Radar target detection method and system based on accumulative permutation entropy weighting

Abstract

The invention discloses a radar target detection method and system based on accumulative permutation entropy weighting, mainly relates to the technical field of radar target detection, and is used for solving the problem that an existing radar target detection method has serious spectrum leakage when the number of pulses is small. And the problem that the scheme characteristic for evaluating the complexity of the time sequence cannot be directly used for distinguishing the pure background noise and the'target signal + background noise 'is solved. Comprising the following steps: calculating a normalized accumulated permutation entropy of an echo pulse sequence; performing modular square on the FFT result of the Doppler channel where the to-be-detected distance unit is located to obtain a modular square calculation result; obtaining a product result by using a modular square calculation result and normalized accumulation permutation entropy; traversing all the to-be-detected distance units, and obtaining a detection statistic corresponding to the current to-be-detected distance unit according to a mean value of a product result of the current to-be-detected distance unit and a product result of the corresponding reference distance unit.

Description

Technical Field

[0001] This application relates to the technical field of radar target detection, and in particular, to a radar target detection method and system based on cumulative permutation entropy weighting. Background Art

[0002] Target detection under low signal-to-noise ratio conditions is an important technical issue in modern radar research. Traditional radar target detection methods based on fast Fourier transform under noise background improve the signal-to-noise ratio through coherent integration. However, when the number of pulses is small, serious spectral leakage occurs, resulting in an unclear spectral peak of the target signal and a decline in detection performance.

[0003] In addition, the permutation entropy feature uses the relative size rather than the absolute size of adjacent points in the time series to evaluate the complexity of the time series, which is a simple and robust method. However, in the noise background, the presence or absence of the target signal does not change the random state of the amplitude and phase of the echo sequence. Therefore, this feature cannot be directly used to distinguish pure background noise from "target signal + background noise".

[0004] Therefore, there is an urgent need for a radar target detection method and system based on cumulative permutation entropy weighting to solve the problems that serious spectral leakage occurs when the number of pulses is small in the existing radar target detection method, resulting in an unclear spectral peak of the target signal, and in addition, the feature of the scheme for evaluating the complexity of the time series cannot be directly used to distinguish pure background noise from "target signal + background noise". Summary of the Invention

[0005] This application provides a radar target detection method and system based on cumulative permutation entropy weighting to solve the problems that serious spectral leakage occurs when the number of pulses is small in the existing radar target detection method, resulting in an unclear spectral peak of the target signal, and in addition, the feature of the scheme for evaluating the complexity of the time series cannot be directly used to distinguish pure background noise from "target signal + background noise".

[0006] In a first aspect, this application provides a radar target detection method based on cumulative permutation entropy weighting. The method includes: Dividing the target detection distance into several distance units, sequentially taking the distance units as the distance units to be detected, and at the same time determining the reference distance units corresponding to the distance units to be detected; Calculating the normalized cumulative permutation entropy of the echo pulse sequence according to the echo pulse sequence of the distance unit; Performing a fast Fourier transform on the echo pulse sequence to obtain an FFT result, and then taking the modulus square of the FFT result of the Doppler channel where the distance unit to be detected is located to obtain a modulus square calculation result; Using the modulus square calculation result and the normalized cumulative permutation entropy to obtain a product result; Traverse all distance cells to be detected, and obtain the detection statistic corresponding to the current distance cell to be detected according to the mean of the product result of the current distance cell to be detected and the product result of the corresponding reference distance cell; Use the detection statistic to determine the target detection threshold, and then perform radar target detection.

[0007] In an implementation manner of the present application, determining the reference distance cell corresponding to the distance cell to be detected specifically includes: When the distance cell to be detected is the first distance cell, determine the adjacent distance cell of the distance cell to be detected as the protection cell, and the R distance cells after the protection cell are the reference distance cells; where R is the total number of preset reference distance cells and R is an even number; When the distance cell to be detected is not the first distance cell and is in the first R / 2 + 1 positions, determine the two adjacent distance cells of the distance cell to be detected as the protection cells, and the distance cells that are not the distance cells to be detected and not the protection cells in the first R + 3 positions are the reference distance cells; When the distance cell to be detected is not in the first R / 2 + 1 positions and is not in the last R / 2 + 1 positions, determine the two adjacent distance cells of the distance cell to be detected as the protection cells, the distance cells in the first R / 2 positions before the first protection cell are the reference distance cells, and the distance cells in the last R / 2 positions after the last protection cell are the reference distance cells; When the distance cell to be detected is in the last R / 2 + 1 positions and is not the last distance cell, determine the two adjacent distance cells of the distance cell to be detected as the protection cells, and the distance cells that are not the distance cells to be detected and not the protection cells in the last R + 3 positions are the reference distance cells; When the distance cell to be detected is the last distance cell, the adjacent distance cell of the distance cell to be detected is the protection cell, and the first R distance cells of the protection cell are the reference distance cells.

[0008] In an implementation manner of the present application, calculating the normalized cumulative permutation entropy of the echo pulse sequence according to the echo pulse sequence of the distance cell specifically includes: For the echo pulse sequence of length Perform Doppler search to obtain the sequence , where ; ; Multiply the sequence by the order upper triangular matrix to achieve the accumulation of sequence elements and obtain the accumulated sequence , ; Calculate the permutation entropy of the sequence .

[0009] In an implementation of the present application, calculating the permutation entropy of a sequence specifically includes: Performing phase space reconstruction on to obtain , , wherein, is the embedding dimension, is the number of reconstructed vectors; Using each element in the sequence to label the sequence with the serial numbers of the corresponding position elements; Sorting the elements in the sequence in ascending order, and then sorting the sequence in the sorting order to obtain the index sequence ; wherein, the element ; If has equal elements, then arrange them in the original order; For each , obtaining the corresponding index sequence , and in the index sequence set , forming a new index sequence set with different index sequences, denoted as ; wherein, represents the number of different index sequences; Through the formula: , calculating the probability of each index sequence appearing; wherein, ; Through the formula: , calculating the cumulative permutation entropy ; When , and , , the cumulative permutation entropy reaches the maximum value ; Using the maximum value of the cumulative permutation entropy to perform normalization to obtain the normalized cumulative permutation entropy .

[0010] In an implementation of the present application, performing a fast Fourier transform on the echo pulse sequence to obtain the FFT result, and then taking the modulus square of the FFT result of the Doppler channel where the distance cell to be detected is located to obtain the modulus square calculation result, specifically including: Perform a fast Fourier transform on the echo pulse sequence to obtain the FFT result: ; where x represents the echo pulse sequence ; Take the modulus square of the result of the Doppler channel where the distance cell to be detected is located : to obtain the modulus square calculation result.

[0011] In an implementation manner of the present application, a product result is obtained by using the modulus square calculation result and the normalized cumulative permutation entropy, specifically including: Through the formula: , obtain the product result t; where, represents the normalized cumulative permutation entropy, represents the modulus square calculation result.

[0012] In an implementation manner of the present application, according to the mean of the product result of the current distance cell to be detected and the product result of the corresponding reference distance cell, the detection statistic corresponding to the current distance cell to be detected is obtained, specifically including: Through the formula: , calculate the detection statistic T; where, represents the product result of the current distance cell to be detected, represents the product result of the rth reference distance cell, and R represents the total number of reference distance cells.

[0013] In an implementation manner of the present application, the target detection threshold is determined by using the detection statistic, specifically including: Enter N Monte Carlo simulation experiments through MATLAB software to generate the distance cells to be detected and the reference distance cells under the condition of no target, and obtain N detection statistics, where, , is the false alarm probability set by the system, represents rounding down; sort the N detection statistics from largest to smallest, and take the 100th detection statistic as the target detection threshold.

[0014] In a second aspect, the present application provides a radar target detection system based on weighted cumulative permutation entropy. The system includes: A determination module, configured to divide a target detection distance into a plurality of distance units, sequentially use the distance units as distance units to be detected, and simultaneously determine reference distance units corresponding to the distance units to be detected; A calculation module, configured to calculate the normalized cumulative permutation entropy of the echo pulse sequence according to the echo pulse sequence of the distance unit; perform a fast Fourier transform on the echo pulse sequence to obtain an FFT result, and then take the modulus square of the FFT result of the Doppler channel where the distance unit to be detected is located to obtain a modulus square calculation result; use the modulus square calculation result and the normalized cumulative permutation entropy to obtain a product result; A statistic module, configured to traverse all distance units to be detected, and obtain a detection statistic corresponding to the current distance unit to be detected according to the mean value of the product result of the current distance unit to be detected and the product result of the corresponding reference distance unit; A detection module, configured to use the detection statistic to determine a target detection threshold, and then perform radar target detection.

[0015] In an implementation manner of the present application, the detection module includes a detection unit, configured to enter N Monte Carlo simulation experiments through MATLAB software, generate distance units to be detected and reference distance units under the condition of no target, and obtain N detection statistics, where, , is the false alarm probability set for the system, represents rounding down; sort the N detection statistics from largest to smallest, and take the 100th detection statistic as the target detection threshold.

[0016] From the above technical solutions, it can be seen that the present application has the following advantages: ‌‌1. Suppress spectrum leakage at low pulse numbers and enhance the target signal peak: By calculating the cumulative permutation entropy of the echo pulse sequence, the complexity of the time series is quantified. The presence of the target signal will change the periodic structure of the sequence, resulting in a decrease in the permutation entropy value, while the randomness of pure noise is higher and the entropy value is larger. Through normalization, the entropy value can be converted into a weight coefficient within the range of 0-1.

[0017] Traditional FFT has spectrum leakage due to insufficient sampling at low pulse numbers, and the target peak is submerged by noise. By multiplying the FFT modulus square result (reflecting the frequency domain energy) by the normalized cumulative permutation entropy (reflecting the complexity of the time domain structure), the frequency points where the target is located will be significantly amplified because they simultaneously satisfy high energy and low entropy value (low complexity), while the product of the noise frequency points is suppressed due to high entropy value. This weighting strategy effectively highlights the target spectrum peak and reduces the impact of spectrum leakage on detection.

[0018] By suppressing noise frequency points through entropy weighting, the energy of the target frequency points is selectively enhanced, making the peak easier to be recognized by the detection algorithm. Under the condition of few pulses, the energy diffusion caused by spectrum leakage is compensated, reducing the risk of missed detection.

[0019] 2. Distinguish between pure noise and "target + noise" scenarios: Traditional permutation entropy only depends on the relative magnitudes of adjacent points and cannot directly distinguish target signals from noise. In this application, by accumulating permutation entropy (time windows or segmented accumulation may be introduced), the sensitivity to the periodic characteristics of the sequence is enhanced. The periodic echoes of the target will result in entropy values significantly lower than the highly random sequences of pure noise.

[0020] Use the mean of the product results of reference cells (usually adjacent cells, assuming no target) as the background noise estimate. By comparing the statistics of the distance cell to be detected with those of the reference cells, the baseline interference of environmental noise can be eliminated, and target signals and noise can be further distinguished.

[0021] The accumulation operation captures the local structural changes caused by the target, making the entropy feature a key indicator for distinguishing signals from noise. Dynamically setting the detection threshold based on the statistics of the reference cells improves the generalization ability of the algorithm in complex noise environments.

[0022] 3. Improve detection performance by integrating time-frequency domain information: FFT provides the frequency-domain energy distribution, and accumulated permutation entropy reflects the time-domain structural characteristics. The product of the two integrates time-frequency dual-domain information and breaks through the limitations of single-domain analysis.

[0023] Construct a detection statistic through local comparison of the product results (the mean of the distance cell to be detected and the reference cells). This design suppresses the influence of non-stationary noise while retaining the joint characteristics of the target signal.

[0024] The combination of frequency-domain energy and time-domain complexity reduces the dependence on a single feature and improves detection reliability. Under low signal-to-noise ratio conditions, noise may exhibit local high energy or low complexity in the time domain or frequency domain, but the probability of both occurring simultaneously is low, and the joint criterion effectively reduces false alarms. Description of the Drawings

[0025] To more clearly illustrate the technical solutions of the present invention, the drawings required for description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0026] Figure 1 It is a flowchart of a radar target detection method based on weighted accumulated permutation entropy provided by an embodiment of this application.

[0027] Figure 2 It is a schematic diagram of constructing a detection statistic provided by an embodiment of this application.

[0028] Figure 3 is a detection threshold graph under different noise power levels provided by an embodiment of the present application.

[0029] Figure 4 is a spectrogram of an 8-pulse pulse sequence of the distance unit to be detected provided by an embodiment of the present application.

[0030] Figure 5 is a spectrogram of a 32-pulse pulse sequence of the distance unit to be detected provided by an embodiment of the present application.

[0031] Figure 6 is a spectrogram of a 128-pulse pulse sequence of the distance unit to be detected provided by an embodiment of the present application.

[0032] Figure 7 is a comparison graph of the detection performance curves of two methods with 128 pulses provided by an embodiment of the present application.

[0033] Figure 8 is a comparison graph of the detection performance curves of two methods with 32 pulses provided by an embodiment of the present application.

[0034] Figure 9 is a comparison graph of the detection performance curves of two methods with 8 pulses provided by an embodiment of the present application.

[0035] Figure 10 is a schematic diagram of the internal structure of a radar target detection system based on the weighted accumulation permutation entropy provided by an embodiment of the present application. Detailed implementation manners

[0036] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0037] Those skilled in the art should understand that the embodiments described below are only the preferred embodiments of the present disclosure, which does not mean that the present disclosure can only be implemented through these preferred embodiments. These preferred embodiments are only used to explain the technical principles of the present disclosure and are not used to limit the protection scope of the present disclosure. Based on the preferred embodiments provided by the present disclosure, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall still fall within the protection scope of the present disclosure.

[0038] It should also be noted that the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, commodity or device including a series of elements not only includes those elements but also includes other elements not expressly listed, or further includes elements inherent to such process, method, commodity or device. Without further limitation, an element defined by the statement "including one..." does not exclude the presence of another identical element in the process, method, commodity or device including the element.

[0039] The technical solutions proposed in the embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0040] The embodiment provides a radar target detection method based on the weighted accumulation permutation entropy, as Figure 1 shown. The method provided in the embodiments of the present application mainly includes the following steps: Step 110: Divide the target detection distance into several distance units, sequentially use the distance units as the distance units to be detected, and at the same time determine the reference distance units corresponding to the distance units to be detected.

[0041] In some embodiments, determining the reference distance unit corresponding to the distance unit to be detected may specifically be: When the distance unit to be detected is the first distance unit, determine the adjacent distance unit of the distance unit to be detected as the protection unit, and the R distance units after the protection unit are the reference distance units; where R is the total number of preset reference distance units and R is an even number; When the distance unit to be detected is not the first distance unit and is in the first R / 2 + 1 positions, determine the two adjacent distance units of the distance unit to be detected as the protection units, and the distance units of the non-distance units to be detected and non-protection units in the first R + 3 positions are the reference distance units; When the distance unit to be detected is not in the first R / 2 + 1 positions and is not in the last R / 2 + 1 positions, determine the two adjacent distance units of the distance unit to be detected as the protection units, the distance units in the first R / 2 positions before the first protection unit are the reference distance units, and the distance units in the last R / 2 positions after the last protection unit are the reference distance units; When the distance unit to be detected is in the last R / 2 + 1 positions and is not the last distance unit, determine the two adjacent distance units of the distance unit to be detected as the protection units, and the distance units of the non-distance units to be detected and non-protection units in the last R + 3 positions are the reference distance units; When the distance unit to be detected is the last distance unit, the adjacent distance unit of the distance unit to be detected is the protection unit, and the first R distance units of the protection unit are the reference distance units.

[0042] Those skilled in the art can understand that, regardless of the position of the distance unit to be detected, protection units are arranged at its adjacent positions (such as the adjacent unit after the first unit, the adjacent units on both sides of the middle unit, etc.).

[0043] The selection of the reference unit avoids the protection unit and the distance unit to be detected itself, ensuring that the reference sample only contains background noise or irrelevant interference. This step avoids contaminating the reference unit when the target signal energy spreads to adjacent units. For example, if the target occupies an adjacent unit, directly selecting the adjacent unit as the reference in the traditional method will result in a high estimated value of background noise, thereby raising the detection threshold and causing missed detections. This step ensures the statistical independence of the reference unit through the isolation of the protection unit. In a strong clutter or multi-target scenario, it reduces the influence of the non-uniform environment on the reference unit and improves the accuracy of the background noise power estimation.

[0044] When the distance unit to be detected is the first or the last distance unit, there are only valid distance units on one side. By adjusting the selection direction of the reference unit (such as selecting backward for the first unit and forward for the last unit) and expanding the number of reference units (R + 3), the sample deficiency in the boundary region is compensated.

[0045] For the non-boundary middle unit, a symmetric reference unit selection method for the front and back (the first R / 2 and the last R / 2) is adopted to ensure that the number of reference samples is sufficient and the distribution is balanced.

[0046] In the traditional method, the noise estimation deviation may occur at the boundary unit due to insufficient number of reference units (such as only being able to select on one side). This step makes the number of reference samples of the first and last units the same as that of the middle unit by dynamically adjusting the number and direction of the reference units, avoiding the fluctuation of the detection performance due to position differences.

[0047] By combining symmetric sampling (for middle units) and extended sampling (for boundary units), the adaptability of the algorithm to the non-stationary noise in the radar range dimension is enhanced.

[0048] It is preset that the total number R of reference units is an even number: By fixing the value of R (such as R = 8, 10, 16, etc.), the scale of the reference unit is limited to avoid the computational burden brought by unrestricted search.

[0049] Step 120: According to the echo pulse sequence of the distance unit, calculate the normalized cumulative permutation entropy of the echo pulse sequence; perform a fast Fourier transform on the echo pulse sequence to obtain the FFT result, and then take the modulus square of the FFT result of the Doppler channel where the distance unit to be detected is located to obtain the modulus square calculation result; use the modulus square calculation result and the normalized cumulative permutation entropy to obtain the product result.

[0050] Among them, according to the echo pulse sequence of the distance unit, calculating the normalized cumulative permutation entropy of the echo pulse sequence can specifically be: Perform a Doppler search on the echo pulse sequence with a length of to obtain a sequence , where ; ; Multiply the sequence by the -order upper triangular matrix to achieve the accumulation of sequence elements and obtain the accumulated sequence , ; Calculate the permutation entropy of the sequence .

[0051] Among them, calculating the permutation entropy of the sequence specifically includes: Perform phase space reconstruction on to obtain , , where is the embedding dimension, and is the number of reconstructed vectors; Use each element in the sequence to label the serial numbers of the corresponding position elements in the sequence ; Arrange the elements in the sequence in ascending order, and then sort the sequence in the order of arrangement to obtain the index sequence ; among them, the element ; If has equal elements, then arrange them in the original order; For each , obtain the corresponding index sequence . In the index sequence set , different index sequences form a new index sequence set, denoted as ; among them, represents the number of different index sequences; Through the formula: , calculate the probability that each index sequence appears; Among them, ; Through the formula: , calculate the cumulative permutation entropy ; When , and , When the cumulative permutation entropy reaches its maximum value ; Using the maximum value of the cumulative permutation entropy to perform normalization to obtain the normalized cumulative permutation entropy .

[0052] Among them, the echo pulse sequence is subjected to fast Fourier transform to obtain the FFT result, and then the modulus square of the FFT result of the Doppler channel where the distance cell to be detected is located is taken to obtain the modulus square calculation result. Specifically, it can be: The echo pulse sequence is subjected to fast Fourier transform to obtain the FFT result: where x represents the echo pulse sequence ; Taking the modulus square of the result of the Doppler channel where the distance cell to be detected is located: to obtain the modulus square calculation result.

[0053] Among them, using the modulus square calculation result and the normalized cumulative permutation entropy to obtain the product result. Specifically, it can be: Through the formula: to obtain the product result t; where represents the normalized cumulative permutation entropy, represents the modulus square calculation result.

[0054] Step 130, traverse all the distance cells to be detected, and obtain the detection statistic corresponding to the current distance cell to be detected according to the mean value of the product result of the current distance cell to be detected and the product result of the corresponding reference distance cell; use the detection statistic to determine the target detection threshold, and then perform radar target detection.

[0055] In some embodiments, according to the mean value of the product result of the current distance cell to be detected and the product result of the corresponding reference distance cell, the detection statistic corresponding to the current distance cell to be detected is obtained. Specifically, it can be: Through the formula: to calculate the detection statistic T; where represents the product result of the current distance cell to be detected, represents the product result of the r-th reference distance cell, and R represents the total number of reference distance cells.

[0056] Such as Figure 2As shown in the figure, taking the distance unit to be detected that is not in the first R / 2 + 1 positions and not in the last R / 2 + 1 positions as an example, the two adjacent distance units of the distance unit to be detected are protection units. The distance units in the first R / 2 positions of the front protection unit are reference distance units, and the distance units in the last R / 2 positions of the rear protection unit are reference distance units. For the reference distance units in the first R / 2 positions and the reference distance units in the last R / 2 positions, find the product results of the reference distance units. The sum of the reference distance units in the first R / 2 positions is X, and the sum of the reference distance units in the last R / 2 positions is Y. Add them and divide by R to obtain the mean value Z of the product results of the reference distance units. Obtain the detection statistic corresponding to the current distance unit to be detected through the ratio of the product result of the current distance unit to be detected and the mean value of the product results of the corresponding reference distance units.

[0057] Among them, using the detection statistic to determine the target detection threshold, specifically, it can be: Enter N Monte Carlo simulation experiments through MATLAB software to generate the distance units to be detected and reference distance units under the condition of no target, and obtain N detection statistics. Among them, , is the false alarm probability set by the system, represents rounding down; sort the N detection statistics from large to small, and take the 100th detection statistic as the target detection threshold.

[0058] In summary, as a specific embodiment: The determined parameters are , , . The noise background is complex Gaussian noise, the noise standard deviation is 1, and the normalized Doppler of the target is 0.2. When the number of pulses is 8, the signal-to-noise ratio is set to -10~10dB; when the number of pulses is 32, the signal-to-noise ratio is set to -10~5dB; when the number of pulses is 128, the signal-to-noise ratio is set to -15~0dB; Through MATLAB software, conduct 10 6 Monte Carlo simulation experiments, generate the data of the distance units to be detected and the reference distance units under the condition of no target, and calculate 10 6 detection statistics according to this. Sort these detection statistics from large to small, and take the 100th value as the detection threshold. Repeat the above simulation experiments at different background noise power levels to obtain the change of the detection threshold, as shown in Figure 3 . This figure shows that as the noise power level changes, the detection threshold of the CPE-FFT method fluctuates very little and is approximately unchanged. Therefore, it shows that the CPE-FFT method has the characteristic of constant false alarm.

[0059] Through Monte Carlo simulation experiments, data of the distance cell to be detected containing the target plus noise and data of the reference cell of pure noise are generated respectively, and the number of pulses is set to 8, 32, and 128 respectively. When the signal-to-noise ratio (SNR) is 0 dB, the spectrum of the pulse sequence in the distance cell to be detected is as shown in Figure 4 (spectrum of the pulse sequence of the distance cell to be detected with 8 pulses), Figure 5 (spectrum of the pulse sequence of the distance cell to be detected with 32 pulses), Figure 6 (spectrum of the pulse sequence of the distance cell to be detected with 128 pulses), which shows that when the number of pulses is small, the spectrum leakage is serious and the signal-to-noise ratio of the Doppler cell where the target is located decreases; when the number of pulses is large, there is also spectrum leakage, but the coherent pulse accumulation is large, and the signal-to-noise ratio of the Doppler cell where the target is located can still remain large. Figures 4 - 6

[0060] Use Monte Carlo simulation experiments to compare the detection performances of the FFT method and the CPE-FFT method. When the number of pulses is 128, the simulation results are as shown in Figure 7 . Compared with the FFT method, the detection probability of the CPE-FFT method is slightly higher, but the advantage is not obvious. When the detection probability is 0.8, the CPE-FFT method requires about 0.5 dB less SNR than the FFT method. When the number of pulses is 32, the simulation results are as shown in Figure 8 . The detection performance of the CPE-FFT method is better than that of the FFT method. When the detection probability is 0.8, the SNR requirement is reduced by about 1.5 dB. When the number of pulses is 8, the simulation results are as shown in Figure 9 . The CPE-FFT method has an obvious advantage over the FFT method. When the detection probability is 0.8, the SNR requirement is reduced by about 3 dB.

[0061] In addition, this application Figure 10 provides a radar target detection system based on weighted cumulative permutation entropy according to an embodiment of this application. As shown in Figure 10 , the system provided by the embodiment of this application mainly includes: A determination module 210, configured to divide the target detection distance into several distance cells, sequentially use the distance cells as the distance cells to be detected, and at the same time determine the reference distance cells corresponding to the distance cells to be detected.

[0062] A calculation module 220, configured to calculate the normalized cumulative permutation entropy of the echo pulse sequence according to the echo pulse sequence of the distance cell; perform a fast Fourier transform on the echo pulse sequence to obtain the FFT result, and then take the modulus square of the FFT result of the Doppler channel where the distance cell to be detected is located to obtain the modulus square calculation result; use the modulus square calculation result and the normalized cumulative permutation entropy to obtain the product result.

[0063] The statistic module 230 is configured to traverse all distance units to be detected, and obtain the detection statistic corresponding to the current distance unit to be detected according to the mean value of the product result of the current distance unit to be detected and the product result of the corresponding reference distance unit.

[0064] The detection module 240 is configured to determine a target detection threshold by using the detection statistic, and then perform radar target detection.

[0065] The detection module 240 includes a detection unit. The detection unit is configured to enter N Monte Carlo simulation experiments through MATLAB software, generate the distance units to be detected and the reference distance units under the condition of no target, and obtain N detection statistics, where , is the false alarm probability set for the system, represents rounding down; sort the N detection statistics from largest to smallest, and take the 100th detection statistic as the target detection threshold.

[0066] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but will be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A radar target detection method based on weighted cumulative permutation entropy, characterized in that The method includes: Dividing the target detection distance into several distance units, sequentially taking the distance units as the to-be-detected distance units, and simultaneously determining the reference distance units corresponding to the to-be-detected distance units; Calculating the normalized cumulative permutation entropy of the echo pulse sequence according to the echo pulse sequence of the distance unit; performing a fast Fourier transform on the echo pulse sequence to obtain the FFT result, and then taking the modulus square of the FFT result of the Doppler channel where the to-be-detected distance unit is located to obtain the modulus square calculation result; using the modulus square calculation result and the normalized cumulative permutation entropy to obtain the product result; Traversing all the to-be-detected distance units, obtaining the detection statistic corresponding to the current to-be-detected distance unit according to the mean value of the product result of the current to-be-detected distance unit and the product result of the corresponding reference distance unit; using the detection statistic to determine the target detection threshold, and then performing radar target detection.

2. The radar target detection method based on the weighted accumulated permutation entropy according to claim 1, wherein, Determining the reference distance unit corresponding to the to-be-detected distance unit specifically includes: When the to-be-detected distance unit is the first distance unit, determining the adjacent distance unit of the to-be-detected distance unit as the protection unit, and the R distance units after the protection unit as the reference distance units; where R is the total number of preset reference distance units and R is an even number; When the to-be-detected distance unit is not the first distance unit and is in the first R / 2 + 1 positions, determining the two adjacent distance units of the to-be-detected distance unit as the protection units, and the distance units that are not the to-be-detected distance units and not the protection units in the first R + 3 positions as the reference distance units; When the to-be-detected distance unit is not in the first R / 2 + 1 positions and not in the last R / 2 + 1 positions, determining the two adjacent distance units of the to-be-detected distance unit as the protection units, the distance units in the first R / 2 positions before the front protection unit as the reference distance units, and the distance units in the last R / 2 positions after the back protection unit as the reference distance units; When the to-be-detected distance unit is in the last R / 2 + 1 positions and is not the last distance unit, determining the two adjacent distance units of the to-be-detected distance unit as the protection units, and the distance units that are not the to-be-detected distance units and not the protection units in the last R + 3 positions as the reference distance units; When the to-be-detected distance unit is the last distance unit, the adjacent distance unit of the to-be-detected distance unit is the protection unit, and the first R distance units of the protection unit are the reference distance units.

3. The radar target detection method based on the weighted cumulative permutation entropy according to claim 1, wherein Calculating the normalized cumulative permutation entropy of the echo pulse sequence according to the echo pulse sequence of the distance unit specifically includes: Perform a Doppler search on an echo pulse sequence of length to obtain a sequence , where ; ​ Multiply the sequence by the upper triangular matrix of order to achieve the accumulation of the sequence elements and obtain the accumulated sequence , ; Calculate the permutation entropy of the sequence.

4. The radar target detection method based on weighted cumulative permutation entropy according to claim 3, characterized in that Calculation sequence The permutation entropy of which specifically includes: Pair Perform phase space reconstruction to obtain , ， Among them, is the embedding dimension, is the number of reconstructed vectors; Label the sequence numbers of the elements at the corresponding positions in the sequence with each element in ; Sort the elements in the sequence in ascending order, and then sort the sequence in the sorted order to obtain the index sequence ; where the element ; If there are equal elements, they shall be arranged in the original order of precedence; For each , obtain the corresponding index sequence . From the index sequence set , form a new index sequence set with different index sequences, denoted as ; among them, represents the number of different index sequences in Through the formula: , calculate the probability of occurrence of each index sequence ; Among them, ; Through the formula: , calculate the cumulative permutation entropy ; When and , the cumulative permutation entropy reaches the maximum value ; Using the maximum value of the cumulative permutation entropy Normalize to obtain the normalized cumulative permutation entropy .

5. The radar target detection method based on weighted cumulative permutation entropy according to claim 1, wherein Performing a fast Fourier transform on the echo pulse sequence to obtain the FFT result, and then taking the modulus square of the FFT result of the Doppler channel where the to-be-detected distance unit is located to obtain the modulus square calculation result, specifically includes: Performing a fast Fourier transform on the echo pulse sequence to obtain the FFT result: ; where x represents the echo pulse sequence ; Take the modulo square of the result of the Doppler channel where the distance unit to be detected is located : , obtain the modular square calculation result.

6. The radar target detection method based on weighted cumulative permutation entropy according to claim 1, wherein Using the modulus square calculation result and the normalized cumulative permutation entropy to obtain the product result, specifically includes: Through the formula: , obtain the product result t; Among them, represents the normalized cumulative permutation entropy, represents the result of the modulus square calculation.

7. The radar target detection method based on weighted accumulation permutation entropy according to claim 1, wherein Obtaining the detection statistic corresponding to the current to-be-detected distance unit according to the mean value of the product result of the current to-be-detected distance unit and the product result of the corresponding reference distance unit, specifically includes: Through the formula: , calculate the detection statistic T; Among them, represents the product result of the currently to-be-detected distance unit, represents the product result of the r-th reference distance unit, and R represents the total number of reference distance units.

8. The radar target detection method based on weighted cumulative permutation entropy according to claim 1, wherein Using the detection statistic to determine the target detection threshold, specifically includes: Enter the N - time Monte Carlo simulation experiment through MATLAB software to generate the distance cells to be detected and reference distance cells under the condition without a target, and obtain N detection statistics. Among them, , is the false - alarm probability set by the system, represents rounding down; sort the N detection statistics from large to small, and take the 100th detection statistic as the target detection threshold.

9. A radar target detection system based on weighted cumulative permutation entropy, characterized in that, The system includes: A determination module, configured to divide a target detection distance into a plurality of distance units, sequentially use the distance units as distance units to be detected, and simultaneously determine reference distance units corresponding to the distance units to be detected; A calculation module, configured to calculate the normalized cumulative permutation entropy of an echo pulse sequence according to the echo pulse sequence of a distance unit; perform a fast Fourier transform on the echo pulse sequence to obtain an FFT result, and then take the modulus square of the FFT result of the Doppler channel where the distance unit to be detected is located to obtain a modulus square calculation result; use the modulus square calculation result and the normalized cumulative permutation entropy to obtain a product result; A statistic module, configured to traverse all distance units to be detected, and obtain a detection statistic corresponding to the current distance unit to be detected according to the mean value of the product result of the current distance unit to be detected and the product result of the corresponding reference distance unit; A detection module, configured to use the detection statistic to determine a target detection threshold, and then perform radar target detection.

10. The radar target detection system based on the weighted cumulative permutation entropy according to claim 9, characterized in that, The detection module includes a detection unit Used to enter the N - th Monte Carlo simulation experiment through MATLAB software, generate the distance cells to be detected and reference distance cells under the condition of no target, and obtain N detection statistics. Among them, , is the false alarm probability set by the system, represents rounding down; sort the N detection statistics from large to small, and take the 100 - th detection statistic as the target detection threshold.

Images