A Mean-Type CFAR Detection Method Using Prior Information

Abstract

The present invention relates to a mean - type CFAR detection method applying prior information, belonging to the field of radar target detection. The steps include: collecting the clutter echo time series of multiple clutter cells within a period of time, and extracting the mean - type CFAR test statistic as training information; performing power transformation Gaussianization on the test statistic in the training information to obtain the optimal power; extracting the test statistic and performing power transformation on the clutter echo time series of multiple clutter cells in the training information to obtain the test statistic output by the clutter cells, and using the Monte Carlo method and false alarm rate to obtain the threshold; extracting the test statistic and performing power transformation on the echo time series of the cell under test and reference cells, then obtaining a new test statistic, and comparing it with the threshold to obtain the decision result. The present invention combines Bayesian estimation with the distribution information of the test statistic, and uses the continuously corrected prior distribution to extract a new test statistic and form a new detection method.

Description

Technical Field

[0001] The present invention relates to a mean - type CFAR detection method applying prior information, belonging to the field of radar target detection. Background Art

[0002] Detecting small targets at sea is one of the main tasks of sea - detecting radars. Above sea state 3, the non - uniform and non - stationary characteristics of sea clutter can easily affect the test statistic of traditional mean - type CFAR, resulting in a large number of false alarms. When sea spikes appear, this type of test statistic also has no ability to suppress them, greatly affecting the performance of the detector.

[0003] In traditional detection methods, the test statistic often obtains information from the echoes in the current time period. However, the echo information in the current time period is relatively limited. Although the detection efficiency of the radar will be affected by increasing the accumulation time to make the test statistic contain more echo information, in order to improve the separability of the test statistic, it is necessary to make full use of historical echo information. It can be predicted that there is a certain connection between the test statistics in the historical time period and the current time period within a few seconds. Currently, the published literature has used this connection to provide prior information for the test statistic in the current time period, thereby supporting the test statistic in the current time period and making the test statistic have better separability.

[0004] In the current processing methods based on prior distribution information, the usual approach is to use the detection statistics of historical frames to estimate the probability density function of the test statistic within a certain time, and select the test statistic with the maximum probability density from the probability densities of multiple test statistics in the current frame as the test statistic of the current frame. Although the above - mentioned method applies the prior distribution, it is not combined with the Bayesian estimation closely related to the prior distribution. As an existing parameter estimation method, Bayesian estimation can fuse the information of historical frames and current frames and make a more reasonable estimation of the distribution parameters. When the Bayesian estimation method is not used and prior information is used to estimate the features of the current frame, the insufficient application of prior information will lead to the fact that the feature estimation value cannot efficiently obtain the prior information of the historical frame, and then lead to insufficient separability of the feature estimation value.

[0005] In view of this characteristic, combined with the mean - type CFAR detection statistic, a target detection method with stronger separability and sea - spike suppression ability can be obtained, improving the detection performance for small targets. Summary of the Invention

[0006] To meet the above requirements, the present invention combines Bayesian estimation with the distribution information of the test statistic, fuses the information contained in the historical - frame test statistic with the current frame, and uses the continuously corrected prior distribution to extract a new test statistic and form a new detection method.

[0007] A mean - type CFAR detection method applying prior information in the present invention is characterized by including the following steps:

[0008] Step 1: Collect the clutter echo time series of multiple clutter cells within a period of time, extract the mean - type CFAR test statistic, and use it as training information;

[0009] Step 2: Perform power - transformation Gaussianization on the test statistic in the training information to obtain the optimal power;

[0010] Step 3: In the training stage, perform test statistic extraction and power - transformation on the clutter echo time series of multiple clutter cells in the training information, and then execute steps a - c to obtain the test statistic output by the clutter cells. Use the Monte Carlo method and false - alarm rate to obtain the threshold;

[0011] Step a: Take the time series of some clutter cells, divide the series into multiple sequence segments with a length of N, calculate the mean of the sequence segments, and then use maximum - likelihood estimation to estimate the mean and variance of the mean sequence as the mean and variance of the prior distribution to complete the setting of prior information;

[0012] Step b: Use the time series of another part of the clutter cells to delimit the cell to be detected and the reference cells;

[0013] Step c: For the time series of the reference cells and the cell to be detected, use the iterative perception algorithm of the Bayesian estimator to perform iteration;

[0014] Step 4: In the detection stage, extract the test statistic and perform power - transformation on the echo time series of the cell to be detected and the reference cells, and then execute step c in step 3 to obtain a new test statistic, and compare it with the threshold to obtain the decision result;

[0015] Step 5: Use the measured data to verify the proposed method.

[0016] Preferably, the specific steps of step 1 are as follows:

[0017] Adopt the selection strategy of the test statistic of the SO - CFAR method. First, perform square - law detection on the IQ echo signal:

[0018] (1);

[0019] In the formula, represents the radar echo signal after square - law detection, represents the real part of the radar IQ echo signal, represents the imaginary part of the radar IQ echo signal;

[0020] Secondly, traverse in the fast time dimension, select the unit to be detected and the corresponding reference unit, and calculate the test statistic of the unit to be detected according to the following formula:

[0021] (2);

[0022] wherein represents the time series of the test statistic of the unit to be referenced, represents the echo time series of the unit to be detected after square-law detection, , represent the echo time series of the reference units on both sides of the unit to be detected after square-law detection.

[0023] Preferably, the specific steps of step 2 are as follows:

[0024] First, Gaussianize the data to ensure the stable iteration of the subsequent prior distribution. The power transformation can be carried out according to the following formula:

[0025] (3);

[0026] wherein, refers to the test statistic sequence after the power the time series after power transformation.

[0027] The method of seeking the optimal power is carried out by using maximum likelihood estimation and chi-square test; when traversing to find the optimal power, first use maximum likelihood estimation to estimate the mean and variance of the test statistic under the current power; secondly, use the data after power transformation and the estimated mean and variance to perform chi-square normality test, and output the power that can make the test statistic closest to the Gaussian distribution as the optimal power.

[0028] Preferably, the specific steps of step 3 are as follows:

[0029] Denote the time series of the test statistic of the unit to be detected after power transformation as , and the time series of the test statistic of the reference unit as :

[0030] Step a: Take the time series of some clutter units. The total number of the series is K. Divide the series into K / N series segments with a length of N, and count the mean of the series segments to obtain K / N mean series. Use maximum likelihood estimation to estimate the mean and variance of these K / N mean series as the mean and variance of the prior distribution, and complete the setting of prior information;

[0031] Step b: Define the unit to be detected and the reference unit based on the time series of another part of the clutter units;

[0032] Step c: For the time series of the reference unit and the unit to be detected , use the iterative sensing algorithm of the Bayesian estimator to perform iterations to obtain the test statistic. The specific steps are as follows: I to VI:

[0033] I: Perform a sliding window on the time series of the unit. The length of the sliding window is N , and the sliding window step size is step = 1. The length of the time series of each unit is M . Unbiased estimates are made for the mean and variance of the test statistic within the current time window;

[0034] (4);

[0035] II: Secondly, using the mean and variance obtained from the unbiased estimate and combining with the prior distribution, obtain the Bayesian estimate of the mean of the test statistic within the current time window, which is also the mean of the posterior probability density function. Similarly, the variance of the posterior probability density function can also be calculated :

[0036] (5);

[0037] In the formula, , refer to the mean and variance of the prior distribution, refers to the unbiased estimate of the variance of the current frame, refers to the unbiased estimate of the mean of the current frame.

[0038] III: Based on the Bayesian estimate value obtained from the current sliding window, perform sequential estimation on the mean and variance of the Bayesian estimate value to update the mean and variance of the Bayesian estimate value:

[0039] (6);

[0040] In the formula, k refers to the number of iterations, refers to the Bayesian estimate value obtained from the current frame.

[0041] IV: Based on the Bayesian estimate value of the current sliding window, the mean and variance of the Bayesian estimate value obtained from the sequential estimation, and the variance of the current frame, correct the mean and variance of the prior distribution. The correction method is as follows:

[0042] (7);

[0043] Wherein, refers to the unbiased estimate of the sample variance of the current frame (the k th frame, which is also the k th iteration).

[0044] Ⅴ: If the current cell is the cell to be detected, execute this step; otherwise, enter step Ⅵ;

[0045] Collect the Bayesian estimates of multiple reference cells at the same time as the current sliding window, and calculate the mean of these Bayesian estimates The probability density under the posterior density function of the current cell , and output it as a new test statistic; when the iteration is stable, the Bayesian estimate is approximately regarded as sampled from the prior distribution of the corresponding cell; since most of the reference cells are clutter cells, when the cell to be detected is also a clutter cell, the Bayesian estimates of the two can be regarded as sampled from the same distribution, so the Bayesian estimate of the reference cell has a high probability density under the probability density function of the cell to be detected; while when the cell to be detected is a target cell, there are differences in their prior distributions, resulting in a low probability density of the Bayesian estimate of the reference cell under the probability density function of the cell to be detected. This phenomenon makes the new test statistic extracted by this method have a higher discrimination degree compared with the original test statistic:

[0046] (8);

[0047] Wherein, , are the mean and variance of the prior distribution

[0048] Ⅵ: Return to step Ⅰ until the sliding window reaches the end of the sequence;

[0049] Step d: Sort the obtained test statistics from small to large, and select the th test statistic as the decision threshold , and step 3 ends.

[0050] Preferably, the specific steps of step 4 are as follows:

[0051] In the detection stage, use the time echo sequences of the cell to be detected and the reference cells , to extract the test statistic according to formula (2), and obtain the test statistic of the cell to be detected and the test statistic Perform a power transformation according to Equation (10) to obtain the test statistic after the power transformation of the unit to be detected and the test statistic after the power transformation of the unit to be detected After step c, the probability density function of the unit to be detected can be obtained (see Equation (8)) and the mean value of the Bayesian estimate of the reference unit , and calculate as the output of the new test statistic. Pass Compare with the threshold . If it indicates that the unit to be detected is the target unit, otherwise it is clutter

[0052] Preferably, the specific steps of step 5 are as follows: To verify the effectiveness, use the radar staring data of HH polarization and VV polarization for verification. The target to be detected in the radar staring data is the channel buoy. Select the unit where the channel buoy is located as the unit to be detected, and the distance units around the unit where the channel buoy is located as the reference units. Calculate the test statistic of the unit to be detected and the test statistic of the reference unit . Refer to step 4 to obtain the new test statistic , and compare it with the threshold . If it indicates that the unit to be detected is the target unit, otherwise it is clutter; according to the judgment result, verify the effectiveness

[0053] The beneficial effects of the present invention are as follows

[0054] First, make full use of the test statistics of historical frames, which not only completes the correction of the prior distribution, but also completes the fusion of the test statistics of historical frames and current frames

[0055] Second, according to the prior distribution containing historical frame information, extract a more discriminative test statistic, improving the performance of the detector

[0056] Third, after the power transformation, the mean-type CFAR statistic that was originally sensitive to sea spikes becomes more concentrated, having a certain inhibitory effect on sea spikes

[0057] After verification with HH and VV measured data, it can very effectively improve the small target detection ability of the detector Description of the Drawings

[0058] Figure 1 : Changes in the distribution histogram of the test statistic before and after the power transformation

[0059] Figure 2 : Prior probability density curves of the unit to be detected and the reference unit and the probability density map of the Bayesian estimate value

[0060] Figure 3 : Binary comparison graph of the detection results of the present invention and the existing method in the HH polarization and five-level sea state data sets;

[0061] Figure 4 : Comparison graph of the detection results of the present invention and the original mean-type CFAR method on 9 data sets. Among them, serial numbers 1 to 2 are HH polarization four-level sea state data, serial number 3 is HH polarization five-level sea state data, and serial numbers 4 to 9 are VV polarization five-level sea state data. Detailed implementation manners

[0062] The following further describes the detailed implementation manners of the present invention in combination with the accompanying drawings and technical solutions.

[0063] Step 1:

[0064] Adopt the selection strategy of the test statistic of the SO-CFAR method. First, perform square-law detection on the IQ echo signal:

[0065] (1);

[0066] In the formula, represents the radar echo signal after square-law detection, represents the real part of the radar IQ echo signal, represents the imaginary part of the radar IQ echo signal.

[0067] Secondly, traverse in the fast time dimension, select the cell to be detected and the corresponding reference cells, and calculate the test statistic of the cell to be detected according to the following formula:

[0068] (2);

[0069] In the formula represents the time series of the test statistic of the reference cell to be detected, represents the echo time series of the cell to be detected after square-law detection, , represent the echo time series after square-law detection of the reference cells on both sides of the cell to be detected.

[0070] Step 2:

[0071] First, Gaussianize the data to ensure the stable iteration of the subsequent prior distribution. The power transformation can be performed according to the following formula:

[0072] (3);

[0073] In the formula, refers to the test statistic sequence after the power The time series after power transformation.

[0074] The method for seeking the optimal power is carried out by using the maximum likelihood estimation and chi-square test; when traversing to find the optimal power, first, the mean and variance of the test statistic under the current power are estimated by using the maximum likelihood estimation; secondly, the chi-square normality test is carried out by using the data after power transformation and the estimated mean and variance, and the power that can make the test statistic closest to the Gaussian distribution is output as the optimal power. The distribution histogram of the test statistic without power transformation is as Figure 1 shown in (a) in Figure 1 and the distribution histogram of the test statistic after power transformation is as

[0075] Step 3: Denote the time series of the test statistic of the unit to be detected after power transformation as and the time series of the test statistic of the reference unit as :

[0076] Step a: Take the time series of some clutter units, the total number of sequences is K, divide the sequences into K / N sequence segments with a length of N, count the mean of the sequence segments, obtain K / N mean sequences, and use the maximum likelihood estimation to estimate the mean and variance of these K / N mean sequences as the mean and variance of the prior distribution to complete the setting of prior information;

[0077] Step b: Use the time series of another part of clutter units to delimit the unit to be detected and the reference unit;

[0078] Step c: For the time series of the reference unit and the unit to be detected , use the iterative sensing algorithm of the Bayesian estimator to perform iteration to obtain the test statistic. The specific steps are as follows: I~VI:

[0079] I: Perform a sliding window on the time series of the unit, the sliding window length is N , the sliding window step size is step = 1, the length of the time series of each unit is M , and perform an unbiased estimation on the mean and variance of the test statistic within the current time window;

[0080] (4);

[0081] II: Secondly, use the mean and variance obtained by the unbiased estimation, combined with the prior distribution, to obtain the Bayesian estimation , that is, the mean of the posterior probability density function. Similarly, the variance of the posterior probability density function can also be calculated. :

[0082] (5);

[0083] In the formula, , refer to the mean and variance of the prior distribution, refers to the unbiased estimate of the variance of the current frame, refers to the unbiased estimate of the mean of the current frame.

[0084] Ⅲ: According to the Bayesian estimate obtained from the current sliding window, perform sequential estimation on the mean and variance of the Bayesian estimate, and update the mean of the Bayesian estimate. and variance :

[0085] (6);

[0086] In the formula, k refers to the number of iterations, refers to the Bayesian estimate obtained from the current frame.

[0087] Ⅳ: According to the Bayesian estimate of the current sliding window, the mean and variance of the Bayesian estimate obtained from sequential estimation, and the variance of the current frame, correct the mean and variance of the prior distribution. The correction method is as follows:

[0088] (7);

[0089] In the formula, refers to the unbiased estimate of the sample variance of the current frame (the k th frame, and also the k th iteration).

[0090] Ⅴ: If the current unit is the unit to be detected, execute this step; otherwise, enter step Ⅵ;

[0091] Collect the Bayesian estimates of multiple reference units at the same time as the current sliding window moment, and calculate the mean of these Bayesian estimates, which is the probability density under the posterior density function of the current unit. , output as the new test statistic; when the iteration is stable, the Bayesian estimate is approximately regarded as sampled from the prior distribution of the corresponding cell; since the reference cells are mostly clutter cells, when the cell to be detected is also a clutter cell, the Bayesian estimates of the two can be regarded as sampled from the same distribution, so the Bayesian estimate of the reference cell has a high probability density under the probability density function of the cell to be detected; while when the cell to be detected is a target cell, there are differences in their prior distributions, resulting in the Bayesian estimate of the reference cell having a low probability density under the probability density function of the Bayesian estimate of the cell to be detected, as Figure 2 shown, this phenomenon makes the new test statistic extracted by this method have a higher discrimination degree compared with the original test statistic:

[0092] (8);

[0093] In the formula, and are the mean and variance of the prior distribution

[0094] Ⅵ: Return to step Ⅰ until the sliding window reaches the end of the sequence;

[0095] Step d: Sort the obtained test statistics from small to large, and select the th test statistic as the decision threshold , and step 3 ends.

[0096] Step 4: In the detection stage, use the time echo sequences of the cell to be detected and the reference cells , to extract the test statistic according to formula (2), and obtain the test statistic of the cell to be detected and the test statistic of the reference cell. Perform power transformation according to formula (10) to obtain the test statistic after power transformation of the cell to be detected and the test statistic after power transformation of the cell to be detected. Through step c, the probability density function (see formula (8)) of the cell to be detected and the mean of the Bayesian estimate of the reference cell can be obtained, and calculate as the new test statistic output. By comparing with the threshold , if , it indicates that the cell to be detected is a target cell, otherwise it is clutter.

[0097] Step 5: To verify the effectiveness of the present invention, the radar staring data with HH polarization and VV polarization are used to verify the present invention. The target to be detected in the radar staring data is a channel buoy. The unit where the channel buoy is located is selected as the unit to be detected, and the range units around the unit where the channel buoy is located are used as reference units. The test statistic of the unit to be detected is calculated by formula (2). and the test statistic of the reference unit . Referring to Step 4, the new test statistic can be obtained and compared with the threshold . If , it indicates that the unit to be detected is the target unit; otherwise, it is clutter. According to the judgment result, the effectiveness of the present invention is verified. The binary images of the detection results of the original method and the present invention are as shown in Figure 3 . The small target detection probability of the original method is only 11.59%, while the small target detection probability of the present invention is 93.87%, which can clearly prove the performance superiority of the present invention.

[0098] The detection results in the remaining radar staring data sets with HH polarization and VV polarization are as shown in Figure 4 , which can also prove the superiority of the present invention compared with the traditional method.

Claims

1. A mean - type CFAR detection method applying prior information, characterized in that It includes the following steps: Step 1: Collect the clutter echo time series of multiple clutter cells over a period of time, and extract the mean - type CFAR test statistic as training information; Step 2: Perform a power transformation and Gaussianization on the test statistic in the training information to obtain the optimal power; Step 3: In the training stage, perform test statistic extraction and power transformation on the clutter echo time series of multiple clutter cells in the training information, and then execute Steps a - c to obtain the test statistic output by the clutter cell. Use the Monte Carlo method and the false - alarm rate to obtain the threshold; Step a: Take the time series of some clutter cells, divide the series into multiple segments of length N, calculate the mean of the segments, and then use maximum - likelihood estimation to estimate the mean and variance of the mean series as the mean and variance of the prior distribution to complete the setting of prior information; Step b: Use the time series of another part of the clutter cells to demarcate the cell to be detected and the reference cells; Step c: For the time series of the reference cells and the cell to be detected, perform iteration using the iterative sensing algorithm of the Bayesian estimator; Step 4: In the detection stage, extract the test statistic and perform power transformation on the echo time series of the cell to be detected and the reference cells, and then execute Step c in Step 3 to obtain a new test statistic, and compare it with the threshold to obtain the decision result; Step 5: Use the measured data to verify the proposed method.

2. A mean - type CFAR detection method using prior information according to claim 1, characterized in that: The specific steps of Step 1 are as follows: Adopt the selection strategy of the test statistic of the SO - CFAR method. First, perform square - law detection on the IQ echo signal: （1）； In the formula, represents the radar echo signal after square-law detection, represents the real part of the radar IQ echo signal, represents the imaginary part of the radar IQ echo signal; Secondly, traverse in the fast - time dimension, select the cell to be detected and the corresponding reference cells, and calculate the test statistic of the cell to be detected according to the following formula: （2）； where denotes the time series of the test statistic of the unit to be referenced, denotes the echo time series of the unit to be detected after square-law detection, , denotes the echo time series of the reference units on both sides of the unit to be detected after square-law detection.

3. A mean - type CFAR detection method using prior information according to claim 1, characterized in that: The specific steps of Step 2 are as follows: First, Gaussianize the data to ensure the stable iteration of the subsequent prior distribution. The power transformation can be carried out according to the following formula: （3）； In the formula, denotes the sequence of test statistics after the power the time series after the power transformation; The method of seeking the optimal power is carried out using maximum - likelihood estimation and chi - square test; when traversing to find the optimal power, first use maximum - likelihood estimation to estimate the mean and variance of the test statistic under the current power; Secondly, use the data after power transformation and the estimated mean and variance to perform a chi - square normality test, and output the power that can make the test statistic closest to the Gaussian distribution as the optimal power.

4. A mean - type CFAR detection method using prior information according to claim 3, characterized in that: The specific steps of Step 3 are as follows: Denote the time series of the test statistic of the unit to be detected after power transformation as , and denote the time series of the test statistic of the reference unit as : Step a: Take the time series of some clutter cells , the total number of sequences is K. Divide the sequences into K / N sequence segments with a length of N, and calculate the mean of each sequence segment to obtain K / N mean sequences. Use the maximum likelihood method to estimate the mean and variance of these K / N mean sequences as the mean and variance of the prior distribution, and complete the setting of prior information; Step b: Use the time series of another part of the clutter cells to demarcate the cell to be detected and the reference cells; Step c: For the time series of the reference unit and the unit to be detected , use the iterative perception algorithm of the Bayesian estimator to iterate and obtain the test statistic Step d: Sort the obtained test statistics in ascending order, and select the th test statistic as the decision threshold , and step 3 ends.

5. A mean - type CFAR detection method using prior information according to claim 4, characterized in that: The specific steps of Step c are as follows in Ⅰ - Ⅵ: Ⅰ: Slide a window over the time series of the unit, with the window length being N , the window step being step = 1, and the length of the time series of each unit being M . Unbiased estimates are made for the mean and variance of the test statistic within the current time window; （4）； Ⅱ: Secondly, using the mean and variance obtained from the unbiased estimation and combining with the prior distribution, the Bayesian estimation of the mean of the test statistic within the current time window is obtained , that is, the mean of the posterior probability density function. Similarly, the variance of the posterior probability density function can also be calculated : （5）； wherein, , denote the mean and variance of the prior distribution, denotes the unbiased estimate of the current frame variance, denotes the unbiased estimate of the current frame mean; Ⅲ: Sequentially estimate the mean and variance of the Bayesian estimate based on the Bayesian estimate value obtained from the current sliding window, and update the mean of the Bayesian estimate value and variance : （6）； In the formula, k denotes the number of iterations, denotes the Bayesian estimation value obtained for the current frame; Ⅳ: Based on the Bayesian estimate value of the current sliding window, the mean and variance of the Bayesian estimate values obtained by sequential estimation, and the variance of the current frame, correct the mean and variance of the prior distribution. The correction method is as follows: （7）； In the formula, refers to the unbiased estimate of the sample variance of the current frame; Ⅴ: If the current cell is the cell to be detected, then execute this step, otherwise enter Step Ⅵ; Collect the Bayesian estimates of multiple reference cells at the same time as the current sliding window moment, and calculate the mean of these Bayesian estimates The probability density under the posterior density function of the current cell , and output it as a new test statistic; when the iteration is stable, the Bayesian estimate is approximately regarded as sampled from the prior distribution of the corresponding cell; since most reference cells are clutter cells, when the cell to be detected is also a clutter cell, the Bayesian estimates of the two can be regarded as sampled from the same distribution, so the Bayesian estimate of the reference cell has a high probability density under the probability density function of the cell to be detected; while when the cell to be detected is a target cell, there are differences in their prior distributions, resulting in a low probability density of the Bayesian estimate of the reference cell under the probability density function of the cell to be detected. This method uses the differences between the clutter cell and the cell to be detected as described above to extract a new test statistic, as shown in Equation (8). This test statistic has a higher discrimination degree compared to the original test statistic: （8）； wherein, and are the mean and variance of the prior distribution; Ⅵ: Return to Step Ⅰ until the sliding window reaches the end of the sequence.

6. The mean - type CFAR detection method using prior information according to claim 4, characterized in that: The specific steps of Step 4 are as follows: In the detection stage, the unit to be detected is used and the reference unit 、 to extract the test statistic according to formula (2) for the time echo sequences of the unit to be detected and the reference unit, obtaining the test statistic of the unit to be detected and the test statistic of the reference unit. Perform a power transformation according to formula (3) to obtain the test statistic after power transformation of the unit to be detected and the test statistic after power transformation of the unit to be detected. Through step c, obtain the probability density function of the unit to be detected, refer to formula (8) and the mean value of the Bayesian estimate of the reference unit, and calculate as the output of the new test statistic. Compare it with the threshold through . If then it indicates that the unit to be detected is the target unit, otherwise it is clutter.

7. A mean - type CFAR detection method using prior information according to claim 3, characterized in that: The specific steps of Step 5 are as follows: To verify the effectiveness, the radar staring data of HH polarization and VV polarization are used for verification. The target to be detected in the radar staring data is the channel buoy. The unit where the channel buoy is located is selected as the unit to be detected, and the range cells around the unit where the channel buoy is located are used as reference cells. The test statistic of the unit to be detected is calculated by formula (2). and the test statistic of the reference cell . Referring to step 4, the new test statistic is obtained and compared with the threshold . If , it indicates that the unit to be detected is the target unit; otherwise, it is clutter. According to the decision result, the effectiveness is verified.

Images