 15 (including 15), a maximum error observation value is taken for every 15 points (rounded up if less than 15) and replaced with the fitting value and refitted. When the number of trajectory points is less than 15, no gross error correction is performed and no refitting is performed. The number 15 is obtained based on data processing experience and is the experience value of the patent holder. The Matlab library function called is the sorting function sort, and the Matlab code of the specific implementation method is as follows:

[0042]

[0043] 3) Calculate the standard deviation between the fitted value of the distance / azimuth sequence and the observed value after gross error correction, denoted as σ R ,σ A , the Matlab library function called is the standard deviation solution function std, and the Matlab code of the specific implementation method is as follows:

[0044] σ R =std(R2-R X );

[0045] σ A =std(A2-A X );

[0046] 4) Calculate the ratio between the peak-to-peak change value (steady-state variable) and the standard deviation on the fitting curve, and record them as the distance change index C R , Azimuth change index C A The Matlab library functions called are the maximum value function max and the minimum value function min. The Matlab code of the specific implementation method is as follows:

[0047]

[0048] 5) Distance change index C R , Azimuth change index C A The larger value of is taken as the target movement index M, that is, M = max(C R ,C A ).

[0049] By applying the above implementation method, the movement index M of any tracked target can be calculated, and the quantified value can be used as the dynamic and static attribute identification feature of the tracked target to support the recognition of low-speed moving targets.

[0050] The active sonar tracking target contains moving targets and stationary clutter echoes. The range / azimuth observation sequence of the tracking target ( Figure 2-Figure 5 The quadratic polynomial least squares curve fitting ( Figure 2-Figure 5 The upper and lower error limits of each fitting value are shown in the figure according to the three times mean square deviation value. Figure 2-Figure 5 In order to compare the differences between the moving target and the stationary clutter in the azimuth / range observation sequence and the fitting curve, the azimuth / range observation ranges of the moving target and the stationary clutter in the drawing are unified with reference to the larger change, and the time is taken as the starting point of each trajectory as the zero time.

[0051] Figure 2-Figure 5 The distribution of the azimuth / distance observation curve and fitting curve of a typical moving target VS stationary clutter interference when being continuously tracked is presented. From the graphical results, we can directly and concretely feel that the azimuth / distance changes of moving targets and stationary clutter are significantly different, that is, the ratio between the steady-state change (peak-to-peak change on the fitting curve) and the transient uncertainty (the error value between the observed sampling value and the fitting curve) is higher for moving targets and lower for stationary clutter.

[0052] According to the method proposed in the present invention, the mobile index value of the tracked target is calculated, and the mobile index-sample number distribution curve of the moving targets and stationary clutter with labels in the sonar marine acquisition database is statistically analyzed, such as Figure 6 As shown, the movement index is the horizontal axis, and the number of tracked target samples with a certain movement index is the vertical axis. Since the sonar observation data has random errors, the distribution of the statistical curve is random. Figure 6 It shows that the mobility index of moving targets and stationary clutter is significantly different, and the distribution curve of mobility index-sample number is approximately a unilateral Rayleigh distribution curve.

[0053] Since moving targets and stationary clutter have different distributions in the mobility index, a certain value can be set as the decision threshold of the mobility index. When the mobility index of a target is higher than the threshold, it is judged as a moving target. Conversely, when the mobility index of the target is lower than the threshold, it is judged as a stationary target.

[0054] Figure 7 The variation curves of missed alarm rate, false alarm rate and confidence level when the moving index decision threshold changes are given. Missed alarm means that it is actually a moving target but is judged as stationary clutter; false alarm means that it is actually stationary clutter but is judged as a moving target. The missed alarm rate is defined as the number of missed alarms divided by the total number of real moving target samples; the false alarm rate is defined as the number of false alarms divided by the total number of reported moving targets; the confidence level is defined as the number of correct reports in the reported moving targets divided by the total number of reported moving targets. The false alarm rate and confidence level are complementary, and their sum is always 100%. In engineering applications, efforts are made to simultaneously take into account low false alarms and low missed alarms, or high confidence and low missed alarms. Sonar identification of underwater moving targets is a current global problem. It is of positive significance if a certain feature has a confidence level of more than 50% in the moving / static target recognition rate. The confidence level of the method proposed in the present invention far exceeds 50%. Figure 7 As shown in the figure, it is proved that the mobility index is feasible and effective as a feature for distinguishing moving targets from stationary clutter. The selection of the specific mobility index threshold should be considered comprehensively based on the missed alarm rate and false alarm rate required in the actual project.

[0055] Figure 2-Figure 7 The results show that the mobile index measurement method proposed in the present invention can quantitatively express the difference characteristics between moving targets and stationary clutter interference, and the use of mobile index is helpful for moving target / stationary clutter identification. The present invention is innovative and has great engineering value.

[0056] The above description is only a description of the preferred embodiments of the present invention, and is not intended to limit the scope of the present invention. Any changes or modifications made by a person skilled in the art in the field of the present invention based on the above disclosure shall fall within the scope of protection of the claims.

Claims

1. A method for extracting dynamic and static attribute recognition features of active sonar targets based on tracking trajectory azimuth and distance information, characterized in that: The steps include: Step S1: firstly perform quadratic polynomial least squares fitting on the azimuth and distance information on the tracking trajectory respectively; Step S2: Eliminate the gross observation error and then evaluate the change index of the azimuth and distance information on the tracking trajectory; Step S3: Finally, the target movement index is extracted by integrating the direction and distance change index as the target dynamic and static attribute identification feature.

2. The method for extracting dynamic and static attribute recognition features of active sonar targets based on tracking trajectory azimuth distance information as claimed in claim 1, characterized in that: The step S1 specifically includes: using Matlab software to perform adaptive quadratic polynomial least squares fitting on the observation sequence, and the Matlab library functions called are polyfit and polyval.

3. The method for extracting dynamic and static attribute recognition features of active sonar targets based on tracking trajectory azimuth distance information as claimed in claim 2, characterized in that: The step S1 also includes: defining an observation distance sequence vector R=[r1, r2, ..., r N ], observation position sequence A = [a1, a2, ..., a N ] and the corresponding observation time sequence T = [t1, t2, ..., t N ], where N is the number of observation sequence points; the observation time sequence refers to its own starting time, that is, t1=0; the active sonar tracking target refers to the position of the ship at the time of sonar emission with reference to the same point.

4. The method for extracting dynamic and static attribute recognition features of active sonar targets based on tracking trajectory azimuth distance information as claimed in claim 1, characterized in that: Eliminating gross observation errors in step S2 specifically includes: when the number of trajectory points is greater than or equal to 15 points, taking a maximum error observation value for every 15 points and replacing it with a fitting value and refitting, and when the remaining trajectory points are less than 15 points, rounding is performed; when the number of trajectory points is less than 15, no gross error correction processing is performed, and no refitting is performed; the Matlab library function called is the sorting function sort.

5. The method for extracting dynamic and static attribute recognition features of active sonar targets based on tracking trajectory azimuth distance information as claimed in claim 1, characterized in that: The evaluation of the azimuth and distance information change index on the tracking trajectory in step S2 specifically includes: calculating the standard deviation between the distance / azimuth sequence fitting value and the observed value after gross error correction, denoted as σ R ,σ A , the Matlab library function called is the standard deviation solution function std; The peak change value on the fitting curve is calculated as the ratio between the steady-state variable and the standard deviation, which are recorded as the distance change index C R , Azimuth change index C A The Matlab library functions called are the maximum value function max and the minimum value function min.

6. The method for extracting dynamic and static attribute recognition features of active sonar targets based on tracking trajectory azimuth distance information as claimed in claim 5, characterized in that: The step S3 specifically includes: using the distance change index C R , Azimuth change index C A The larger value of is taken as the target movement index M, that is, M = max(C R ,C A ).

7. The method for extracting dynamic and static attribute recognition features of active sonar targets based on tracking trajectory azimuth distance information as claimed in claim 1, characterized in that: When performing quadratic polynomial least squares curve fitting, the upper and lower error limits corresponding to each fitting value are indicated according to 3 times the mean square deviation value while fitting.

8. The method for extracting dynamic and static attribute recognition features of active sonar targets based on tracking trajectory azimuth distance information as claimed in claim 1, characterized in that: A value is set as a decision threshold of the movement index. When the movement index of a target is higher than the decision threshold, it is judged as a moving target. Conversely, when the movement index of the target is lower than the decision threshold, it is judged as a stationary target.

9. The method for extracting dynamic and static attribute recognition features of active sonar targets based on tracking trajectory azimuth distance information as claimed in claim 8, characterized in that: When the decision threshold of the mobile index changes, it also includes missed alarm rate, false alarm rate and confidence parameters; wherein the missed alarm means that the target is actually a moving target but is judged as stationary clutter; the false alarm means that the target is actually stationary clutter but is judged as a moving target; the missed alarm rate is defined as the number of missed alarms divided by the total number of real moving target samples; the false alarm rate is defined as the number of false alarms divided by the total number of reported moving targets; the confidence is defined as the number of correct reports in the reported moving targets divided by the total number of reported moving targets; the false alarm rate and the confidence are complementary, and the sum of the two is always 100%.

Images