Anti-interference fine target positioning method of millimeter wave radar in power transmission line measurement

Abstract

The invention discloses an anti-interference fine target positioning method of a millimeter wave radar in power transmission line measurement, and relates to the technical field of target positioning, and the method comprises the steps: carrying out the optimization of a radar echo signal through time domain filtering and matched filtering, carrying out the multi-scale analysis through wavelet transformation, effectively recognizing different types of interference sources, and carrying out the positioning of a small target. The method comprises the following steps of: realizing self-adaptive filtering by combining a machine learning algorithm, realizing high-precision target positioning by adopting a phase difference positioning method based on an interference-removed signal, dynamically tracking a target by utilizing particle filtering, ensuring stable tracking of the target in a complex environment, and realizing high-precision target positioning by combining infrared and optical sensor and millimeter wave radar data. And deep feature extraction and fusion are carried out through a data fusion algorithm and a convolutional neural network, so that the accuracy and robustness of target identification and positioning are further improved, the defects of a traditional method in the aspects of fine target positioning and interference resistance are overcome, and the precision and reliability of target positioning in power transmission line measurement are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of target positioning, and more specifically, it relates to a method for positioning small targets with anti-interference in transmission line measurement using millimeter-wave radar. Background Art

[0002] In the power system, transmission lines are key facilities for power transmission. Ensuring the safe operation of transmission lines is crucial for guaranteeing power supply. For this reason, power companies need to conduct regular inspections on transmission lines to detect possible problems such as damage, looseness, corrosion, and external interference in the lines. If these problems are not discovered and addressed in a timely manner, they may lead to serious safety hazards and even power outages. Traditional methods for inspecting transmission lines, such as manual inspections and manual remote sensing equipment, are inefficient, costly, and difficult to achieve real-time dynamic monitoring.

[0003] With the development of technology, millimeter-wave radar has been widely used in the target positioning and detection of transmission lines. Especially under complex weather and environmental conditions, millimeter-wave radar has significant advantages, such as the ability to penetrate obstacles such as haze, rain, and snow. However, the application of millimeter-wave radar still faces some challenges, especially in the positioning of small targets and anti-interference capabilities.

[0004] In current transmission line detection, millimeter-wave radar is often used to identify the position and status of targets through the reflection characteristics of target echoes. However, since there may be small targets in transmission lines (such as metal objects on wires, small obstacles on the lines, etc.), the radar reflection signals of these targets are weak and are easily affected by strong interference sources, resulting in difficulty in accurately identifying and positioning. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for positioning small targets with anti-interference in transmission line measurement using millimeter-wave radar, aiming to solve the technical problem that the radar reflection signals of current small targets are weak and are easily affected by strong interference sources, resulting in difficulty in accurately identifying and positioning.

[0006] The above technical purpose of the present invention is achieved through the following technical solutions:

[0007] In a first aspect, the present application provides a method for positioning small targets with anti-interference in transmission line measurement using millimeter-wave radar, including the following specific steps:

[0008] Based on the millimeter-wave radar antenna array, receive high-frequency millimeter-wave signals from small targets, perform time-domain filtering processing on the millimeter-wave signals, and obtain radar echo data corresponding to the millimeter-wave signals and after filtering processing;

[0009] Match-filter the radar echo data with the transmitted pulse signal, and obtain target signal data from the signal after match-filtering through a preset target detection algorithm;

[0010] Perform multi-scale analysis on the target signal data using wavelet transform to identify different types of interference sources in the target signal data. Obtain the existing interference types from the analysis of historical data and real-time signals based on a machine learning algorithm, and perform adaptive filtering processing on the target signal data using the interference types to obtain target signal data with each interference source removed;

[0011] Use phase difference positioning and particle filtering to obtain the target positioning result and the dynamic tracking result respectively from the target signal data with each interference source removed;

[0012] Use a data fusion algorithm to fuse the target positioning result, the dynamic tracking result, and the data of the preset infrared sensor and optical sensor, and use a convolutional neural network to extract and fuse features from the fused multi-modal data to obtain the target positioning information of the millimeter-wave signal.

[0013] On the basis of the above technical solution, the present invention can also be improved as follows.

[0014] Further, the above time-domain filtering processing uses a low-pass filter, specifically:

[0015] Set the cut-off frequency: Preset the cut-off frequency f c of the low-pass filter. Assume the frequency range of the signal is [f min , f max , where the cut-off frequency f c is above f max and lower than the main frequency component of the noise;

[0016] Pulse response: Use the cut-off frequency f c to generate the pulse response h(t) of the low-pass filter:

[0017]

[0018] In the formula: t represents the time variable, which is the time characteristic of the pulse response;

[0019] Convolution operation: Perform a convolution operation on the millimeter-wave signal x(t) and the pulse response h(t) of the low-pass filter to obtain the radar echo data y(t) after filtering processing:

[0020]

[0021] Where: y(t) represents the radar echo data after filtering; x(τ) represents the millimeter-wave signal; τ represents the integration variable, which is the value of the signal at different time points; t represents the time index of the convolution operation result, which is the position of the output signal in the time domain; dτ represents the integration of the input signal and the impulse response h(t - τ) on the time axis; t - τ represents the time domain delay.

[0022] Further, the above-mentioned matching filtering of the radar echo data and the transmitted pulse signal is specifically as follows:

[0023] The impulse response of the matched filter:

[0024] h match (t) = p(-t);

[0025] Where: p(-t) represents the time reversal of the transmitted pulse signal; h match (t) represents the impulse response of the matched filter;

[0026] The convolution operation of the matched filtering:

[0027]

[0028] Where: y match (t) represents the signal after the matched filtering; p(t - τ) represents the time domain reversal of the transmitted pulse signal p(t); dτ represents the integration of the input signal and the impulse response on the time axis, and x(τ) represents the millimeter-wave signal.

[0029] Further, the above-mentioned obtaining of the target signal data from the signal after the matched filtering through the preset target detection algorithm is specifically as follows:

[0030] Signal energy calculation: The energy of the signal at time t is obtained by the sum of the squares of the signal:

[0031] E(t) = |y match (t)| 2 ;

[0032] Where: E(t) represents the signal energy at time t; y match (t) represents the signal after the matched filtering;

[0033] Target detection: Set a threshold γ to judge whether the signal energy reaches the target detection standard, where the threshold γ is set based on the estimated value of the noise power as follows:

[0034]

[0035] Where: γ represents the threshold of the energy detection; N represents the gain factor; represents the estimated value of the noise power;

[0036] When E(t) > γ, there is target signal data at time t;

[0037] Time window extraction of target signal data: After detecting the target signal data, extract the target signal data by setting a duration window Δt:

[0038] y target (t) = y match (t), t ∈ t start , t start +Δt;

[0039] In the formula: t start represents the time when the energy exceeds the threshold γ, and y target (t) represents the target signal data, and y match (t) represents the signal after matched filtering.

[0040] Furthermore, the above uses wavelet transform to perform multi-scale analysis on the target signal data, identify different types of interference sources in the target signal data, obtain the existing interference types from the analysis of historical data and real-time signals based on machine learning algorithms, and use the interference types to perform adaptive filtering processing on the target signal data to obtain the target signal data with each interference source removed. Specifically:

[0041] Wavelet transform:

[0042]

[0043] In the formula: W(a, b) represents the wavelet transform coefficient, which is the local feature of the signal at different scales a and positions b; y target (t) represents the target signal data; represents the complex conjugate mother wavelet function; a represents the scale parameter; b represents the translation parameter;

[0044] Feature extraction: Based on the results of wavelet transform, extract frequency components and time-frequency distribution features;

[0045] Machine learning algorithm: Based on the extracted frequency components and time-frequency distribution features, use a support vector machine to judge the interference sources in the target signal data;

[0046] Adaptive filtering processing:

[0047] w(k + 1) = w(k) + μ·e(k)·x(k);

[0048] Where: w(k) represents the coefficient of the adaptive filter at the kth iteration; μ represents the step size factor; e(k) = d(k) - y(k) represents the error signal, where d(k) represents the desired output of the adaptive filter and y(k) represents the actual output of the adaptive filter; x(k) represents the input signal;

[0049] Target signal recovery: The target signal data after removing each interference source is expressed as follows:

[0050] y target (t) ′ =y target (t)-y interference (t);

[0051] Where: y target (t) ′ Represents the target signal data after removing each interference source; y interference (t) represents the interference signal removed by the adaptive filter, y target (t) represents target signal data.

[0052] Furthermore, the above target positioning results are obtained by:

[0053] Azimuth positioning: The radar signal received by each antenna array is converted from analog to digital to form discrete signals r1(k), r2(k),…, r N (k), where N is the number of antennas and k is the time step;

[0054] For each pair of antennas, calculate the phase difference Δφ i :Δφ i =arg(r i (k))-arg(r1(k));

[0055] Where: arg(r i (k)) represents the phase of the received signal;

[0056] Estimate the azimuth:

[0057] Where: Δφ i represents the phase difference between the i-th antenna and the reference antenna; λ represents the wavelength of the radar signal; d represents the spacing between two adjacent antennas in the antenna array;

[0058] Locating the target: Combining the position of the antenna array and the estimated azimuth, the actual position of the target is calculated:

[0059] x=x0+R·cos(θ);

[0060] y=y0+R·sin(θ);

[0061] where: (x0, y0) represents the coordinates of the center of the antenna array; R represents the distance from the antenna array to the target; (x, y) represents the actual position coordinates of the target.

[0062] Furthermore, the above dynamic tracking result is obtained in the following way:

[0063] Initialize particles: Set the initial state of the target as x0 = [x0, y0, v x , v y T , where x0 and y0 respectively represent the initial position of the target, and v x and v y respectively represent the initial velocity of the target;

[0064] The particle filter represents the distribution of the target state through a set of particles , where N represents the number of particles;

[0065] Prediction: Based on the motion model of the target, the state of the particles is updated at each moment. The motion of the target in the plane is represented by the following state equation:

[0066] x k+1 = F · x k + v k ; In the formula: x k = [x k , y k , v x , v y T represents the state of the target at time k; F represents the state transition matrix; v k represents the process noise;

[0067] Update: After receiving the observation data z k of the target signal, calculate the weight of each particle according to the observation model which represents the probability that the particle conforms to the current observation data:

[0068] z k = h(x k ) + v k ; In the formula: h(x k ) represents the observation model of the target; v k represents the observation noise;

[0069] In the formula: σ 2 represents the variance of the observation noise; represents the predicted position of the i-th particle;

[0070] Resampling: Resample the particles according to the weights of the particles to generate a new set of particles ​​Particles with larger weights are sampled more frequently;

[0071] Estimate the target state: Through weighted averaging, obtain the estimated position of the target: In the formula: represents the estimated position of the target;

[0072] Based on the target's motion model and observation data, gradually update the target's position and velocity through particle filtering, and adjust the target state estimation in real time.

[0073] Furthermore, the above multi-modal data is obtained through the following methods:

[0074] Weighted fusion of the target localization result and the dynamic tracking result: The target localization data obtains the target localization result (x, y) through the phase difference method, and the dynamic tracking result (x k , y k , v x , v y ) provided by the particle filtering method. Then, the target state vector is obtained after weighted fusion:

[0075] x fusion = w pos ·(x, y) + w track ·(x k , y k , v x , v y );

[0076] In the formula: w pos represents the weight of the localization data; w track represents the weight of the particle wave dynamic tracking data; x, y represent the target localization result obtained through phase difference localization; (x k , y k , v x , v y ) represents the dynamic tracking result obtained by particle filtering;

[0077] Image data fusion: Concatenate the infrared image I IR of the infrared sensor and the optical image I optical of the optical sensor to obtain the fused image:

[0078] I fusion = Concat(I IR , I optical );

[0079] In the formula: I fusion represents the fused image tensor, which combines the number of channels of the infrared image and the optical image, with a shape of h fusion × w fusion × (c + c′);

[0080] Fuse the target state vector x fusion with the image tensor I fusion as follows:

[0081] X final = Concat(x fusion , I fusion );

[0082] In the formula: X final represents the multimodal data obtained after fusion, with a shape of h fusion × w fusion × (c + c ′ + 4) tensor, where 4 is the dimension of the target state (x, y, v x , v y ).

[0083] In a second aspect, the present application provides a small target positioning system for anti-interference in transmission line measurement using a millimeter-wave radar, which applies the anti-interference small target positioning method of the millimeter-wave radar in transmission line measurement according to any one of the first aspects, including:

[0084] A time-domain filtering module, configured to perform time-domain filtering on the millimeter-wave signals received from small targets based on the millimeter-wave radar antenna array, and obtain radar echo data corresponding to the millimeter-wave signals and filtered;

[0085] A target detection module, configured to perform matched filtering on the radar echo data and the transmitted pulse signal, and obtain target signal data from the signal after the matched filtering through a preset target detection algorithm;

[0086] An interference removal module, configured to perform multi-scale analysis on the target signal data using wavelet transform, identify different types of interference sources in the target signal data, obtain the existing interference types from the analysis of historical data and real-time signals based on a machine learning algorithm, and perform adaptive filtering on the target signal data using the interference types to obtain target signal data with each interference source removed;

[0087] A positioning and tracking module, configured to respectively obtain a target positioning result and a dynamic tracking result from the target signal data with each interference source removed using phase difference positioning and particle filtering;

[0088] A fusion processing module, configured to use a data fusion algorithm to fuse the target positioning result, the dynamic tracking result, and the data of a preset infrared sensor and optical sensor, and perform feature extraction and fusion on the multimodal data obtained after fusion using a convolutional neural network to obtain target positioning information of the millimeter-wave signal.

[0089] In a third aspect, the present application provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the method according to any one of the first aspect is implemented.

[0090] In a fourth aspect, the present application provides a non-transitory computer-readable storage medium storing computer instructions that cause a computer to execute the method according to any one of the first aspect.

[0091] Compared with the prior art, the present invention has at least the following beneficial effects:

[0092] In the present application, first, time-domain filtering processing is performed on the millimeter-wave signal, and the radar echo data after filtering processing is subjected to matched filtering with the transmitted pulse signal, and then the target signal data is processed to remove each interference source; second, phase difference positioning and particle filtering are used to respectively obtain the target positioning result and the dynamic tracking result from the target signal data after removing each interference source; finally, the target positioning result, the dynamic tracking result, and the data of the preset infrared sensor and optical sensor are subjected to data fusion through a data fusion algorithm, and a convolutional neural network is used to extract and fuse features from the multimodal data obtained after fusion, and finally the target positioning information of the millimeter-wave signal is obtained, solving the technical problem that the radar reflection signal of a small target is weak at present and is easily affected by strong interference sources, resulting in difficult accurate identification and positioning.

[0093] In the present application, first, the radar echo signal is optimized by time-domain filtering and matched filtering, significantly improving the quality and signal-to-noise ratio of the target signal and overcoming the challenge of weak signals; then, multi-scale analysis is performed through wavelet transform to effectively identify different types of interference sources, and combined with machine learning algorithms to achieve adaptive filtering, successfully removing interference signals in a complex environment and ensuring the purity of the target signal; then, based on the signal after removing interference, a phase difference positioning method is used to achieve high-precision target positioning, and particle filtering is used to dynamically track the target, ensuring stable tracking of the target in a complex environment; finally, combined with infrared, optical sensor and millimeter-wave radar data, deep feature extraction and fusion are performed through a data fusion algorithm and a convolutional neural network (CNN), further improving the accuracy and robustness of target recognition and positioning; overall, this solution solves the deficiencies of traditional methods in small target positioning and anti-interference through means such as multi-level signal processing, intelligent interference suppression, high-precision positioning and dynamic tracking, and multimodal data fusion, improving the accuracy and reliability of target positioning in transmission line measurement. Description of the Drawings

[0094] The accompanying drawings described herein are used to provide a further understanding of the embodiments of the present invention, form a part of this application, and do not limit the embodiments of the present invention. In the drawings:

[0095] Figure 1 It is a flowchart of the positioning method in the embodiment of the present invention;

[0096] Figure 2 It is a schematic connection diagram of the positioning system in the embodiment of the present invention;

[0097] Figure 3 It is a schematic connection diagram of the electronic device in the embodiment of the present invention. Detailed implementation manners

[0098] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Usually, the components of the embodiments of the present invention described and illustrated in the accompanying drawings herein can be arranged and designed in various different configurations.

[0099] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed present invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0100] It should be noted that: Similar reference numerals and letters denote similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0101] In the description of the embodiments of the present invention, "a plurality" represents at least two.

[0102] Embodiment 1: To solve the technical problem that the radar reflection signal of small targets is currently weak and is easily affected by strong interference sources, resulting in difficult accurate identification and positioning, this embodiment provides a method for positioning small targets with anti-interference in transmission line measurement using a millimeter-wave radar, as Figure 1 shown, including the following specific steps:

[0103] S1, Based on the millimeter-wave radar antenna array, receive high-frequency millimeter-wave signals from small targets, and perform time-domain filtering processing on the millimeter-wave signals to obtain radar echo data corresponding to the millimeter-wave signals and filtered.

[0104] Among them, the above time-domain filtering processing uses a low-pass filter, and the specific steps can be:

[0105] S11, Set the cut-off frequency: Preset the cut-off frequency f of the low-pass filter c , and set the frequency range of the signal as [f min , f max , where the cut-off frequency f c is above f max and lower than the main frequency component of the noise.

[0106] S12, Pulse response: Generate the pulse response h(t) of the low-pass filter using the cut-off frequency f c :

[0107]

[0108] In the formula: t represents the time variable and is the time characteristic of the pulse response.

[0109] S13, Convolution operation: Perform a convolution operation on the millimeter-wave signal x(t) and the pulse response h(t) of the low-pass filter to obtain the radar echo data y(t) after filtering:

[0110]

[0111] In the formula: y(t) represents the radar echo data after filtering; x(τ) represents the millimeter-wave signal; τ represents the integration variable and is the value of the signal at different time points; t represents the time index of the convolution operation result and is the position of the output signal in the time domain; dτ represents the integration of the input signal and the pulse response h(t - τ) on the time axis; t - τ represents the time domain delay.

[0112] S2, Perform a matched filter on the radar echo data and the transmitted pulse signal, and obtain the target signal data from the signal after the matched filter through a preset target detection algorithm.

[0113] Among them, the above-mentioned matched filter of the radar echo data and the transmitted pulse signal can specifically be:

[0114] S21, Pulse response of the matched filter:

[0115] h match (t) = p(-t);

[0116] In the formula: p(-t) represents the time reversal of the transmitted pulse signal; h match (t) represents the pulse response of the matched filter.

[0117] S22, Convolution operation of the matched filter:

[0118]

[0119] In the formula: ymatch (t) represents the signal after matched filtering; p(t - τ) represents the time-domain inversion of the transmitted pulse signal p(t); dτ represents the integration of the input signal and the impulse response on the time axis, and x(τ) represents the millimeter-wave signal.

[0120] Specifically, based on the high-frequency millimeter-wave signal received by the millimeter-wave radar antenna array, time-domain filtering is performed to remove low-frequency noise and unnecessary interference, ensuring the clarity of the radar echo data; then, the radar echo signal is optimized through the matched filtering method, enabling the target signal to be more accurately aligned with the received echo signal, thereby improving the signal-to-noise ratio. The processing in this stage greatly enhances the quality of the target signal and provides a more reliable data basis for subsequent target detection and positioning.

[0121] Optionally, obtaining the target signal data from the signal after matched filtering through the preset target detection algorithm is specifically as follows:

[0122] S23, Signal energy calculation: The energy of the signal at time t is obtained by the sum of the squares of the signal:

[0123] E(t) = |y match (t)| 2 ;

[0124] In the formula: E(t) represents the signal energy at time t; y match (t) represents the signal after matched filtering.

[0125] S24, Target detection: Set a threshold γ to determine whether the signal energy reaches the target detection standard, where the threshold γ is set based on the estimated value of the noise power as follows:

[0126]

[0127] In the formula: γ represents the threshold of energy detection; N represents the gain factor; represents the estimated value of the noise power; when E(t) > γ, there is target signal data at time t.

[0128] S25, Time window extraction of target signal data: When the target signal data is detected, the target signal data is extracted by setting a duration window Δt:

[0129] y target (t) = y match (t), t ∈ t start , t start + Δt;

[0130] In the formula: t start represents the moment when the energy exceeds the threshold γ, and y target(t) represents the target signal data, y match (t) represents the signal after matched filtering.

[0131] S3. Perform multi-scale analysis on the target signal data using wavelet transform to identify different types of interference sources in the target signal data. Obtain the existing interference types from the analysis of historical data and real-time signals based on machine learning algorithms, and use the interference types to perform adaptive filtering on the target signal data to obtain the target signal data with each interference source removed.

[0132] Among them, the above-mentioned performing multi-scale analysis on the target signal data using wavelet transform to identify different types of interference sources in the target signal data, obtaining the existing interference types from the analysis of historical data and real-time signals based on machine learning algorithms, and using the interference types to perform adaptive filtering on the target signal data to obtain the target signal data with each interference source removed can be specifically expressed as:

[0133] S31. Wavelet transform:

[0134]

[0135] In the formula: W(a, b) represents the wavelet transform coefficient, which is the local feature of the signal at different scales a and positions b; y target (t) represents the target signal data; represents the complex conjugate mother wavelet function; a represents the scale parameter; b represents the translation parameter.

[0136] S32. Feature extraction: Based on the result of wavelet transform, extract the frequency components and time-frequency distribution features.

[0137] S33. Machine learning algorithm: Based on the extracted frequency components and time-frequency distribution features, use a support vector machine to judge the interference sources in the target signal data.

[0138] S34. Adaptive filtering processing: w(k + 1) = w(k) + μ·e(k)·x(k);

[0139] In the formula: w(k) represents the coefficient of the k-th iteration adaptive filter; μ represents the step size factor; e(k) = d(k) - y(k) represents the error signal, where d(k) represents the expected output of the adaptive filter, and y(k) represents the actual output of the adaptive filter; x(k) represents the input signal.

[0140] S35. Target signal recovery: The target signal data with each interference source removed is expressed as follows:

[0141] y target (t) ′ = y target (t) - y interference(t);

[0142] where: y target (t) ′ represents the target signal data after removing each interference source; y interference (t) represents the interference signal removed by the adaptive filter, and y target (t) represents the target signal data.

[0143] Specifically, by performing multi-scale analysis on the target signal using wavelet transform, the characteristics of the interference sources can be identified at different frequency bands and time scales. The multi-resolution characteristic of wavelet transform helps the system effectively distinguish the target signal from the interference signal, especially in an environment with strong interference. Then, combined with machine learning algorithms, by analyzing the characteristics of historical data and real-time signals, the system can automatically identify the interference type and adopt corresponding adaptive filtering strategies to effectively remove the interference signal and ensure the purity of the target signal. This process effectively solves the problem that the target signal is overwhelmed by interference sources in a complex environment.

[0144] S4. Use phase difference positioning and particle filtering to separately obtain the target positioning result and the dynamic tracking result from the target signal data after removing each interference source.

[0145] Among them, the above-mentioned target positioning result is obtained through the following method:

[0146] S41. Azimuth angle positioning: The radar signals received by each antenna array are converted from analog to digital to form discrete signals r1(k), r2(k), …, r N (k), where N represents the number of antennas and k represents the time step; for each pair of antennas, calculate the phase difference Δφ i : Δφ i = arg(r i (k)) - arg(r1(k));

[0147] where: arg(r i (k)) represents the phase of the received signal;

[0148] S42. Estimate the azimuth angle:

[0149] where: Δφ i represents the phase difference between the i-th antenna and the reference antenna; λ represents the wavelength of the radar signal; d represents the spacing between adjacent antennas in the antenna array.

[0150] S43. Locate the target: Combine the position of the antenna array and the estimated azimuth angle to calculate the actual position of the target:

[0151] x = x0 + R·cos(θ);

[0152] y = y0 + R·sin(θ);

[0153] Where: (x0, y0) represents the coordinates of the center of the antenna array; R represents the distance from the antenna array to the target; (x, y) represents the actual position coordinates of the target.

[0154] Optionally, the above dynamic tracking result is obtained in the following way:

[0155] S44, Initialize particles: Set the initial state of the target as x0 = [x0, y0, v x , v y T , where x0 and y0 respectively represent the initial positions of the target, and v x and v y respectively represent the initial velocities of the target; the particle filter represents the distribution of the target state through a set of particles , where N represents the number of particles.

[0156] S45, Prediction: Based on the motion model of the target, the state of the particles is updated at each moment. The motion of the target in the plane is represented by the following state equation:

[0157] x k+1 = F·x k + v k ; Where: x k = [x k , y k , v x , v y T represents the state of the target at time k; F represents the state transition matrix; v k represents the process noise.

[0158] S46, Update: After receiving the observation data z k of the target signal, calculate the weight of each particle according to the observation model indicating the probability that the particle conforms to the current observation data:

[0159] z k = h(x k ) + v k ; Where: h(x k ) represents the observation model of the target; v k represents the observation noise;

[0160] Where: σ 2 represents the variance of the observation noise; represents the predicted position of the i-th particle.

[0161] ​​S47, Resampling: Resample the particles according to the weights of the particles to generate a new set of particles. Particles with larger particle weights are sampled more.

[0162] S48, Estimating the target state: Obtain the estimated position of the target through weighted average: In the formula: represents the estimated position of the target.

[0163] S49, According to the motion model of the target and the observation data, gradually update the position and velocity of the target through particle filtering, and adjust the state estimation of the target in real time.

[0164] Specifically, for the target signal data after removing interference, the phase difference positioning method is used for accurate target positioning. Through the phase difference in the radar echo signal, the system can accurately calculate the position of the target, overcoming the error problem of traditional positioning methods in complex environments. At the same time, the particle filter algorithm is used to dynamically track the target, which can update the position information of the target in real time, adapt to the motion characteristics of the target, and ensure stable tracking of the target in a dynamically changing environment. [[ID=!7]]

[0165] S5, Use the data fusion algorithm to fuse the target positioning results, dynamic tracking results, and the data of the preset infrared sensor and optical sensor, and use the convolutional neural network to extract and fuse the features of the fused multi-modal data to obtain the target positioning information of the millimeter wave signal.

[0166] Among them, the above multi-modal data is obtained in the following way:

[0167] S51, Weighted fusion of the target positioning result and the dynamic tracking result: The target positioning data obtains the target positioning result (x, y) through the phase difference method, and the dynamic tracking result (x k , y k , v x , v y ) provided by the particle filter method, then the target state vector is obtained after weighted fusion:

[0168] x fusion = w pos ·(x, y) + w track ·(x k , y k , v x , v y );

[0169] In the formula: w pos represents the weight of the positioning data; w track represents the weight of the particle wave dynamic tracking data; x, y represent the target positioning result obtained by phase difference positioning; (x k , yk , v x , v y ) represents the dynamic tracking result obtained by particle filtering.

[0170] S52, Image data fusion: Concatenate the infrared image I of the infrared sensor IR and the optical image I of the optical sensor optical to obtain the fused image:

[0171] I fusion = Concat(I IR , I optical );

[0172] In the formula: I fusion represents the fused image tensor, which combines the number of channels of the infrared image and the optical image, with a shape of h fusion × w fusion × (c + c').

[0173] S53, Fuse the target state vector x fusion with the image tensor I fusion :

[0174] X final = Concat(x fusion , I fusion );

[0175] In the formula: X final represents the multimodal data obtained after fusion, with a shape of h fusion × w fusion × (c + c ′ + 4) tensor, where 4 is the dimension of the target state (x, y, v x , v y ).

[0176] Specifically, the specific structure of the above convolutional neural network is as follows:

[0177] Input layer: The input size is h fusion × w fusion × (c + c ′ + 4), and the input data includes image information (infrared image and optical image) and the dynamic information of the target (position, speed);

[0178] First convolutional layer: The convolutional kernel size is 3×3, the number of convolutional kernels is 64, the stride is 1, the padding is same, the activation function is ReLU, and the output size is h fusion × w fusion × 64;

[0179] First pooling layer: The pooling window size is 2×2, the stride is 2, and the output size is

[0180] Second Convolutional Layer: The convolutional kernel size is 3×3, the number of convolutional kernels is 128, the stride is 1, the padding is same, the activation function is ReLU, and the output size is

[0181] Second Pooling Layer: The pooling window size is 2×2, the stride is 2, and the output size is

[0182] Third Convolutional Layer: The convolutional kernel size is 3×3, the number of convolutional kernels is 256, the stride is 1, the padding is same, the activation function is ReLU, and the output size is

[0183] Third Pooling Layer: The pooling window size is 2×2, the stride is 2, and the output size is

[0184] Fourth Convolutional Layer: The convolutional kernel size is 3×3, the number of convolutional kernels is 512, the stride is 1, the padding is same, the activation function is ReLU, and the output size is

[0185] Fourth Pooling Layer: The pooling window size is 2×2, the stride is 2, and the output size is

[0186] First Fully Connected Layer: The number of neurons is 1024, the activation function is ReLU, and the output size is 1024;

[0187] Second Fully Connected Layer: The number of neurons is 512, the activation function is ReLU, and the output size is 512;

[0188] Output Layer: The output size is 4 (the final position and velocity of the target: x, y, v x , v y ) The activation function uses Linear.

[0189] In this embodiment, in order to further improve the positioning accuracy, the system combines the data of infrared, optical sensors, and millimeter-wave radars, and effectively integrates the data of different sensors through a data fusion algorithm. This process of multi-modal data fusion can synthesize the advantages of various sensors, make up for the deficiencies of a single sensor, and improve the robustness of target positioning. Especially by using a convolutional neural network (CNN) to extract features from the fused multi-modal data, the system can deeply mine the hidden information in various data and further improve the accuracy of target recognition and positioning.

[0190] Embodiment 2: The embodiment of the present application provides a small target positioning system for anti-interference in transmission line measurement using a millimeter-wave radar, which is applied to the small target positioning method for anti-interference in transmission line measurement using the millimeter-wave radar in Embodiment 1. As Figure 2 shown, it includes:

[0191] A time-domain filtering module, which is used to receive high-frequency millimeter-wave signals from small targets based on the millimeter-wave radar antenna array, perform time-domain filtering on the millimeter-wave signals, and obtain radar echo data corresponding to the millimeter-wave signals and filtered;

[0192] A target detection module, which is used to perform matched filtering on the radar echo data and the transmitted pulse signal, and obtain target signal data from the signal after matched filtering through a preset target detection algorithm;

[0193] An interference removal module, which is used to perform multi-scale analysis on the target signal data using wavelet transform, identify different types of interference sources in the target signal data, obtain the existing interference types from the analysis of historical data and real-time signals based on a machine learning algorithm, and perform adaptive filtering on the target signal data using the interference types to obtain target signal data with each interference source removed;

[0194] A positioning and tracking module, which is used to obtain target positioning results and dynamic tracking results respectively from the target signal data with each interference source removed using phase difference positioning and particle filtering;

[0195] A fusion processing module, which is used to fuse the target positioning results, dynamic tracking results, and data of preset infrared sensors and optical sensors using a data fusion algorithm, and perform feature extraction and fusion on the fused multi-modal data using a convolutional neural network to obtain target positioning information of the millimeter-wave signals.

[0196] Embodiment 3: The embodiment of the present application provides an electronic device. As Figure 3 shown, it includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements the method in Embodiment 1.

[0197] Embodiment 4: The embodiment of the present application provides a non-transitory computer-readable storage medium. The non-transitory computer-readable storage medium stores computer instructions, and the computer instructions cause the computer to execute the method in Embodiment 1.

[0198] The specific embodiments described above further elaborate on the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only for the specific embodiments of the present invention and is not used to limit the protection scope of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for positioning small targets with anti-interference in transmission line measurement by millimeter-wave radar, characterized in that It includes the following specific steps: Based on the millimeter-wave radar antenna array to receive high-frequency millimeter-wave signals from small targets, perform time-domain filtering processing on the millimeter-wave signals to obtain radar echo data corresponding to the millimeter-wave signals and processed by filtering; Perform matched filtering on the radar echo data and the transmitted pulse signal, and obtain target signal data from the signal after matched filtering through a preset target detection algorithm; Use wavelet transform to perform multi-scale analysis on the target signal data, identify different types of interference sources in the target signal data, obtain the existing interference types from the analysis of historical data and real-time signals based on machine learning algorithms, and use the interference types to perform adaptive filtering processing on the target signal data to obtain target signal data with each interference source removed; Use phase difference positioning and particle filtering to respectively obtain target positioning results and dynamic tracking results from the target signal data with each interference source removed; Use a data fusion algorithm to fuse the target positioning results, the dynamic tracking results, and the data of preset infrared sensors and optical sensors, and use a convolutional neural network to extract and fuse features from the fused multi-modal data to obtain the target positioning information of the millimeter-wave signal.

2. The method for positioning small targets with anti-interference in transmission line measurement using the millimeter-wave radar according to claim 1, characterized in that, The time-domain filtering processing uses a low-pass filter, specifically: Set the cut-off frequency: Preset the cut-off frequency f of the low-pass filter c , and set the frequency range of the signal as [f min , f max . Among them, the cut-off frequency f c is above f max and lower than the main frequency component of the noise; Impulse response: Using the cut-off frequency f c Generate the impulse response h(t) of the low-pass filter: In the formula: t represents the time variable, which is the time characteristic of the impulse response; Convolution operation: Perform convolution operation on the millimeter-wave signal x(t) and the impulse response h(t) of the low-pass filter to obtain the radar echo data y(t) processed by filtering: In the formula: y(t) represents the radar echo data processed by filtering; x(τ) represents the millimeter-wave signal; τ represents the integration variable, which is the value of the signal at different time points; t represents the time index of the convolution operation result, which is the position of the output signal in the time domain; dτ represents the integration of the input signal and the impulse response h(t - τ) on the time axis; t - τ represents the time delay in the time domain.

3. The method for positioning small targets with anti-interference in transmission line measurement by the millimeter-wave radar according to claim 1, characterized in that Perform matched filtering on the radar echo data and the transmitted pulse signal, specifically: The impulse response of the matched filter: h match h(t) = p( -t); where: p(-t) represents the time reversal of the transmitted pulse signal; h match (t) represents the impulse response of the matched filter; The convolution operation of the matched filtering: where: y match (t) represents the signal after matched filtering; p(t - τ) represents the time-domain inversion of the transmitted pulse signal p(t); dτ represents the integration of the input signal and the impulse response on the time axis, and x(τ) represents the millimeter-wave signal.

4. The method for positioning small targets with anti-interference in transmission line measurement by the millimeter-wave radar according to claim 1, characterized in that Obtain target signal data from the signal after matched filtering through a preset target detection algorithm, specifically: Signal energy calculation: The energy of the signal at time t is obtained by the sum of the squares of the signal: E(t) = |y match (t)| 2 ; where: E(t) represents the signal energy at time t; y match (t) represents the signal after matched filtering; Target detection: Set a threshold γ to determine whether the signal energy meets the target detection standard, where the threshold γ is set based on the estimated value of the noise power Set as follows: Where: γ represents the threshold of energy detection; N represents the gain factor; represents the estimated value of the noise power; When E(t) > γ, there is target signal data at time t; Time window extraction of target signal data: When target signal data is detected, extract the target signal data by setting a duration window Δt: y target y(t) = y match (t), t ∈ t start , t start + Δt; where: t start represents the moment when the energy exceeds the threshold γ, y target (t) represents the target signal data, y match (t) represents the signal after matched filtering.

5. The method for positioning small targets with anti-interference in transmission line measurement by the millimeter-wave radar according to claim 1, characterized in that The use of wavelet transform to perform multi-scale analysis on the target signal data, identify different types of interference sources in the target signal data, obtain the existing interference types from the analysis of historical data and real-time signals based on machine learning algorithms, and use the interference types to perform adaptive filtering processing on the target signal data to obtain target signal data with each interference source removed, specifically: Wavelet transform: Where: \(W(a, b)\) represents the wavelet transform coefficient, which is the local feature of the signal at different scales \(a\) and positions \(b\); \(y\) target (t) represents the target signal data; represents the complex conjugate mother wavelet function; \(a\) represents the scale parameter; \(b\) represents the translation parameter; Feature extraction: Based on the results of wavelet transform, extract frequency components and time-frequency distribution features; Machine learning algorithm: Based on the extracted frequency components and time-frequency distribution features, use a support vector machine to judge the interference sources in the target signal data; Adaptive filtering processing: w(k + 1) = w(k) + μ·e(k)·x(k); Where: w(k) represents the coefficient of the adaptive filter at the k-th iteration; μ represents the step size factor; e(k) = d(k) - y(k) represents the error signal, where d(k) represents the desired output of the adaptive filter, and y(k) represents the actual output of the adaptive filter; x(k) represents the input signal; Target signal recovery: The target signal data after removing each interference source is represented as follows: y target (t) ′ = y target (t) - y interference (t); Where: y target (t) ′ represents the target signal data after removing each interference source; y interference (t) represents the interference signal removed by the adaptive filter, and y target (t) represents the target signal data.

6. The method for positioning small targets with anti-interference in transmission line measurement by the millimeter-wave radar according to claim 1, characterized in that The target positioning result is obtained in the following way: Azimuth angle positioning: After the radar signals received by each antenna array are converted from analog to digital, discrete signals r1(k), r2(k), …, r N (k) are formed, where N represents the number of antennas and k represents the time step; For each pair of antennas, calculate the phase difference Δφ i : Δφ i = arg(r i (k)) - arg(r1(k)); where: arg(r i (k)) represents the phase of the received signal; Estimate azimuth angle: where: Δφ i represents the phase difference between the i-th antenna and the reference antenna; λ represents the wavelength of the radar signal; d represents the spacing between two adjacent antennas in the antenna array; Positioning the target: Combining the position of the antenna array and the estimated azimuth angle, calculate the actual position of the target: x = x0 + R·cos(θ); y = y0 + R·sin(θ); Where: (x0, y0) represents the coordinates of the center of the antenna array; R represents the distance from the antenna array to the target; (x, y) represents the actual position coordinates of the target.

7. The method for positioning small targets with anti-interference in transmission line measurement by the millimeter-wave radar according to claim 1, characterized in that The dynamic tracking result is obtained in the following way: Initialize particles: Set the initial state of the target as x0 = [x0, y0, v x , v y T , where x0 and y0 respectively represent the initial positions of the target, and v x and v y respectively represent the initial velocities of the target;​ The particle filter represents the distribution of the target state through a set of particles where N represents the number of particles; Prediction: Based on the motion model of the target, the particles update their states at each moment. The motion of the target in the plane is represented by the following state equation: x k+1 = F·x k + v k ; In the formula: x k = [x k , y k , v x , v y T represents the state of the target at time k; F represents the state transition matrix; v k represents the process noise;​ Update: After receiving the observed data z of the target signal k the weight of each particle is calculated according to the observation model indicating the probability that the particle conforms to the current observed data: z k = h(x k ) + v k ; where: h(x k ) represents the observation model of the target; v k represents the observation noise; where: σ 2 represents the variance of the observation noise; represents the predicted position of the i-th particle; Resampling: Resample particles according to their weights to generate a new set of particles Particles with larger weights are sampled more frequently; Estimate the target state: Obtain the estimated position of the target through weighted average: Where: represents the estimated position of the target; According to the motion model of the target and the observation data through particle filtering, gradually update the position and velocity of the target, and adjust the state estimation of the target in real time.

8. The method for positioning small targets with anti-interference in transmission line measurement by the millimeter-wave radar according to claim 1, characterized in that The multi-modal data is obtained in the following way: Weighted fusion of the target localization result and the dynamic tracking result: The target localization data obtains the target localization result (x, y) through the phase difference method, and the dynamic tracking result (x k , y k , v x , v y ) provided by the particle filter method. Then, the target state vector is obtained after weighted fusion: x fusion = w pos ·(x,y) + w track ·(x k , y k , v x , v y ); where: w pos represents the weight of the positioning data; w track represents the weight of the particle wave dynamic tracking data; x, y represent the target positioning results obtained by phase difference positioning; (x k , y k , v x , v y ) represents the dynamic tracking result obtained by particle filtering; Image data fusion: Infrared image I of the infrared sensor IR and optical image I of the optical sensor optical are stitched together to obtain the fused image: I fusion = Concat(I IR , I optical ); Where: I fusion represents the fused image tensor, which combines the number of channels of the infrared image and the optical image, and has a shape of h fusion × w fusion × (c + c'); Fuse the target state vector x fusion with the image tensor I fusion as follows: X final = Concat(x fusion , I fusion ); Where: X final represents the multi-modal data obtained after fusion, with a shape of h fusion ×w fusion ×(c + c ′ + 4) tensor, where 4 is the dimension of the target state (x, y, v x , v y ).

9. A fine target positioning system for anti-interference in transmission line measurement using a millimeter-wave radar, which is applied to the method for anti-interference fine target positioning in transmission line measurement using the millimeter-wave radar according to any one of claims 1-8, characterized in that, Including: A time-domain filtering module, which is used to receive high-frequency millimeter-wave signals from small targets based on a millimeter-wave radar antenna array, perform time-domain filtering processing on the millimeter-wave signals, and obtain radar echo data corresponding to the millimeter-wave signals and after filtering processing; A target detection module, which is used to perform matched filtering on the radar echo data and the transmitted pulse signal, and obtain target signal data from the signal after matched filtering through a preset target detection algorithm; An interference removal module, which is used to perform multi-scale analysis on the target signal data by using wavelet transform, identify different types of interference sources in the target signal data, obtain the existing interference types from the analysis of historical data and real-time signals based on a machine learning algorithm, and perform adaptive filtering processing on the target signal data by using the interference types to obtain target signal data after removing each interference source; A positioning and tracking module, which is used to obtain the target positioning result and the dynamic tracking result from the target signal data after removing each interference source by using phase difference positioning and particle filtering; A fusion processing module, which is used to fuse the target positioning result, the dynamic tracking result, and the data of a preset infrared sensor and optical sensor by using a data fusion algorithm, and perform feature extraction and fusion on the multi-modal data obtained after fusion by using a convolutional neural network to obtain the target positioning information of the millimeter-wave signal.

10. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements the method described in any one of claims 1-8.

Images