Cluster target staring detection method based on distance-Doppler two-dimensional sequential filtering

Abstract

The invention provides a cluster target staring detection method based on distance-Doppler two-dimensional sequential filtering, belongs to the technical field of radar detection, and is mainly used for solving the problem of detecting cross-distance-Doppler unit dense group targets by a staring radar. According to the method, firstly, cluster target distance migration is eliminated based on Keystone transformation, and then a potential distance unit of a cluster target is extracted through low-threshold coherent detection, so that the number of subsequent long-time accumulation processing times is reduced; secondly, positioning a highest-energy target on a potential distance unit of the group target with a high detection threshold by using a long-time coherent accumulation method jointly constrained by operation complexity and detection probability, and completing estimation of the distance, the speed and the acceleration of the target; further, a distance-Doppler two-dimensional combined wave gate is designed to eliminate the shielding effect of the target on space adjacent targets in the group; and finally, realizing detection and parameter estimation of all cluster individuals in a sequential filtering mode.

Description

Technical Field

[0001] The present invention relates to the technical field of radar detection. Background Art

[0002] Cluster target detection technology is a key technology that needs to be overcome in the radar field. Generally speaking, the radar should dynamically adjust the system's time, space, frequency and energy resources according to the changes in physical factors such as the RCS, motion characteristics and environment of the cluster target to achieve the best detection performance. The staring method is used to detect individuals in the cluster target group with high resolution in the Doppler domain, which provides a physical basis for the robust detection of dense group targets. However, the radar CPI cycle is long during staring detection, and maneuvering targets can easily cross multiple range-Doppler resolution units within one CPI. Multiple targets are seriously crossed in the range-Doppler domain. The range-Doppler two-dimensional sequential filtering can achieve reliable detection and parameter estimation of multiple aliased targets.

[0003] Typical long-term coherent accumulation methods such as RFT, LVD, FrFT, etc. can accumulate multi-target peaks in the velocity-acceleration domain through parameter search. However, in order to achieve dense multi-target detection, the shielding effect of adjacent multi-target detection must be further considered in subsequent processing, and the computational complexity of such algorithms is extremely high, making them difficult to implement in engineering. Patent CN116559808A discloses a long-term coherent accumulation method with joint constraints on computational complexity and detection probability. The algorithm obtains the optimal parameter search space by optimizing the computational complexity, significantly reducing the algorithm complexity. However, this method is only applicable to single-target gaze detection scenarios. Summary of the invention

[0004] The present invention proposes a cluster target staring detection method based on range-Doppler two-dimensional sequential filtering, which is mainly used to solve the problem of staring radar detecting dense cluster targets across range-Doppler units, and can improve the radar's detection capability for cluster targets.

[0005] The present invention first eliminates the distance migration of cluster targets based on Keystone transformation, and then extracts the potential distance unit of the cluster targets through low-threshold coherent detection to reduce the number of subsequent long-term accumulation processing times; secondly, the long-term coherent accumulation method with joint constraints of computational complexity and detection probability is used to locate the highest energy target on the potential distance unit of the cluster target with a high detection threshold, and complete the distance, speed, and acceleration estimation of the target; further, a two-dimensional range-Doppler joint wave gate is designed to eliminate the shielding effect of the target on the adjacent targets in the space within the cluster; finally, the detection and parameter estimation of all cluster individuals are realized in a sequential filtering manner. The specific steps for realizing the technical solution of the present invention are:

[0006] Step 1: Eliminate cluster target distance migration based on Keystone transformation;

[0007] Step 2: Extract cluster target potential distance unit samples through low threshold coherent detection;

[0008] Step 3: Use a high-threshold gaze coherent detection algorithm to compensate for target crossing loss in the range-Doppler domain, complete the detection of the highest energy target in the potential range unit, and estimate the range, velocity, and acceleration parameters;

[0009] Step 4: 2D joint filtering of the highest energy target in the range-Doppler domain and restoring the signal to the range-pulse domain;

[0010] Step 5: Repeat steps 3 to 4 until all targets within the potential distance unit are filtered out sequentially and the cluster target parameter information is output.

[0011] Compared with the prior art, the present invention has the following significant advantages: it adopts range-Doppler two-dimensional sequential filtering to eliminate the shielding effect of multiple nearby targets in a group, and robustly estimates motion parameters and the number of dense targets; at the same time, it implements Doppler filtering based on a low computational complexity and long-term accumulation algorithm, which is easy to implement in engineering.

[0012] The present invention is described in further detail below in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 is a block diagram of an embodiment of the present invention;

[0014] Figure 2 This is the energy distribution diagram of multiple targets in the range-velocity domain after traditional FFT processing;

[0015] Figure 3 This is the point trace distribution diagram of multiple targets in the range-velocity domain after being processed by the algorithm proposed in the present invention;

[0016] Figure 4 This is the point trace distribution diagram of multi-target velocity-acceleration domain after being processed by the algorithm proposed in the present invention. DETAILED DESCRIPTION

[0017] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific implementation methods, but the protection scope of the present invention is not limited by the implementation regulations.

[0018] The present invention proposes a range-Doppler two-dimensional sequential filtering method to eliminate the shielding effect of multiple nearby targets during the detection process, which has the characteristics of low computational complexity and dense multi-target detection, and can solve the problem of robust detection and parameter estimation of cluster targets. Figure 1 As shown, the present invention comprises the following steps:

[0019] Step 1: Eliminate cluster target distance migration based on Keystone transformation;

[0020] Step 2: Extract cluster target potential distance unit samples through low threshold coherent detection;

[0021] Step 3: Use a high-threshold gaze coherent detection algorithm to compensate for target crossing loss in the range-Doppler domain, complete the detection of the highest energy target in the potential range unit, and estimate the range, velocity, and acceleration parameters;

[0022] Step 4: 2D joint filtering of the highest energy target in the range-Doppler domain and restoring the signal to the range-pulse domain;

[0023] Step 5: Repeat steps 3 to 4 until all targets within the potential distance unit are filtered out sequentially and the cluster target parameter information is output.

[0024] Preferably, step 2 comprises the following steps:

[0025] Step 2-1: According to the upper limit of the accelerometer range a max , calculate the number of low threshold coherent detection pulses N l :

[0026]

[0027] In the formula, is the rounding operation, lb(·) is the logarithm operation with base 2, and f r is the pulse repetition frequency, λ is the radar operating wavelength, N l The value of should accumulate the target energy as much as possible and prevent the target from experiencing range-Doppler migration.

[0028] Step 2-2: According to the number of gaze coherent pulses N h , the radar echo distance-pulse domain data x(τ (n) ,r (m) ) is divided into P segments along the pulse domain. For any p-th segment data x(τ (p,n) ,r (m) ), both along the pulse domain do N l Point coherent accumulation processing:

[0029]

[0030] Where P = N h / N l , n is the pulse domain sample index, r (m) is the distance value corresponding to the mth distance unit, is the frequency value corresponding to the nth Doppler unit, YesN l Point Fast Fourier Transform;

[0031] Step 2-3: Preset the low threshold false alarm probability P lfas, calculate the detection threshold T lfas :

[0032]

[0033] In the formula, x Noise is the range-pulse domain data x(τ (n) ,r (m) )The average amplitude of noise in

[0034] Step 2-4: Traverse All sample points in the search are searched for sample points whose amplitude exceeds the detection threshold T lfas If there is no sample point with a lower threshold, go to step 5. If there is a sample point with a lower threshold, then select the distance unit sequence from the distance-pulse domain data x(τ (n) ,r (m) ) to extract the low threshold detection data Among them, r l is the under-threshold detection data distance unit index;

[0035] Preferably, step 3 comprises the following steps:

[0036] Step 3-1: Detect data that is too low threshold In all distance units, the long-term coherent integration method with joint constraints on computational complexity and detection probability is used to achieve long-term coherent integration along the pulse domain to obtain And the target velocity estimation matrix Acceleration Estimation Matrix in, is the frequency value corresponding to the nth Doppler unit, It is the rth l The distance value corresponding to each distance unit;

[0037] Step 3-2: Set the false alarm probability P to the preset high threshold hfas , calculate the detection threshold T hfas :

[0038]

[0039] In the formula, x′ Noise Is too low threshold distance-pulse domain data The average amplitude of the noise in the phase coherent integration algorithm is α, which is the dynamic adjustment factor of the noise power. This value is used to adjust the influence of nonlinear operation on the noise power in the long-term coherent integration algorithm.

[0040] Step 3-3: Traversal All sample points in the search are searched for sample points whose amplitude exceeds the detection threshold T hfasIf there is no sample point with a high threshold, go to step 5. If there is a sample point with a high threshold, go to step 5 according to the distance unit sequence. Extracting over-threshold detection data Among them, r h is the over-threshold detection data distance unit index;

[0041] Step 3-4: For each distance unit r h , the Doppler unit index where the spectrum peak searches for the highest energy target Record speed estimates separately Acceleration estimate

[0042] Step 3-5: For each range cell, find the Doppler cell where the highest energy target is located. Set the distance-related gate D r , for the data in the range gate that meets the association criteria, calculate the target distance estimate:

[0043]

[0044] In the formula, is the distance correlation gate D r The index of the dth distance unit in the ;

[0045] Preferably, step 4 comprises the following steps:

[0046] Step 4-1: For each distance unit r h , with the distance to the highest energy target - Doppler index As the center, calculate the distance to filter the gate index boundary and the Doppler filter gate size Where Δr is the distance resolution unit, Δr s is the distance sampling unit, Δf d is the Doppler resolution unit, Δf ds is the Doppler sampling unit, β r , β fd are the distance, Doppler filter gate expansion factor, β r , The value is a compromise between range-Doppler resolution and target filtering effect;

[0047] Step 4-2: For each distance unit r h , with range-Doppler index As the center, fill the distance and Doppler filter gate data with noise to obtain the filtered data

[0048] Step 4-3: For each distance unit rh , do N along the Doppler domain h Point IFFT processing restores the signal to the range-pulse domain;

[0049] Preferably, step 5 comprises the following steps:

[0050] If the number of filtered targets in step 4-2 has reached the maximum processable target limit, or no target has passed the detection threshold in steps 2-4 and 3-3, the distance, speed, and acceleration estimates of all detected targets are output; otherwise, go to step 3.

[0051] Example:

[0052] Assume that the pulse width of the radar transmission waveform is 10μs, the bandwidth is 15MHz, the wavelength is 0.01m, the complex sampling rate is 30Msps, the number of coherent pulses is 2048, the pulse repetition period is 200μs, and the simulation sets 4 targets with radial distances of [2500, 2510, 2520, 2530]m, radial velocities of [5, 5.1, 5.2, 5.3]m / s, and radial accelerations of [5, 5, 5, 5]m / s 2 , the signal-to-noise ratio before accumulation is 0dB, the number of low threshold detection pulses is N l is 128, the low threshold false alarm probability P lfas For 10 -3 , high threshold false alarm probability P hfas For 10 -6 , the noise power dynamic adjustment factor α is 0.676, and the distance filter gate expansion factor β r =1, Doppler filter gate expansion factor β fd is 1.

[0053] The energy distribution of multiple targets in the range-velocity domain after conventional FFT processing is as follows: Figure 2 As shown in the figure, it can be seen that the energy of the four targets is seriously aliased and difficult to detect effectively; after being processed by the algorithm proposed in the present invention, the distribution of multi-target distance-velocity domain points is as follows Figure 3 The velocity-acceleration domain point distribution is shown as Figure 4 As shown, it can be seen that the algorithm proposed in the present invention can realize reliable detection of dense multi-targets and accurate estimation of motion parameters and the number of cluster individuals.

Claims

1. A cluster target gaze detection method based on range-Doppler two-dimensional sequential filtering, characterized in that: The steps include: Step 1: Eliminate cluster target distance migration based on Keystone transformation; Step 2: Extract cluster target potential distance unit samples through low threshold coherent detection; Step 3: Use a high-threshold gaze coherent detection algorithm to compensate for target crossing loss in the range-Doppler domain, complete the detection of the highest energy target in the potential range unit, and estimate the range, velocity, and acceleration parameters; Step 4: 2D joint filtering of the highest energy target in the range-Doppler domain and restoring the signal to the range-pulse domain; Step 5: Repeat steps 3 to 4 until all targets within the potential distance unit are filtered out sequentially and the cluster target parameter information is output.

2. The method for cluster target gaze detection based on range-Doppler two-dimensional sequential filtering according to claim 1, characterized in that: The step 2 comprises the following steps: Step 2-1: According to the upper limit of the accelerometer range a max , calculate the number of low threshold coherent detection pulses N l : In the formula, is the rounding operation, lb(·) is the logarithm operation with base 2, and f r is the pulse repetition frequency, λ is the radar operating wavelength; Step 2-2: According to the number of gaze coherent pulses N h , the radar echo distance-pulse domain data x(τ (n) ,r (m) ) is divided into P segments along the pulse domain. For any p-th segment data x(τ (p,n) ,r (m) ), along the pulse domain do N l Point coherent accumulation processing: Where P = N h / N l , n is the pulse domain sample index, r (m) is the distance value corresponding to the mth distance unit, is the frequency value corresponding to the nth Doppler unit, YesN l Point fast discrete Fourier transform; Step 2-3: Preset the low threshold false alarm probability P lfas , calculate the detection threshold T lfas : In the formula, x Noise is the range-pulse domain data x(τ (n) ,r (m) ) is the average noise amplitude; Step 2-4: Traverse All sample points in the search are searched for sample points whose amplitude exceeds the detection threshold T lfas If there is no sample point with a lower threshold, go to step 5. If there is a sample point with a lower threshold, then select the distance unit sequence from the distance-pulse domain data x(τ (n) ,r (m) ) to extract the under-threshold detection data x(τ (n) ,r (rl) ), where r l It is the index of the distance unit of the data detected too low.

3. The method for cluster target gaze detection based on range-Doppler two-dimensional sequential filtering according to claim 2, characterized in that: The step 3 comprises the following steps: Step 3-1: Detect data at too low a threshold In all distance units, the long-term coherent integration method with joint constraints on computational complexity and detection probability is used to achieve long-term coherent integration along the pulse domain to obtain And the target velocity estimation matrix Acceleration Estimation Matrix in, is the frequency value corresponding to the nth Doppler unit, It is the rth l The distance value corresponding to each distance unit; Step 3-2: Set the false alarm probability P to the preset high threshold hfas , calculate the detection threshold T hfas : In the formula, x′ Noise Is too low threshold distance-pulse domain data The average amplitude of the noise in the medium, α is the dynamic adjustment factor of the noise power, which adjusts the impact of nonlinear operation on the noise power in the long-term coherent integration algorithm; Step 3-3: Traversal All sample points in the search are searched for sample points whose amplitude exceeds the detection threshold T hfas If there is no sample point with a high threshold, go to step 5. If there is a sample point with a high threshold, go to step 5 according to the distance unit sequence. Extracting over-threshold detection data Among them, r h is the over-threshold detection data distance unit index; Step 3-4: For each distance unit r h , the Doppler unit index where the spectrum peak searches for the highest energy target Record speed estimates separately Acceleration estimate Step 3-5: For each range cell, find the Doppler cell where the highest energy target is located. Set the distance-related gate D r , for the data in the range gate that meets the association criteria, calculate the target distance estimate: In the formula, is the distance correlation gate D r The index of the dth distance unit within.

4. The method for cluster target gaze detection based on range-Doppler two-dimensional sequential filtering according to claim 3, characterized in that: The step 4 comprises the following steps: Step 4-1: For each distance unit r h , with the distance to the highest energy target - Doppler index As the center, calculate the distance to filter the gate index boundary and the Doppler filter gate size Where Δr is the distance resolution unit, Δr s is the distance sampling unit, Δf d is the Doppler resolution unit, Δf ds is the Doppler sampling unit, β r , β fd are the distance and Doppler filtering gate expansion factor respectively; Step 4-2: for each distance unit r h , with range-Doppler index As the center, fill the distance and Doppler filter gate data with noise to obtain the filtered data Step 4-3: For each distance unit r h , do N along the Doppler domain h The IFFT processing restores the signal to the range-pulse domain.

5. The method for cluster target gaze detection based on range-Doppler two-dimensional sequential filtering according to claim 4, characterized in that: The step 5 includes: if the number of filtered targets in step 4-2 has reached the upper limit of the maximum processable targets, or no target has passed the detection threshold in steps 2-4 and 3-3, then output the distance, speed, and acceleration estimation values ​​of all detected targets; otherwise, go to step 3.

Images