Improved variable index constant false alarm rate detection algorithm

Abstract

The invention belongs to the field of radar signal processing, and particularly relates to an improved variable index constant false alarm rate detection algorithm. According to the property of the reference window before and after constant false alarm detection, the corresponding constant false alarm processing method is adaptively selected, and the detection performance of the detector in a multi-target background environment can be effectively improved. Aiming at the situation that a traditional variable index constant false alarm detector (VI-CFAR) has targets in front and back reference windows and the detection performance is poor when a detector selection unit averagely selects a small constant false alarm detector (SO-CFAR), the detector is improved by adopting an adaptive deletion selection small constant false alarm detector (ACSO-CFAR), and when a target exists in a single-side reference window, the detection performance of the detector selection unit is improved by adopting an adaptive deletion selection small constant false alarm detector (ACSO-CFAR). An improved unit average maximum constant false alarm rate detector (CA-CFAR) is selected, and compared with a traditional CA-CFAR, the detection performance is improved under the multi-target background.

Description

Technical Field

[0001] The present invention belongs to the field of radar signal processing, and specifically relates to an improved variable exponent constant false alarm detection algorithm. Background Art

[0002] In practical radar applications, a complex noise background is often faced, and echo signals are often accompanied by clutter and interference signals. Therefore, in signal processing, it is necessary to set threshold values adapted to different environments. A constant false alarm (CFAR) detector estimates the background clutter power based on the reference cells on both sides of the detection cell and adaptively adjusts the threshold value to achieve a constant false alarm rate. Common CFAR detectors include cell average constant false alarm detector (CA-CFAR), greatest of cell averages CFAR detector (GO-CFAR), smallest of cell averages CFAR detector (SO-CFAR), and ordered statistic CFAR detector (OS-CFAR). However, these detectors perform well in specific clutter background environments, but their detection performance will significantly decline in complex environments. To solve this problem, foreign scholars proposed an adaptive constant false alarm algorithm based on variable exponents (VI-CFAR) in 2000. This algorithm judges the clutter background of the left and right reference cells by evaluating the second-order parameters of the reference cells, and selects appropriate CA, GO, and SO detectors according to the characteristics of the clutter background. This method can more accurately estimate the clutter background in complex environments, thereby improving the detection probability of the system.

[0003] The variable exponent constant false alarm detector adaptively selects a specific CFAR detector according to the clutter characteristics, and has certain advantages. However, when there are targets in both the left and right reference cells, since the VI-CFAR selects the SO-CFAR detector, it will affect the detection performance. To solve this problem, the combined mean level detector (CMLD-CFAR) is combined with VI-CFAR to propose an adaptive combined mean level smallest of cell averages detector (ACSO-CFAR). At the same time, CA-CFAR is improved, and finally an improved variable exponent constant false alarm detector is obtained, which can improve the detection probability in a multi-target background and has a more flexible background estimation ability. Summary of the Invention

[0004] The purpose of the present invention is to provide an improved variable exponent constant false alarm detection algorithm, which selects a corresponding constant false alarm detection threshold by judging the clutter properties of the left and right reference cells, and obtains a constant false alarm detection judgment result.

[0005] The technical solution for achieving the object of the present invention is: A method for implementing an improved variable exponent constant false alarm detector based on FPGA, including the following steps:

[0006] Step 1: Establish a detection cell, a protection cell, and a reference cell;

[0007] Step 2: Segment the reference unit to calculate the mean value, compare it with the set threshold and perform replacement;

[0008] Step 3: Sort the samples in the reference unit and adopt the cyclic deletion algorithm to obtain the clutter power probability level;

[0009] Step 4: Calculate the clutter state according to the values in the left and right reference units;

[0010] Step 5: Select the corresponding logic according to the clutter background estimation situation of the reference unit;

[0011] Step 6: Obtain the decision threshold according to the selected logic, compare it with the value of the detection unit, and output the constant false alarm detection result.

[0012] Compared with the prior art, the significant advantages of the present invention are as follows: 1) Compared with the traditional single-type detector, it can adapt to complex clutter environments; 2) The ACSO-CFAR detector is used to replace the traditional SO-CFAR, improving the problem that the detection performance of the variable exponent constant false alarm detector decays significantly when there are multi-target interferences in both the left and right reference units; 3) The CA-CFAR is improved, enhancing the detection probability of this algorithm in the face of multi-targets. Description of the Drawings

[0013] Figure 1 It is the schematic diagram of the improved VI-CFAR algorithm of the present invention.

[0014] Figure 2 It is the schematic diagram of the reference unit used in the present invention. Detailed Embodiment

[0015] The following further describes the present invention in detail with reference to the drawings.

[0016] An improved variable exponent constant false alarm detection algorithm of the present invention, as Figure 1 shown, the method includes:

[0017] Step 1: Delay and beat the input sequence to establish the left and right reference units:

[0018] As Figure 2 shown, delay the input sequence, establish a dynamic sliding window, the center of the sliding window is the detection unit, protection units are set on both sides, and reference units are on both sides of the protection units. The left and right reference units are respectively denoted as x i , y i , set the number of detection units to 1, the number of protection units to 2, and the number of left and right reference units to 16.

[0019] Step 2 sums up the left and right reference units and selects the larger one, then multiplies them by the normalization factor respectively to obtain the thresholds T CA and T GO , specifically as follows:

[0020] Step 2.1: Calculate the sum of the clutter powers of the left and right reference units and the total clutter power of the reference units respectively. The calculation formula is as follows:

[0021]

[0022] Step 2.2: Select the larger one of the clutter powers of the left and right reference units;

[0023] Step 2.3: Multiply them by the normalization factor respectively to obtain the thresholds. The calculation formula is as follows:

[0024] T CA = C N * Z

[0025] T GO = C N / 2 * max(∑Z L , ∑Z R )

[0026] where C N = (P fa ) -1 / N - 1, P fa is the false alarm probability and N is the number of reference units.

[0027] Step 3: Calculate the mean value of each segment of the reference units, compare it with the overall mean value, and replace it after comparing with the set threshold:

[0028] Step 3.1: Divide the left and right reference units into m equal parts respectively, and calculate the mean value of each sub-unit. The calculation formula is as follows;

[0029]

[0030]

[0031] where m = 4

[0032] Step 3.2: Compare the mean value of the sub-reference unit with the threshold K. If it exceeds the threshold, replace it with the mean clutter power Z / N, and reassign the sub-reference unit. The specific judgment is as follows:

[0033]

[0034]

[0035] The threshold K in the formula is: K = αT0Z, T0 = (Pfa ) -1 / 2n -1, select α as 0.5;

[0036] Step 3.3: Accumulate and sum the re-assigned sub-reference units to obtain the final clutter power estimate as:

[0037]

[0038] Step 3.4: Multiply by the normalization factor to obtain the threshold T CA_better = C N / 2 * Z all ;

[0039] Step 4: Arrange the samples in the reference unit, and then perform the cyclic deletion algorithm to obtain the clutter power probability level:

[0040] Step 4.1: Arrange the samples in the left and right reference units in ascending order. The specific formula is as follows:

[0041] q (1) ≤ q (2) ≤ … q (N)

[0042] Step 4.2: According to the set false deletion probability, the normalization factors in the cyclic deletion can be obtained. The specific formula is as follows:

[0043]

[0044] In the formula, select N = 8.

[0045] Step 4.3: Execute the cyclic deletion algorithm, and the specific performance of this algorithm is:

[0046]

[0047] The above cyclic deletion is a binary hypothesis testing problem. T in the formula k is the normalization factor at the k-th step, and Z k is the sum of k relatively low-ordered samples. When the sampled sample value q(k + 1) of the reference unit is greater than the threshold T k Z k , it is the H1 hypothesis; when the sampled value q(k + 1) is less than the threshold T k Z k , it is the H0 hypothesis. When the H1 hypothesis holds, it means that the sampled sample value and the samples q (k+1) , … q (N) contain interfering targets. At this time, delete q (k+1) , … q (N) , and stop the algorithm. The final clutter power level estimate is:

[0048]

[0049] Step 4.4: Multiply the clutter power level by the normalization factor T k , compare the left and right reference cells, and select the smaller one.

[0050] Step 5: Calculate the clutter state based on the values of the reference cells within the left and right reference windows, specifically as follows:

[0051] Step 5.1: Calculate the second-order statistic that describes the clutter sample state within the reference cells;

[0052] Parameter V VIA , V VIB are indicators reflecting whether the clutter environment within the left and right reference cells is uniform, and are second-order parameters describing the clutter sample state. Their specific calculation formulas are as follows:

[0053]

[0054] Step 5.2: Calculate the mean ratio of the left and right reference cells;

[0055] Parameter V MR is the mean ratio of the left and right reference cells, reflecting whether the means of the left and right reference cells are the same. The specific calculation formula is

[0056]

[0057] Step 6: Select the corresponding logic based on the clutter background estimation of the reference window, specifically as follows:

[0058] Step 6.1: Clutter level estimation:

[0059] By comparing V VI with the threshold K VI to determine whether the clutter within the reference cells is uniform, where K VI = 4.76. The specific decision criterion is as follows,

[0060]

[0061]

[0062] Step 6.2: Determine whether the clutter means of the left and right reference cells are the same:

[0063] By comparing V MR with the threshold K MR to determine whether the clutter means within the left and right reference cells are the same, where K MR = 1.806. The specific decision criterion is as follows,

[0064]

[0065]

[0066] Step 6.3: Selection of VI-CFAR Adaptive Detector

[0067] According to the required false alarm rate P fa , obtain the threshold of the corresponding CFAR detector; according to the estimated clutter background of the reference window, select the CFAR threshold of the corresponding detection unit by referring to the following table

[0068] Table 1 Processing Methods of Detectors Corresponding to Improved VI-CFAR

[0069]

[0070] Step 7: Compare the detection threshold with the values of the detection unit and output the constant false alarm detection result, specifically:

[0071]

[0072] where S is the detection threshold of the above-mentioned improved variable exponent detector, and the decision H0 indicates that the target does not exist, and H1 indicates that the target exists.

Claims

1. An improved variable exponent constant false alarm detection algorithm, characterized in that, Including the steps: Step 1: Establish a detection unit, a protection unit, and left and right reference units; Step 2: Accumulate and sum the left and right reference units, select the larger one, and multiply it by the normalization factor respectively to obtain the thresholds S CA and S GO ; Step 3: Segment the reference unit to calculate the average value, compare it with the set threshold, replace it if necessary, and multiply it by the normalization factor to obtain the threshold S CA-better ; Step 4: Sort the samples in the reference unit, and then perform the loop deletion algorithm. Select the smaller value from the left and right reference units to obtain the threshold S ACSO ; Step 5: Calculate the clutter state based on the specific values within the left and right reference windows; Step 6: Select the corresponding logic according to the clutter background estimation of the reference units; Step 7: Based on the corresponding logic in Step 6, select the corresponding decision threshold in Steps 2 - 4, compare it with the value of the detection unit, and output the constant false alarm detection result.

2. The improved variable exponent constant false alarm detection algorithm according to claim 1, wherein, In Step 1, establish a detection unit, a protection unit, and left and right reference units, specifically: Delay the input sequence, establish a dynamic sliding window, with the detection unit at the center of the sliding window, protection units are set on both the left and right sides, and reference units are on both sides of the protection units. The left and right reference units are denoted as x i ,y i .

3. The improved variable exponent constant false alarm detection algorithm according to claim 1, wherein, Step 2 accumulates and sums the left and right reference units, selects the larger one, and multiplies them by the normalization factor respectively to obtain the thresholds T ca and T GO , specifically: Step 2.1: Calculate the sum of the clutter powers of the left and right reference units and the total clutter power of the reference units respectively. The calculation formula is as follows: Step 2.2: Select the larger one of the clutter powers of the left and right reference units; Step 2.3: Multiply them by the normalization factor respectively to obtain the threshold. The calculation formula is as follows: S CA = C N * Z S GO = C N / 2 * max(∑Z L , ∑Z R ) where C N = (P fa ) -1 / N - 1, and N is the number of reference units.

4. The improved variable exponent constant false alarm detection algorithm according to claim 1, wherein, In Step 3, segment the reference units to find the mean value, compare it with the set threshold, and then replace it. Specifically: Step 3.1: Perform m - equal - division operations on the left and right reference units respectively, calculate the mean value of each sub - unit. The calculation formula is as follows; Step 3.2: Compare the mean value of the sub - reference unit with the threshold K. If it exceeds the threshold, replace it with the mean clutter power Z / N, and re - assign the sub - reference unit. The specific judgment is as follows: The threshold K in the formula is expressed as: K = αT0Z, T0 = (P fa ) -1 / 2n -1; Step 3.3: Accumulate and sum the re - assigned sub - reference units to obtain the final clutter power estimate as: Step 3.4: Multiply by the normalization factor to obtain the threshold S ca_better = C N / 2 * Z all。 5. The improved variable exponent constant false alarm detection algorithm according to claim 1, characterized in that In Step 4, arrange the samples within the reference units, and then perform the cyclic deletion algorithm to obtain the clutter power probability level. Specifically: Step 4.1: Arrange the sampling samples in the left and right reference units in ascending order. The specific formula is as follows: q (1) ≤q (2) ≤…q (N) Step 4.2: According to the set false deletion probability, obtain the normalization factors in the cyclic deletion. The specific formula is as follows: Step 4.3: Execute the cyclic deletion algorithm, and the specific manifestation of this algorithm is: The above cyclic deletion is a binary hypothesis testing problem. T in the formula k is the normalization factor at the k-th step, and Z k is the sum of k relatively low-ordered samples. When the sampled value q(k + 1) of the reference unit is greater than the threshold T k Z k , it is the H1 hypothesis; when the sampled value q(k + 1) is less than the threshold T k Z k , it is the H0 hypothesis. When the H1 hypothesis holds, it means that the sampled value and the samples q (k+1) ,…q (N) greater than itself contain interfering targets. At this time, delete q (k+1) ,…q (N) , and stop the algorithm. The final clutter power level estimate is: Step 4.4: Multiply the clutter power level by the normalization factor T k , compare the left and right reference cells, and select the smaller one among them.

6. The improved variable exponent constant false alarm detection algorithm according to claim 1, wherein, In Step 5, calculate the clutter state based on the reference unit values within the left and right reference windows. Specifically: Step 5.1: Calculate the second - order statistic describing the clutter sample state within the reference unit; Parameter V VLA , V VIB is an index reflecting whether the clutter environment in the left and right reference units is uniform. It is a second-order parameter describing the state of clutter samples, and its specific calculation formula is as follows: Step 5.2: Calculate the mean ratio of the left and right reference units; Parameter V MR is the mean ratio of the left and right reference units, reflecting whether the means of the left and right reference units are the same. The specific calculation formula is 7. The improved variable exponent constant false alarm detection algorithm according to claim 1, characterized in that, In Step 6, select the corresponding logic according to the clutter background estimation of the reference units. Specifically: Step 6.1: Clutter level estimation: By comparing V VI with the threshold K VI it is determined whether the clutter in the reference unit is uniform Step 6.2: Clutter judgment of the left and right reference units: By comparing V MR with threshold K MR to determine whether the clutter means in the reference unit are the same: or Step 6.3: VI - CFAR adaptive detector selection According to the required false alarm rate P fa = 10 -6 , the threshold of the corresponding CFAR detector is obtained; according to the estimation of the clutter background in the reference cell, the CFAR threshold of the corresponding detection cell is selected by referring to the following table Table 1 Detector processing methods corresponding to the improved VI - CFAR 8. The improved variable exponent constant false alarm detection algorithm according to claim 1, characterized in that, In Step 7, based on the corresponding logic in Step 6, select the corresponding decision threshold in Steps 2 - 4, compare it with the value of the detection unit, and output the constant false alarm detection result. Specifically: Where S is the decision threshold of the above - mentioned improved adaptive detector, and the decision H0 indicates that the target does not exist, and H1 indicates that the target exists.

Images