Technical method for detecting hovering helicopter facing rotor wing composite material

Abstract

According to the hovering helicopter detection method for the rotor wing composite material, the blades and the hub of the helicopter rotor wing can be detected in a mode of combining low-threshold CFAR detection and high-threshold CFAR detection, and energy of the micro-motion Doppler characteristic of the rotor wing can be accumulated by utilizing energy-weighted plot condensation; the method combines a grid density estimation method to perform preliminary screening on rotor targets, reduces non-rotor target false alarms, and finally estimates rotor Doppler modulation features through dual RANSAC fitting to complete hovering helicopter detection. Compared with a traditional hovering helicopter detection method, the technology can effectively extract weak Doppler features generated by a modern helicopter rotor made of a composite material on a radar frequency spectrum, and especially has a better effect on hovering helicopter detection in a strong clutter environment under an air-to-air downward-looking condition.

Description

Technical Field

[0001] This invention relates to the field of radar detection technology, and in particular to a method for detecting hovering helicopters using rotor composite materials. Background Technology

[0002] Detecting aerial rotorcraft helicopters using radar is an important research direction in target detection and identification technology, and it has wide application value in both military and civilian fields.

[0003] The metal fuselage and rotor rotation of a helicopter have strong reflective properties, generating a strong radar echo signal. The high-speed rotation of the helicopter rotor produces a significant Doppler frequency shift in the radar echo signal spectrum. On both sides of the strong echo peak generated by the helicopter fuselage, multiple discrete peaks spanning the Doppler domain appear, which are the micro-doppler characteristics generated by the helicopter rotor.

[0004] However, to effectively reduce flight energy consumption and improve flight performance, modern helicopters use lightweight and highly durable composite materials such as carbon fiber reinforced plastics in their rotors. These materials have smaller RCS and weaker micro-motion characteristics. The rotors of the American Apache helicopter mainly use titanium alloys and carbon fiber composites, while the rotors of the Russian Ka-52 helicopter mainly use titanium alloys and glass fiber composites.

[0005] Currently, traditional hovering helicopter detection methods primarily utilize the periodic modulation characteristics of helicopter rotor rotation in the time or frequency domain, extracting the micro-Doppler features of the rotor in the radar signal for detection. However, this method requires a high signal-to-noise ratio from the helicopter rotor in the radar echo. If the helicopter rotor is made of composite materials, traditional methods cannot fully extract the rotor's micro-Doppler features, leading to helicopter detection failure. Moreover, airborne radar detection of helicopters faces the challenge of numerous false alarms due to ground clutter under look-down conditions. Traditional detection methods are not suitable for airborne radar detection of composite rotor helicopters under strong ground clutter conditions. Summary of the Invention

[0006] To address the shortcomings of traditional hovering helicopter detection methods, which struggle to extract weak Doppler features due to the low spectral modulation signal-to-noise ratio of rotor composite materials, a new hovering helicopter detection method for rotor composite materials is proposed. This method utilizes the different micro-motion Doppler features exhibited by the rotor hub and blades in the frequency spectrum, and employs methods such as dual-threshold CFAR detection, point clustering, mesh density estimation, and RANSAC fitting to detect hovering helicopters.

[0007] This invention is achieved through the following scheme:

[0008] A method for detecting hovering helicopters using rotor composite materials includes the following steps:

[0009] Step 1: Data Preprocessing and Dual Threshold Detection

[0010] a) Perform pulse compression and coherent accumulation on the raw echo data, and convert the time-domain signal into a frequency-domain signal using FFT;

[0011] b) The frequency domain signal is fed into a low-threshold SO-CFAR detector for constant false alarm rate (CFAR) detection to obtain the low-threshold detection result.

[0012] Det_cfar1;

[0013] c) The low-threshold detection result is sent to the high-threshold SO-CFAR detector for constant false alarm detection, and the detection result Det_cfar2 is obtained.

[0014] Step 2: Energy-weighted point aggregation

[0015] a) Extract energy information from dual-threshold CFAR detection points;

[0016] b) Weighted average of adjacent points, with points of higher energy having greater weight;

[0017] c) Calculate the energy-weighted target location;

[0018] Step 3: Grid-based point density estimation and rotor target selection

[0019] a) After condensing the energy-weighted point traces, divide the result into equal-area grids based on distance gates and frequency gates;

[0020] b) Count the number of dots in each grid and calculate the dot density;

[0021] c) Determine if the spot density is greater than the threshold T_density. If the spot density is greater than the threshold, retain it as a rotor target; otherwise, discard the target.

[0022] Step 4: RANSAC algorithm fitting and hovering helicopter modulation spectrum detection

[0023] a) Use the RANSAC algorithm to fit the low-threshold detection result Det_cfar1, and iteratively solve the model fitting parameters to obtain the fitting result Fit_line1;

[0024] b) Use the RANSAC algorithm to fit the high-threshold detection result Det_cfar2, iteratively solve the model fitting parameters, and obtain the fitting result Fit_line2;

[0025] c) For the targets preserved by the dot density estimation, calculate the line length Dop_line1 and the mean amplitude Amp_line1 of Fit_line1, and calculate the line length Dop_line2 and the mean amplitude Amp_line2 of Fit_line2.

[0026]

[0027] Where R x_max and R x_min The maximum and minimum x-coordinates of the interior point R are given by R. y_max and R y_min These are the maximum and minimum y-coordinates of the interior point R;

[0028] d) Calculate the amplitude difference Amp_diff between Amp_line1 and Amp_line2. If Amp_diff is greater than T_amp and Dop_line1 is greater than the threshold, then a hovering helicopter is detected.

[0029] Preferably, the step of performing pulse compression and coherent accumulation on the original echo data, and converting the time-domain signal to a frequency-domain signal using FFT, specifically involves: performing pulse compression on the original echo data by using an encoded signal and a matched filter to improve resolution; then performing coherent accumulation to coherently superimpose multiple pulse echo signals; and finally converting the time-domain signal to a frequency-domain signal using FFT processing.

[0030] Y = FFT{Y pc (t)}

[0031] Where y k (t) is the post-pulse compression signal; Y acc (t) represents the result of coherent accumulation; FFT represents the fast Fourier transform.

[0032] Preferably, the step of sending the frequency domain signal into a low-threshold SO-CFAR detector for constant false alarm rate (CFAR) detection to obtain the low-threshold detection result Det_cfar1 specifically involves:

[0033] b1) SO-CFAR detection parameter settings

[0034] b11) Guard window length setting: When estimating target clutter background, the target itself does not participate in the detection threshold estimation, so the guard window length N on one side is set. g =1 to 4;

[0035] b12) Background window length setting: The samples selected within the background window should reflect the clutter changes in the local area of ​​the target. The length N of the single-sided guard window should be set. b =4~32;

[0036] b13) False alarm probability setting: Setting the false alarm probability P for low threshold detection fa1 =1e -2 ;

[0037] b2) The frequency domain signal is sent to a low-threshold SO-CFAR detector for constant false alarm rate detection to obtain the low-threshold detection value Det_cfar1.

[0038] Preferably, the step of sending the low-threshold detection result to the high-threshold SO-CFAR detector for constant false alarm rate (CFAR) detection, and obtaining the detection result Det_cfar2, specifically involves:

[0039] c1) SO-CFAR detection parameter settings

[0040] c11) Guard window length setting: When estimating target clutter background, the target itself does not participate in the detection threshold estimation, so the guard window length N on one side is taken. g =1 to 4;

[0041] c12) Background window length setting: The samples selected within the background window should reflect the clutter changes in the local area of ​​the target. The length N of the single-sided guard window should be set. b =4~32;

[0042] c13) False alarm probability setting: Setting the false alarm probability P for high threshold detection fa1 =1e -4 ;

[0043] c2) The low threshold detection result is sent to the high threshold SO-CFAR detector for constant false alarm detection to obtain the high threshold detection value Det_cfar2.

[0044] Preferably, the calculation of the energy-weighted target location specifically involves:

[0045]

[0046] Where x i E is the position coordinate of the i-th point. i X is the energy of the i-th point, and X is the position coordinate of the merged target.

[0047] Preferably, the grid divided into equal-area units based on distance gates and frequency gates is R×F, where R is the width of the distance gate in the grid and F is the width of the frequency gate in the grid.

[0048] Preferably, the formula for calculating the dot density D is:

[0049] Where N is the number of dots in each grid.

[0050] Preferably, the step of using the RANSAC algorithm to fit the low-threshold detection result Det_cfar1 and iteratively solving the model fitting parameters to obtain the fitting result Fit_line1 is specifically as follows:

[0051] a1) Model estimation

[0052] The coordinates of the point in the detection result Det_cfar1 are (x i ,y i Given a linear model y = mx + b, select two points (x1, y1) and (x2, y2) and calculate the slope m and intercept b:

[0053]

[0054] b = y1 - mx1

[0055] a2) Error Calculation

[0056] Detection point (x) i ,y i The perpendicular distance from the line is:

[0057]

[0058] If the vertical distance d is less than the threshold T_ran, then the point is an interior point; otherwise, it is an exterior point.

[0059]

[0060] a3) Iterative estimation

[0061] Record the number of inliers in the current model. If the number of inliers is greater than the threshold, stop the iteration; otherwise, continue executing a1) and a2).

[0062] Preferably, the step of using the RANSAC algorithm to fit the high-threshold detection result Det_cfar2 and iteratively solving the model fitting parameters to obtain the fitting result Fit_line2 is specifically as follows:

[0063] b1) Model estimation

[0064] The coordinates of the point in the detection result Det_cfar1 are (x i ,y i Given a linear model y = mx + b, select two points (x1, y1) and (x2, y2) and calculate the slope m and intercept b:

[0065]

[0066] b = y1 - mx1

[0067] b2) Error calculation:

[0068] Detection point (x) i ,y i The perpendicular distance from the line is:

[0069]

[0070] If the vertical distance d is less than the threshold T_ran, then the point is an interior point; otherwise, it is an exterior point.

[0071]

[0072] b3) Iterative estimation:

[0073] Record the number of interior points in the current model. If the number of interior points is greater than the threshold, stop the iteration; otherwise, continue executing b1) and b2).

[0074] Compared with the prior art, the significant advantages of this invention are:

[0075] This invention proposes a hovering helicopter detection method for rotor composite materials. By combining low-threshold and high-threshold CFAR detection, the blades and hub of the helicopter rotor can be detected separately. Energy-weighted point accumulation ensures the accumulation of energy for the rotor's micro-doppler characteristics. A grid density estimation method is used for initial screening of rotor targets and reduces false alarms from non-rotor targets. Finally, dual RANSAC fitting is used to estimate the rotor's Doppler modulation characteristics, completing the hovering helicopter detection. Compared to traditional hovering helicopter detection methods, this technology can effectively extract the weak Doppler characteristics generated by modern helicopter rotors made of composite materials in the radar spectrum, especially showing better results in hovering helicopter detection under strong clutter environments with air-to-air look-down conditions. Attached Figure Description

[0076] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0077] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings.

[0078] Reference Figure 1 The implementation steps of this invention are as follows:

[0079] Step 1: Data Preprocessing and Dual Threshold Detection

[0080] a) Pulse compression is performed on the raw echo data to improve resolution by using coded signals and matched filters;

[0081] Then, coherent accumulation is performed to coherently superimpose multiple pulse echo signals, and FFT processing is used to convert the time-domain signal into a frequency-domain signal.

[0082] Y = FFT{Y pc (t)}

[0083] Where y k (t) is the post-pulse compression signal; Y acc (t) represents the result of coherent accumulation; FFT represents the fast Fourier transform.

[0084] b) The frequency domain signal is fed into a low-threshold SO-CFAR detector for constant false alarm rate (CFAR) detection to obtain the low-threshold detection result.

[0085] Det_cfar1;

[0086] b1) SO-CFAR detection parameter settings

[0087] b11) Guard window length setting: When estimating target clutter background, the target itself does not participate in the detection threshold estimation. Typically, the guard window length N on one side is set. g =1 to 4;

[0088] b12) Background window length setting: The samples selected within the background window can reflect the clutter changes in the local area of ​​the target. Typically, the length N of the single-sided guard window is set. b =4~32;

[0089] b13) False alarm probability setting: Setting the false alarm probability P for low threshold detection fa1 =1e -2 ;

[0090] b2) The frequency domain signal is fed into a low-threshold SO-CFAR detector for constant false alarm rate (CFAR) detection to obtain a low-threshold detection point. c) The low-threshold detection result is fed into a high-threshold SO-CFAR detector for CFAR detection to obtain the detection result Det_cfar2.

[0091] c1) SO-CFAR detection parameter settings

[0092] c11) Protection window length setting: When estimating target clutter background, the target itself does not participate in the detection gate.

[0093] For limited estimation, the length N of a single-sided protective window is usually taken. g =1 to 4;

[0094] c12) Background window length setting: The sample selected within the background window can reflect the clutter variation in the local area of ​​the target.

[0095] For example, the length N of a single-sided protective window is typically taken. b=4~32;

[0096] c13) False alarm probability setting: Setting the false alarm probability P for high threshold detection fa1 =1e -4 ;

[0097] c2) The low-threshold detection result is sent to the high-threshold SO-CFAR detector for constant false alarm detection to obtain the high-threshold detection point.

[0098] Step 2: Energy-weighted point aggregation

[0099] a) Extracting the energy information E of dual-threshold CFAR detection points i ;

[0100] b) Weighted average of adjacent points, with points of higher energy having greater weight;

[0101] c) Calculate the energy-weighted target location;

[0102]

[0103] Where x i E is the position coordinate of the i-th point. i X is the energy (amplitude) of the i-th point, and X is the position coordinate of the merged target.

[0104] Step 3: Grid-based point density estimation and rotor target selection

[0105] a) The energy-weighted point trace condensation result is divided into equal-area grids (R) based on distance and frequency gates.

[0106] ×F), the distance gate width R in the grid is generally 1, 3, or 5, and the frequency gate width F covers all frequency gates in the spectrum, generally 128, 256, or 512;

[0107] b) Count the points within each grid cell, calculate the number of points N in each grid cell, and then calculate the point density D using the formula...

[0108] The formula is:

[0109]

[0110] c) Determine if the point density D is greater than or equal to the threshold T_density. If the point density is greater than the threshold, retain it as...

[0111] Rotorcraft target; otherwise, remove the target.

[0112] Step 4: RANSAC Algorithm Fitting and Helicopter Modulation Spectrum Detection a) Use the RANSAC algorithm to fit the low-threshold detection result Det_cfar1, and iteratively solve the model fitting parameters.

[0113] The fitting result Fit_line1 is obtained;

[0114] a1) Model estimation

[0115] The coordinates of the point in the detection result Det_cfar1 are (x i ,y i Given a linear model y = mx + b, select two points (x1, y1) and (x2, y2) and calculate the slope m and intercept b:

[0116]

[0117] b = y1 - mx1

[0118] a2) Error Calculation

[0119] Detection point (x) i ,y i The perpendicular distance from the line is:

[0120]

[0121] If the vertical distance d is less than the threshold T_ran (usually taken as 1 to 3), then the point is an interior point; otherwise, it is an exterior point.

[0122]

[0123] a3) Iterative estimation

[0124] Record the number of inliers in the current model. If the number of inliers is greater than the threshold (usually 10-20), stop the iteration.

[0125] Otherwise, continue with steps a1) and a2).

[0126] b) Use the RANSAC algorithm to fit the high-threshold detection result Det_cfar2, iteratively solve the model fitting parameters, and obtain the fitting result Fit_line2;

[0127] b1) Model estimation

[0128] The coordinates of the point in the detection result Det_cfar1 are (x i ,y i Given a linear model y = mx + b, select two points (x1, y1) and (x2, y2) and calculate the slope m and intercept b:

[0129]

[0130] b = y1 - mx1

[0131] b2) Error calculation:

[0132] Detection point (x) i ,y i The perpendicular distance from the line is:

[0133]

[0134] If the vertical distance d is less than the threshold T_ran (usually taken as 1 to 3), then the point is an interior point; otherwise, it is an exterior point.

[0135]

[0136] b3) Iterative estimation:

[0137] Record the number of inliers in the current model. If the number of inliers is greater than the threshold (usually 10 to 15), stop the iteration.

[0138] Otherwise, continue with steps b1) and b2).

[0139] c) For the targets preserved from the dot density estimation, calculate the line length (Dop_line1) and mean amplitude (Amp_line1) of Fit_line1, and calculate the line length (Dop_line2) and mean amplitude (Amp_line2) of Fit_line2.

[0140]

[0141] Where R x_max and R x_min The maximum and minimum x-coordinates of the interior point R are given by R. y_max and R y_min These are the maximum and minimum y-coordinates of the interior point R.

[0142] d) Calculate the amplitude difference Amp_diff between Amp_line1 and Amp_line2. If Amp_diff is greater than T_amp (generally taken as 5 to 20 dB) and Dop_line1 is greater than the threshold (generally taken as 10 to 20), then it is determined that a hovering helicopter has been detected.

[0143] The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which this invention pertains may make various modifications, additions, or similar substitutions to the described specific embodiments without departing from the spirit of the invention or exceeding the scope defined by the appended claims.

Claims

1. A method for detecting hovering helicopters using rotor composite materials, characterized in that, Includes the following steps: Step 1: Data Preprocessing and Dual Threshold Detection a) Perform pulse compression and coherent accumulation on the raw echo data, and convert the time-domain signal into a frequency-domain signal using FFT; b) The frequency domain signal is fed into a low-threshold SO-CFAR detector for constant false alarm detection to obtain the low-threshold detection result Det_cfar1; c) The low-threshold detection result is sent to the high-threshold SO-CFAR detector for constant false alarm detection, and the detection result Det_cfar2 is obtained. Step 2: Energy-weighted point aggregation a) Extract energy information from dual-threshold CFAR detection points; b) Weighted average of adjacent points, with points of higher energy having greater weight; c) Calculate the energy-weighted target location; Step 3: Grid-based point density estimation and rotor target selection a) After condensing the energy-weighted point traces, divide the result into equal-area grids based on distance gates and frequency gates; b) Count the number of dots in each grid and calculate the dot density; c) Determine if the spot density is greater than the threshold T_density. If the spot density is greater than the threshold, retain it as a rotor target; otherwise, discard the target. Step 4: RANSAC algorithm fitting and hovering helicopter modulation spectrum detection a) Use the RANSAC algorithm to fit the low-threshold detection result Det_cfar1, and iteratively solve the model fitting parameters to obtain the fitting result Fit_line1; b) Use the RANSAC algorithm to fit the high-threshold detection result Det_cfar2, iteratively solve the model fitting parameters, and obtain the fitting result Fit_line2; c) For the targets preserved by the dot density estimation, calculate the line length Dop_line1 and the mean amplitude Amp_line1 of Fit_line1, and calculate the line length Dop_line2 and the mean amplitude Amp_line2 of Fit_line2. Where R x_max and R x_min The maximum and minimum x-coordinates of the interior point R are given by R. y_max and R y_min These are the maximum and minimum y-coordinates of the interior point R; d) Calculate the amplitude difference Amp_diff between Amp_line1 and Amp_line2. If Amp_diff is greater than T_amp, and If Dop_line1 is greater than the threshold, it is determined that a hovering helicopter has been detected.

2. The hovering helicopter inspection method for rotor composite materials according to claim 1, characterized in that, The process of pulse compression, coherent accumulation, and FFT conversion of the time-domain signal to the frequency-domain signal from the original echo data specifically involves: pulse compression of the original echo data by using an encoded signal and a matched filter to improve resolution; then coherent accumulation to coherently superimpose multiple pulse echo signals; and finally, FFT processing to convert the time-domain signal to the frequency-domain signal. Y=FFT{Y pc (t)} Where y k (t) is the post-pulse compression signal; Y acc (t) represents the result of coherent accumulation; FFT represents the fast Fourier transform.

3. The hovering helicopter inspection method for rotor composite materials according to claim 1, characterized in that, The step of sending the frequency domain signal into a low-threshold SO-CFAR detector for constant false alarm rate (CFAR) detection, and obtaining the low-threshold detection result Det_cfar1, specifically involves: b1) SO-CFAR detection parameter settings b11) Guard window length setting: When estimating target clutter background, the target itself does not participate in the detection threshold estimation, so the guard window length N on one side is set. g =1 to 4; b12) Background window length setting: The samples selected within the background window should reflect the clutter changes in the local area of ​​the target. The length N of the single-sided guard window should be set. b =4~32; b13) False alarm probability setting: Setting the false alarm probability P for low threshold detection fa1 =1e -2 ; b2) The frequency domain signal is sent to a low-threshold SO-CFAR detector for constant false alarm rate detection to obtain the low-threshold detection value Det_cfar1.

4. The hovering helicopter inspection method for rotor composite materials according to claim 1, characterized in that, The process of sending the low-threshold detection result into the high-threshold SO-CFAR detector for constant false alarm rate (CFAR) detection, and obtaining the detection result Det_cfar2, is as follows: c1) SO-CFAR detection parameter settings c11) Guard window length setting: When estimating target clutter background, the target itself does not participate in the detection threshold estimation, so the guard window length N on one side is taken. g =1 to 4; c12) Background window length setting: The samples selected within the background window should reflect the clutter changes in the local area of ​​the target. The length N of the single-sided guard window should be set. b =4~32; c13) False alarm probability setting: Setting the false alarm probability P for high threshold detection fa1 =1e -4 ; c2) The low threshold detection result is sent to the high threshold SO-CFAR detector for constant false alarm detection to obtain the high threshold detection value Det_cfar2.

5. The hovering helicopter inspection method for rotor composite materials according to claim 1, characterized in that, The calculation of the energy-weighted target location is specifically as follows: Where x i E is the position coordinate of the i-th point. i X is the energy of the i-th point, and X is the position coordinate of the merged target.

6. The hovering helicopter inspection method for rotor composite materials according to claim 1, characterized in that, The grid, divided into equal-area units by distance gates and frequency gates, is R×F, where R is the width of the distance gate in the grid and F is the width of the frequency gate in the grid.

7. The hovering helicopter detection method for rotor composite materials according to claim 6, characterized in that, The formula for calculating the dot density D is: Where N is the number of dots in each grid.

8. The hovering helicopter inspection method for rotor composite materials according to claim 1, characterized in that, The process of using the RANSAC algorithm to fit the low-threshold detection result Det_cfar1 and iteratively solving the model fitting parameters to obtain the fitting result Fit_line1 is as follows: a1) Model estimation The coordinates of the point in the detection result Det_cfar1 are (x i ,y i Given a linear model y = mx + b, select two points (x1, y1) and (x2, y2) and calculate the slope m and intercept b: b = y1 - mx1 a2) Error Calculation Detection point (x) i ,y i The perpendicular distance from the line is: If the vertical distance d is less than the threshold T_ran, then the point is an interior point; otherwise, it is an exterior point. a3) Iterative estimation Record the number of inliers in the current model. If the number of inliers is greater than the threshold, stop the iteration; otherwise, continue executing a1) and a2).

9. The hovering helicopter inspection method for rotor composite materials according to claim 1, characterized in that, The process of using the RANSAC algorithm to fit the high-threshold detection result Det_cfar2, iteratively solving the model fitting parameters, and obtaining the fitting result Fit_line2 is as follows: b1) Model estimation The coordinates of the point in the detection result Det_cfar1 are (x i ,y i Given a linear model y = mx + b, select two points (x1, y1) and (x2, y2) and calculate the slope m and intercept b: b = y1 - mx1 b2) Error calculation: Detection point (x) i ,y i The perpendicular distance from the line is: If the vertical distance d is less than the threshold T_ran, then the point is an interior point; otherwise, it is an exterior point. b3) Iterative estimation: Record the number of inliers in the current model. If the number of inliers is greater than the threshold, stop the iteration; otherwise, continue executing b1) and b2).

Images