Low false alarm target detection method based on modal decomposition

Abstract

The invention provides a low false alarm target detection method based on modal decomposition, and the method comprises the steps: carrying out the coherent demodulation of a continuous pulse radar echo signal, obtaining a two-dimensional data matrix formed by baseband data, carrying out the obtaining of fast and slow time dimension variables of the two-dimensional data matrix, and carrying out the detection of a moving target; the OS-CFAR is adopted to detect target echoes, and a range gate where the detected target is located is indexed; and step S300, processing the signal based on a complex mode decomposition method to obtain a sub-mode.

Description

Technical Field

[0001] The present invention relates to a radar tracking technology, in particular to a low false alarm target detection method based on modal decomposition. Background Art

[0002] Conventional clutter suppression algorithms are mainly moving target indication (MTI) and moving target detection (MTD) algorithms. The core idea of MTI for clutter suppression is to utilize the phase consistency of stationary target echoes, subtract clutter information by canceling successive pulses, and form a notch at zero frequency to achieve clutter suppression. Its advantage lies in the simple and effective implementation of the algorithm, while its drawback is that the notch at zero frequency has a certain width, which will attenuate slow targets near zero frequency. Therefore, MTI is not suitable for detecting slow and weak targets such as unmanned aerial vehicles. The clutter suppression algorithm based on subspace decomposition can achieve better suppression effects, but it must be based on the correct estimation of the clutter subspace, which requires the support of certain prior information. Otherwise, the algorithm performance will be severely degraded. Summary of the Invention

[0003] The present invention provides a low false alarm target detection method based on modal decomposition, including:

[0004] Step S100: For a two-dimensional data matrix formed by baseband data obtained after coherent demodulation of continuous pulse radar echo signals, obtain variables in fast and slow time dimensions and then perform moving target detection;

[0005] Step S200: Use OS-CFAR to detect target echoes and index the range gates where the detected targets are located;

[0006] Step S300: Process the signal based on the complex modal decomposition method to obtain sub-modalities.

[0007] Further, step S300 includes:

[0008] Step S301: Set the conditions satisfied by the intrinsic mode function:

[0009] Step S302: Set the number of loops K, let the iteration number j = 1, and determine the minimum and maximum values of the iterative signal h j (t) in different directions;

[0010] Step S303: Use cubic spline interpolation functions to connect the maximum value points in the same direction to form an upper envelope Connect the minimum value points to form a lower envelope

[0011] Step S304: Obtain the mean value m j (t)

[0012]

[0013] Step S305, subtract the mean value m j (t) from the iterative signal h j (t) to obtain a new signal h j+1 (t)

[0014] h j+1 (t) = h j (t) - m j (t) (2)

[0015] Step S306, if h j (t) does not meet the establishment conditions of the intrinsic mode function, go to step 302, j = j + 1, h j+1 (t) ← h j (t); if the conditions are met, go to step S307;

[0016] Step S307, calculate the remaining residue r i (t)

[0017] r i (t) = r i-1 (t) - h j (t) (3)

[0018] where i = 1, 2,..., N, and N is the number of sub - modes included in the original signal; if the number of extreme points of r i (t) in one direction is not less than 3, go to step S302, j = j + 1, and let h j+1 (t) ← r i (t); if the number of extreme points of r i (t) in all directions is less than 3, or j = K, output r i (t).

[0019] Furthermore, the conditions that the intrinsic mode function satisfies in step S301 include:

[0020] (1) When the complex signal completes one screening of the IMF, in at least one direction, the number of its extreme points does not exceed 2;

[0021] (2) In all directions, the mean value of the upper envelope line formed by local maximum points and the lower envelope line formed by local minimum points tends to zero.

[0022] Furthermore, for the first iteration in step S302, let the iterative signal h 1 (t) = x(t), where x(t) is the original signal, and initialize the residue r 0 (t) = x(t).

[0023] The present invention can complete clutter suppression and target detection without any prior information and built-in library functions, can greatly reduce the number of clutter traces, maintain a high detection probability in complex scenarios, and always maintain the lowest false alarm probability, reducing the subsequent data processing pressure.

[0024] The present invention will be further described below in conjunction with the accompanying drawings of the specification. Description of the Drawings

[0025] Figure 1 It is a schematic diagram of the method flow of the present invention.

[0026] Figure 2 It is a comparison diagram of the detection performance when there is interference only in the Doppler dimension for each algorithm, where (a) is the target detection probability curve diagram and (b) is the clutter false alarm probability curve diagram.

[0027] Figure 3 It is a comparison diagram of the detection performance when there is interference in both the range dimension and the Doppler dimension for each algorithm, where (a) is the target detection probability curve diagram and (b) is the clutter false alarm probability curve diagram.

[0028] Figure 4 It is a schematic diagram of the sorting result of the detection probability for each algorithm.

[0029] Figure 5 It is a schematic diagram of the sorting result of the false alarm probability for each algorithm. Detailed Embodiment

[0030] Combined with Figure 1 , a low false alarm target detection method based on modal decomposition, includes:

[0031] Step S100, for a two-dimensional data matrix formed by baseband data obtained after coherent demodulation of continuous pulse radar echo signals, obtain the variables of the fast and slow time dimensions of the two-dimensional data matrix and then perform moving target detection (MTD);

[0032] Step S200, use OS-CFAR to detect the target echo and index the range gate where the target is located after detection;

[0033] Step S300, process the signal based on the complex empirical mode decomposition (CEMD) method to obtain sub-modalities.

[0034] Conventional detection algorithms cannot distinguish between clutter and targets, resulting in a large number of false alarm points. By separating the clutter and target sub-modalities through modal decomposition, the clutter components can be suppressed, and only the target components are retained, achieving low false alarms.

[0035] In step S100, the fast time dimension variable (FastTime) refers to the time when the signal processor samples, processes, and demodulates the signal after the radar receives the echo signal; in the fast time variable dimension, the radar system usually measures in nanoseconds or microseconds. Fast time is mainly used to measure the distance between the target and the radar or range finding. The slow time dimension variable (SlowTime) refers to the time interval between multiple echo signals received by the radar system within a period of time; in the slow time dimension, the radar system usually measures in milliseconds or seconds; slow time is mainly used to observe the motion characteristics of the target, such as the speed, acceleration, and motion direction of the target. When processing the fast time dimension and slow time dimension data, the data is first windowed and then Fourier transformed. During the processing of the fast time dimension variable, each row in the two-dimensional data matrix obtained after processing the pulsed radar echo signal corresponds to a continuous sampling of a pulsed echo, that is, a continuous range gate.

[0036] In step S100, MTD is a technique that uses a set of narrowband Doppler filters to separate targets with different speeds, and its main basis is that the Doppler frequency shifts generated by different speeds are different. When the signal after pulse compression processing passes through the Doppler filter bank, that is, when performing velocity dimension FFT, due to the different Doppler frequency shifts generated by different speeds, the targets at each speed will fall into the corresponding Doppler channels, and then multiplied by the velocity resolution, from which the target velocity can be calculated.

[0037] In step S200, the purpose of indexing the range gate where the detected target is located is to reduce the number of subsequent complex modal decomposition operations and reduce the computational amount. After the first-level detection, the range position of the detected target is cached, and indexing specific range cells is used to detect the target velocity in subsequent complex modal decomposition, rather than doing it for each range cell.

[0038] The specific process of step S300 includes:

[0039] Step S301, set the conditions that the intrinsic mode function (IMF) satisfies:

[0040] (1) When the complex signal completes one screening of the IMF, in at least one direction, the number of extreme points does not exceed 2;

[0041] (2) In all directions, the mean of the upper envelope line formed by local maximum points and the lower envelope line formed by local minimum points tends to zero;

[0042] For condition (2) which is the main judgment condition for screening the mode, in this embodiment, it is set that both conditions (2) are satisfied;

[0043] Step S302, set the number of loops K, let the iteration number j = 1, and determine the iteration signal h j(t) Minima and maxima in different directions. Let the number of directions be M, then the signals in each direction are m = 1, 2, ..., M, representing the real - signal projection of the complex signal in the m - th direction; for the first iteration, let the iterative signal h 1 (t) = x(t), where x(t) is the original signal, and initialize the residual r 0 (t) = x(t);

[0044] Step S303, use the cubic spline interpolation function to connect the maximum points in the same direction to form an upper envelope Connect the minimum points to form a lower envelope

[0045] Step S304, calculate the mean value m j (t)

[0046]

[0047] Step S305, subtract the mean value m j (t) from the iterative signal h j (t) to obtain a new signal h j+1 (t)

[0048] h j+1 (t) = h j (t) - m j (t) (2)

[0049] Step S306, if m j (t) does not tend to 0 and the number of extreme - value points of h j (t) in each direction is greater than 2, it means that h j (t) does not meet the establishment conditions of the intrinsic mode function (IMF), go to step 302, j = j + 1, h j+1 (t) ← h j (t); if the conditions are met, go to step S307;

[0050] Step S307, calculate the remaining residual r i (t)

[0051] r i (t) = r i-1 (t) - h j (t) (3)

[0052] where i = 1, 2, ..., N, and N is the number of sub - modes included in the original signal; if r i(t) There are at least 3 extreme points, go to step S302, j = j + 1, and let h j+1 (t) ← r i (t); if the number of extreme points of r i (t) in all directions is less than 3, or j = K, output r i (t).

[0053] Comparative example

[0054] This comparative example mainly compares the traditional constant false alarm detection algorithm and the statistical ordered cascaded complex modal decomposition algorithm, and evaluates the algorithm performance with the detection probability curve, false alarm probability curve and the number of clutter traces in the measured data. For the theoretical performance of the algorithm, the detection probability and the corresponding false alarm probability of a single target in a simulated clutter environment are analyzed and compared. There are 7 constant false alarm algorithms participating in the comparison, namely one-dimensional OS, CA, GO, OS-CA cascaded algorithm and two-dimensional OS, OS + Doppler peak search and the OS-CEMD detection algorithm of this embodiment. All algorithms are subjected to 200 Monte Carlo experiments. From Figure 2 、 Figure 3 It can be seen that when there is only clutter interference in the Doppler dimension adjacent to the target, the target detection probability of OS-CEMD is similar to that of one-dimensional CFAR, with only partial performance loss, but the false alarm probability remains the lowest. In the case where there is clutter interference in both the range dimension and the Doppler dimension, the target detection probability of OS-CEMD is the highest and the false alarm probability is also the lowest, with good target detection performance and clutter suppression performance.

[0055] Sample 100 groups of UAV target data frames, and the target signal-to-clutter ratio ranges from 13dB to 22dB. Analyze the measured target detection probability of each algorithm, and the statistical results are as Figure 4 、 Figure 5 shown. From the statistical results, it can be seen that among the 100 groups of data, the detection probability of OS-CEMD-CFAR is 94%, which is 3% lower than the 97% detection probability of OS-CFAR; while the detection probability of OS-Doppler peak search is 87%, and the detection probability of two-dimensional OS is 78%, both lower than OS-CEMD-CFAR. In terms of clutter suppression rate, based on the number of clutter traces detected by OS-CA cascaded CFAR, the average clutter suppression probability of OS-CEMD-CFAR is as high as 88.12%, with a suppression rate of 79.09% relative to one-dimensional OS-CFAR and a suppression rate of 40.60% relative to two-dimensional OS-CFAR.

Claims

1. A low false alarm target detection method based on modal decomposition, characterized in that: include: Step S100, after coherent demodulation of the continuous pulse radar echo signal, a two-dimensional data matrix formed by baseband data is obtained, and the fast and slow time dimension variables of the two-dimensional data matrix are obtained to perform moving target detection; Step S200, using OS-CFAR to detect target echo, and indexing the range gate where the detected target is located; Step S300: Process the signal based on a complex mode decomposition method to obtain sub-modes.

2. The method according to claim 1, characterized in that: Step S300 includes: Step S301, setting the conditions satisfied by the intrinsic mode function: Step S302, set the number of loops K, set the number of iterations j = 1, and determine the iteration signal h j (t) Minima and maxima in different directions; Step S303: Use the cubic spline interpolation function to connect the maximum points in the same direction to form an upper envelope. Connect the minimum points to form the lower envelope Step S304, calculate the mean value m of all upper and lower envelopes j (t) Step S305, using the iteration signal h j (t) minus the mean m j (t), and get a new signal h j+1 (t) h j+1 (t)=h j (t)-m j (t) (2) Step S306, if h j (t) The condition for the establishment of the intrinsic mode function is not satisfied, go to step 302, j = j + 1, h j+1 (t)←h j (t); If the conditions are met, go to step S307; Step S307, step S307, calculate the remaining residual r i (t) r i (t)=r i-1 (t)-h j (t) (3) Where i = 1, 2, ..., N, N is the number of sub-modes contained in the original signal; if there is a direction r i (t) If there are no less than 3 extreme points, go to step S302, j = j + 1, and let h j+1 (t)←r i (t); if all directions r i (t) has less than 3 extreme points, or j = K, output r i (t).

3. The method according to claim 2, characterized in that The conditions satisfied by determining the intrinsic mode function in step S301 include: (1) After the complex signal completes an IMF screening, the number of its extreme points in at least one direction does not exceed 2; (2) In all directions, the mean of the upper envelope formed by the local maximum points and the lower envelope formed by the local minimum points tends to zero.

4. The method according to claim 3, characterized in that: In step S302, for the first iteration, let the iterated signal h1(t)=x(t), x(t) is the original signal, and initialize the residual r0(t)=x(t).

Images