Millimeter wave radar calibration method and application based on natural point target search

By defining ideal point targets and establishing feature databases, we automatically discover natural point targets for millimeter-wave radar calibration, solving the problems of cumbersome and difficult automation of traditional radar calibration methods, and achieving automatic calibration and performance improvement of radar in various application scenarios.

CN118604758BActive Publication Date: 2024-11-15UNIV OF SCI & TECH OF CHINA
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410773491.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-06-17
Publication Date
2024-11-15
Estimated Expiration
2044-06-17

AI Technical Summary

Technical Problem

Traditional radar calibration methods are cumbersome for commercial radar users and cannot be automated, and cannot effectively handle antenna phase errors, limiting the use of radar in various application scenarios.

Method used

By defining ideal point targets and establishing their characteristic databases, using mathematical modeling and fingerprint matching technology to automatically discover natural point targets, perform fully automatic calibration of millimeter wave radar, and realize automatic calibration of physical arrays and synthetic aperture arrays.

Benefits of technology

It realizes fully automatic calibration of radar in various application scenarios, effectively compensates for antenna manufacturing errors and phase errors caused by radar motion, improves radar perception and imaging performance, and reduces dependence on manual calibration.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118604758B_ABST
    Figure CN118604758B_ABST
Patent Text Reader

Abstract

The present disclosure provides a millimeter wave radar calibration method and application based on natural point target search, including: defining an ideal point target that can be used as a calibration source and performing mathematical modeling; establishing a feature database of the ideal point target; evaluating the scene measured radar echo data according to the feature database; obtaining a natural point target that can be used as a calibration source in the current scene according to the evaluation result; and calibrating and evaluating through the obtained natural point target.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present disclosure relates to the field of radar technology, and in particular to a millimeter wave radar calibration method based on natural point target search. Background Art

[0002] In recent years, millimeter-wave radar has been widely used in commercial fields such as autonomous driving and human perception due to its high resolution and high precision. Most of these emerging civilian radars use Multi-Input Multi-Output (MIMO) technology to achieve high angular resolution and high precision millimeter-wave radar perception through the coherence between multiple antennas, which requires accurate phase calibration between multiple antennas.

[0003] However, traditional calibration methods are impractical for long-term and daily use by end users. Specifically, traditional calibration methods require professionals to manually deploy corner reflectors in specially designed scenarios every few months and use the unique properties of corner reflectors to complete calibration. This traditional calibration method is feasible for dedicated radars. But for commercial radars, it is very cumbersome to require ordinary end users (such as car owners and commercial radar owners) to use professional corner reflectors to regularly calibrate between radar antennas. Although there are some radar self-calibration methods based on antenna coupling, the existing self-calibration methods have high scene restrictions and cannot handle situations where the antenna phase error exceeds the convergence threshold, which limits their application.

[0004] Therefore, there is an urgent need for a method that can automatically calibrate radar anytime and anywhere to meet the needs of millimeter-wave radar in various application scenarios. Summary of the invention

[0005] Based on the above problems, the present disclosure provides a millimeter wave radar calibration method based on natural point target search to alleviate the above technical problems in the prior art.

[0006] One aspect of the present disclosure provides a millimeter-wave radar calibration method based on natural point target search, including: defining an ideal point target that can serve as a calibration source and performing mathematical modeling; establishing a feature database of the ideal point target; evaluating the scene measured radar echo data according to the feature database; obtaining a natural point target in the current scene that can serve as a calibration source according to the evaluation result; and calibrating and evaluating using the obtained natural point target.

[0007] According to an embodiment of the present disclosure, an ideal point target is defined as a reflector that satisfies the following conditions: any point on the reflector has the same radar cross-sectional area relative to different antennas and does not change due to differences in the shape of the reflector itself.

[0008] According to an embodiment of the present disclosure, establishing a feature database of an ideal point target includes: simulating and generating intermediate frequency echo data of a single ideal point target at different positions; performing distance estimation and spatial angle estimation on the intermediate frequency echo data obtained for each ideal point target to obtain a spatial spectrum corresponding to different distance units of each ideal point target; and establishing a feature database of the ideal point target based on the spatial spectrum.

[0009] According to an embodiment of the present disclosure, distance estimation and spatial angle estimation are performed on the intermediate frequency echo data of each ideal point target to obtain the spatial spectrum corresponding to different distance units of each ideal point target, including: utilizing the scene invariance of the ideal point target, converting the echo data from the antenna-frequency domain to the distance-angle domain, and extracting the spatial spectrum at different distance units as feature data.

[0010] According to an embodiment of the present disclosure, evaluating the scene measured radar echo data based on the feature database includes: selecting a candidate point target set based on the scene measured radar echo data; and matching and evaluating the echo data of the candidate point targets in the candidate point target set with the feature data of the ideal point targets through a comprehensive scoring mechanism to obtain a scoring result.

[0011] According to an embodiment of the present disclosure, the comprehensive scoring mechanism includes: a matching score reflecting the degree of matching between the candidate point target and the ideal point target features; and a position score reflecting the spatial position of the candidate point target relative to the antenna.

[0012] According to an embodiment of the present disclosure, for each point target in the candidate point target set, the cosine similarity between the spatial spectrum of each point target and the feature data of the ideal point target is calculated as a matching score.

[0013] According to an embodiment of the present disclosure, for a candidate point target at any position, its position score measures the degree to which the candidate point target deviates from the z axis in the xy plane, where the x and y directions constitute the radar plane, and z represents the distance direction perpendicular to the radar.

[0014] According to an embodiment of the present disclosure, natural point targets in the current scene that can serve as calibration sources are obtained based on the evaluation results, including: obtaining a comprehensive score for each candidate point target based on the matching score and the position score; sorting the candidate point targets according to the comprehensive score; and selecting a set number of candidate point targets with the highest ranking as natural point targets that can serve as calibration sources.

[0015] Another aspect of the present disclosure provides an application of the above-mentioned millimeter-wave radar calibration method based on natural point target search in a radar calibration project, wherein the radar includes a MIMO radar, a synthetic aperture radar, and a physiological perception radar; the calibration project includes at least one of a physical array calibration, a spatial synthetic array calibration, and a temporal synthetic array calibration.

[0016] It can be seen from the above technical solutions that the millimeter-wave radar calibration method based on natural point target search disclosed in the present invention can not only find hidden point targets from natural scenes, but also use these point targets to realize automatic calibration of physical arrays and synthetic aperture arrays, effectively compensate for antenna manufacturing errors and phase errors caused by radar motion, and improve radar perception and imaging performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 The present invention is a flowchart of a millimeter wave radar calibration method based on natural point target search according to an embodiment of the present invention.

[0018] Figure 2 A schematic diagram of the principle of a millimeter wave radar calibration method based on natural point target search according to an embodiment of the present disclosure.

[0019] Figure 3 This is a schematic diagram of point targets that can be used as calibration sources found in different scenarios by the millimeter wave radar calibration method based on natural point target search in an embodiment of the present disclosure.

[0020] Figure 4 Schematic diagram of the comparison of phase curves of point targets, traditional strong reflection targets and ideal point targets found by the millimeter wave radar calibration method based on natural point target search in an embodiment of the present disclosure.

[0021] Figure 5 This is a schematic diagram of comparing the phase errors of a radar before and after calibration using a millimeter-wave radar calibration method based on natural point target search in an embodiment of the present disclosure.

[0022] Figure 6 It is a schematic diagram comparing the imaging results of the millimeter wave radar calibration method based on natural point target search according to an embodiment of the present disclosure and the calibration method based on a traditional artificial corner reflector.

[0023] Figure 7 A schematic diagram comparing whether the millimeter wave radar calibration method based on natural point target search in an embodiment of the present disclosure performs respiratory motion compensation. DETAILED DESCRIPTION

[0024] The present disclosure provides a millimeter-wave radar calibration method based on natural point target search, which can automatically discover point targets in a scene that can be used for calibration, and use these point targets to complete fully automatic millimeter-wave radar calibration.

[0025] The key to millimeter-wave radar phase calibration is to have a reference target that meets the reflection characteristics of a point target. This reference target can make the echo phase depend only on the phase error of the antenna itself and is independent of other factors such as the shape of the target, thereby achieving phase calibration. This is also the essential reason why the traditional calibration method requires the placement of a corner reflector as a reference target. Based on the above key findings of the inventors of the present disclosure, the present disclosure first mathematically models the point target that can be used as a reference object and derives its two properties: (1) Each point target can be regarded as a virtual isolated point (VIP); (2) The characteristics of the point target exhibit scene invariance (SI). Using these properties, the present disclosure first simulates virtual isolated points at different locations and establishes a fingerprint database, and then proposes a fingerprint matching technology for point targets, which can identify point targets from raw radar measurement data without any manual labeling.

[0026] The present disclosure performs well in discovering and locating point targets in natural scenes. In static scenes, despite the coarse range resolution (0.06m) and angle resolution (1°) of the radar, the present disclosure can still achieve accurate positioning. In dynamic scenes, the phase trajectory similarity between the discovered point target and the ideal point target is 14.3 times higher than that of the traditional strong reflective target.

[0027] In practical applications, the present disclosure can realize automatic calibration of physical arrays and synthetic arrays (such as motion compensation) anytime and anywhere. For the calibration of physical arrays of commercial radars, the present disclosure realizes fully automatic calibration. For handheld and vehicle-mounted millimeter-wave imaging (which can be regarded as a spatial synthetic array calibration project), the present disclosure can achieve high-quality millimeter-wave handheld imaging without corner reflectors. In addition, the present disclosure can also compensate for motion in respiratory perception (which can be regarded as a temporal synthetic array calibration project). From physical array calibration to synthetic array calibration, the present disclosure verifies the feasibility of replacing traditional corner reflectors with abundant natural point targets in the scene.

[0028] By demonstrating that the environment itself can provide a rich source of calibration in the form of natural point targets, the present disclosure breaks the reliance on manual calibration and enables fully automatic implementation anytime and anywhere in a variety of applications.

[0029] In order to make the objectives, technical solutions and advantages of the present disclosure more clearly understood, the present disclosure is further described in detail below in combination with specific embodiments and with reference to the accompanying drawings.

[0030] In an embodiment of the present disclosure, a millimeter wave radar calibration method based on natural point target search is provided. Figure 1 and Figure 2 As shown, the calibration method includes:

[0031] Operation S1: defining an ideal point target that can be used as a calibration source and performing mathematical modeling;

[0032] Operation S2: establishing a feature database of an ideal point target;

[0033] Operation S3: evaluating the scene measured radar echo data according to the feature database;

[0034] Operation S4: obtaining a natural point target in the current scene that can be used as a calibration source according to the evaluation result; and

[0035] Operation S5: Calibrate and evaluate using the obtained natural point target.

[0036] According to the embodiment of the present disclosure, first, an ideal point target is defined as a reflector that satisfies the following conditions: any point on the reflector has the same radar cross-sectional area relative to different antennas and does not change due to differences in the shape of the reflector itself.

[0037] According to an embodiment of the present disclosure, establishing a feature database of an ideal point target includes:

[0038] Simulate and generate intermediate frequency echo data of a single ideal point target at different positions;

[0039] Performing distance estimation and spatial angle estimation on the intermediate frequency echo data of each ideal point target obtained, so as to obtain the spatial spectrum corresponding to different distance units of each ideal point target; and

[0040] A feature database of an ideal point target is established according to the spatial spectrum.

[0041] Specifically, according to the requirements of the calibration task for the reference target, a mathematical model is performed on the point target that can be used as the calibration source. Assume that a space Ω contains each point P ω =( x , y , z ) is a reflector. n =( x´ , y´ , z´ ) n The millimeter wave measurement value of each antenna can be expressed as:

[0042] .

[0043] in, and Respectively represent n An antenna and a reflector ω The position vector of . express and The propagation distance between ) represents attenuation, Φ( ) represents the phase delay introduced by distance. is the point on the reflector relative to the nth antenna ω Radar cross section (RCS), F( ) represents the mapping relationship between propagation distance, RCS and mmWave measurement values. The key feature of a point target is that it can modulate the antenna signal independently of its shape, which means that a point target should always have Based on this analysis, the present disclosure can formally define a point target as a reflector that satisfies the following conditions:

[0044] .

[0045] in, Is the antenna location function, so the above formula represents the value of any point on the reflector. ω, Relative to The partial derivative of ) are all 0. This definition reveals the essential properties of point targets: the antenna signal under a point target depends only on the position of the antenna relative to the point target, and has nothing to do with the shape of the point target itself. Through the above analysis, the present disclosure obtains two key properties of point targets: (1) Each point target can be regarded as a virtual isolated point (VIP) because the signals from different antennas depend only on the distance between the antenna and the point target. This property makes the point target appear as an isolated point in space. (2) Point targets have scene invariance (SI). Regardless of the surrounding environment, the conditions in the above formula always hold. This means that an ideal point target has a fixed set of feature data, similar to unique "fingerprint" data, which is independent of the surrounding environment. These two key properties show that by establishing a feature database based on the virtual isolated point property and matching the radar measurements in different scenes with the database based on the scene invariance property of the point target, natural point targets can be automatically discovered.

[0046] According to an embodiment of the present disclosure, distance estimation and spatial angle estimation are performed on the intermediate frequency echo data of each ideal point target to obtain the spatial spectrum corresponding to different distance units of each ideal point target, including: utilizing the scene invariance of the ideal point target, converting the echo data from the antenna-frequency domain to the distance-angle domain, and extracting the spatial spectrum at different distance units as feature data.

[0047] In order to establish a feature database of point targets, the present disclosure first creates a virtual isolated point simulator for generating points at different locations. A single point target ω The intermediate frequency (IF) echo signal s(n,k) The simulator takes into account the FMCW radar parameters, MIMO antenna configuration, and the propagation delay difference between the radar antenna and the target. For the n The antenna and k sampling points, intermediate frequency echo signal s ( n , k ) can be expressed as:

[0048] .

[0049] in A represents the amplitude, f c is the carrier frequency, S is the frequency modulation slope, T s is the sampling interval. By simulating point targets at various locations, the present disclosure obtains a set of diverse VIP echoes as the basis of the fingerprint database. Afterwards, unique features that can characterize reflections that satisfy backscattering and isolation characteristics are extracted from the simulated echoes. Utilizing the scene invariance of point targets, a fingerprint feature extractor is designed to convert the echo from the antenna-frequency domain to the distance-angle domain, and extract the spatial spectrum of the corresponding distance unit as the fingerprint feature. This method can capture the essential feature data of an ideal point target while minimizing the impact of the surrounding environment.

[0050] According to an embodiment of the present disclosure, evaluating the scene measured radar echo data according to the feature database includes:

[0051] Selecting a set of candidate point targets based on the scene measured radar echo data; and

[0052] The echo data of the candidate point targets in the candidate point target set are matched and evaluated with the feature data of the ideal point targets through a comprehensive scoring mechanism to obtain a scoring result.

[0053] Optionally, the comprehensive scoring mechanism includes:

[0054] A matching score reflecting the degree of matching between the candidate point target and the ideal point target features; and

[0055] The position score that reflects the spatial position of the candidate point target relative to the antenna.

[0056] According to an embodiment of the present disclosure, for each point target in the candidate point target set, the cosine similarity between the spatial spectrum of each point target and the feature data of the ideal point target is calculated as a matching score.

[0057] According to an embodiment of the present disclosure, for a candidate point target at any position, its position score measures the degree to which the candidate point target deviates from the z axis in the xy plane, where the x and y directions constitute the radar plane, and z represents the distance direction perpendicular to the radar.

[0058] Specifically, the process of discovering natural point targets in complex environments aims to identify a set of candidate point targets that exhibit ideal backscattering and isolation properties. In order to select the candidate point target set To select the best point target or ideal point target, the present disclosure introduces a comprehensive scoring mechanism. This mechanism evaluates and ranks each candidate point target based on two key factors: the match between the candidate point target and the ideal point target fingerprint, and the spatial position of the candidate point target relative to the radar. Measures the consistency between the candidate point target and the ideal point target features. For each point in the candidate point target set , the present disclosure calculates its spatial spectrum and the corresponding ideal point target feature data The cosine similarity between is used as the matching score:

[0059] .

[0060] Position score Measure the spatial position of the candidate point target relative to the radar. For each candidate point target , whose position score is defined as:

[0061] .

[0062] in is a small positive number, such as 0< <1e-7 to avoid division by zero. The position score measures how far the candidate point target deviates from the Z axis (radar line of sight) in the xy plane. x i ,y i ,z i is the coordinate of the candidate point target in space, where the x and y directions constitute the radar plane, and z represents the distance direction perpendicular to the radar; the higher the score, the closer the candidate point target is to the ideal position facing the radar. Based on the matching score and the position score, the present disclosure can obtain a comprehensive score for the candidate point target. :

[0063] .

[0064] In the embodiment of the present disclosure, a natural point target that can be used as a calibration source in the current scene is obtained according to the evaluation result, including:

[0065] Obtain the comprehensive score of each candidate point target based on the matching score and position score;

[0066] Sort each candidate point target according to the comprehensive score; and

[0067] A set number of candidate point targets with the highest ranking are selected as natural point targets that can be used as calibration sources.

[0068] With the comprehensive scores of the candidate point targets, selecting the best point target becomes simple and straightforward. All candidate point targets are sorted according to the comprehensive scores and the top candidate point targets with the highest scores are selected. K point as the optimal point target. Here, K It can be flexibly set according to the specific perception task and scene characteristics, and can usually be set to 30.

[0069] The present disclosure also provides an application of the millimeter-wave radar calibration method based on natural point target search as described above in radar calibration projects, such as MIMO radar, synthetic aperture radar, and physiological perception radar calibration projects. The calibration projects may include at least one of physical array calibration, spatial synthetic array calibration, and temporal synthetic array calibration.

[0070] This paper verifies the accuracy of the natural point targets found by the above method that can be used as calibration sources in static scenes. Experiments are conducted in different static scenes, including outdoor open areas, parking lots, and complex indoor environments such as laboratories and halls. Corner reflectors are placed in the scene as real point targets, and measurement data is collected using a commercial cascade radar. The position error of the point target found by this paper is compared with the real point target (corner reflector) to evaluate the positioning accuracy.

[0071] like Figure 3 The results shown show that in all scenarios, the present disclosure can effectively locate the corner reflector, and the point targets found (orange point group) are well consistent with the real 3D scene captured by the lidar (blue point group). This example confirms the accuracy of the present disclosure in locating point targets, despite the limited range resolution (6 cm) and angular resolution ( ) is relatively rough.

[0072] The results under the above static scenes show that the present disclosure not only identifies artificially placed corner reflectors, but also discovers point targets that exist naturally in common object structures, such as corners formed by the edges of trees and vehicles, the edges of doors and ladders, and pillars and glass walls. The ubiquity of these natural point targets in everyday scenes expands the scope of application of the present disclosure. By utilizing these existing point target structures, electromagnetic scattering characteristics similar to those of corner reflectors can be obtained without the use of artificial reflectors. This discovery highlights the importance and potential of the present disclosure in enhancing radar perception capabilities in actual scenes, such as radar parameter calibration and motion compensation demonstrated in subsequent embodiments.

[0073] Furthermore, the disclosed embodiment also verifies the accuracy of the natural point targets found as calibration sources in dynamic scenarios. Experiments are conducted with the aid of a track platform to simulate radar motion scenarios. The actual phase trajectory of the candidate point target is compared with the theoretically expected phase trajectory to evaluate the accuracy of dynamic point target search. By comparing the actual phase trajectory of the candidate point target with the theoretically expected phase trajectory, the accuracy of dynamic point target search can be evaluated.

[0074] like Figure 4 As shown, the phase curves of the natural point targets (T3 and T4) found are highly similar to the phase curve of the ideal point target, while the phase curves of the strong reflection targets (T1 and T2) are significantly different. This shows that the method of the present invention can effectively find point targets, while using strong targets as approximations is not reliable. In addition, the point targets found show a high degree of similarity with the ideal point targets within a local range (within 20 mm), and maintain a high consistency even within 100 mm. In contrast, the phase trajectory of the strongly reflecting target deviates significantly from that of the ideal point target. Specifically, the normalized phase is used to measure local similarity. It has been verified that within 20 mm, the normalized phase deviation of the natural point target is 0.6% and 0.3%, within 100 mm, it is 0.7% and 1.1%, while the normalized phase deviation of the strong reflection point is within 20 mm. and , within 100mm and The normalized phase deviation of a point target over the entire 380 mm range is and , and the normalized phase deviation of the strong reflection point is and It can be seen that the local similarity of the natural point targets found by the present disclosure is 14.3 times higher than that of the traditional strong reflection points. This embodiment shows that the phase trajectories of the point targets found by the present disclosure are highly similar to the ideal points, confirming that they are real point targets and verifying the effectiveness of the present disclosure.

[0075] Furthermore, the disclosed embodiment also verifies the calibration of the physical array using the natural point target found as a calibration source. Physical array calibration is a crucial step in MIMO radar signal processing, which aims to estimate and compensate for the amplitude and phase errors between different antenna channels, thereby improving the radar's angle estimation and imaging performance. The basic principle of calibration is to estimate the frequency and phase errors of each channel using the echo of a reference target at a known distance.

[0076] Assume that the original ADC data received by the radar is S ,but:

[0077] .

[0078] in, , and Respectively represent the number of data frames, the number of sampling points per frame and the number of antenna channels, represents the complex space. The calibration process starts with a preprocessing step where the raw ADC data is loaded into this three-dimensional matrix. The phase deviation between different channels is estimated using the peak range cells and their corresponding complex values ​​obtained from the range fast Fourier transform results. For each channel, the average peak value of all frames is first calculated. :

[0079] .

[0080] in It is i Frame j Then, by taking the inverse of the average peak value and extracting the phase information, we obtain the phase calibration vector :

[0081] .

[0082] in represents the complex angle operator. An important improvement of the present disclosure is the ability to accommodate point targets at different angles. This is critical in real-world scenarios, as an ideal point target directly facing the radar phase center may not always be available. To achieve this, the present disclosure introduces a phase calibration scheme that incorporates the theoretical phase gradient in the antenna array. Assume θ Represents the angle of the point target relative to the center of the radar beam, d Represents the spacing between antenna elements. Theoretical phase gradient It can be calculated as:

[0083] .

[0084] Where λ is the wavelength of the radar signal. The present disclosure then compares the measured phase value of the point target in the antenna array with the expected phase gradient. The phase calibration vector is refined as ,but:

[0085] .

[0086] in represents the refined phase calibration vector taking into account the point target angle. Indicates j The accumulated theoretical phase of each channel is obtained by applying the estimated frequency and phase calibration vectors to the raw ADC data. The calibration process can be expressed as:

[0087] .

[0088] in, Indicates i Frame, j Channels and k The ADC sample value after calibration indexed by the sampling point, represents the exponential function, and the positive form j represents the imaginary unit. represents the exponential function, and the positive form j represents the imaginary unit.

[0089] The present disclosure demonstrates its effectiveness in fully automatic calibration of widely used commercial MIMO radars. For 77GHz millimeter-wave radars, ideally the antenna spacing should be half a wavelength, or 1.95mm, but errors are inevitably introduced in actual manufacturing. Traditionally, the calibration process requires the use of corner reflectors to collect echoes before the radar is used for the first time, and to estimate the phase error of each antenna based on the echoes. However, in actual use, due to factors such as vibration and temperature changes, the phase error of the antenna may change and require recalibration. The present disclosure provides a new online calibration method for millimeter-wave radars, which can automatically find point targets in complex scenes, such as without the use of corner reflectors. Figure 5As shown. Specifically, the radar can be calibrated using two different principles: traditional manual calibration using corner reflectors and automatic calibration using natural point targets found using the method proposed in the present disclosure. This step results in two sets of phase error vectors, one from the corner reflector and the other from the natural point target. The experimental results demonstrate the effectiveness of the present disclosure in calibrating the radar using natural point targets. Before calibration, the phase estimation error showed significant variation, ranging from -1.76 rad to 1.43 rad. In contrast, after applying the calibration method of the present disclosure, the phase estimation error was greatly reduced and became more consistent, ranging from -0.29 rad to 0.46 rad, with the error variance reduced by 86.6% and the MAE reduced by 35%, highlighting the effectiveness and reliability of the present disclosure in calibrating millimeter-wave radar systems.

[0090] The disclosed embodiment also verifies the calibration of the spatial synthetic array using the natural point target found as a calibration source. Synthetic aperture radar (SAR) imaging is a technology that uses radar to acquire target echo signals during motion and achieves high-resolution imaging through signal processing methods. However, due to the motion trajectory error of the handheld device or carrier, the obtained intermediate frequency echo signal cannot be directly used for effective imaging. Therefore, it is necessary to use natural point targets to perform phase compensation and calibration on SAR imaging. First, use the present disclosure to identify the point target P(x', y', z') from the echo signal. Then, set and represent the theoretical phase history and measured phase history at the point target P, respectively, where n represents the pulse index. The phase error at point P It can be obtained by:

[0091] .

[0092] Assuming that the phase error at point P is mainly caused by the translation error of the radar, then It can be approximated as the phase error introduced by radar motion. Therefore, the compensated intermediate frequency echo signal It can be expressed as:

[0093] .

[0094] in, represents the original intermediate frequency signal, k Indicates the fast time sample index.

[0095] Figure 6The experimental results prove the effectiveness of the present disclosure in improving the radar imaging performance by using point targets naturally existing in the environment. It can be seen that the use of natural point targets can effectively compensate for phase errors, whether for handheld imaging or vehicle-mounted imaging. In addition, the imaging results obtained using natural point targets are comparable to those obtained using corner reflectors. This shows that the present disclosure can effectively and fully automatically compensate for the phase error of near-field synthetic aperture radar imaging.

[0096] The disclosed embodiment also verifies that the natural point targets found can be used as calibration sources for radar calibration of time-synthesized arrays. In radar perception applications, such as respiratory monitoring, the respiratory signal may be buried due to the movement of the radar, and the waveform obtained by direct processing cannot reflect the actual respiratory situation. In order to eliminate the phase error caused by radar movement, the natural point targets in the environment found by the present disclosure can be used to calibrate the radar of the time-synthesized array, perform least squares estimation and phase compensation, effectively calibrate the time-synthesized array radar, eliminate the phase error caused by radar movement, and achieve accurate respiratory monitoring in mobile scenarios. This fully automatic phase compensation method greatly improves the robustness and practicality of radar perception and expands the application scope of radar in mobile applications. Figure 7 The experimental results are shown, and it can be seen that the method proposed in the present disclosure can effectively find point targets in natural scenes that can be used to calibrate the time synthesis array, so that it can be used for new radar applications such as indoor health detection and human body radar.

[0097] Based on the application and verification of the above calibration methods, it can be seen that the millimeter-wave radar calibration method based on natural point target search proposed in the present invention can not only find hidden point targets from natural scenes, but also use these point targets to realize automatic calibration of physical arrays and synthetic aperture arrays, avoiding professional operators to place corner reflectors for regular calibration, effectively compensating for antenna manufacturing errors and phase errors caused by radar movement, and improving radar perception and imaging performance.

[0098] So far, the embodiments of the present disclosure have been described in detail in conjunction with the accompanying drawings. It should be noted that the implementation methods not shown or described in the drawings or the body of the specification are all forms known to ordinary technicians in the relevant technical field and are not described in detail. In addition, the above definitions of each element and method are not limited to the various specific structures, shapes or methods mentioned in the embodiments, and ordinary technicians in the field can simply change or replace them.

[0099] In addition, unless the steps are specifically described or must occur in sequence, the order of the above steps is not limited to the above list, and can be changed or rearranged according to the required design. And the above embodiments can be mixed and matched with each other or with other embodiments based on design and reliability considerations, that is, the technical features in different embodiments can be freely combined to form more embodiments.

[0100] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present disclosure. It should be understood that the above description is only a specific embodiment of the present disclosure and is not intended to limit the present disclosure. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present disclosure should be included in the protection scope of the present disclosure.

Claims

1. A millimeter wave radar calibration method based on natural point target search, comprising: Define and mathematically model an ideal point target that can serve as a calibration source; Establish a feature database of ideal point targets; The scene measured radar echo data is evaluated according to the feature database, including selecting a candidate point target set according to the scene measured radar echo data; and matching and evaluating the echo data of the candidate point targets in the candidate point target set with the feature data of the ideal point target through a comprehensive scoring mechanism to obtain a scoring result; the comprehensive scoring mechanism includes a matching score reflecting the degree of matching between the features of the candidate point target and the ideal point target, and a position score reflecting the spatial position of the candidate point target relative to the antenna; Obtaining natural point targets in the current scene that can be used as calibration sources according to the evaluation results, including obtaining a comprehensive score of each candidate point target according to the matching score and the position score; sorting each candidate point target according to the comprehensive score; and selecting a set number of candidate point targets with the highest ranking as natural point targets that can be used as calibration sources; as well as Calibration is performed using the obtained natural point targets and evaluation is performed.

2. According to the millimeter wave radar calibration method of claim 1, the ideal point target is defined as a reflector that meets the following conditions: Any point on the reflector has the same radar cross-sectional area relative to different antennas and will not change due to differences in the shape of the reflector itself.

3. According to the millimeter wave radar calibration method of claim 1, establishing a feature database of an ideal point target comprises: Simulate and generate intermediate frequency echo data of a single ideal point target at different positions; The obtained intermediate frequency echo data of each ideal point target are used for distance estimation and spatial angle estimation to obtain the spatial spectrum of each ideal point target corresponding to different distance units; as well as A feature database of an ideal point target is established according to the spatial spectrum.

4. The millimeter wave radar calibration method according to claim 3, wherein the step of performing distance estimation and spatial angle estimation on the intermediate frequency echo data of each ideal point target to obtain a spatial spectrum corresponding to different distance units of each ideal point target comprises: By utilizing the scene invariance of ideal point targets, the echo data is converted from the antenna-frequency domain to the range-angle domain, and the spatial spectrum at different range units is extracted as feature data.

5. The millimeter wave radar calibration method according to claim 1, for each point target in the candidate point target set, the cosine similarity between the spatial spectrum of each point target and the feature data of the ideal point target is calculated as the matching score.

6. According to the millimeter wave radar calibration method of claim 5, for a candidate point target at any position, its position score measures the degree to which the candidate point target deviates from the z axis in the xy plane, where the x and y directions constitute the radar plane, and z represents the distance direction perpendicular to the radar.

7. An application of the millimeter-wave radar calibration method based on natural point target search as described in any one of claims 1 to 6 in a radar calibration project, wherein the radar includes a MIMO radar, a synthetic aperture radar, and a physiological perception radar; the calibration project includes at least one of a physical array calibration, a spatial synthetic array calibration, and a temporal synthetic array calibration.