Missile-borne weapon ground target high-resolution one-dimensional range profile identification method

A high-resolution one-dimensional range profile recognition method for ground targets of missile-borne weapons Technical field

[0001] The present invention relates to the field of radar target detection and recognition, and in particular to a high-resolution one-dimensional range image recognition method for ground targets of missile-borne weapons.

Background technique

[0002] The main purpose of the seeker's precise guidance is to accurately detect and track targets from a complex background environment and a variety of active and passive interference. In order to improve the hit rate and effective kill range of guided weapons, the carrier frequency and emission bandwidth of the active seeker are getting higher and higher to obtain high-resolution images of the detection area, including one-dimensional range profile (HRRP). ) And two-dimensional SAR (Synthetic Aperture Radar) images to achieve the purpose of target information extraction and target recognition. Two-dimensional SAR images use the relative movement between the target and the radar to obtain detailed information such as target size, shape, structure and attitude. It has a good ability to identify vehicles, ships and other targets. However, it is affected by the disturbance of the missile-borne platform. SAR imaging Faced with complex motion compensation problems, and the time-frequency domain transformation and reference signal complex multiplication in the imaging process make the imaging process very complicated, which requires high hardware resources and is difficult to meet real-time processing requirements.

Summary of the invention

[0003] The purpose of the present invention is to provide a high-resolution one-dimensional range profile recognition method for ground targets of missile-borne weapons, which solves the problems of strong interference signals, low recognition efficiency, and difficulty in precise target guidance in the complex ground clutter environment of the missile-borne platform.

[0004] In order to achieve the above objective, the present invention provides a high-resolution one-dimensional range image recognition method for ground targets of missile-borne weapons, which includes the following steps:

[0005] Step S1, extract the multi-dimensional features of the one-dimensional range profile of the target to be recognized;

[0006] Step S2, screening the multi-dimensional features of the target to be identified to obtain the preliminary screening target;

[0007] Step S3, analyze the queue characteristics of the preliminary screening target;

[0008] Step S4, establish a suspected target credibility evaluation model according to the multi-dimensional features of the target to be recognized and the preliminary screening target, and output the target credibility value as the target recognition result.

[0009] Before extracting the multi-dimensional features, the one-dimensional range profile of the target to be identified is normalized and aligned in the range.

[0010] The multi-dimensional features include: target distance dimension, target scattering point number, scattering point aggregation degree, second-order normalized center distance, third-order normalized center distance, energy extreme value, signal-to-noise ratio;

[0011] Target distance dimension: the continuous one-dimensional image length that crosses the detection threshold in the one-dimensional distance image;

[0012] The number of strong scattering points of the target: the number of the extreme value of the one-dimensional range image that exceeds the specified threshold within the target size range;

[0013] Scattering point aggregation degree: T = [square (left from j (left a (Adj tBuf, Y 3 represents this)) 1 I f shows the k-th strong scattering point amplitude, kjkR and 4 respectively represent one The starting, ending, and middle scattering point numbers of the dimensional distance image, Δ square is the interval between the i-th and i+1-th scattering points;

[0014] Second Order Return

<img file="CN113687328A_D0001.tif" />

Hand (/7) m = o,..., double-l is the average one-dimensional range profile, n is the number of the one-dimensional range profile of the same target, if N=1, it means that only a single one-dimensional range profile is recognized;

N-1 _ / M-1 _

[0015] Three-order normalized center distance: 5 = Σ (s-ten thousand lines such as / of 3)) 4 κ = ο Ν / "ο *

[0016] Energy extreme value: the maximum value of the amplitude of the target part in the one-dimensional range image;

[0017] Signal-to-noise ratio: the signal-to-noise ratio corresponding to the maximum amplitude of the target part in the one-dimensional range profile.

[0018] In the step S2, the nearest neighbor classifier is used to screen the multi-dimensional features of the target to be identified, and set the class ω. The decision function is:

X) = 1,2, ... ,Μλ ¹⁼¹ ,

[0019] 4(χ)=cut off the feet(ί = 1,2,···,υ)

[0020] The judgment rule is:

[0021] 4 "(core) = then η, (χ) (E = L2, -·, C)

[0022] Bay Ij Judge x6 3₁₁₁, Means * belongs to the 111th category, 1116 {1,2,...,c}.

[0023] In the step S3, the queue characteristics of the target include: the positional relationship between the targets, the positional relationship between the target and the queue, and the number of targets contained in the queue;

[0024] The positional relationship between the targets: if the radial distance between the centroids of the two targets is within an allowable range, the two targets are in the same queue;

[0025] The positional relationship between the target and the queue: if the distance between a target and any target in the known queue is less than the allowable radial distance range, the target is considered to be a target in the known queue;

[0026] The number of targets contained in the queue: the number of targets contained in the queue is confirmed as a valid queue if the number of targets contained in the queue is within the threshold.

[0027] In the step S3, if there are other targets around the target and simultaneously satisfy the positional relationship between the targets, the positional relationship between the targets and the queue, and the number of targets contained in the queue, the target is determined to be a queue target, and the queue information is counted.

[0028] In the step S4, the target credibility calculation formula is: "cccci, (() (Dis-minf/Jii')) ¥

[0029] Prank = ~------------------^-+Per.

, Max(Dfs)-min(Z)is) j

[0030] Among them, Per^x and Perg. To set the maximum and minimum values of reliability, generally select 0 · 99 and 0. ObDis is the feature vector similarity, Dis = Heng Erzhu JL is the default value of the multi-dimensional feature of the target one-dimensional range profile

K The multi-dimensional features of the one-dimensional range profile to be identified extracted in step S1, κ is the influence coefficient.

[0031] The target to be recognized with the highest target credibility value is taken as the confirmation target, and the position of the target is output.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] 1. The present invention can solve the problems of single target recognition method and low recognition rate of the missile-borne platform.

[0034] 2. Compared with the matching degree classification method, the support vector machine classification method, the restricted Boltzmann machine classification method, and the convolutional neural network classification method, the present invention requires fewer training samples, high data processing efficiency, and can Quickly implement engineering applications.

[0035] 3. Starting from the application requirements of precision guidance for missile-borne weapons, the present invention comprehensively considers the requirements for real-time, robustness, and accuracy of target recognition, and proposes a target recognition method based on a suspected target credibility evaluation model. Target category and similarity degree, give the strike target ranking, improve weapon guidance efficiency.

Description of the drawings

[0036] FIG. 1 is a flow chart of a high-resolution one-dimensional range image recognition method for ground targets of a missile-borne weapon of the present invention.

[0037] FIG. 2 is an HRRP image of the object 1 of the present invention from different angles.

[0038] FIG. 3 is an HRRP image of the object 2 of the present invention from different angles.

Detailed ways

[0039] Hereinafter, a preferred embodiment of the present invention will be described in detail based on FIGS. 1 to 3.

[0040] Relatively speaking, high-resolution HRRP reflects the radial distribution of target scattering points, which can characterize the structure, center distance, and transform domain information of the target, and is simpler in acquisition and processing methods. Therefore, target recognition based on HRRP has Better feasibility and practicality. Based on the above, it is really necessary to develop a high-resolution one-dimensional range profile recognition method for ground targets of missile-borne weapons.

[0041] As shown in FIG. 1, the present invention provides a high-resolution one-dimensional range profile recognition method for ground targets of missile-borne weapons, which realizes ground target recognition based on multi-dimensional features, queue features, and reliability features, which specifically includes the following steps:

[0042] Step S1, preprocessing and target multi-dimensional feature extraction;

[0043] Step S2, target multi-dimensional feature threshold screening to determine the target type;

[0044] Step S3, target queue feature analysis;

[0045] Step S4, establish a credibility evaluation model for suspected targets, and give a ranking of value target tags;

[0046] Step S5, confirm the target location information.

[0047] The multi-dimensional characteristics of the target one-dimensional range profile HRRP include: the target distance dimension, the number of target scattering points, the degree of scattering point aggregation, the second-order normalized center distance, the third-order normalized center distance, the extreme energy value, and the information Noise ratio, etc.; the target's queue feature is based on the formation layout formed by cooperating with surrounding targets in actual combat; the target's credibility feature describes the degree of similarity between the detected target and the real target; the combat scenario is realized through step-by-step feature recognition and judgment Accurate and rapid identification of medium targets.

[0048] In the step S1, the purpose of the preprocessing is to normalize and align the HRRP to ensure that the HRRP amplitude envelope of the target is at the same level and the maximum amplitude is in the middle position to facilitate subsequent feature extraction.

[0049] Target multi-dimensional feature extraction is the difficulty of correct classification and recognition, and its core is to use multi-dimensional feature vectors to reflect the essential characteristics of the target.

[0050] The target multi-dimensional feature of the one-dimensional range profile HRRP includes:

[0051] Target distance dimension: the length of a continuous one-dimensional image that crosses the detection threshold in the one-dimensional distance image;

[0052] The number of strong scattering points of the target: the number of extreme values of the one-dimensional range profile that exceed the specified threshold within the target size range;

110,000) 2 net wandering)/(Ganzhi bid 3, of which,

[0053] Scattering point aggregation degree J +&gt; -h) 2 y (Adjg pay attention) [where, y (k) represents If)/ I k = long J represents the k-th strong scattering point amplitude, krkR and q respectively represent one The starting, ending and middle scattering point numbers of the dimensional distance image, Δ is also the interval between the i-th and i+1th scattering points;

N-\

[0054] The second-order normalized center distance: %=2(=..., 2-1 is the average one-dimensional range profile, n is the number of the one-dimensional range profile of the same target, if N=1, it means only for a single one Dimensional distance profile recognition;

MT _ / NT _

[0055] Three-order normalized center distance: %=Σ(J) /(Σ(called 4; "=qn /...₌₀

[0056] Energy extreme value: the maximum value of the amplitude of the target part in the one-dimensional range image;

[0057] Signal-to-noise ratio: the signal-to-noise ratio corresponding to the maximum amplitude of the target part in the one-dimensional range profile.

[0058] In the step S2, according to the multi-dimensional feature vector extracted in the above step S1, the nearest neighbor classifier is used to discriminate the point target and the volume target, and the target within the specific feature range is preliminarily screened to remove interference.

[0059] For the quantity problem of type c, suppose that there are N training samples Xj(j=1,2,...,NJ,

C n = ΣΜ·. The idea of nearest neighbor classifier classification is to calculate the distance between a pattern X to be recognized and the training sample of each known class E=1, and distinguish it as the class to which the nearest sample belongs. Under this classification idea, the decision function of% class is:

[0060] d_t(X)= min Buyouyiyi[i =

[0061] The judgment rules are:

[0062] (x)=2 is called 4 div)

[0063] Bay 1J Judge X6 3₁₁₁, Means * belongs to the 111th category, 1116 {1,2,...,c}.

[0064] In the step S3, the queue characteristics of the target are analyzed based on the body target preliminary screened in the step S2, and false alarms and ground clutter interference are further removed.

[0065] The target queue characteristics include:

[0066] Feature 1, the positional relationship between the targets;

[0067] If the radial distance between the centroids of the two volume targets is within an allowable range, the two volume targets are in the same queue. Allowable values Set different parameter values according to different recognition target scenarios. The vehicle target can be set within 301n according to the recognition scene.

[0068] Feature 2. The positional relationship between the target and the queue;

[0069] If the distance between an individual target and any target in the known queue is less than the allowable range of the radial distance, the individual target is considered to be a target in the known queue. Allowable values are set differently according to different recognition target scenarios

The parameter value. The vehicle target can be set within 301n according to the recognition scene.

[0070] Feature 3. The number of body targets contained in the queue;

[0071] The number of targets contained in the queue is within a certain range (generally 370) to be confirmed as a valid queue.

[0072] If an area contains too few volume targets, it may be composed of isolated targets or clutter. If an area contains too many volume targets, it may be composed of artificial clutter false alarms. .

[0073] If there are other targets around the target and meet the above three queue characteristic conditions, the target is determined to be a queue target and the queue information is counted: the queue radius and the number of targets, etc. The queue feature is used as a bonus item for target recognition, and it is not used as a criterion for target judgment (that is, if the target has high credibility, but it is not a formation target, it cannot be considered that the target is not a recognition target in this case).

[0074] S4, based on the target multi-dimensional features extracted in the above step S1 and the volume target initially screened in step S2, establish a credibility evaluation model for the suspected target.

[0075] The HRRP features involved in the calculation of the target credibility model include: target distance dimension, number of target scattering points, scattering point aggregation, second-order normalized center distance, third-order normalized center distance, energy extreme, Signal-to-noise ratio, etc. Considering the real-time requirements of the seeker's HRRP target recognition, a single-frame or two-frame HRRP recognition method is generally used to ensure the high-speed output of target discrimination. At the same time, the recognition results need to be correlated and confirmed.

[0076] The credibility of the target is characterized based on the similarity of feature vectors, which can be written as:

[0078] Wherein, HXf is the preset value of the target HRRP multi-dimensional feature vector, that is, the combat target, obtained from prior information; Hy, is the multi-dimensional feature vector of the target HRRP to be identified extracted in step S1, and K is the influence coefficient.

[0079] The target credibility calculation formula is as follows:

[0080] p also becomes / j (% Fei "corpse throat) (plus 5 interviews (g)) + p Ming j, max (Dis)-min (Qis) j

[0081] Among them, Permx and Perm are the set maximum and minimum reliability values, and 0.99 and 0.01 are generally selected.

[0082] The target credibility value is output as the target recognition result. The higher the credibility, the greater the similarity between the detected target and the combat target, and then the target is completed according to the target queue information and multi-frame correlation in step S3. confirm.

[0083] In step S5, according to the target confirmed in step S4, the position of the target is output, and guidance information is provided. The target position includes: distance, pitch angle and azimuth angle. This information is refreshed in real time on the seeker signal processing board, and the corresponding distance, azimuth angle and pitch angle are output according to the maximum HRRP energy.

[0084] Simulation experiment

[0085] First, use the parameter simulation of Table 1 to generate the original echo data, and use the high-resolution one-dimensional range profile recognition method for ground targets of missile-borne weapons proposed by the present invention to identify two vehicle targets to verify the proposed algorithm of the present invention. Effectiveness.

[0086] Table 1 Simulation parameter settings

<td>parameter name</td><td>Parameter value</td>

<td>Carrier frequency</td><td>17GHz</td>

<td>Sampling points</td><td>81920</td>

<td>Pulse duration</td><td>200us</td>

<td>Distance to door length</td><td>800m</td>

<td>Emission bandwidth</td><td>150MHz</td>

<td>M/m</td><td>2000m/s</td>

<td>Bullet distance</td><td>5Km</td>

<td>Accumulated frames</td><td>1</td>

[0088] Illustratively, the HRRP of target 1 at different angles is shown in FIG. 2. Figure 3 shows the HRRP of target 2 under different angles. It can be seen from the two figures that the HRRP collected by the same target at different times is different, but the amplitude distribution is basically the same, with only some differences in the scattering details; the HRRP of different targets has smaller differences in energy and signal-to-noise ratio. It is difficult to distinguish the two targets, which has a great influence on the single pulse guidance method of the traditional seeker. However, the two targets are in the target distance dimension, the number of target scattering points, the scattering point aggregation degree, the normalized center distance, etc. There are certain differences.

[0089] Simulation experiment analysis

[0090] The HRRP recognition method based on the multi-dimensional feature credibility ranking proposed by the present invention is used to classify and recognize 1000 sets of test data. The recognition accuracy of target 1 is 100%, and the recognition accuracy of target 2 is 99.6%. The frame discrimination method can increase the recognition rate of the two targets to 100%, and the recognition results can fully meet the guidance requirements of the seeker.

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

[0092] 1. The present invention can solve the problem of single target recognition method and low recognition rate of the missile-borne platform.

[0093] 2. Compared with the matching degree classification method, the support vector machine classification method, the restricted Boltzmann machine classification method, and the convolutional neural network classification method, the present invention requires fewer training samples, high data processing efficiency, and can Quickly implement engineering applications.

[0094] 3. Starting from the application requirements of precision guidance for missile-borne weapons, the present invention comprehensively considers the requirements for real-time, robustness, and accuracy of target recognition, and proposes a target recognition method based on a suspected target credibility evaluation model. Target category and similarity degree, give the order of strike targets, improve weapon guidance efficiency.

[0095] It should be noted that in the embodiments of the present invention, the terms "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", " "Front", "Back", "Left", "Right", "Vertical", "Horizontal", "Top", "Bottom", "Inner", "Outer", "Clockwise", "Counterclockwise", " The orientation or positional relationship indicated by "axial", "radial", "circumferential", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the embodiment, and does not indicate or imply the device or The element must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of the present invention. In addition, the terms "first", "second", and "third" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance.

[0096] In the present invention, unless otherwise clearly defined and limited, the terms "installed", "connected", "connected", "fixed" and other terms should be understood in a broad sense, for example, it may be a fixed connection or It can be detachably connected or integrated; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the internal communication of two components or the interaction of two components. For those of ordinary skill in the art, the specific meanings of the above-mentioned terms in the present invention can be understood according to specific situations.

[0097] Although the content of the present invention has been described in detail through the above preferred embodiments, it should be recognized that the above description should not be considered as limiting the present invention. After those skilled in the art have read the above content, various modifications and substitutions to the present invention will be obvious. Therefore, the protection scope of the present invention should be defined by the appended claims.