Radar ranging method for low-altitude-low-speed-small target, computer equipment and storage medium

Abstract

The invention provides a radar ranging method for a low-altitude-low-speed-small target. The method is applied to a multi-radar (2-3) system, and the radars are respectively located at different spatial positions of a machine body. According to the model, a machine body and a radar can be abstracted into lattice points in a space dot matrix, the lattice points in the dot matrix can be correlated from an originally irrelevant state, so that symmetry transformation is carried out on the lattice points by utilizing a group theory in a Hilbert space, finally, signal waves and noise waves are separated, and according to the signal waves obtained at the moment, the noise waves are extracted. And high-precision radar ranging can be realized. The dot matrix model shows excellent accuracy and stability in the process of processing multi-radar ranging, the accuracy is almost not affected by spatial distribution changes or radar ranging error changes, and the characteristics enable the dot matrix model to have potential application prospects in the field of low-slow-small target airborne radars and the like.

Description

Technical Field

[0001] The present invention relates to the field of radar detection technology, and in particular to a radar ranging method for a low-altitude, slow-speed, small-sized target, a computer device and a storage medium. Background Art

[0002] In the field of radar detection, the detection of low-altitude, slow-speed, small (LSS) targets is a hot and difficult issue in airborne radar research. In particular, the rapid development of artificial intelligence in recent years has led to an increasing demand for high-precision radar detection of low-slow-small targets such as self-driving cars and drones. For these low-slow-small targets, due to their lower altitude, ground wave interference is more likely to affect the signal wave received by the radar, thereby reducing the signal-clutter-noise ratio (SCR). In addition, since the target reflection cross-section is small and the spectrum is almost submerged in the clutter, and the slow speed causes the spectrum of the clutter to become wider per unit time, making it easier for the signal wave unit and the clutter unit to mix with each other, it becomes very difficult to accurately extract the signal wave from the sensor data.

[0003] In order to solve this problem, researchers usually use noise suppression methods to improve the wave-to-noise ratio, such as Kalman filtering algorithm, dimensionality reduction method, sparse matrix, etc. Although these methods have made some progress, when facing LSS targets with low wave-to-noise ratio and difficult Doppler distinction between targets and clutter, the clutter suppression method itself has limited improvement on the wave-to-noise ratio, making the existing suppression decision methods have the characteristics of slow speed, unstable effect, and lack of intelligence. In addition, deep learning has also made some progress in radar detection, such as support vector machine, high-order fusion, neural network, etc., but the performance improvement is still limited. The method of using micro-Doppler effect to improve the detection accuracy of LSS targets has attracted much attention in recent years. This method helps to distinguish specific LSS targets and has a certain effect on extracting signal waves. Unfortunately, the micro-Doppler effect technology that currently uses the micro-motion characteristics of LSS targets for analysis has a large application range limitation.

[0004] In order to successfully separate the signal wave from the clutter, we propose a new airborne radar detection method based on the lattice model. This method uses a multi-radar system and gives the multi-radar system symmetry through the lattice model, so as to use the unique group theory to perform symmetry transformation on the multi-radar data, and finally separate the signal wave from the noise. Summary of the invention

[0005] The purpose of the present invention is to propose a radar ranging method for low altitude, slow speed and small targets, aiming to solve the problems of low detection accuracy and poor robustness of traditional low altitude, slow speed and small (low-slow-small LSS) targets, so as to enable intelligent machinery (such as robots, drones, self-driving cars and self-driving ships, etc.) to obtain high-precision detection target data during mission execution.

[0006] To achieve the above object, the technical solution of the present invention is:

[0007] A radar ranging method for a low-altitude, slow-speed, small target, specifically comprises the following steps:

[0008] S1. Based on a lattice model framework, the intelligent machine is abstracted into a space lattice with spatial distribution symmetry, wherein the intelligent machine includes a body and a plurality of radars distributed on the body;

[0009] S2, obtaining ranging data sets with different spatial distribution characteristics measured by multiple radars, and the ranging data sets have spatial symmetry;

[0010] S3. The distance measurement data set uses the symmetry of the space lattice to perform a linear coupling-decoupling process in the Hilbert space, separates the distance measurement data set into an eigenstate representing an accurate value and a multiplet state representing noise, and obtains a distance measurement result according to the eigenstate.

[0011] Preferably, the low-altitude, slow-speed, and small targets include self-driving cars, robots, and drones.

[0012] Preferably, the types of spatial distribution symmetry include body-centered symmetry, translational symmetry, and spatial inversion.

[0013] Preferably, the process of abstracting the intelligent machine into a spatial lattice with spatial distribution symmetry based on the lattice model framework specifically comprises the following steps:

[0014] The space group structure of the body is determined based on the space group theory in crystal physics; the body is abstracted into a corresponding shape; a coordinate system is established with the center of mass of the body as the origin; the body and multiple radars are abstracted into a space lattice with spatial distribution symmetry.

[0015] Preferably, S3 is specifically:

[0016] The radar's ranging value data is divided into eigenstates and multiplets:

[0017] {X j}={X j0}+Σ{ΔX j} (2)

[0018] Among them, {X j} is the ranging value data of a certain radar in the space dot matrix, {X j0} is the eigenstate, Σ{ΔX j} is a multiple state;

[0019] The generic higher-order term method is used to expand the first-order terms of the multiple states and transform equation (2) into equation (3):

[0020] {X j}={X j0}+{ΔX j} (3)

[0021] Considering that the eigenstate has central symmetry and the multiplet has translational symmetry; for {X j} for linear changes:

[0022]

[0023] Among them, {Y j}={Y j0}+{ΔY j} is the distance measurement data of another radar in the space dot matrix, is the central inversion symmetry operation, It is a translation symmetry operation. Through the symmetry operation, the ranging value data of N radars are coupled and decoupled, and finally the average value of the eigenstate of the N radar ranging value data is obtained:

[0024]

[0025] The average value {D0} of the eigenstates of the N radar ranging value data is a ranging result that does not contain noise.

[0026] Preferably, the machine body is a car, and the car is abstracted into a space lattice with spatial distribution symmetry based on the lattice model framework as follows:

[0027] Based on the space group theory in crystal physics, the space group structure of the car is determined to be a tetragonal crystal structure, and the physical car is abstracted into a tetrahedron shape; with the center of mass of the car as the origin O, an O-xyz coordinate system is established to obtain the tetrahedron lattice and determine the positions of multiple radars in the tetrahedron lattice, thus obtaining a space lattice with spatial distribution symmetry.

[0028] Preferably, the vehicle is equipped with 2 or 3 radars.

[0029] Preferably, the calculation of the ranging result obtained based on the ranging data sets of the two radars is specifically as follows:

[0030] The range data of radar S1 is expressed as {X j}, the ranging value data of radar S2 is expressed as {Y j}, the ranging data of radar S1 and radar S2 are divided into eigenstates and multiplets, and the first-order terms of the multiplets are expanded using the generic high-order term method:

[0031] {X j}={X j0}+{ΔX j} (4)

[0032] {Y j}={Y j0}+{ΔY j} (5)

[0033] The data of radar S1 and radar S2 are coupled by using the symmetry operation in the lattice model. Since the eigenstate has central symmetry and the multiplet has translational symmetry, the symmetry transformation formula is:

[0034]

[0035] in, is the central inversion symmetry operation, is a translation symmetry operation; solving equations (4)-(7) yields the following equation:

[0036]

[0037] Wherein, {D0} represents the average value of the eigenstates of radar S1 and radar S2, and {D0} is the ranging result without noise.

[0038] A computer device comprises a processor, a memory and a computer program stored in the memory, wherein when the processor executes the computer program, the steps in any one of the above-mentioned radar ranging methods are specifically performed.

[0039] A computer-readable storage medium having a computer program stored thereon, characterized in that when the program is executed by a processor, any of the above-mentioned radar ranging methods is implemented.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] High accuracy: The radar ranging result of the traditional classic Kalman filter algorithm is more than 3 times higher, reaching the 0.20m level;

[0042] Low cost: This method does not require expensive precision radar equipment. The radar commonly used in intelligent machinery can also achieve centimeter-level high-precision ranging;

[0043] Broad application scenarios: It can be applied to all application scenarios that require radar, including but not limited to robots, drones, self-driving cars and ships. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a schematic diagram of modeling the lattice model of the present invention;

[0045] Figure 2 An automatic driving simulation platform based on Matlab in one embodiment of the present invention;

[0046] Figure 3 1 is a diagram showing the ranging situation of three non-identical radars in a typical position distribution according to an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The technical solution of the present invention is described in detail below in conjunction with the accompanying drawings.

[0048] The present invention proposes a radar ranging method for low altitude, slow speed and small targets, as shown in the attached Figure 1 As shown in the figure, a new lattice model is adopted. First, the intelligent machine (including the body and radar) is abstracted into a space lattice based on the lattice model framework. Then the radar system measures the distance of the target to obtain a ranging data set with different spatial distribution characteristics. Then, the data set uses the symmetry of the space lattice and performs a linear coupling-decoupling process in the Hilbert space. The Hilbert space has complete inner product operations and can be operated by linear algorithms. Finally, the data set can be separated into eigenstates representing precise values ​​and multiple states representing noise. According to the results of the eigenstates, accurate ranging can be achieved.

[0049] exist Figure 1 In , the modeling rules of the lattice model are as follows:

[0050] (1) Using the space group theory in crystal physics (this theory is recorded in various books related to crystal physics), find the space group structure of the car in the corresponding figure. In this figure, the space group structure is a tetragonal crystal structure. Therefore, Figure 1 The physical car in (a) will be transformed into Figure 1 Tetrahedral shape in (b).

[0051] (2) Figure 1 In (b), the O-xyz coordinate system is established, where the origin O is the center of mass of the car, but not necessarily the center of the tetrahedron. The two are the same only when the center of mass of the car coincides with the center of the tetrahedron.

[0052] (3) Figure 1 In (b), s1 and s2 are the positions of the two radars used in this implementation case in the tetrahedral lattice.

[0053] After the car is abstracted into a tetrahedron according to the dot matrix model, the radar changes from its original state of no symmetry to a state of symmetry, so that calculations can be performed using the following formula.

[0054]

[0055] Formula (1) can expand the data set collected by the radar into a determinant using a linear operator. The data set is used instead of data because the data collected by the radar includes different types of data, such as the radar range value, phase, temperature, time, etc. in this embodiment. Among them, is a linear operator, ν is the number of combinations, k is the proportionality coefficient, {X ν} is defined as a function basis or basis function, {X ij} is the data set collected by the radar, and i and j represent the number of rows and columns that make up the data set.

[0056] Dataset {X ij It consists of two parts: the first part represents the true value, which is called the eigenstate in theoretical physics; the other part is caused by system errors, measurement errors, various noises and other factors, which are called multiple states. Usually, these two parts are mixed together in the radar ranging results, especially for the detection of low-slow-small targets, the eigenstate is completely submerged in the multiple states and is difficult to extract. Therefore, the main task of data processing is to distinguish these two parts.

[0057] For the data set {X ij}, we can expand it as follows:

[0058] {X ij}={X ij0}+Σ{ΔX ij} (2)

[0059] Formula (2) divides the data set into two parts, where {X ij0} is the eigenstate, Σ{ΔX ij} is a multiplet. However, the multiplets in this equation are very complex and the solution process is difficult to converge. Therefore, in this case, we use the generic "higher-order term" method to expand the multiplets into first-order terms, so that formula (2) becomes formula (3):

[0060] {X ij}={X ij0}+{ΔX ij} (3)

[0061] The above formula still cannot solve the analytical values ​​of eigenstates and multiplets, but the lattice model gives these data spatial symmetry. Therefore, the group theory in linear space (Hilbert space) can be used to change these data to obtain analytical solutions.

[0062] In order to simplify the calculation process of the equation, we only consider the radar ranging value in the data set, and other data such as phase and temperature are not included in the data set. In this way, the radar ranging value is only a single row or column of data. In other words, i in equation (3) can be simplified and removed to become {X j}.

[0063] In this case, we select two radars for display, and more radars will be displayed in the same way. j}, we represent the dataset of S1 as {X j}, while the S2 dataset uses {Y j}, then equation (3) becomes two equations, namely:

[0064] {X j}={X j0}+{ΔX j} (4)

[0065] {Y j}={Y j0}+{ΔY j} (5)

[0066] We use the symmetry operation in the lattice model to couple the data of the S1 radar and the S2 radar. Considering that the eigenstate has central symmetry and the multiplet has translational symmetry, the symmetry transformation formula is:

[0067]

[0068] in, is the central inversion symmetry operation, is a translation symmetry operation. Solving equations (4)-7), we can obtain the following equation:

[0069]

[0070] Where {D0} represents the average value of the eigenstates of the S1 radar and S2 radar ranging. For N radars, equation (10) will be expanded to:

[0071]

[0072] From then on, the signal waves and noise that were originally mixed together are completely separated, and {D0} is the average value of the signal waves of N radar ranging, which no longer contains noise.

[0073] The effectiveness of the algorithm is verified using the automatic driving simulation platform in Matlab. Figure 2As shown in (a) of Figure 1, we establish an O-xyz body coordinate system on the self-driving car, where the origin O

[000] is located at the center of mass. The blue car equipped with a radar detector is called the detection car, and the truck in front of it on the x-axis is called the target car. Their initial distance is 14.00m, as shown in Figure 1. Figure 2 As shown in (b), assuming that their speeds are both v x =10m / s, then its distance will remain unchanged (x=14.00m). Usually, autonomous vehicles require multiple radars to perform different tasks. For example, the front radar is used to measure the distance of the vehicle in front, the side radar is used to detect the distance of the side vehicle, and the rear radar is used to detect the distance of the rear vehicle. In this case, once a radar has a significant error or failure, the detection result in that direction will have a serious negative impact. Therefore, it is crucial to maintain two or more radars covering the same tasks to ensure the safe driving of autonomous vehicles. Considering the cost and space structure constraints, 2-3 radars are usually required to meet this requirement. In view of this, in this experiment we will use 3 radars for research and investigation. In the experiment, the ranging performance based on the lattice model under different combinations of radars was studied and compared with the results of the mean filter algorithm and the Kalman filter algorithm. In the experiment, except for the differences in the position and distance resolution deviation scores, the other parameters of all radars are consistent, which are shown in Table 1.

[0074] Table 1 Radar parameters in the experiment

[0075]

[0076] When there are three radars, we mainly study the distribution of non-identical parameter radars in some typical positions, as shown in Table 2. These typical positions are respectively distributed along the x-axis, y-axis and symmetrically along the y-axis.

[0077] Table 2 Ranging results of three non-identical radars at typical locations

[0078]

[0079] The experimental results are as follows Figure 3As shown, (ac) x-axis position distribution (

[000] ,

[100] ,

[200] ); (df) y-axis position distribution (

[000] , [00.50], [0-0.50]); (gi) symmetrical distribution along the y-axis (000], [20.50], [2-0.50]); the mean filter ranging accuracy can reach up to 0.38m; the Kalman filter accuracy can reach 0.51m; the lattice model accuracy can reach 0.19m. It can be observed that in the three typical position distributions, the lattice model can maintain a stable accuracy of about 0.20m, while the mean filter and Kalman filter show changes under different radar position distributions. In addition, under the conditions of the three radars, the Kalman filter shows the lowest ranging accuracy.

[0080] In summary, this embodiment describes in detail the physical model and mathematical representation of the lattice model, and verifies the model using the MatlabSimulink simulation system. Experimental results show that the lattice model exhibits excellent accuracy and stability in processing multi-radar ranging, and its ranging accuracy can reach 0.20m, which is more than 3 times higher than the 0.73m of the traditional Kalman filter technology; this accuracy is hardly affected by changes in spatial distribution or radar ranging errors. These characteristics make it have potential application prospects in the field of low-slow-small target airborne radar.

[0081] The above are preferred embodiments of the present invention. Any changes made according to the technical solution of the present invention, as long as the resulting functions do not exceed the scope of the technical solution of the present invention, belong to the protection scope of the present invention.

Claims

1. A radar ranging method for low altitude, slow speed and small targets, characterized in that: The specific steps include: S1. Based on a lattice model framework, the intelligent machine is abstracted into a space lattice with spatial distribution symmetry, wherein the intelligent machine includes a body and a plurality of radars distributed on the body; S2, obtaining ranging data sets with different spatial distribution characteristics measured by multiple radars, and the ranging data sets have spatial symmetry; S3. The distance measurement data set uses the symmetry of the space lattice to perform a linear coupling-decoupling process in the Hilbert space, separates the distance measurement data set into an eigenstate representing an accurate value and a multiplet state representing noise, and obtains a distance measurement result according to the eigenstate.

2. A radar ranging method for low altitude, slow speed and small targets according to claim 1, characterized in that: The low-altitude, slow-speed, and small targets include self-driving cars, robots, and drones.

3. The radar ranging method for low altitude, slow speed and small targets according to claim 1, characterized in that: The types of spatial distribution symmetry include body-centered symmetry, translational symmetry, and spatial inversion.

4. The radar ranging method for low altitude, slow speed and small targets according to claim 1, characterized in that: The method of abstracting the intelligent machinery into a spatial lattice with spatial distribution symmetry based on the lattice model framework specifically includes the following steps: The space group structure of the body is determined based on the space group theory in crystal physics; the body is abstracted into a corresponding shape; a coordinate system is established with the center of mass of the body as the origin; the body and multiple radars are abstracted into a space lattice with spatial distribution symmetry.

5. The radar ranging method for low altitude, slow speed and small targets according to claim 1, characterized in that: The S3 is specifically: The radar's ranging value data is divided into eigenstates and multiplets: {X j }={X j0 }+Σ{ΔX j } (2) Among them, {X j } is the ranging value data of a certain radar in the space dot matrix, {X j0 } is the eigenstate, Σ{ΔX j } is a multiple state; The generic higher-order term method is used to expand the first-order terms of the multiple states and transform equation (2) into equation (3): {X j }={X j0 }+{ΔX j } (3) Considering that the eigenstate has central symmetry and the multiplet has translational symmetry; for {X j } for linear changes: Among them, {Y j }={Y j0 }+{ΔY j } is the distance measurement data of another radar in the space dot matrix, is the central inversion symmetry operation, It is a translation symmetry operation. Through the symmetry operation, the ranging value data of N radars are coupled and decoupled, and finally the average value of the eigenstate of the N radar ranging value data is obtained: The average value {D0} of the eigenstates of the N radar ranging value data is a ranging result that does not contain noise.

6. A radar ranging method for low altitude, slow speed and small targets according to claim 4, characterized in that: The body is a car, and based on the lattice model framework, the car is abstracted into a spatial lattice with spatial distribution symmetry: Based on the space group theory in crystal physics, the space group structure of the car is determined to be a tetragonal crystal structure, and the physical car is abstracted into a tetrahedron shape; with the center of mass of the car as the origin O, an O-xyz coordinate system is established to obtain the tetrahedron lattice and determine the positions of multiple radars in the tetrahedron lattice, thus obtaining a space lattice with spatial distribution symmetry.

7. A radar ranging method for low altitude, slow speed and small targets according to claim 6, characterized in that: The car is equipped with 2 or 3 radars.

8. The radar ranging method for low altitude, slow speed and small targets according to claim 7, characterized in that: The calculation of the ranging results based on the ranging data sets of two radars is as follows: The range data of radar S1 is expressed as {X j }, the ranging value data of radar S2 is expressed as {Y j }, the ranging data of radar S1 and radar S2 are divided into eigenstates and multiplets, and the first-order terms of the multiplets are expanded using the generic high-order term method: {X j }={X j0 }+{ΔX j } (4) {Y j }={Y j0 }+{ΔY j } (5) The data of radar S1 and radar S2 are coupled by using the symmetry operation in the lattice model. Since the eigenstate has central symmetry and the multiplet has translational symmetry, the symmetry transformation formula is: in, is the central inversion symmetry operation, is a translation symmetry operation; solving equations (4)-(7) yields the following equation: Wherein, {D0} represents the average value of the eigenstates of radar S1 and radar S2, and {D0} is the ranging result without noise.

9. A computer device, characterized in that: The method comprises a processor, a memory and a computer program stored in the memory. When the processor executes the computer program, the method specifically performs the steps in the radar ranging method according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the radar ranging method as described in any one of claims 1 to 8 is implemented.

Images