High resolution automatic reconnaissance method based on time-frequency analysis

Abstract

The application discloses a high-resolution automatic reconnaissance method based on time-frequency analysis, comprising the following steps: detecting pulse segments of the normalized power spectrum of each beam; according to the characteristic that the azimuth and frequency segment continuously change between adjacent segments of the same pulse signal, splicing the detected pulse segments to obtain spliced pulse segments; correcting the pulse segments; obtaining the characteristic parameters of the spliced pulse segments as reference characteristic parameters; comparing the characteristic parameters of the new pulse with the reference characteristic parameters; the high-resolution automatic reconnaissance method based on time-frequency analysis provided by the application completes splicing according to the information relationship of the pulse segments to form spliced pulse segments, and the frequency parameters and time parameters of the pulse segments are corrected, so that the time domain and frequency domain resolutions are ensured at the same time; the characteristic parameters of the spliced pulse segments are obtained as the reference characteristic parameters, the characteristic parameters of the new pulse are compared with the reference characteristic parameters, and the tracking and form discrimination of the new pulse are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of sonar pulse detection, in particular to a high-resolution automatic detection method based on time-frequency analysis. BACKGROUND

[0002] In the prior art, the start and end time of the pulse is detected by time-domain energy accumulation, and the pulse signal obtained by detection is subjected to Fourier transform (FFT) to obtain pulse frequency domain information. The detection method based on time-domain energy detection has low time-domain resolution, and although accurate frequency domain parameter estimation can be obtained, the time-frequency variation characteristics of the pulse cannot be obtained. The detection method based on short-time Fourier transform (STFT) segments the input time-domain signal and performs STFT transform, detects the start and end segment of the pulse to determine the start and end of the pulse, and then obtains the frequency information of each segmented pulse at the corresponding time through the STFT spectrum, and combines to obtain the time-frequency characteristics of the complete pulse. Although the complete time-frequency characteristics of the pulse can be obtained, the time-domain and frequency-domain resolution cannot be considered at the same time. SUMMARY

[0003] The main purpose of the present application is to provide a high-resolution automatic detection method based on time-frequency analysis, which aims to solve the problem that in the pulse detection method, the complete time-frequency characteristics of the pulse can be obtained, but the time-domain and frequency-domain resolution cannot be considered at the same time.

[0004] In order to achieve the above-mentioned purpose, the present application provides a high-resolution automatic detection method based on time-frequency analysis, comprising:

[0005] Step 1: performing fast Fourier transform on the received beam-time domain data to obtain beam-frequency domain data, and performing preprocessing to obtain the normalized power spectrum of each beam;

[0006] Step 2: detecting the pulse segment of the normalized power spectrum of each beam;

[0007] Step 3: according to the characteristics of the continuous change of the azimuth and frequency segment between adjacent segments of the same pulse signal, splicing the detected pulse segment to obtain a spliced pulse segment;

[0008] Step 4: correcting the pulse segment;

[0009] Step 5: obtaining the feature parameters of the spliced pulse segment as reference feature parameters;

[0010] Step 6: comparing the feature parameters of the new pulse with the reference feature parameters to realize the tracking and form discrimination of the new pulse.

[0011] Further, the step 1 comprises:

[0012] Step 1.1: receiving MxN point beam-time domain data X m(n), where n = 0,...,N-1, m = 0,...,M-1;

[0013] Step 1.2: Fast Fourier transform is performed on the beam-time domain data;

[0014] Step 1.3: Power spectrum estimation is performed on the fast Fourier transformed beam-time domain data;

[0015] Step 1.4: Normalized power spectrum of each beam is calculated according to the power spectrum estimation.

[0016] Further, in the step 1.4, the normalized power spectrum of each beam is calculated in the following way:

[0017]

[0018] where CPX m (n) is the mth beam noise power spectrum estimation, PX i (n) is the current power spectrum, which is calculated in the following way: PX m (n) = |FXm(n)| 2 and FX m (n) = FFT(X m (n)).

[0019] Further, the criterion for detection in the step 2 is:

[0020] The current signal-to-noise ratio estimation of each beam is obtained:

[0021]

[0022] The pulse segment detection criterion of each beam is run:

[0023]

[0024] Further, the step 2 further includes:

[0025] If the detection fails, the beam noise power spectrum estimation corresponding to the beam is updated, where the updating way is:

[0026]

[0027] Further, the step 3 includes:

[0028] The pulse segments detected in the current batch are matched with the pulse segments in the previous batch, taking the pulse segment corresponding beam Nb(d) and frequency point Nf(d) as the features: if the matching passes, the pulse segments are stored in the same pulse splicing group, and the information of the pulse splicing group is updated; if the matching fails, the pulse segments are stored in a new pulse splicing group, and the information of the new pulse splicing group is updated.

[0029] Further, the information of the pulse stitching group in step 3 includes the azimuth parameter of the pulse segment, and the calculation method of the azimuth parameter of the pulse segment is as follows:

[0030] The input pulse segment parameters frequency point Nf(d), beam Nb(d) and beam power spectrum PX m (n) are used to obtain the corresponding frequency point power PD2(d) and the corresponding azimuth θ2(d) of each pulse segment, and the power PD1(d), PD3(d) and the azimuth θ1(d), θ3(d) of the corresponding frequency point of the front and rear beams, and the parabolic interpolation method is used to obtain the azimuth θ d of the pulse segment.

[0031] Further, the correction method of the frequency parameter in step 4 is as follows:

[0032] The frequency point to which the pulse segment belongs is M, and the corresponding time domain data is S d (n), n=0,...,N-1;

[0033] A narrowband filter with a length of N / 2 is constructed:

[0034]

[0035] After filtering the time domain data S d (n), the SF d (k) is obtained.

[0036]

[0037] The pulse frequency d is obtained by using the rotation invariance of SF .

[0038]

[0039] Further, step 7 is included after step 6:

[0040] If the characteristic parameters of the new pulse match the reference characteristic parameters, the reference characteristic parameters are updated with the characteristic parameters of the new pulse.

[0041] Further, step 8 is also included after step 6:

[0042] If the new pulse is tracked only once, the pulse disappearance determination time is set to 100 seconds.

[0043] If the new pulse is tracked more than twice, the pulse disappearance determination time is set to 3 times the pulse period.

[0044] The high-resolution automatic reconnaissance method based on time-frequency analysis provided by the application converts the signal of a beam, then detects a pulse segment, and then completes splicing to form a spliced pulse segment according to the information relationship of the pulse segment, and corrects the frequency parameter and the time parameter of the pulse segment, so as to realize the time domain and frequency domain resolution at the same time. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 is a step schematic diagram of the high-resolution automatic reconnaissance method based on time-frequency analysis of an embodiment of the application;

[0046] Figure 2 is a step 1 schematic diagram of the high-resolution automatic reconnaissance method based on time-frequency analysis of an embodiment of the application;

[0047] Figure 3 is a flowchart of steps 2 to 4 of the high-resolution automatic reconnaissance method based on time-frequency analysis of an embodiment of the application;

[0048] Figure 4 is a flowchart of step 3 of the high-resolution automatic reconnaissance method based on time-frequency analysis of an embodiment of the application;

[0049] Figure 5 is a flowchart of step 4 of the high-resolution automatic reconnaissance method based on time-frequency analysis of an embodiment of the application;

[0050] Figure 6 is a flowchart of the high-resolution automatic reconnaissance method based on time-frequency analysis of an embodiment of the application.

[0051] The implementation, functional features and advantages of the application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0052] It should be understood that the specific embodiments described herein are merely intended to explain the application, and are not intended to limit the application.

[0053] It is to be understood that the singular forms "a," "an," and "the" include plural referents unless the context clearly dictates otherwise. It is further understood that the terms "comprise" and "comprising" and the like, when used in the specification, include all the following conditions: there is a, there are at least one, and the like. It is to be understood that when we say that an element is "connected" or "coupled" to another element, it can be directly connected or coupled to the other element, or intervening elements can be present. Further, "connected" or "coupled" as used herein can include wirelessly connected or wirelessly coupled. As used herein, the term "and / or" includes all combinations of one or more of the associated listed items.

[0054] It is to be understood that the terms so used are intended to encompass common and accepted meanings in the art, unless otherwise explicitly provided. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.

[0055] Reference Figure 1 In one embodiment of the present application, a high-resolution automatic reconnaissance method based on time-frequency analysis includes:

[0056] Step 1: Fast Fourier transform is performed on the received beam-time domain data to obtain beam-frequency domain data, and pre-processing is performed to obtain the normalized power spectrum of each beam;

[0057] Step 2: Pulse segment detection is performed on the normalized power spectrum of each beam;

[0058] Step 3: According to the characteristics of the continuous change of the azimuth and frequency segment between adjacent segments of the same pulse signal, the detected pulse segments are spliced to obtain spliced pulse segments;

[0059] Step 4: The pulse segments are corrected;

[0060] Step 5: The characteristic parameters of the spliced pulse segments are obtained as reference characteristic parameters;

[0061] Step 6: The characteristic parameters of the new pulse are compared with the reference characteristic parameters to realize the tracking and form discrimination of the new pulse.

[0062] In the prior art, the starting and ending time of the pulse is detected by time domain energy accumulation, the pulse signal obtained by detection is subjected to Fourier transform (FFT) to obtain pulse frequency domain information, the reconnaissance method based on time domain energy detection has low time domain resolution, although accurate frequency domain parameter estimation can be obtained, but the pulse time-frequency variation characteristics cannot be obtained. The reconnaissance method based on short-time Fourier transform (STFT) segments the input time domain signal to perform STFT transform, detects the starting and ending segment of the pulse to determine the starting and ending of the pulse, and then obtains the frequency information of each segmented pulse at the corresponding time through the STFT spectrum, and the time-frequency characteristics of the complete pulse are obtained after combination, although the complete time-frequency characteristics of the pulse can be obtained, but the time domain and frequency domain resolution cannot be considered at the same time.

[0063] In the present application, after receiving the beam-time domain data in steps 1-2, fast Fourier transform is performed to obtain beam-frequency domain data, and preprocessing is performed to obtain the normalized power spectrum of each beam, and the pulse segment is selected from the normalized power spectrum of each beam.

[0064] In step 3, the splicing step is performed through the azimuth and frequency band information in the pulse segment; for example, the pulse segment detected in the current batch is matched with the pulse segment in the previous batch by taking the pulse segment corresponding to the beam Nb(d) and the frequency point Nf(d) as the characteristics: if the matching is passed, it is stored in the same pulse splicing group, and the information of the pulse splicing group is updated; if the matching is not passed, it is stored in a new pulse splicing group, and the information of the new pulse splicing group is updated.

[0065] In step 4, the frequency parameter and time parameter can be corrected by using the rotation invariant characteristic, so as to consider the time domain and frequency domain resolution, which is beneficial to the matching and tracking process in the subsequent steps.

[0066] In steps 5-6, the characteristic parameters of the spliced pulse segment are obtained as reference characteristic parameters, and then the characteristic parameters of the new pulse are compared with the reference characteristic parameters to realize the tracking and form discrimination of the new pulse. For example, the criterion for single pulse matching tracking is: Wherein, f c is the center frequency of the new pulse, is the center frequency of the single pulse in the tracker; PB is the wide / narrow band type of the newly detected pulse, is the wide / narrow band type of the single pulse in the tracker; θ w is the azimuth of the newly detected single pulse, is the azimuth of the single pulse in the tracker, is the distance between the two nearest beams, and the unit is degree.

[0067] In summary, after converting the signals of the beams, the pulse segments are detected, then the splicing is completed according to the information relationship of the pulse segments to form spliced pulse segments, the frequency parameters and time parameters of the pulse segments are corrected, so that the time domain and frequency domain resolutions are ensured at the same time; the characteristic parameters of the spliced pulse segments are obtained as reference characteristic parameters, the characteristic parameters of the new pulse are compared with the reference characteristic parameters, and the tracking and form discrimination of the new pulse are realized.

[0068] Reference Figure 2 In one embodiment, the step 1 comprises:

[0069] Step 1.1: receiving MxN point beam-time domain data X m (n), wherein n = 0,..., N-1, m = 0,..., M-1;

[0070] Step 1.2: performing fast Fourier transform on the beam-time domain data;

[0071] Step 1.3: performing power spectrum estimation on the beam-time domain data after the fast Fourier transform;

[0072] Step 1.4: calculating the normalized power spectrum of each beam according to the power spectrum estimation.

[0073] In the embodiment, a sequence for processing the beam-time domain data is proposed, the input beam-time domain data is subjected to Fourier transform (FFT) to convert into beam-frequency domain data; then the power spectrum of each input beam is calculated, the normalized power spectrum is obtained by using the cumulative power spectrum of each beam, and the beam-normalized power spectrum data is output to a single pulse detection and parameter estimation module for subsequent processing.

[0074] In one embodiment, in the step 1.4, the calculation method of the normalized power spectrum of each beam is:

[0075]

[0076] wherein CPX m (n) is the noise power spectrum estimation of the mth beam, PX i (n) is the current power spectrum, and the calculation method is: PX m (n) = |FXm(n)| 2 and FX m (n) = FFT(X m (n)).

[0077] In the embodiment, a specific calculation method for calculating the normalized power spectrum of each beam is given.

[0078] In one embodiment, the criterion for detection in the step 2 is:

[0079] Obtain current SNR estimate of each beam:

[0080]

[0081] Run pulse segment detection decision criterion for each beam:

[0082]

[0083] In this embodiment, a calculation formula is proposed to detect the normalized power spectrum of each beam in the form of SNR, to more accurately distinguish pulse segments and noise. For the passing beam, output the information of the beam to which all pulse segments belong and the frequency point to which the maximum normalized power belongs:

[0084] Beam to which pulse segment belongs: Nb(d), d = 1,...,D

[0085] Frequency point to which pulse segment belongs: Nf(d), d = 1,...,D

[0086] Wherein: D is the total number of pulse segments detected in the current batch data, for use in step 3.

[0087] In one embodiment, the step 2 further comprises:

[0088] If the detection fails, update the beam noise power spectrum estimate corresponding to the beam, wherein the update method is:

[0089]

[0090] In this embodiment, if the detection fails, the beam noise power spectrum estimate is corrected, forming feedback, to reduce the problem of excessive screening.

[0091] Reference Figure 3 and 4 In one embodiment, the step 3 comprises:

[0092] Characterized by the pulse segment corresponding beam Nb(d) and frequency point Nf(d), the pulse segments detected in the current batch are matched with the pulse segments in the previous batch: if the matching passes, store in the same pulse splicing group and update the information of the pulse splicing group; if the matching fails, store in a new pulse splicing group and update the information of the new pulse splicing group.

[0093] In this embodiment, the pulse segment matching criterion is:

[0094]

[0095]

[0096] Wherein, and are the beam information and frequency information of the i-th pulse splicing group, Nb(d) and frequency Nf(d) are the beam information and frequency information corresponding to the new pulse segment, N i The number of pulse segments added to the pulse splicing group.

[0097] If the d-th pulse segment is matched by the i-th pulse splicing group, the information of the i-th pulse splicing group is updated to the beam information and frequency information of the new pulse segment:

[0098] Number of pulse segments: N i =N i +1;

[0099] Belongs to Nth i Pulse frequency point number:

[0100] Belongs to Nth i Pulse segment beam number:

[0101] Belongs to Nth i Pulse segment position:

[0102] Pulse termination batch time domain data:

[0103] If the matching criteria are not met, it is determined that a new pulse has started, and the I+1th pulse splicing group is added, where I is the number of current pulse splicing groups:

[0104] Number of pulses in the splicing group: I=I+1;

[0105] Number of pulse segments: N I =1;

[0106] The frequency point number of the first pulse segment:

[0107] Beam number of the first pulse segment:

[0108] The position of the first pulse segment:

[0109] Pulse start batch time domain data:

[0110] Pulse termination batch time domain data:

[0111] In one embodiment, if the i-th pulse stitching group does not add new pulse segments in the data of the last three batches, it is determined that the pulse stitching of the i-th pulse stitching group is completed. The pulse stitching of the pulse stitching group is determined to be completed by not adding new pulse segments in a certain time.

[0112] In one embodiment, the information of the pulse stitching group in step 3 includes the azimuth parameter of the pulse segment, and the azimuth parameter of the pulse segment is calculated by:

[0113] The input pulse segment parameters frequency point Nf(d), beam Nb(d) and beam power spectrum PX m (n) are used to obtain the corresponding frequency point power PD2(d) and the corresponding azimuth θ2(d) of each pulse segment, and the frequency point power PD1(d) and PD3(d) and the azimuth θ1(d) and θ3(d) of the front and rear beams, and the parabolic interpolation method is used to obtain the azimuth θ d of the pulse segment.

[0114] In this embodiment, a method for calculating the azimuth of the pulse segment is provided, which can accurately realize the calculation and acquisition of the azimuth.

[0115] In one embodiment, the reference feature parameters in step 5 include:

[0116] The average value of the azimuths of the pulse segments in the pulse stitching group is taken as the pulse azimuth , which is calculated as follows:

[0117]

[0118] Referring to Figure 5 , in one embodiment, the correction method of the frequency parameter in step 4 is:

[0119] The frequency point to which the pulse segment belongs is M, and the corresponding time domain data is S d (n), n=0,...,N-1;

[0120] A narrowband filter with a length of N / 2 is constructed:

[0121]

[0122] After filtering the time domain data S d (n), the SF d (k) is obtained:

[0123]

[0124] The pulse frequency is estimated by using the rotation invariance of SF d (k):

[0125]

[0126] In this embodiment, after the filter is constructed, the time domain data S d (n) is smoothed, and then a more accurate pulse frequency d (k) is obtained by using the rotation invariance of SF (k), so that the frequency parameter of the pulse segment is corrected accurately. After the frequency parameter is corrected, the time parameter information, such as the start time and the end time of the pulse segment, can be calculated by using the information of the frequency parameter. For example, the amplitude operation is performed on SF d (k) to obtain:

[0127] ASF d (k), k = 0, 1,..., N / 2 Equation (3)

[0128]

[0129] The first position greater than E / 2 in the ASF d (k) is searched, and is recorded as Ts, which is the start position of the corrected pulse segment; the last position greater than E / 2 is recorded as Te, which is the end position of the corrected pulse segment. The frequency point corresponding to the start pulse segment and the time domain data (k) are input, and the start frequency of the corrected pulse (k) is obtained by using Equations (1) to (4), and the start position Ts of the pulse correction is obtained. (k) are input, and the end frequency of the corrected pulse (k) is obtained by using Equations (1) to (4), and the end position Te of the pulse correction is obtained.

[0130] In one embodiment, the reference characteristic parameters in step 5 include the following single pulse parameters:

[0131] Single pulse start frequency:

[0132] Single pulse end frequency:

[0133] Single pulse center frequency:

[0134] Single pulse wide / narrow band determination:

[0135]

[0136] Pulse width: in seconds;

[0137] Pulse azimuth:

[0138] Pulse start time: PTs = T + Ts; T is system time;

[0139] Pulse end time: PTe = PTs + PW; T is system time.

[0140] In the embodiment, a single pulse parameter acquisition method is provided, which is convenient to implement and has good use effect.

[0141] In one embodiment, the step 6 is followed by step 7:

[0142] If the characteristic parameters of the new pulse match the reference characteristic parameters, the reference characteristic parameters are updated with the characteristic parameters of the new pulse.

[0143] In the embodiment, the criterion for single pulse matching tracking is:

[0144]

[0145]

[0146]

[0147] Wherein: f c is the center frequency of the new single pulse, is the center frequency of the single pulse in the tracker; PB is the wide / narrow band type of the new single pulse, is the wide / narrow band type of the single pulse in the tracker; θ w is the azimuth of the new single pulse, is the azimuth of the single pulse in the tracker, is the distance between the two beams closest to the new single pulse, is the angular interval between the two beams closest to the new single pulse, in degrees.

[0148] If matched, the matching pulse parameters (i.e. reference characteristic parameters) in the tracker are updated, and the following are output:

[0149] Pulse center frequency in the tracker:

[0150] Pulse wide / narrow band in the tracker:

[0151] Pulse pulse width in the tracker:

[0152] Pulse period in the tracker:

[0153] Pulse azimuth in the tracker:

[0154] Pulse end time in the tracker:

[0155] Besides the single pulse matching, the matching of two combined pulses can also be determined according to the matching degree between the sub-pulses constituting the two combined pulses. By using the single pulse matching criterion, if all the sub-pulses contained in one of the combined pulses pass the matching determination, it is determined that the two combined pulses are matched, and each parameter in the combined pulse tracker is updated, and the following is output:

[0156] Number of sub-pulses:

[0157] Center frequency of the i-th sub-pulse:

[0158] i-th sub-wide / narrow band:

[0159] i-th sub-pulse pulse width:

[0160] Combined pulse pulse width:

[0161] Combined pulse azimuth:

[0162] The above updates the setting of the reference characteristic parameters with the characteristic parameters of the new pulse, and good tracking effect can be achieved.

[0163] In one embodiment, the step 6 is followed by a step 8;

[0164] If the new pulse is tracked only once, the pulse disappearance determination time is set to 100 seconds;

[0165] If the new pulse is tracked more than twice, the pulse disappearance determination time is set to 3 times the pulse period.

[0166] In this embodiment, the number of times the new pulse is tracked is used to set the corresponding pulse disappearance determination time; for example, if it is tracked only once, the pulse disappearance determination time is set to 100 seconds, i.e. if the pulse is not detected for 100 consecutive seconds, it is determined to be disappeared and removed from the pulse tracker; if it is tracked more than twice, the pulse disappearance determination time is set to 3 times the pulse period; the above targeted manner can improve the tracking efficiency.

[0167] The high-resolution automatic reconnaissance method based on time-frequency analysis provided by the present application converts the signals of the beam, detects the pulse segments, then completes splicing to form spliced pulse segments according to the information relationship of the pulse segments, corrects the frequency parameters and time parameters of the pulse segments, thereby realizing the simultaneous guarantee of the time domain and frequency domain resolution; the characteristic parameters of the spliced pulse segments are taken as the reference characteristic parameters, the characteristic parameters of the new pulse are compared with the reference characteristic parameters, and the tracking and form discrimination of the new pulse are realized.

[0168] The above merely describes preferred embodiments of the present application, and is not intended to limit the patent scope of the present application, and any equivalent structure or equivalent process conversion made by using the content of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A high-resolution automatic reconnaissance method based on time-frequency analysis, characterized in that: include: Step 1: Perform fast Fourier transform on the received beam-time domain data to obtain beam-frequency domain data, and perform preprocessing to obtain the normalized power spectrum of each beam; Step 2: Perform pulse segment detection on the normalized power spectrum of each beam; Step 3: Based on the characteristics of continuous changes in azimuth and frequency between adjacent segments of the same pulse signal, the pulse segments that have passed the detection are spliced ​​to obtain spliced ​​pulse segments; Step 4: Correct the pulse segment; Step 5: Obtain characteristic parameters of the spliced ​​pulse segment as reference characteristic parameters; Step 6: Compare the characteristic parameters of the new pulse with the reference characteristic parameters to achieve tracking and form discrimination of the new pulse; The step 3 comprises: Using the pulse segment corresponding beam Nb(d) and frequency point Nf(d) as features, the pulse segments detected in the current batch are matched with the pulse segments in the previous batch. If the match is successful, they are stored in the same pulse splicing group and the information of the pulse splicing group is updated. If the match fails, they are stored in a new pulse splicing group and the information of the new pulse splicing group is updated. The frequency parameter correction method in step 4 is: The frequency point of the pulse segment is M, and the corresponding time domain data is S d (n), n=0,...,N-1; Construct a narrowband filter with length N / 2: For time domain data S d (n) Obtain SF after filtering d (k): Using SF d (k) The rotation invariance of the pulse frequency is obtained estimate: 。 2. The high-resolution automatic reconnaissance method based on time-frequency analysis according to claim 1 is characterized in that: The step 1 comprises: Step 1.1: Receive M×N spot beam-time domain data X m (n), where n=0,...,N-1, m=0,...,M-1; Step 1.2: Perform fast Fourier transform on the beam-time domain data; Step 1.3: Perform power spectrum estimation on the beam-time domain data after fast Fourier transformation; Step 1.4: Calculate the normalized power spectrum of each beam based on the power spectrum estimate.

3. The high-resolution automatic reconnaissance method based on time-frequency analysis according to claim 2 is characterized in that: In step 1.4, the normalized power spectrum of each beam is calculated as follows: Among them, CPX m (n) is the noise power spectrum estimate of the mth beam, PX i (n) is the current power spectrum, calculated as: PX m (n)=|FXm(n)| 2 , while FX m (n) = FFT(X m (n)).

4. The high-resolution automatic reconnaissance method based on time-frequency analysis according to claim 3 is characterized in that: The criteria for detection in step 2 are: Get the current signal-to-noise ratio estimate for each beam: Run the detection and decision criteria for each beam pulse segment: 。 5. The high-resolution automatic reconnaissance method based on time-frequency analysis according to claim 4 is characterized in that: After step 2, the following steps are also included: If the test fails, the beam noise power spectrum estimate corresponding to the beam is updated as follows: 。 6. The high-resolution automatic reconnaissance method based on time-frequency analysis according to claim 1, characterized in that: The information of the pulse splicing group in step 3 includes the azimuth parameters of the pulse segment, and the azimuth parameters of the pulse segment are calculated as follows: Using the input pulse segment parameters frequency Nf(d), beam Nb(d) and beam power spectrum PX m (n), obtain the power PD2(d) and azimuth θ2(d) of each pulse segment, as well as the power PD1(d), PD3(d) and azimuth θ1(d), θ3(d) of the corresponding frequency points of the front and rear beams, and use the parabolic interpolation method to obtain the pulse segment azimuth θ d .

7. The high-resolution automatic reconnaissance method based on time-frequency analysis according to claim 1, characterized in that: The step 6 includes the following steps: If the characteristic parameters of the new pulse match the reference characteristic parameters, the reference characteristic parameters are updated with the characteristic parameters of the new pulse.

8. The high-resolution automatic reconnaissance method based on time-frequency analysis according to claim 1, characterized in that: After step 6, step 8 is also included; If the new pulse is tracked only once, the pulse disappearance determination time is set to 100 seconds; If the new pulse is tracked more than twice, the pulse disappearance judgment time is set to 3 times the pulse period.

Images