An adaptive constant false alarm detection method for extended range targets based on improved VI

Abstract

The present invention relates to an adaptive constant false alarm detection technology for distance-extended targets, and belongs to the field of radar target detection. Aiming at the problem of adaptive constant false alarm detection of distance-extended targets when interference targets appear in both reference windows on both sides, the present invention provides an adaptive constant false alarm detection method for distance-extended targets based on improved VI. First, the VI and MR, which measure the degree of uniformity and consistency of the background, are calculated; then, the spatial distribution type of the background is divided according to the VI value and the MR value; then, different detection methods are selected according to the different spatial distribution types of the background, and the key is to solve the problem of determining the modified k value in the OS-CFAR detection method. Compared with the classical VI method, the present invention jointly utilizes the phase difference characteristics of the target and the sea clutter, the modified OS-CFAR method and the classical VI method, and has the advantages of strong anti-interference target capability and accurate clutter background estimation.

Description

Technical Field

[0001] The invention belongs to the field of radar target detection, and in particular relates to an adaptive constant false alarm detection technology for a range-extended target when interference targets appear in reference windows on both sides. Background Art

[0002] For marine target detection, in order to obtain the best constant false alarm detection performance, it is necessary to use the reference unit near the detection unit to estimate the average power of the background, so as to adaptively set the detection threshold. Different background power estimation methods correspond to different constant false alarm detection methods. When the background is uniform, the CA-CFAR detector has the best detection performance, but when the background is non-uniform, the performance of the CA-CFAR detector will drop sharply. In order to combat the clutter edge, the GO-CFAR detector is generally used; and in order to deal with multiple adjacent interferences in the reference sliding window, the SO-CFAR detector and the OS-CFAR detector are generally used; but the performance improvement of these detectors under a certain non-uniform background will lead to additional CFAR loss or masking effect in other backgrounds. VI-CFAR makes comprehensive use of the advantages of CA, GO, and SO. It has a smaller constant false alarm loss under uniform background and better robustness under non-uniform background. As the radar resolution increases, the target echo appears as a distance-extended target. If there are multiple adjacent targets at this time, the clutter background power estimation of OS-CFAR will be too large, thus losing its excellent performance in a multi-target environment. This requires accurate calculation of the k value to prevent the clutter background power estimation from being too large.

[0003] Aiming at the problem of adaptive constant false alarm detection of range-extended targets when interference targets appear in reference windows on both sides, the present invention provides an adaptive constant false alarm detection method for range-extended targets based on improved VI. Summary of the invention

[0004] The purpose of the present invention is to provide a method for adaptive constant false alarm detection of extended range targets by using the phase difference characteristics of the target and sea clutter, a modified OS-CFAR method and a classic VI-CFAR method, in order to solve the problem that the detection performance of the traditional VI-CFAR method is seriously reduced when performing adaptive constant false alarm detection of extended range targets when interference targets appear in both reference windows. The technical problems to be solved include:

[0005] (1) When an interfering target appears in the single-side reference window, the classic VI-CFAR adopts the half-window CA-CFAR detection strategy; when an interfering target appears in the double-side reference window, VI-CFAR adopts the SO-CFAR detection strategy; however, these detection strategies all have the problem of background power estimation error, which leads to the failure of the detection algorithm.

[0006] (2) The classic OS-CFAR method has the defect of incorrectly selecting the k value. At this time, the remaining interference targets in the reference window will cause serious performance degradation.

[0007] The adaptive constant false alarm detection method for extended range targets based on improved VI of the present invention comprises the following steps:

[0008] S1, calculate the background uniformity and consistency metrics VI and MR;

[0009] S2. Classify the spatial distribution type of the background according to the VI value and MR value;

[0010] S3. Select different detection methods based on the different spatial distribution types of the background.

[0011] Preferably, the step S1, calculating the measures VI and MR of the background uniformity and consistency, specifically includes:

[0012] The VI value measures the uniformity of the reference window sampling value, and the MR value measures the consistency of the power mean of the front and back reference windows. The front reference window is denoted as A, and the observation samples in it are denoted as {x A,1 ,…,x A,n}, n is the number of samples, and the VI value of the previous reference sliding window is recorded as VI A , the sample mean is recorded as The back reference window is denoted as B, and the observation samples in it are denoted as {x B,1 ,…,x B,n}, the VI value of the back reference sliding window is recorded as VI B , the sample mean is recorded as Therefore, the calculation method of VI value and MR value is as follows:

[0013]

[0014] Preferably, the step S2, dividing the spatial distribution type of the background according to the VI value and the MR value, specifically includes:

[0015] S21, the calculated VI A Value, VI B The values ​​are respectively related to the decision threshold K VI Compare, if VI A <K VI , then the sampling values ​​of the front reference window are uniform, otherwise they are non-uniform; if VI B <K VI , then the sampling values ​​of the back reference window are uniform, otherwise they are non-uniform;

[0016] S22, the calculated MR value and K MR For comparison, if Then the power means of the front and rear reference windows are the same, otherwise the means are different;

[0017] S23, determine K according to the preset false alarm rate VI and K MR Generally, the probability α of misjudging a uniform environment as a non-uniform environment should be of the same order of magnitude as the preset false alarm rate, and the probability β of misjudging the same mean of the front and rear reference windows as different is required to be no greater than 0.1. Therefore, K VI =4.76, K MR =1.806 respectively corresponds to α=4×10 -4 , β = 0.1;

[0018] S24. According to the VI-CFAR decision criterion, the spatial distribution types of the reference window clutter background are divided into the following three cases:

[0019] (1) If the front reference window is uniform, the rear reference window is uniform, and the power mean of the front and rear reference windows is the same, then the background is uniform;

[0020] (2) If the front reference window is uniform, the rear reference window is uniform, and the power means of the front and rear reference windows are different, it is a clutter edge background;

[0021] (3) If the front reference window or the rear reference window is uneven, it is judged that an interference target appears in the reference window.

[0022] Preferably, the step S3 selects different detection methods according to the spatial distribution types of different backgrounds, specifically including:

[0023] S31, if the background is uniform, the CA-CFAR detection method is used;

[0024] S32, if it is a clutter edge background, the GO-CFAR detection method is used;

[0025] S33. If it is determined that there is an interference target in the reference window, the OS-CFAR detection method with a modified k value is adopted, wherein the method for determining the modified k value and the estimation of the clutter background power is as follows:

[0026] (S331) Take out the pulse train echo corresponding to each distance unit in the reference window, use the linear regression function to calculate the phase linearity estimation of this pulse train echo, and use 0.85 as the threshold. When the phase linearity estimation is greater than 0.85, then

[0027] Determine that the observation sample of the range unit is an interference target, otherwise it is clutter;

[0028] (S332) Count the number of distance cells in the reference window that are determined to be interference targets, recorded as m; then, calculate as follows:

[0029] The k value of OS-CFAR detection method,

[0030]

[0031] (S333) The kth ordered sampling value is used as the clutter background power estimation of OS-CFAR.

[0032] The principle of the present invention is as follows:

[0033] Calculate the background uniformity and consistency metrics VI and MR. The VI value represents the uniformity of the reference window sampling value, and the MR value represents the consistency of the power mean of the previous and next reference windows;

[0034] According to the VI value and MR value, the spatial distribution type of the background is divided. The calculated VI value and MR value are compared with the judgment threshold K VI and K MR Compare, if VI<K VI , then the reference window sampling values ​​are uniform, otherwise they are non-uniform; if Then the power means of the front and rear windows are the same, otherwise the means are different. According to the VI-CFAR decision criterion, the spatial distribution type of the clutter background in the reference window is divided into the following three cases: (1) uniform background; (2) clutter edge background; (3) the case where an interference target appears in the reference window;

[0035] Different detection methods are selected according to the different spatial distribution types of the background. (1) If it is a uniform background, the CA-CFAR detection method is used; (2) If it is a clutter edge background, the GO-CFAR detection method is used; (3) If there is an interference target in the reference window, the OS-CFAR detection method with a modified k value is used. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a principle block diagram of an adaptive constant false alarm detection method for extended-range targets based on improved VI;

[0037] Figure 2 It is the principle diagram of the OS-CFAR detection method with modified k value;

[0038] Figure 3 is the result of distinguishing the spatial distribution type of clutter background in the reference window Figure 1 ;

[0039] Figure 4 is the result of distinguishing the spatial distribution type of clutter background in the reference window Figure 2 ;

[0040] Figure 5 is the result of distinguishing the spatial distribution type of clutter background in the reference window Figure 3 ;

[0041] Figure 6 The results before and after screening Figure 1 ;

[0042] Figure 7 The results before and after screening Figure 2 ;

[0043] Figure 8 There is one interference target image in each of the front and rear reference sliding windows;

[0044] Fig. 9 There are two interference target images in the front and rear reference sliding windows. DETAILED DESCRIPTION

[0045] The present invention will be described in further detail below in conjunction with the accompanying drawings. Figure 1 The specific implementation of the present invention is divided into the following steps:

[0046] (1) Calculate the background uniformity and consistency metrics VI and MR;

[0047] (2) Based on the VI and MR values, the spatial distribution types of the background are divided;

[0048] (3) Select different detection methods based on the different spatial distribution types of the background.

[0049] The following is a detailed description of the above steps:

[0050] (1) Calculate the background uniformity and consistency metrics VI and MR.

[0051] The VI value measures the uniformity of the reference window sampling value, and the MR value measures the consistency of the power mean of the front and back reference windows. The front reference window is denoted as A, and the observation samples in it are denoted as {x A,1 ,…,x A,n}, n is the number of samples, and the VI value of the previous reference sliding window is recorded as VI A , the sample mean is recorded as The back reference window is denoted as B, and the observation samples in it are denoted as {x B,1 ,…,x B,n}, the VI value of the back reference sliding window is recorded as VI B , the sample mean is recorded as Therefore, the calculation method of VI value and MR value is as follows:

[0052]

[0053] (2) According to the VI value and MR value, the spatial distribution type of the background is divided.

[0054] ① The calculated VI A Value, VIB The values ​​are respectively related to the decision threshold K VI Compare, if VI A <K VI , then the sampling values ​​of the front reference window are uniform, otherwise they are non-uniform; if VI B <K VI , then the sampling values ​​of the back reference window are uniform, otherwise they are non-uniform;

[0055] ②Compare the calculated MR value with K MR For comparison, if Then the power means of the front and rear reference windows are the same, otherwise the means are different;

[0056] ③Determine K based on the preset false alarm rate VI and K MR Generally, the probability α of misjudging a uniform environment as a non-uniform environment should be of the same order of magnitude as the preset false alarm rate, and the probability β of misjudging the same mean of the front and rear reference windows as different is required to be no greater than 0.1. Therefore, K VI =4.76, K MR =1.806 respectively corresponds to α=4×10 -4 , β=0.1.

[0057] ④ According to the VI-CFAR decision criterion, the spatial distribution type of the reference window clutter background is divided into the following three cases: (a) If the front reference window is uniform, the rear reference window is uniform, and the power means of the front and rear reference windows are the same, it is a uniform background; (b) If the front reference window is uniform, the rear reference window is uniform, and the power means of the front and rear reference windows are different, it is a clutter edge background; (c) If the front reference window or the rear reference window is not uniform, it is judged that an interference target appears in the reference window.

[0058] For the 100th distance unit where the target to be measured is located in the figure below, calculate its corresponding VI according to the calculation formula in step (1): A VI B and MR value, where n = 16; then combined with K given in step (2) VI =4.76, K MR

[0059] =1.806, using the discriminant given in step (2) ① and ②, the discriminant results of the reference window clutter background spatial distribution type in the following three figures are obtained as follows:

[0060] Figure 3 Medium, VI A =2.1314,VI B =2.3464, MR = 1.4937; because VI A <K VI VI B <KVI , Therefore, we can judge Figure 1 The clutter background spatial distribution type of the reference window corresponding to the No. 100 unit under test is a uniform environment;

[0061] Figure 4 Medium, VI A =1.5426,VI B =2.0805, MR = 0.1472; because VI A <K VI VI B <K VI Established, and Therefore, judge Figure 2 The clutter background spatial distribution type of the reference window corresponding to the No. 100 unit under test is a clutter edge environment;

[0062] Figure 5 Medium, VI A =15.5632,VI B =10.7013, MR = 2.4995; because VI A >K VI VI B >K VI , so it is judged Figure 3 The clutter background spatial distribution type of the reference window corresponding to the No. 100 unit under test is a multi-target environment; (MR judgment is not required at this time).

[0063] (3) Select different detection methods based on the spatial distribution types of different backgrounds.

[0064] ① If the background is uniform, the CA-CFAR detection method is used;

[0065] ② If the background is clutter edge, the GO-CFAR detection method is used;

[0066] ③ If it is determined that there is an interference target in the reference window, the OS-CFAR detection method with a corrected k value is used, where the method for determining the corrected k value and the estimated clutter background power is as follows:

[0067] (a) Take out the pulse train echo corresponding to each range unit in the reference window, use the linear regression function to calculate the phase linearity estimation of this series of pulse echoes, and use 0.85 as the threshold. When the phase linearity estimation is greater than 0.85, the observation sample of the range unit is judged as an interference target, otherwise it is clutter. According to this step, the measured data shown in the figure below are screened to obtain the results before and after screening. Figure 6-7 as follows.

[0068] (b) Count the number of distance cells judged as interference targets in the reference window, denoted as m. Then, calculate the k value of the OS-CFAR detection method as follows:

[0069]

[0070] (c) The kth ordered sampling value is used as the clutter background power estimation of OS-CFAR.

[0071] (4) After steps (1), (2) and (3) to determine the spatial distribution type of the clutter background in the reference window and screen the reference unit, the background power level estimate of the unit to be tested is obtained, denoted as Z, and the sample of the unit to be tested is denoted as D. The threshold factor is obtained by Monte-Carlo simulation under a uniform background, denoted as T. Thus, the following detector is obtained, denoted as VIMOS-CFAR detector,

[0072] D<>TZ

[0073] Attached Figure 8 , attached Fig. 9 Taking a multi-target environment with interference on both sides as an example, a performance simulation comparison between the VIMOS-CFAR detector and the classic VI-CFAR detector is given. It can be seen that in this environment, the VIMOS-CFAR detector is significantly better than the VI-CFAR detector.

Claims

1. An adaptive constant false alarm detection method for extended range targets based on improved VI, characterized in that The following steps are involved: S1, calculate the background uniformity and consistency metrics VI and MR; S2. Classify the spatial distribution type of the background according to the VI value and MR value; S3. Select different detection methods according to different spatial distribution types of the background; The step S3, according to the spatial distribution type of different backgrounds, selects different detection methods, specifically including: S31, if the background is uniform, the CA-CFAR detection method is used; S32, if it is a clutter edge background, the GO-CFAR detection method is used; S33. If it is determined that there is an interference target in the reference window, the OS-CFAR detection method with a modified k value is adopted, wherein the method for determining the modified k value and the estimation of the clutter background power is as follows: S331 takes out the pulse train echo corresponding to each range unit in the reference window, calculates the phase linearity estimation of this series of pulse echoes by using the linear regression function, and takes 0.85 as the threshold. When the phase linearity estimation is greater than 0.85, it is judged that the observation sample of the range unit is an interference target, otherwise it is clutter; S332 counts the number of distance cells judged as interference targets in the reference window, recorded as m; then, the k value of the OS-CFAR detection method is calculated according to the following formula: S333 uses the kth ordered sampling value as the clutter background power estimation of OS-CFAR.

2. A method for adaptive constant false alarm detection of extended range targets based on improved VI according to claim 1, wherein step S1, calculating the metrics VI and MR of background uniformity and consistency, specifically comprises: The VI value measures the uniformity of the reference window sampling value, and the MR value measures the consistency of the power mean of the front and back reference windows. The front reference window is denoted as A, and the observation samples in it are denoted as {x A,1 ,…,x A,n }, n is the number of samples, and the VI value of the previous reference sliding window is recorded as VI A , the sample mean is recorded as The back reference window is denoted as B, and the observation samples in it are denoted as {x B,1 ,…,x B,n }, the VI value of the back reference sliding window is recorded as VI B , the sample mean is recorded as Therefore, the calculation method of VI value and MR value is as follows:

3. According to a method for adaptive constant false alarm detection of extended range targets based on improved VI according to claim 1, step S2, dividing the spatial distribution type of the background according to the VI value and the MR value, specifically comprises: S21, the calculated VI A Value, VI B The values ​​are respectively related to the decision threshold K VI Compare, if VI A <K VI , then the sampling values ​​of the front reference window are uniform, otherwise they are non-uniform; if VI B <K VI , then the sampling values ​​of the back reference window are uniform, otherwise they are non-uniform; S22, the calculated MR value and K MR For comparison, if Then the power means of the front and rear reference windows are the same, otherwise the means are different; S23, determine K according to the preset false alarm rate VI and K MR Generally, the probability α of misjudging a uniform environment as a non-uniform environment should be of the same order of magnitude as the preset false alarm rate, and the probability β of misjudging the same mean of the front and rear reference windows as different is required to be no greater than 0.

1. Therefore, K VI =4.76, K MR =1.806 respectively corresponds to α=4×10 -4 , β = 0.1; S24. According to the VI-CFAR decision criterion, the spatial distribution types of the reference window clutter background are divided into the following three cases: (1) If the front reference window is uniform, the rear reference window is uniform, and the power mean of the front and rear reference windows is the same, then the background is uniform; (2) If the front reference window is uniform, the rear reference window is uniform, and the power means of the front and rear reference windows are different, it is a clutter edge background; (3) If the front reference window or the rear reference window is uneven, it is judged that an interference target appears in the reference window.

Images