CFAR detection method and system based on multiple screening modules and adaptive weighting

Abstract

The invention discloses a CFAR detection method and system based on multiple screening modules and adaptive weighting, and mainly solves the problems of poor detection performance and difficult configuration in a complex clutter environment in the prior art. The method comprises the following implementation steps of: eliminating other interference signals in a processing window by a phase characteristic screening module; the repeated eliminating module is used for eliminating obvious noise or clutter peaks; the result set establishing module establishes a result set by utilizing environmental characterization VI and MR; the weighted CFAR detector initialization and weighted distribution module is used for carrying out corresponding weighted OS-CFAR detector initialization and weighted distribution on various complex clutter environments; and the detection output module calculates a detection threshold according to the weight and outputs a detection result. According to the method, the weights of the VI and MR parameters are calculated, and the OS-CFAR detectors are subjected to weighted output through the weights, so that the method still has the advantage of keeping relatively high detection performance in various clutter environments while the configuration complexity and the operation complexity are reduced.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of radar target detection, and further relates to a CFAR detection method and system based on multiple screening modules and adaptive weighting in the technical field of radar target near-field detection. The present application can be used for target detection in different clutter environments. BACKGROUND

[0002] With the increasingly complex radar working environment, the requirement for target detection effect is continuously improved, and the traditional CFAR detection method has shown more and more problems, especially in the detection performance in complex and variable clutter environment. The adaptive CFAR detection method is a technical means for detecting complex clutter environment, which automatically selects algorithm and parameters according to different background environment. The classical adaptive CFAR has HCE-CFAR and VI-CFAR, however, HCE-CFAR can only estimate the clutter edge environment, and VI-CFAR will greatly reduce the detection performance in the case of uneven front and rear windows, and cannot meet the requirement of high detection performance in complex clutter environment.

[0003] Nanjing University of Technology discloses a millimeter wave water surface flow velocity measurement radar CFAR detection method and system in its applied patent document (application date: 2025-4-24, application number: 202510524808.8, application publication number: CN 120490999 A). The scheme removes outliers in advance through local outlier factor, and estimates the sample quartile range using the screened data, thereby obtaining the sample truncation threshold, and finally obtains the background clutter power estimation value through maximum likelihood estimation. The introduction of quartile range optimizes the selection of truncation threshold, and the local outlier factor is used to screen the reference window samples in advance, which enhances the robustness of the quartile range and the truncation threshold. Although this method enhances the environmental adaptability of the algorithm, it still has the following deficiencies: it only removes the local outlier factor of the echo, and does not fully utilize the environment representation, so the subsequent single-sided truncation by quartile range is required, resulting in high configuration and operation complexity.

[0004] Northwestern Polytechnical University disclosed an adaptive CFAR detection method based on scene classification in its patent application "An Adaptive CFAR Detection Method Based on Scene Classification" (application date: 2024-12-13, application number: 202411843035.1, publication number: CN 119540738 A). The method's implementation steps include: using K-means clustering to classify the input SAR image into scenes; using VI-CFAR to detect targets in the classified pure sea areas; and using a saliency model to perform CFAR detection in coastal areas. While this method solves the problems of high false alarm rates and unstable detection performance of traditional CFAR algorithms in complex backgrounds such as ports and land, it still has shortcomings. It uses VI-CFAR for detection, which has poor performance at clutter edges and in environments with uneven lateral surfaces, making it unsuitable for various environments. Furthermore, it utilizes K-means clustering for pre-training and classification of sea and land environments, which requires collecting a large dataset for training, resulting in high configuration and computational complexity. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of the prior art by proposing a CFAR detection method and system based on multiple screening modules and adaptive weighting, which solves the problems of high system configuration complexity and low detection performance in complex clutter environments in the prior art.

[0006] The technical approach to achieving the objective of this invention is that the method calculates weights for the VI and MR parameters, and then uses these weights to weight and output multiple OS-CFAR detectors. Because only OS-CFAR is used for weighting, it overcomes the shortcomings of existing CFAR methods, which suffer from high configuration and computational complexity. The system of this invention uses a phase feature filtering module and a repeated deletion module to filter out other interference signals outside the detection unit in the echo, and to repeatedly delete interference signals outside the detection unit in the echo, strong clutter, and other interference factors. During the detection stage, VI and MR parameters are used for weighting, fully utilizing environmental characterization. This allows the system of this invention to maintain good detection performance in complex clutter environments while maintaining low configuration complexity.

[0007] To achieve the above objectives, the technical solution adopted by the present invention includes the following steps:

[0008] The CFAR detection method of the present invention includes the following steps:

[0009] Step 1: Construct a CFAR processing window, remove other interference signals from the processing window, and obtain the CFAR processing window after phase feature filtering;

[0010] Step 2: Using an iterative approach, remove reference cells that exceed the detection threshold in the CFAR processing window after phase feature screening;

[0011] Step 3: Calculate the half-window consistency characterization of the left reference cell in the detection unit. Consistency characterization of the right reference unit half-window Differential representation and according to , , Establish three corresponding result sets;

[0012] Step 4, according to , , The result set is used for CFAR detector initialization and weighting.

[0013] Step 5: Calculate the detection threshold based on the weights and output the detection results.

[0014] Furthermore, constructing the CFAR processing window refers to randomly selecting one range unit from multiple range units of the received signal as a detection unit, selecting M / 2 range units on each side of the selected detection unit as protection units, and selecting N / 2 range units on each side of each protection unit as reference units. The selected detection unit, protection units, and reference units are then combined to form a CFAR processing window. , M represents the total number of protection cells in a CFAR processing window, and N represents the total number of reference cells in a CFAR processing window.

[0015] Furthermore, the step of removing other interference signals from the processing window is as follows:

[0016] The first step is to select an unselected reference cell and perform univariate linear regression on the echo phase information of the selected reference cell to obtain the correlation coefficient and significance probability corresponding to the selected reference cell.

[0017] The second step is to determine whether the correlation coefficient of the selected reference unit is greater than or equal to the correlation coefficient threshold. And at the same time, the significance probability is also greater than or equal to the default significance level. If so, the phase characteristic of the selected reference cell is considered linear, and the selected reference cell will be removed from the CFAR processing window; otherwise, no processing is performed. and All of these are thresholds set based on experience;

[0018] The third step is to determine whether all reference cells have been selected. If so, obtain the CFAR processing window after phase feature filtering; otherwise, proceed to the first step.

[0019] Furthermore, the step of iteratively removing reference cells exceeding the detection threshold in the CFAR processing window after phase feature screening is as follows:

[0020] The first step is to calculate the threshold factor corresponding to the current CFAR processing window: ,in, This indicates the total number of reference cells in the current CFAR processing window. This represents the set constant false alarm probability;

[0021] The second step is to calculate the estimated background clutter power for the current CFAR processing window: ,in, Indicates the current CFAR processing window. One reference unit, , Indicates the first The echo intensity of each reference element;

[0022] The third step is to calculate the culling threshold for the current CFAR processing window: ;

[0023] The fourth step is to match the echo intensity of each reference cell with the rejection threshold. Compare the reference cells and remove those that exceed the threshold from the current CFAR processing window; otherwise, do nothing.

[0024] Fifth, repeat steps one through four until no more reference cells need to be removed from the CFAR processing window, thus obtaining the removed CFAR processing window.

[0025] Furthermore, the half-window consistency characterization of the left reference unit in the detection unit is calculated separately. Consistency characterization of the right reference unit half-window Differential representation It is obtained from the following formula:

[0026] ;

[0027] ;

[0028] ;

[0029] in, This represents the total number of all reference cells to the left of the detection unit. This indicates the currently selected leftmost [number]. The echo intensity of each reference element, , This represents the average echo intensity of all reference cells to the left of the detection unit. This represents the total number of all reference cells to the right of the detection unit. This indicates the currently selected rightmost [number]. The echo intensity of each reference element, , This represents the average echo intensity of all reference units to the right of the detection unit.

[0030] Furthermore, the aforementioned according to , , Establishing the three corresponding result sets refers to: , , respectively with and Compare and build a result set accordingly:

[0031] if Then set the first result set: ,otherwise, ;

[0032] if Then set a second result set: ,otherwise, ;

[0033] if Then, set a third result set: ,otherwise, ,

[0034] in, , All of these are judgment thresholds set based on experience.

[0035] Furthermore, the initialization of the CFAR detector refers to counting the number of reference cells in the current CFAR processing window. The CFAR detector consists of multiple OS-CFAR detectors. M OS-CFAR detectors are set to be weighted, and each detector has its own corresponding selection coefficient. Where M represents the total number of OS-CFAR detectors to be weighted. , Indicates the first The coefficients corresponding to each weighted CFAR detector , .

[0036] Furthermore, the weighted allocation refers to the allocation based on the input... , , The three result sets are then weighted according to the following four scenarios, with each weighted OS-CFAR detector corresponding to a separate weight:

[0037] exist ,and , In this case, the weight allocation is as follows:

[0038] ;

[0039] exist ,and , In this case, the weight allocation is as follows:

[0040] ;

[0041] exist and When the values ​​are different, the weights are allocated as follows:

[0042] ;

[0043] exist and In this case, the weight allocation is as follows:

[0044] ;

[0045] in, Indicates the first The weights corresponding to the weighted OS-CFAR detectors.

[0046] Furthermore, the steps for calculating the detection threshold based on the weights and outputting the detection results are as follows:

[0047] The first step is to sort all reference cells in the current CFAR processing window in ascending order based on their echo intensity.

[0048] The second step is to calculate the threshold factor corresponding to each weighted OS-CFAR detector. :

[0049] ;

[0050] in, This represents the set constant false alarm probability. Represents the gamma function. Indicates the first A weighted threshold factor for the OS-CFAR detector;

[0051] The third step is to sum the weighted values ​​of all weighted OS-CFAR detectors to obtain the detection threshold. :

[0052] ;

[0053] in, Indicates the first The echo intensity of each reference element;

[0054] The fourth step is to determine whether the echo intensity of the detection unit is greater than [the specified value]. If yes, then proceed to step five after the target has been detected; otherwise, it is considered that no target has been detected.

[0055] Fifth step: Output the target detection results.

[0056] The CFAR detection system of the present invention includes a phase feature screening module, a repeated deletion module, a result set establishment module, a weighted CFAR detector initialization and weighting allocation module, and a detection output module; wherein:

[0057] The phase feature filtering module is used to remove other interference signals in the processing window to obtain the CFAR processing window after phase feature filtering.

[0058] The repeated elimination module is used to iteratively eliminate reference cells that exceed the detection threshold in the CFAR processing window after phase feature screening, in order to eliminate significant noise or clutter spikes.

[0059] The result set building module is used to calculate the half-window consistency characterization of the left reference unit in the detection unit respectively. Consistency characterization of the right reference unit half-window With differential representation and according to , , Create a result set;

[0060] The weighted CFAR detector initialization and weighting allocation module is used to, according to , , The result set is used to initialize and assign weights to OS-CFAR detectors in various complex clutter environments.

[0061] The detection output module is used to calculate the detection threshold based on the weights and output the detection results.

[0062] Compared with the prior art, the present invention has the following advantages:

[0063] First, the method of this invention performs adaptive weighting processing on multiple OS-CFAR detectors for detection output. Compared with other adaptive CFAR detection algorithms, it outputs detection results with a simple combination of OS-CFAR detectors, overcoming the high configuration complexity of existing technologies. This gives the invention the advantages of low computational complexity and easy configuration.

[0064] Secondly, in the echo screening stage, the system of this invention employs two modules: phase screening and repeated deletion. These modules respectively screen interference signals and strong clutter outside the detection unit in the echo. During the detection stage, parameters such as VI and MR are used for weighting. This allows the invention to fully utilize environmental characterization, improve detection performance in complex clutter environments, and reduce detector configuration complexity. It maintains high detection performance in various clutter environments while keeping computational complexity low. Attached Figure Description

[0065] Figure 1 This is a system block diagram of the present invention;

[0066] Figure 2 This is a flowchart of the method of the present invention;

[0067] Figure 3 This is a comparison of the detection performance in a uniform background environment and a probability diagram of the weighted allocation scenario in simulation experiment 1 of this invention;

[0068] Figure 4 This is a comparison of detection performance and a probability diagram of weighted allocation when there is an interfering target in the left reference unit half-window of simulation experiment 2 of this invention;

[0069] Figure 5 This is a comparison of detection performance and a probability diagram of weighted allocation scenarios when there is an interfering target in each of the left and right reference unit half-windows in simulation experiment 3 of this invention.

[0070] Figure 6 This is a comparison of detection performance and a probability diagram of weighted allocation when there are two interfering targets in each of the left and right reference unit half-windows in simulation experiment 4 of this invention.

[0071] Figure 7 This is a comparison of the clutter edge environment detection performance and the probability diagram of the weighted allocation case selection in simulation experiment 5 of this invention. Detailed Implementation

[0072] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0073] Reference Figure 1 The present invention will now be described in further detail.

[0074] The system of this invention includes a phase feature screening module, a repeated elimination module, a result set establishment module, a weighted CFAR detector initialization and weighting allocation module, and a detection output module; wherein:

[0075] The phase feature filtering module is used to remove other interference signals in the processing window to obtain the CFAR processing window after phase feature filtering.

[0076] The repeated elimination module is used to iteratively eliminate reference cells that exceed the detection threshold in the CFAR processing window after phase feature screening, in order to eliminate significant noise or clutter spikes.

[0077] The result set building module is used to calculate the half-window consistency characterization of the left reference unit in the detection unit respectively. Consistency characterization of the right reference unit half-window With differential representation and according to , , Create a result set;

[0078] The weighted CFAR detector initialization and weighting allocation module is used to, according to , , The result set is used to initialize and assign weights to OS-CFAR detectors in various complex clutter environments.

[0079] The detection output module is used to calculate the detection threshold based on the weights and output the detection result. (Refer to...) Figure 2 The implementation steps of the embodiments of the method of the present invention will be described in further detail below.

[0080] Reference Figure 2 The implementation steps of embodiments of the method of the present invention will be further described below.

[0081] Step 1: Construct a CFAR processing window and remove distance cells with linear phase features.

[0082] This step uses a phase feature filtering module to remove elements in the reference element whose phase features are linear. The implementation includes the following:

[0083] Step 1.1: Construct the CFAR processing window.

[0084] Select detection unit D, and take M / 2 protection units P on each side of D. P is not included in the background power calculation. Take N / 2 distance units on each side of D as reference units. Detection unit D, protection unit P, reference unit Together, they are referred to as the CFAR processing window. This example sets up, but is not limited to, the CFAR processing window. , ;

[0085] Step 1.2: Select an unselected reference cell, obtain its phase information, and perform univariate linear regression to obtain the correlation coefficient. and significance probability .

[0086] Step 1.3, if the correlation coefficient reaches And the significance probability reaches The default significance level is used to determine if the phase characteristic of the reference cell is linear, and if so, the reference cell is discarded; otherwise, no processing is performed. This example sets, but is not limited to, the following settings. , .

[0087] Step 1.4: Determine whether all reference cells have been selected. If yes, obtain the CFAR processing window after phase feature filtering; otherwise, proceed to step 1.1.

[0088] Step 2: For the CFAR processing window after phase feature screening, an iterative approach is adopted to repeatedly remove reference cells that exceed the detection threshold.

[0089] This step uses an iterative removal module to delete reference cells that exceed the detection threshold. The implementation includes the following:

[0090] Step 2.1: Calculate the current threshold factor based on the principle of the average constant false alarm rate detector. :

[0091] ;

[0092] in, This indicates the total number of reference cells in the current CFAR processing window. This represents the set constant false alarm probability.

[0093] Step 2.2: Sum and average the echo intensities of the reference cells to obtain a new background clutter power estimate. :

[0094] ;

[0095] in, Indicates the current CFAR processing window. One reference unit, , Indicates the first The echo intensity of each reference element;

[0096] Step 2.3, based on the threshold factor and background clutter power estimates Get the deletion threshold :

[0097] ;

[0098] Step 2.4: Compare the echo intensity of each reference cell with the deletion threshold, and remove reference cells that exceed the threshold from the current CFAR processing window; otherwise, do not perform any processing.

[0099] Step 2.5: Repeat steps 2.1 to 2.4 until no more reference cells need to be removed from the CFAR processing window, and obtain the removed CFAR processing window.

[0100] Step 3: Calculate the half-window consistency characterization of the reference cell to the left of the unit to be detected, D. Consistency characterization of the right reference unit half-window With differential representation and according to , , Create a result set .

[0101] Step 3.1, Used to determine whether the half-window echo intensity of the left reference unit is uniform. Used to determine whether the echo intensity of the right reference cell half-window is uniform. This value is used to determine whether the mean values ​​of the left and right reference cell half-windows are consistent. The calculation method is as follows:

[0102] ;

[0103] ;

[0104] ;

[0105] in, This represents the total number of all reference cells to the left of the detection unit. This indicates the currently selected leftmost [number]. The echo intensity of each reference element, , This represents the average echo intensity of all reference cells to the left of the detection unit. This represents the total number of all reference cells to the right of the detection unit. This indicates the currently selected rightmost [number]. The echo intensity of each reference element, , This represents the average echo intensity of all reference units to the right of the detection unit.

[0106] Step 3.2, will , , Respectively with the judgment threshold and In comparison, the embodiments of the present invention are configured, but not limited to: , And based on this, three result sets are established:

[0107] if Then set the first result set: ,otherwise, ;

[0108] if Then set a second result set: ,otherwise, ;

[0109] if Then, set a third result set: ,otherwise, .

[0110] Step 4, according to , , The result set is used for CFAR detector initialization and weighting.

[0111] Step 4.1: Count the number of reference cells in the current CFAR processing window. .

[0112] Step 4.2: Set up M OS-CFAR detectors and set their selection coefficients for each OS-CFAR detector. Where M represents the total number of CFAR detectors to be weighted. Indicates the first The coefficients corresponding to each OS-CFAR detector, which vary with Increase and increase, , The embodiments of the present invention are configured, but not limited to: , , , .

[0113] Step 4.3, based on the input , , The result set is weighted and assigned to obtain the weights of the M OS-CFAR detectors. The weighted allocation is as follows:

[0114] Scenario 1, in ,and , In the case where both reference half-windows are deemed uniform and have the same mean, the background can be approximated as a uniform environment, and the weights are assigned as follows:

[0115] ;

[0116] Scenario 2, in ,and , In the case where both reference half-windows are judged to be uniform but their means are different, it can be approximately considered that clutter edges exist. In this case, the weights are assigned as follows:

[0117] ;

[0118] Scenario 3, in and When the values ​​are different, if one side of the reference half-window is judged to be uniform while the other side is non-uniform, it can be approximately assumed that the non-uniform side is caused by the presence of clutter edges or interfering targets. In this case, only the reference half-window on the uniform side is used for background power estimation, and the weights are assigned as follows:

[0119] ;

[0120] Scenario 4, in and In the case where both reference half-windows are non-uniform, it can be approximated that there may be interfering targets or clutter edges in both reference half-windows. In this case, the weights are assigned as follows:

[0121] .

[0122] Step 5: Calculate the detection threshold based on the weights and output the detection results.

[0123] Step 5.1: For all reference cells in the current CFAR processing window, calculate their echo intensity. Sort in ascending order, that is:

[0124] .

[0125] Step 5.2, based on the selected coefficients Calculate the threshold factor corresponding to each weighted OS-CFAR detector. Threshold factor The method for finding is as follows:

[0126] ;

[0127] in, This represents the set constant false alarm probability. Represents the gamma function. Indicates the first The threshold factor of the weighted OS-CFAR detector.

[0128] Step 5.3, take the first... Echo intensity of each reference unit The detection threshold is obtained by weighting the clutter power level as a detector. As shown in the following formula:

[0129] .

[0130] Step 5.4, if the echo intensity of detection unit D is greater than If the target is detected, it is considered that the target has been detected; otherwise, it is considered that the target has not been detected.

[0131] Step 5.5: Output the target detection results.

[0132] The effectiveness of this invention can be further demonstrated through the following simulation.

[0133] 1. Simulation experimental conditions.

[0134] The software platform for the simulation experiment of this invention is: Windows 10 operating system and Matlab R2021b.

[0135] 2. Simulation content and result analysis.

[0136] The present invention includes five simulation experiments.

[0137] 2.1 Simulation Experiment 1 simulates the detection performance of the present invention under a uniform background environment, obtaining a comparison of detection performance under a uniform background environment and a probability diagram of the weighted allocation scenario, as shown below. Figure 3 As shown.

[0138] The number of OS-CFAR detectors set in simulation experiment 1 of this invention The selected coefficients are respectively , , False alarm probability Number of protection units Number of reference units The signal-to-noise ratio (SNR) ranged from 0 dB to 30 dB, with a clutter power of 20 dB. A total of [number] tests were conducted. Monte Carlo simulation.

[0139] Simulation Experiment 1 of this invention uses the method of this invention and five existing technologies to obtain detection performance at 31 signal-to-noise ratios, and then plots the relationship between the obtained detection probability and signal-to-noise ratio as shown in the figure. Figure 3 The six detection probability curves shown in Figure (a) are used to statistically analyze the weighted allocation of the present invention under 31 signal-to-noise ratios. The relationship between the statistical probability and the signal-to-noise ratio for the four scenarios is plotted as follows. Figure 3 The four curves showing the selection probability of the weighted allocation scenario are shown in Figure (b).

[0140] In simulation experiment 1, the five existing technologies used are:

[0141] The prior art 1 refers to CA-CFAR proposed by FINN HM et al. in their paper "Adaptive detection mode with threshold control as a function of spatially sampled clutter level estimates[J]. RCA Review, 1968, 29(9): 414-464."

[0142] Prior art 2 refers to GO-CFAR proposed by HANSEN VG et al. in their paper "Constant false alarm rateprocessing in search radars[C] / / IEEE International Radar Conference. London:IEEE, 1973: 325-332."

[0143] Prior art 3 refers to SO-CFAR proposed by TRUNK GV et al. in their paper “Range resolution of targets using automatic detectors[J]. IEEE Transactions on Aerospace and Electronic Systems, 1978, 14(5): 750-755.”

[0144] Prior art 4 refers to OS-CFAR proposed by ROHLING H et al. in their paper “Radar CFAR Thresholding in Clutter and Multiple Target Situations[J]. IEEE Transactions on Aerospace and Electronic Systems, 1983, 19(4): 608-621.”

[0145] Prior art 5 refers to VI-CFAR proposed by Smith Me et al. in their paper “VARSHNEY P K. Intelligent CFAR processor based on data variability[J]. IEEE Transactions on Aerospace and Electronic Systems, 2000, 36(3): 837-847.”

[0146] 2.2 Simulation Experiment 2 is a simulation of the detection performance of the present invention when there is an interfering target in the half-window of the left reference unit of the detection unit. Figure 4 As shown.

[0147] The number of OS-CFAR detectors set in simulation experiment 2 of this invention The selected coefficients are respectively , , False alarm probability Number of protection units Number of reference units The signal-to-noise ratio (SNR) ranges from 0 dB to 30 dB, and the interference target's interference-to-noise ratio... The clutter power was 20dB, and a total of Monte Carlo simulation.

[0148] Simulation Experiment 2 of this invention uses the method of this invention and five existing technologies to obtain the detection performance at 31 signal-to-noise ratios, and then plots the relationship between the obtained detection probability and the signal-to-noise ratio as shown in the figure. Figure 4 The six detection probability curves shown in Figure (a) are used to statistically analyze the weighted allocation of the present invention under 31 signal-to-noise ratios. The relationship between the statistical probability and the signal-to-noise ratio for the four scenarios is plotted as follows. Figure 4 The selection probability curves for the four weighted allocation scenarios are shown in (b).

[0149] In simulation experiment 2, the five existing technologies used are the same as those in simulation experiment 1.

[0150] 2.3, Simulation Experiment 3 is a simulation of the detection performance of the present invention when there is an interfering target in each of the half-windows of the left and right reference units of the detection unit. Figure 5 As shown.

[0151] The simulation parameters used in simulation experiment 3 of this invention are the same as those used in simulation experiment 2.

[0152] Simulation Experiment 3 of this invention uses the method of this invention and five existing technologies to obtain the detection performance at 31 signal-to-noise ratios, and then plots the relationship between the obtained detection probability and the signal-to-noise ratio as shown in the figure. Figure 5 The six curves shown in Figure (a) simultaneously statistically represent the weighted allocation scenarios of the present invention under 31 signal-to-noise ratios. The relationship between the statistical probability of the four scenarios and the signal-to-noise ratio is plotted as follows: Figure 5 The four curves shown in (b) are shown in the middle.

[0153] In simulation experiment 3, the five existing technologies used are the same as those in simulation experiment 1.

[0154] 2.4, Simulation Experiment 4 is a simulation of the detection performance of the present invention under the condition that there are two interfering targets in each of the half-windows of the reference units on the left and right sides of the detection unit. Figure 6 As shown.

[0155] The simulation parameters used in simulation experiment 4 of this invention are the same as those used in simulation experiment 2.

[0156] Simulation Experiment 4 of this invention uses the method of this invention and two existing technologies to obtain the detection performance at 31 signal-to-noise ratios, and then plots the relationship between the obtained detection probability and the signal-to-noise ratio as shown in the figure. Figure 6 The three curves shown in (a) simultaneously statistically analyze the weighted allocation of the present invention under 31 signal-to-noise ratios, and plot the relationship between the statistical probability of the four cases and the signal-to-noise ratio as shown in the figure. Figure 6 The four curves shown in (b) are shown in the middle.

[0157] In simulation experiment 4, the two existing technologies used are the same as those used in simulation experiment 1, namely OS-CFAR and VI-CFAR.

[0158] 2.5, Simulation Experiment 5 simulates the detection performance of the present invention in a clutter edge environment, such as... Figure 7 As shown.

[0159] The number of OS-CFAR detectors set in simulation experiment 2 of this invention The selected coefficients are respectively , , False alarm probability Number of protection units Number of reference units The signal-to-noise ratio (SNR) ranged from 0 dB to 30 dB. The power in the weak clutter region at the clutter edge was 20 dB, and the clutter power in the strong clutter region was 30 dB. A total of [number missing] tests were conducted. Monte Carlo simulation.

[0160] Simulation Experiment 5 of this invention uses the method of this invention and five existing technologies to obtain the detection performance at 31 signal-to-noise ratios, and then plots the relationship between the obtained detection probability and the signal-to-noise ratio as shown in the figure. Figure 7 The six curves shown in Figure (a) simultaneously statistically represent the weighted allocation scenarios of the present invention under 31 signal-to-noise ratios. The relationship between the statistical probability of the four scenarios and the signal-to-noise ratio is plotted as follows: Figure 7 The four curves shown in (b) are shown in the middle.

[0161] In simulation experiment 5, the five existing technologies used are the same as those in simulation experiment 1.

[0162] The following is combined with Figure 3 , Figure 4 , Figure 5 , Figure 6 , Figure 7 The simulation diagrams further illustrate the effects of the present invention.

[0163] Figure 3 In Figure (a), the horizontal axis represents the signal-to-noise ratio (SNR) of the detected target, in dB, and the vertical axis represents the detection probability of the detected target. The curve marked with a red circle represents the relationship between the detection probability and SNR obtained by simulation using prior art 1; the curve marked with a blue vertical dot represents the relationship between the detection probability and SNR obtained by simulation using prior art 2; the curve marked with a blue asterisk represents the relationship between the detection probability and SNR obtained by simulation using prior art 3; the curve marked with a pink dashed line represents the relationship between the detection probability and SNR obtained by simulation using prior art 4; the curve marked with a pink asterisk represents the relationship between the detection probability and SNR obtained by simulation using prior art 5; and the curve marked with a black pentagram represents the relationship between the detection probability and SNR obtained by simulation using the method proposed in this invention.

[0164] from Figure 3 As can be seen in (a), the 31 detection probabilities obtained by the method of the present invention increase continuously with the increase of signal-to-noise ratio. Under the same signal-to-noise ratio, the detection probabilities obtained by the method of the present invention closely approximate the CA-CFAR of the prior art 1, which has the best detection performance in a uniform background environment, and are superior to the other four prior art techniques.

[0165] Figure 3In (b), the horizontal axis represents the signal-to-noise ratio (SNR) of the detected target, in dB, and the vertical axis represents the selection probability of the four weighted allocation scenarios of the present invention. The curve marked with a red circle represents the relationship between the selection probability of scenario one and the SNR, the curve marked with a blue vertical dot represents the relationship between the selection probability of scenario two and the SNR, the curve marked with a blue asterisk represents the relationship between the selection probability of scenario three and the SNR, and the curve marked with a pink dashed line represents the relationship between the selection probability of scenario four and the SNR.

[0166] from Figure 3 As can be seen in (b), the present invention mainly uses a combination of Case 1 and Case 2 to perform detection in a uniform background environment, which closely approximates the real clutter environment.

[0167] Figure 4 In (a), the horizontal axis represents the signal-to-noise ratio (SNR) of the detected target, in dB, and the vertical axis represents the detection probability of the detected target. The curve marked with a red circle represents the relationship between the detection probability and SNR obtained by simulation using prior art 1; the curve marked with a blue vertical dot represents the relationship between the detection probability and SNR obtained by simulation using prior art 2; the curve marked with a blue asterisk represents the relationship between the detection probability and SNR obtained by simulation using prior art 3; the curve marked with a pink dashed line represents the relationship between the detection probability and SNR obtained by simulation using prior art 4; the curve marked with a pink asterisk represents the relationship between the detection probability and SNR obtained by simulation using prior art 5; and the curve marked with a black pentagram represents the relationship between the detection probability and SNR obtained by simulation using the method proposed in this invention.

[0168] from Figure 4 As can be seen in (a), the 31 detection probabilities obtained by the method of the present invention increase continuously with the increase of signal-to-noise ratio. Under the same signal-to-noise ratio, the detection probability obtained by the method of the present invention when there is an interfering target in the left reference cell half window is better than the other five existing technologies.

[0169] Figure 4 In (b), the horizontal axis represents the signal-to-noise ratio (SNR) of the detected target, in dB, and the vertical axis represents the selection probability of the four weighted allocation scenarios of the present invention. The curve marked with a red circle represents the relationship between the selection probability of scenario one and the SNR, the curve marked with a blue vertical dot represents the relationship between the selection probability of scenario two and the SNR, the curve marked with a blue asterisk represents the relationship between the selection probability of scenario three and the SNR, and the curve marked with a pink dashed line represents the relationship between the selection probability of scenario four and the SNR.

[0170] from Figure 4As can be seen in (b), when there is an interfering target in the left reference unit half window, the present invention tends to identify the environment as a uniform background when the signal-to-noise ratio is low. However, as the signal-to-noise ratio increases, the power of the interfering target also increases, and the present invention tends to identify a unilateral non-uniform environment, thus achieving the effect of adaptive environment.

[0171] Figure 5 In (a), the horizontal axis represents the signal-to-noise ratio (SNR) of the detected target, in dB, and the vertical axis represents the detection probability of the detected target. The curve marked with a red circle represents the relationship between the detection probability and SNR obtained by simulation using prior art 1; the curve marked with a blue vertical dot represents the relationship between the detection probability and SNR obtained by simulation using prior art 2; the curve marked with a blue asterisk represents the relationship between the detection probability and SNR obtained by simulation using prior art 3; the curve marked with a pink dashed line represents the relationship between the detection probability and SNR obtained by simulation using prior art 4; the curve marked with a pink asterisk represents the relationship between the detection probability and SNR obtained by simulation using prior art 5; and the curve marked with a black pentagram represents the relationship between the detection probability and SNR obtained by simulation using the method proposed in this invention.

[0172] from Figure 5 As can be seen in (a), the 31 detection probabilities obtained by the method of the present invention increase continuously with the increase of signal-to-noise ratio. Under the same signal-to-noise ratio, the detection probabilities obtained by the method of the present invention when there is one interfering target in each half window of the left and right reference units are significantly better than the other five existing technologies.

[0173] Figure 5 In (b), the horizontal axis represents the signal-to-noise ratio (SNR) of the detected target, in dB, and the vertical axis represents the selection probability of the four weighted allocation scenarios of the present invention. The curve marked with a red circle represents the relationship between the selection probability of scenario one and the SNR, the curve marked with a blue vertical dot represents the relationship between the selection probability of scenario two and the SNR, the curve marked with a blue asterisk represents the relationship between the selection probability of scenario three and the SNR, and the curve marked with a pink dashed line represents the relationship between the selection probability of scenario four and the SNR.

[0174] from Figure 5 As can be seen in (b), when there is an interfering target in each half window of the left and right reference units, the present invention tends to identify the environment as a uniform background when the signal-to-noise ratio is low. As the signal-to-noise ratio increases, the power of the interfering target also increases, and the present invention tends to identify a unilateral non-uniform environment or a bilateral non-uniform environment, thus achieving the effect of adaptive environment.

[0175] Figure 6In (a), the horizontal axis represents the signal-to-noise ratio (SNR) of the detected target, in dB, and the vertical axis represents the detection probability of the detected target. The curve marked with a pink dashed line represents the relationship between the detection probability and the SNR obtained by simulation using the prior art 4. The curve marked with a pink asterisk represents the relationship between the detection probability and the SNR obtained by simulation using the prior art 5. The curve marked with a black pentagram represents the relationship between the detection probability and the SNR obtained by simulation using the method proposed in this invention.

[0176] from Figure 6 As can be seen in (a), the 31 detection probabilities obtained by the method of the present invention increase continuously with the increase of signal-to-noise ratio. Under the same signal-to-noise ratio, the detection probabilities obtained by the method of the present invention when there are two interfering targets in each of the left and right reference cell half windows are significantly better than the other two existing technologies.

[0177] Figure 6 In (b), the horizontal axis represents the signal-to-noise ratio (SNR) of the detected target, in dB, and the vertical axis represents the selection probability of the four weighted allocation scenarios of the present invention. The curve marked with a red circle represents the relationship between the selection probability of scenario one and the SNR, the curve marked with a blue vertical dot represents the relationship between the selection probability of scenario two and the SNR, the curve marked with a blue asterisk represents the relationship between the selection probability of scenario three and the SNR, and the curve marked with a pink dashed line represents the relationship between the selection probability of scenario four and the SNR.

[0178] from Figure 6 As can be seen in (b), when there are two interfering targets in each of the left and right reference unit half windows, the present invention tends to identify the environment as a uniform background when the signal-to-noise ratio is low. As the signal-to-noise ratio increases, the power of the interfering targets also increases, and the present invention tends to identify a unilateral non-uniform environment or a bilateral non-uniform environment, thus achieving the effect of adaptive environment.

[0179] Figure 7In (a), the horizontal axis represents the signal-to-noise ratio (SNR) of the detected target, in dB, and the vertical axis represents the detection probability of the detected target. The curve marked with a red circle represents the relationship between the detection probability and SNR obtained by simulation using prior art 1; the curve marked with a blue vertical dot represents the relationship between the detection probability and SNR obtained by simulation using prior art 2; the curve marked with a blue asterisk represents the relationship between the detection probability and SNR obtained by simulation using prior art 3; the curve marked with a pink dashed line represents the relationship between the detection probability and SNR obtained by simulation using prior art 4; the curve marked with a pink asterisk represents the relationship between the detection probability and SNR obtained by simulation using prior art 5; and the curve marked with a black pentagram represents the relationship between the detection probability and SNR obtained by simulation using the method proposed in this invention.

[0180] from Figure 7 As can be seen in (a), the 31 detection probabilities obtained by the method of the present invention increase continuously with the increase of signal-to-noise ratio. Under the same signal-to-noise ratio, although the detection probability of the method of the present invention is not as good as that of the prior art 3 which uses a large number of false alarm probabilities to obtain detection probability in the clutter edge environment, its detection probability is still better than the other four prior art.

[0181] Figure 7 In (b), the horizontal axis represents the signal-to-noise ratio (SNR) of the detected target, in dB, and the vertical axis represents the selection probability of the four weighted allocation scenarios of the present invention. The curve marked with a red circle represents the relationship between the selection probability of scenario one and the SNR, the curve marked with a blue vertical dot represents the relationship between the selection probability of scenario two and the SNR, the curve marked with a blue asterisk represents the relationship between the selection probability of scenario three and the SNR, and the curve marked with a pink dashed line represents the relationship between the selection probability of scenario four and the SNR.

[0182] from Figure 7 As can be seen from (b), the present invention tends to identify the environment as a clutter edge environment in the case of clutter edge environment, which closely approximates the real clutter environment.

[0183] The simulation results above show that the present invention has good environmental adaptability and can achieve good detection performance in various clutter background environments.

[0184] The above description is merely a specific example of the present invention and does not constitute any limitation on the present invention. Obviously, those skilled in the art, after understanding the content and principles of the present invention, may make various modifications and changes in form and details without departing from the principles and structure of the present invention. However, these modifications and changes based on the ideas of the present invention are still within the scope of protection of the claims of the present invention.

Claims

1. A CFAR detection method based on multiple screening modules and adaptive weighting, characterized in that, Includes the following steps: Step 1: Construct a CFAR processing window, remove other interference signals from the processing window, and obtain the CFAR processing window after phase feature filtering; Step 2: Using an iterative approach, remove reference cells that exceed the detection threshold in the CFAR processing window after phase feature screening; Step 3: Calculate the half-window consistency characterization of the left reference cell in the detection unit. Consistency characterization of the right reference unit half-window Differential representation and according to , , Establish three corresponding result sets; Step 4, according to , , The result set is used for CFAR detector initialization and weighting. Step 5: Calculate the detection threshold based on the weights and output the detection results.

2. The CFAR detection method according to claim 1, characterized in that, The construction of the CFAR processing window in step 1 refers to randomly selecting one range cell from multiple range cells of the received signal as a detection unit, selecting M / 2 range cells on each side of the selected detection unit as protection units, and selecting N / 2 range cells on each side of each protection unit as reference units. The selected detection unit, protection units, and reference units are then combined to form a CFAR processing window. , M represents the total number of protection cells in a CFAR processing window, and N represents the total number of reference cells in a CFAR processing window.

3. The CFAR detection method according to claim 2, characterized in that, The steps for removing other interference signals from the processing window in step 1 are as follows: The first step is to select an unselected reference cell and perform univariate linear regression on the echo phase information of the selected reference cell to obtain the correlation coefficient and significance probability corresponding to the selected reference cell. The second step is to determine whether the correlation coefficient of the selected reference unit is greater than or equal to the correlation coefficient threshold. And at the same time, the significance probability is less than or equal to the default significance level. If so, the phase characteristic of the selected reference cell is considered linear, and the selected reference cell will be removed from the CFAR processing window; otherwise, no processing is performed. and All of these are thresholds set based on experience; The third step is to determine whether all reference cells have been selected. If so, obtain the CFAR processing window after phase feature filtering; otherwise, proceed to the first step.

4. The CFAR detection method according to claim 3, characterized in that, The step 2, which involves iteratively removing reference cells exceeding the detection threshold in the CFAR processing window after phase feature filtering, is as follows: The first step is to calculate the threshold factor corresponding to the current CFAR processing window: ,in, This indicates the total number of reference cells in the current CFAR processing window. This represents the set constant false alarm probability; The second step is to calculate the estimated background clutter power for the current CFAR processing window: ,in, Indicates the current CFAR processing window. One reference unit, , Indicates the first The echo intensity of each reference element; The third step is to calculate the culling threshold for the current CFAR processing window: ; The fourth step is to match the echo intensity of each reference cell with the rejection threshold. Compare the reference cells and remove those that exceed the threshold from the current CFAR processing window; otherwise, do nothing. Fifth, repeat steps one through four until no more reference cells need to be removed from the CFAR processing window, thus obtaining the removed CFAR processing window.

5. The CFAR detection method according to claim 1, characterized in that, Step 3 describes calculating the half-window consistency characterization of the left reference unit in the detection unit. Consistency characterization of the right reference unit half-window Differential representation It is obtained from the following formula: ； ； ； in, This represents the total number of all reference cells to the left of the detection unit. This indicates the currently selected leftmost [number]. The echo intensity of each reference element, , This represents the average echo intensity of all reference cells to the left of the detection unit. This represents the total number of all reference cells to the right of the detection unit. This indicates the currently selected rightmost [number]. The echo intensity of each reference element, , This represents the average echo intensity of all reference units to the right of the detection unit.

6. The CFAR detection method according to claim 5, characterized in that, According to step 3 , , Establishing the three corresponding result sets refers to: , , respectively with and Compare and build a result set accordingly: if Then set the first result set: ,otherwise, ; if Then set a second result set: ,otherwise, ; if Then, set a third result set: ,otherwise, , in, , All of these are judgment thresholds set based on experience.

7. The CFAR detection method according to claim 6, characterized in that, Step 4, which involves initializing the CFAR detector, refers to counting the number of reference cells in the current CFAR processing window. The CFAR detector consists of multiple OS-CFAR detectors. M OS-CFAR detectors are set to be weighted, and each detector has its own corresponding selection coefficient. Where M represents the total number of CFAR detectors to be weighted. , Indicates the first The coefficients corresponding to each OS-CFAR detector, which vary with Increase and increase, , .

8. The CFAR detection method according to claim 7, characterized in that, The weighted allocation mentioned in step 4 refers to the allocation based on the input... , , The three result sets are then weighted according to the following four scenarios, with each weighted OS-CFAR detector corresponding to a separate weight: exist ,and , In this case, the weight allocation is as follows: ； exist ,and , In this case, the weight allocation is as follows: ； exist and When the values ​​are different, the weights are allocated as follows: ； exist and In this case, the weight allocation is as follows: ； in, Indicates the first The weights corresponding to the weighted OS-CFAR detectors.

9. The CFAR detection method according to claim 8, characterized in that, The steps in step 5, which involve calculating the detection threshold based on the weights and outputting the detection results, are as follows: The first step is to sort all reference cells in the current CFAR processing window in ascending order based on their echo intensity. The second step is to calculate the threshold factor corresponding to each weighted OS-CFAR detector. : ； in, This represents the set constant false alarm probability. Represents the gamma function. Indicates the first A weighted threshold factor for the OS-CFAR detector; The third step is to sum the weighted values ​​of all weighted OS-CFAR detectors to obtain the detection threshold. : ； in, Indicates the first The echo intensity of each reference element; The fourth step is to determine whether the echo intensity of the detection unit is greater than [the specified value]. If yes, then proceed to step five after the target has been detected; otherwise, it is considered that no target has been detected. Fifth step: Output the target detection results.

10. A CFAR detection system based on multiple screening modules and adaptive weighting according to claim 1, characterized in that, The system includes a phase feature filtering module, an iterative elimination module, a result set establishment module, a weighted CFAR detector initialization and weighting allocation module, and a detection output module; wherein: The phase feature filtering module is used to remove other interference signals in the processing window to obtain the CFAR processing window after phase feature filtering. The repeated elimination module is used to iteratively eliminate reference cells that exceed the detection threshold in the CFAR processing window after phase feature screening, in order to eliminate significant noise or clutter spikes. The result set building module is used to calculate the half-window consistency characterization of the left reference unit in the detection unit respectively. Consistency characterization of the right reference unit half-window With differential representation and according to , , Create a result set; The weighted CFAR detector initialization and weighting allocation module is used to, according to , , The result set is used to initialize and assign weights to OS-CFAR detectors in various complex clutter environments. The detection output module is used to calculate the detection threshold based on the weights and output the detection results.

Images