Space-based Radar Discrete Sidelobe Clutter Recognition Method and System Based on Sliding Window Filtering Loss

Abstract

The present invention discloses a method for identifying discrete sidelobe clutter of space-based radar based on the loss of filtering response of a sliding window. The implementation steps are as follows: 1. Design of sliding window distance segmentation; 2. Space-time adaptive dimensionality reduction processing; 3. Modeling of filtering response loss; 4. Range cell registration; 5. Decision of filtering response loss. The present invention also provides a system for identifying discrete sidelobe clutter of space-based radar based on the loss of filtering response of a sliding window. By combining the wide-area detection capability of space-based radar and the non-uniform scene distribution characteristics of discrete strong scattering points, the present invention improves the fault-tolerant decision-making ability between real targets and discrete sidelobe clutter by statistically analyzing the loss of space-time adaptive processing filtering response corresponding to different segmented data. At the same time, the idea of sample segmented registration is introduced, avoiding the defect of poor clutter suppression ability at the scene edge, with simple implementation and high robustness, providing an effective way for robust suppression of non-uniform environment clutter of space-based radar.

Description

Technical Field

[0001] The present invention belongs to the technical field of radar, and relates to a method and system for identifying discrete sidelobe clutter of a space-based radar based on the loss of sliding window filtering, which is applicable to clutter suppression of a space-based radar in a non-uniform environment. Background Technique

[0002] The space-based radar has a wide field of view, and the distribution of ground objects is complex. Discrete sidelobe clutter in the echo mainly appears in non-uniform regions with discrete strong scattering points. Considering that the sample differences corresponding to different echo distance segments in such non-uniform scenarios are relatively large, for a specific non-uniform sample discrimination criterion, the sample determination results corresponding to different sample sets are also different.

[0003] Currently, the research on identifying discrete sidelobe clutter of space-based radars at home and abroad mainly includes two categories. The first is the synthetic channel gain method, which judges whether a candidate target is a real target by comparing the outputs of a synthetic channel high-gain narrow-beam antenna and an independent channel low-gain wide-beam antenna after space-time adaptive processing. Ideally, if the target is located at the pointing position of the antenna beam center, the synthetic channel gain is approximately equal to 10log 10 N2, where N is the number of antenna channels. If the candidate target comes from the antenna sidelobe, the synthetic channel amplitudes cannot be coherently accumulated, resulting in the synthetic channel gain being lower than the above ideal value. According to the difference between the synthetic channel gain of the candidate target and the independent channel gain, the discrete sidelobe clutter can be removed. The second is the filtering response loss method, which identifies discrete sidelobe clutter by analyzing the energy loss of candidate target points before and after filtering. After space-time adaptive processing, discrete sidelobe clutter and real targets are regarded as candidate targets in the range-Doppler plane. However, due to the space-time steering vector pointing to the main lobe target area, the energy loss of discrete sidelobe clutter is relatively large after clutter suppression processing. On this basis, some scholars use multiple space-time steering vector constraint methods to identify discrete sidelobe clutter. First, two space-time steering vectors pointing to the target and non-target regions are used to construct an optimal weight vector and perform space-time adaptive processing. Then, the energy losses of candidate targets at the same position after processing are compared. The one with a larger loss value is judged as a real target, and vice versa is regarded as discrete sidelobe clutter. The above two methods provide effective means for identifying discrete sidelobe clutter to a certain extent. However, real targets generally do not lie at the center of the antenna beam pointing position, resulting in the synthetic channel gain method not being able to achieve the ideal channel gain. In addition, although discrete sidelobe clutter can form a relatively large loss amount after being processed by the filtering response loss method, the target also has energy loss after filtering.

[0004] Therefore, the existing methods cannot significantly improve the fault-tolerant discrimination ability between real targets and discrete sidelobe clutter, and it is necessary to further study a robust discrete sidelobe clutter identification method applicable to space-based radars with a large field of view. Summary of the Invention

[0005] The technical problem solved by the present invention is: overcoming the deficiencies of the prior art, and proposing a method and system for identifying discrete sidelobe clutter based on filtering loss, so as to solve the problem of robust suppression of non-uniform environmental clutter of space-based radar.

[0006] The technical solution of the present invention is as follows:

[0007] A method for identifying discrete sidelobe clutter of space-based radar based on sliding window filtering loss, comprising:

[0008] Performing sliding window range segmentation on the multi-channel echoes of the space-based radar corresponding to the observation scene;

[0009] Performing space-time adaptive dimensionality reduction processing on the multi-channel echoes of the space-based radar corresponding to the observation scene before and after the sliding window;

[0010] According to the results of the space-time adaptive dimensionality reduction processing, perform range cell registration on the multi-channel echo images of the space-based radar corresponding to the observation scene before and after the sliding window, sequentially count the detection points exceeding the constant false alarm threshold in the two registered images, and add them to the discrete sidelobe clutter candidate set;

[0011] Establish a filtering response loss model, and the filtering response loss model is where E_in is the energy of the cell where the candidate target is located before clutter suppression, and E_out is the energy of the cell where the candidate target is located after clutter suppression;

[0012] For any candidate target in the discrete sidelobe clutter candidate set, calculate its filtering response loss Loss1 before the sliding window and its filtering response loss Loss2 after the sliding window, and judge whether the candidate target is discrete sidelobe clutter according to the filtering response loss decision criterion, so as to realize the identification of discrete sidelobe clutter of space-based radar.

[0013] Preferably, when performing sliding window range segmentation on the multi-channel echoes of the space-based radar corresponding to the observation scene, assume that the multi-channel echo sample set of the space-based radar corresponding to the observation scene before the sliding window is X1, the multi-channel echo sample set of the space-based radar corresponding to the observation scene after the sliding window is X2, the optimal weight vector corresponding to X1 is w1, and the optimal weight vector corresponding to X2 is w2. The space-time adaptive processing results result1 of X1 and result2 of X2 satisfy:

[0014]

[0015]

[0016] Preferably, the method for performing space-time adaptive dimensionality reduction processing is as follows:

[0017] Assume that the antenna array adopts a forward-looking model, the number of radar antenna channels is N, the number of pulses received by each array element within the coherent processing interval is K, and the number of range gates is L. If the l-th range ring includes C l blocks of clutter, the echo data of the l-th range ring is expressed as:

[0018]

[0019] where G i is the echo intensity of the i-th block of clutter, G target is the echo intensity of the target, represents the Kronecker product operation, N l represents noise, and respectively represent the time-domain steering vector and the space-domain steering vector of the i-th block of clutter, and represent the time-domain steering vector and the space-domain steering vector of the target;

[0020] Assume that the dimensionality reduction matrix T is an NK×PQ matrix, and P and Q respectively represent the space-domain and time-domain degrees of freedom of the system after dimensionality reduction. Using the linearly constrained minimum variance criterion, the adaptive optimal weight is expressed as follows:

[0021] w = μ R -1 s

[0022] In the above formula the clutter covariance matrix R and the target space-time steering vector s after dimensionality reduction are expressed as

[0023]

[0024]

[0025] where (·) -1 and (·) H respectively represent the inverse and conjugate transpose operations.

[0026] Preferably, and satisfy:

[0027]

[0028] where f r represents the pulse repetition frequency, V is the flight speed of the radar platform, λ is the carrier wavelength, θ tar is the spatial cone angle between the target and the radar, v is the radial velocity of the target, d is the element spacing, exp and cos respectively represent the exponential operation and the cosine operation, the superscript T is the transpose operation, and j = sqrt(-1).

[0029] Preferably, in the range-Doppler domain, assuming that x represents the echo of the cell where the candidate target is located, then E_in = |x H x|, E_out = |w H x| 2 , where w is the optimal weight vector corresponding to x.

[0030] Preferably,

[0031] The filtering response loss decision criterion is set as follows:

[0032]

[0033] η is the filtering response loss decision threshold.

[0034] A space-based radar discrete sidelobe clutter recognition system based on sliding window filtering loss, including a sliding window range segmentation module, a space-time adaptive dimensionality reduction processing module, a discrete sidelobe clutter candidate set generation module, a filtering response loss calculation module, and a discrete sidelobe clutter recognition module;

[0035] Sliding window range segmentation module: Perform sliding window range segmentation on the multi-channel echoes of the space-based radar corresponding to the observation scene;

[0036] Space-time adaptive dimensionality reduction processing module: Perform space-time adaptive dimensionality reduction processing on the multi-channel echoes of the space-based radar corresponding to the observation scene before and after the sliding window;

[0037] Discrete sidelobe clutter candidate set generation module: According to the processing results of the space-time adaptive dimensionality reduction processing module, perform range cell registration on the multi-channel echo images of the space-based radar corresponding to the observation scene before and after the sliding window, and sequentially count the detection points exceeding the constant false alarm threshold in the two registered images, and add them to the discrete sidelobe clutter candidate set;

[0038] Filtering response loss calculation module: Establish a filtering response loss model, and the filtering response loss model is where E_in is the energy of the cell where the candidate target is located before clutter suppression, and E_out is the energy of the cell where the candidate target is located after clutter suppression;

[0039] Discrete sidelobe clutter recognition module: For any candidate target in the discrete sidelobe clutter candidate set, use the filtering response loss calculation module to calculate its filtering response loss Loss1 before the sliding window and its filtering response loss Loss2 after the sliding window, and judge whether the candidate target is discrete sidelobe clutter according to the filtering response loss decision criterion, so as to realize the recognition of discrete sidelobe clutter of the space-based radar.

[0040] Preferably, the implementation method of the sliding window range segmentation module is as follows:

[0041] Suppose the multi-channel echo sample set of the space-based radar corresponding to the observation scene before the sliding window is X1, and the multi-channel echo sample set of the space-based radar corresponding to the observation scene after the sliding window is X2. The optimal weight vector corresponding to X1 is w1, and the optimal weight vector corresponding to X2 is w2. The space-time adaptive processing results result1 of X1 and result2 of X2 satisfy:

[0042]

[0043]

[0044] Preferably, in the range-Doppler domain, assuming that x represents the echo of the cell where the candidate target is located, then E_in = |x H x|, E_out = |w H x| 2 , where w is the optimal weight vector corresponding to x.

[0045] Preferably, the filtering response loss decision criterion is set as follows:

[0046]

[0047] η is the filtering response loss decision threshold.

[0048] The beneficial effects of the present invention compared with the prior art are as follows:

[0049] 1. The present invention proposes a method for identifying discrete sidelobe clutter of space-based radar based on sliding window filtering response loss. This method combines the wide-area detection ability of space-based radar and the non-uniform scene distribution characteristics of discrete strong scatterers. By statistically analyzing the filtering response loss of space-time adaptive processing corresponding to different segmented data, it improves the fault-tolerant decision-making ability between real targets and discrete sidelobe clutter, and has high engineering practical value.

[0050] 2. The present invention proposes a method for reducing the dimension of space-time adaptive processing based on sample segment registration. By introducing the idea of sample segment registration, this method avoids the defect of poor clutter suppression ability at the scene edge, is simple to implement and has high robustness. The experimental results based on airborne calibration flight measured data verify the effectiveness of this method. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a flow chart of the method for identifying discrete sidelobe clutter of space-based radar based on sliding window filtering loss.

[0052] Figure 2For the processing result of the second - stage data, where (a) is before clutter suppression of the first - stage data; (b) is after clutter suppression of the first - stage data; (c) is the processing result of the synthetic channel gain method for Target 1; (d) is the processing of the synthetic channel gain method for Target 2; (e) is the processing result of the synthetic channel gain method for Target 3; (f) is the processing result of the filtering response loss method for Target 1; (g) is the processing result of the filtering response loss method for Target 2; (h) is the processing result of the filtering response loss method for Target 3.

[0053] Figure 3 For the processing result of the second - stage data, where (a) is before clutter suppression of the first - stage data; (b) is after clutter suppression of the first - stage data; (c) is the processing result of the synthetic channel gain method for Target 1; (d) is the processing of the synthetic channel gain method for Target 2; (e) is the processing result of the synthetic channel gain method for Target 3; (f) is the processing result of the filtering response loss method for Target 1; (g) is the processing result of the filtering response loss method for Target 2; (h) is the processing result of the filtering response loss method for Target 3. Detailed implementation mode

[0054] The following is a further detailed description of the implementation of the present invention.

[0055] The present invention uses a sliding - window design to segment the received echo data. Since the sample sets used to construct the clutter covariance matrix before and after the sliding - window operation are different, the filtering loss degrees corresponding to the discrete sidelobe clutter points after clutter suppression are also different. Using the above two sets of sliding - window data sets, the maximum filtering loss of each discrete sidelobe clutter point can be statistically obtained. However, the real target is located in the main lobe region of the antenna, and the optimal weight vectors used in space - time adaptive processing under different sample set conditions have the same space - time steering vector for the target, so the filtering loss of the real target is small.

[0056] The application scenario of the present invention is:

[0057] The present invention can be applied to the field of clutter suppression of space - based radars. This method combines the wide - area detection ability of space - based radars and the non - uniform scene distribution characteristics of discrete strong scatterers. By statistically analyzing the filtering response losses corresponding to different segmented data in space - time adaptive processing, the fault - tolerance decision - making ability between real targets and discrete sidelobe clutter is improved, providing an efficient way for robust clutter suppression in non - uniform environments of space - based radars. The implementation steps are as follows:

[0058] Step 1: Perform sliding - window range segmentation on the multi - channel echoes of the space - based radar corresponding to the observation scene

[0059] Use a sliding window to segment the observation scene into different range intervals. Assume that the different sample sets before and after the sliding window are X1 and X2, the corresponding clutter covariance matrices are R1 and R2, the optimal weight vectors are w1 and w2, and the STAP processing results corresponding to the echoes of different range segments can be expressed as:

[0060]

[0061]

[0062] Step 2 performs spatio-temporal adaptive dimensionality reduction processing on the space-based radar multi-channel echoes corresponding to the observation scenes before and after the sliding window.

[0063] Assume that the antenna array adopts a squint model. Assume that the number of radar antenna channels is N, the number of pulses received by each array element within the coherent processing interval is K, and the number of range gates is L. If the l-th range ring includes C l blocks of clutter, then the corresponding echo data can be expressed as:

[0064]

[0065] where G i and G target are the echo intensities of the i-th clutter block and the target respectively, represents the Kronecker product operation, N l represents noise, and refer to the time-domain steering vector and the space-domain steering vector of the i-th clutter block, and represent the time-domain steering vector and the space-domain steering vector of the target. The specific expressions are as follows:

[0066]

[0067]

[0068] where f r represents the pulse repetition frequency, V is the flight speed of the radar platform, λ is the carrier wavelength, θ tar is the spatial cone angle between the target and the radar, v is the radial velocity of the target, d is the element spacing, exp and cos represent the exponential operation and the cosine operation respectively, the superscript T is the transpose operation, and j = sqrt(-1).

[0069] Assume that the dimensionality reduction matrix T is an NK×PQ matrix, and P and Q represent the spatial and temporal degrees of freedom of the system after dimensionality reduction respectively. Using the linearly constrained minimum variance criterion, the adaptive optimal weight can be expressed as follows:

[0070] w = μR -1 s (4)

[0071] In the above formula the clutter covariance matrix R and the target spatio-temporal steering vector s after dimensionality reduction can be expressed as

[0072]

[0073]

[0074] where (·) -1 and (·) H represent the inverse and conjugate transpose operations respectively.

[0075] Step 3 Range cell registration.

[0076] Perform range cell registration on the two images result1 and result2 according to the range cell numbers of the segments before and after the sliding window. Sequentially count the detection points exceeding the constant false alarm threshold in the two registered images and add them to the discrete sidelobe clutter candidate set.

[0077] Step 4 Filter response loss modeling.

[0078] Without loss of generality, the subsequent processing all adopts the EFA dimensionality reduction algorithm. Define the filter response loss as

[0079]

[0080] where E_in is the energy of the cell where the candidate target is located before clutter suppression, and E_out is the energy of the cell where the candidate target is located after clutter suppression. In the range-Doppler domain, assuming x represents the cell where the candidate target is located, the above variables are expressed as E_in = |x H x|, E_out = |w H x| 2 .

[0081] Step 5 Filter response loss decision.

[0082] For any candidate target in the image, assuming the filter response losses corresponding to the space-time adaptive processing are Loss1 and Loss2 respectively, and the filter response loss decision threshold is denoted as η, the filter response loss decision criterion is set as follows:

[0083]

[0084] A space-based radar discrete sidelobe clutter recognition system based on sliding window filtering loss, including a sliding window range segmentation module, a space-time adaptive dimensionality reduction processing module, a discrete sidelobe clutter candidate set generation module, a filter response loss calculation module, and a discrete sidelobe clutter recognition module.

[0085] Sliding window range segmentation module: Perform sliding window range segmentation on the multi-channel echoes of the space-based radar corresponding to the observation scene.

[0086] Space-time adaptive dimensionality reduction processing module: Perform space-time adaptive dimensionality reduction processing on the multi-channel echoes of the space-based radar corresponding to the observation scene before and after the sliding window.

[0087] Discrete sidelobe clutter candidate set generation module: According to the processing result of the space-time adaptive dimensionality reduction processing module, perform range cell registration on the multi-channel echo images of the space-based radar corresponding to the observation scenes before and after the sliding window. In the two registered images, sequentially count the detection points exceeding the constant false alarm threshold and add them to the discrete sidelobe clutter candidate set.

[0088] Filtering response loss calculation module: Establish a filtering response loss model, and the filtering response loss model is where \(E_{in}\) is the energy of the cell where the candidate target is located before clutter suppression, and \(E_{out}\) is the energy of the cell where the candidate target is located after clutter suppression.

[0089] Discrete sidelobe clutter identification module: For any candidate target in the discrete sidelobe clutter candidate set, use the filtering response loss calculation module to calculate its filtering response loss Loss1 before the sliding window and its filtering response loss Loss2 after the sliding window, and judge whether the candidate target is discrete sidelobe clutter according to the filtering response loss decision criterion, so as to realize the identification of discrete sidelobe clutter of the space-based radar.

[0090] On the one hand, the present invention combines the wide-area detection ability of the space-based radar and the non-uniform scene distribution characteristics of discrete strong scatterers, and improves the fault-tolerant decision-making ability of real targets and discrete sidelobe clutter by statistically calculating the filtering response loss of space-time adaptive processing corresponding to different segmented data. On the other hand, the idea of sample segmented registration is introduced, which avoids the defect of poor clutter suppression ability at the scene edge, has simple implementation and high robustness, and provides an effective way for robust clutter suppression in the non-uniform environment of space-based radar.

[0091] The effect of the present invention is further described below through simulation data.

[0092] Taking the measured data of a multi-channel airborne radar system as an example to verify the effectiveness of the proposed algorithm. The data comes from a certain province in the eastern part of China, and the radar collected echoes include various ground object information such as complex terrain and discrete strong scatterers. At the same time, in order to simulate the real air target detection working mode of the space-based radar, there is a real target aircraft flying in cooperation within the main lobe of the antenna during the down-looking operation of the carrier aircraft. In the experiment, the Y-12 aircraft is selected as the carrier aircraft and the Cessna aircraft is selected as the target aircraft. The parameter settings of the airborne calibration flight radar system are shown in Table 1.

[0093] Table 1 Parameter settings of the airborne calibration flight radar system

[0094]

[0095]

[0096] The clutter suppression dimensionality reduction algorithm uses the EFA algorithm. Due to the strong non-uniformity of the detection area, in order to achieve robust clutter suppression, non-uniform samples are removed when constructing the covariance matrix, which may result in the coexistence of real targets and discrete sidelobe clutter in the final image after clutter suppression. During the data processing, the number of range segmentation units is set to 400, and the overlapping distance units of the two sections of data before and after the sliding window is 100. Figure 2 and Figure 3 respectively correspond to the clutter suppression processing results of the two sections of data. It can be seen from (b) in Figure 2 and (b) in Figure 3 that two candidate targets appear in both sections of data after clutter suppression. The coordinates of the candidate targets in the first section of data are (140, 215) and (-150, 138), and the coordinates of the candidate targets in the second section of data are (140, 115) and (110, 14). Considering that only one target drone is set in this experiment, it is necessary to identify the candidate targets in the image after clutter suppression. After distance registration of the above candidate targets, each of the two images generates 1 additional point to be detected, which are (110, 114) and (-150, 38) respectively. However, the candidate point (140, 215) in the first image and the candidate point (140, 115) in the second image completely coincide after distance matching and belong to the same target. Therefore, there are three candidate targets in the scene. Figure 2 (c) - (h) in Figure 3 and (c) - (h) in

[0097] respectively represent the processing results of the three candidate targets using the synthetic channel gain method, the filtering response loss method, and the sliding window filtering response loss method of the present invention.

[0097] The statistical results of the measured data processing are shown in Table 2. According to the decision criterion, candidate target 1 is a real target, and candidate targets 2 and 3 are discrete sidelobe clutter.

[0098] Table 2 Processing Results of Candidate Targets by Different Methods

[0099]

[0100] Conclusion of simulation analysis: Among the above three methods, the sliding window filtering response loss method proposed by the present invention has the strongest fault tolerance ability, followed by the filtering response loss method, and the synthetic channel gain method has the worst fault tolerance ability. Here, the fault tolerance ability is defined as the available dynamic range of the decision threshold, that is, the reliable threshold setting interval for distinguishing real targets from discrete sidelobe clutter.

[0101] The parts not detailed in the present invention belong to the common knowledge of those skilled in the art.

Claims

1. A method for identifying discrete sidelobe clutter of space-based radar based on sliding window filtering loss, characterized in that Including: Performing sliding window range segmentation on the multi-channel echoes of the space-based radar corresponding to the observation scene; Performing space-time adaptive dimensionality reduction processing on the multi-channel echoes of the space-based radar corresponding to the observation scene before and after the sliding window; According to the results of the space-time adaptive dimensionality reduction processing, performing range cell registration on the multi-channel echo images of the space-based radar corresponding to the observation scene before and after the sliding window, successively counting the detection points exceeding the constant false alarm threshold in the two registered images, and adding them to the discrete sidelobe clutter candidate set; Build a filtering response loss model, where the filtering response loss model is where E_in is the energy of the cell where the candidate target is located before clutter suppression, and E_out is the energy of the cell where the candidate target is located after clutter suppression; For any candidate target in the discrete sidelobe clutter candidate set, calculating its filtering response loss Loss1 before the sliding window and its filtering response loss Loss2 after the sliding window, and judging whether the candidate target is discrete sidelobe clutter according to the filtering response loss decision criterion, so as to realize the recognition of discrete sidelobe clutter of the space-based radar; When performing sliding window range segmentation on the multi-channel echoes of the space-based radar corresponding to the observation scene, assuming that the multi-channel echo sample set of the space-based radar corresponding to the observation scene before the sliding window is X1, the multi-channel echo sample set of the space-based radar corresponding to the observation scene after the sliding window is X2, the optimal weight vector corresponding to X1 is w1, the optimal weight vector corresponding to X2 is w2, and the space-time adaptive processing results result1 of X1 and result2 of X2 satisfy: The filtering response loss decision criterion is set as follows: η is the filtering response loss decision threshold.

2. A method for identifying discrete sidelobe clutter of a space-based radar based on sliding window filtering loss according to claim 1, characterized in that, The method for performing space-time adaptive dimensionality reduction processing is as follows: Assume that the antenna array adopts a forward-looking model, the number of radar antenna channels is N, the number of pulses received by each array element within the coherent processing interval is K, and the number of range gates is L. If the l-th range bin includes C l blocks of clutter, the echo data of the l-th range bin is expressed as: where G i is the echo intensity of the i-th clutter, G target is the echo intensity of the target, represents the Kronecker product operation, N l denotes the noise, and respectively denote the time-domain steering vector and the space-domain steering vector of the i-th clutter, and represent the time-domain steering vector and the space-domain steering vector of the target; Assuming that the dimensionality reduction matrix T is an NK×PQ matrix, where P and Q respectively represent the spatial and temporal degrees of freedom of the system after dimensionality reduction, using the linearly constrained minimum variance criterion, the adaptive optimal weight is expressed as follows: w = μR -1 s In the above formula The clutter covariance matrix R and the target spatio-temporal steering vector s after dimensionality reduction are expressed as where (·) -1 and (·) H represent the inverse and conjugate transpose operations, respectively.

3. A method for identifying discrete sidelobe clutter of a space-based radar based on sliding window filtering loss according to claim 2, characterized in that and satisfy: where f r represents the pulse repetition frequency, V is the flight speed of the radar platform, λ is the carrier wavelength, θ tar is the spatial cone angle between the target and the radar, v is the radial velocity of the target, d is the element spacing, exp and cos represent the exponential operation and the cosine operation respectively, the superscript T represents the transpose operation, and j = sqrt(-1).

4. A space-based radar discrete sidelobe clutter recognition method based on sliding window filtering loss according to claim 1, characterized in that In the range-Doppler domain, assuming that x represents the echo of the cell where the candidate target is located, then E_in = |x H x|, E_out = |w H x| 2 , where w is the optimal weight vector corresponding to x.

5. A space-based radar discrete sidelobe clutter recognition system based on sliding window filtering loss, characterized in that, Including a sliding window range segmentation module, a space-time adaptive dimensionality reduction processing module, a discrete sidelobe clutter candidate set generation module, a filtering response loss calculation module, and a discrete sidelobe clutter recognition module; Sliding window range segmentation module: Performing sliding window range segmentation on the multi-channel echoes of the space-based radar corresponding to the observation scene; Space-time adaptive dimensionality reduction processing module: Performing space-time adaptive dimensionality reduction processing on the multi-channel echoes of the space-based radar corresponding to the observation scene before and after the sliding window; Discrete sidelobe clutter candidate set generation module: According to the processing results of the space-time adaptive dimensionality reduction processing module, performing range cell registration on the multi-channel echo images of the space-based radar corresponding to the observation scene before and after the sliding window, successively counting the detection points exceeding the constant false alarm threshold in the two registered images, and adding them to the discrete sidelobe clutter candidate set; Filter response loss calculation module: Establish a filter response loss model, and the filter response loss model is where E_in is the energy of the unit where the candidate target is located before clutter suppression, and E_out is the energy of the unit where the candidate target is located after clutter suppression; Discrete sidelobe clutter recognition module: For any candidate target in the discrete sidelobe clutter candidate set, using the filtering response loss calculation module to calculate its filtering response loss Loss1 before the sliding window and its filtering response loss Loss2 after the sliding window, and judging whether the candidate target is discrete sidelobe clutter according to the filtering response loss decision criterion, so as to realize the recognition of discrete sidelobe clutter of the space-based radar; The implementation method of the sliding window range segmentation module is as follows: Suppose the space-based radar multi-channel echo sample set corresponding to the observation scene before the sliding window is X1, the space-based radar multi-channel echo sample set corresponding to the observation scene after the sliding window is X2, the optimal weight vector corresponding to X1 is w1, the optimal weight vector corresponding to X2 is w2, and the space-time adaptive processing results result1 of X1 and result2 of X2 satisfy: The filtering response loss decision criterion is set as follows: η is the filtering response loss decision threshold.

6. The space-based radar discrete sidelobe clutter recognition system based on sliding window filtering loss according to claim 5, wherein, In the range-Doppler domain, assuming that x represents the echo of the cell where the candidate target is located, then E_in = |x H x|, E_out = |w H x| 2 , where w is the optimal weight vector corresponding to x.

Images