A battlefield detection radar person-vehicle classification and identification method

Abstract

The application discloses a battlefield detection radar person-vehicle classification and identification method, comprising the following steps: step 1, acquiring a radar receiving signal, performing dynamic target detection analysis on the receiving signal, and extracting the speed and RCS value of the dynamic target; step 2, establishing a Gaussian probability distribution function according to the target speed and target RCS, setting the mean value and standard deviation corresponding to the person and the vehicle, performing target classification probability statistics through the Gaussian distribution function, and calculating the target category score according to the target classification probability and the target characteristic value proportion. The application establishes a Gaussian probability distribution function through two aspects of the radar reflection cross section (RCS) and the speed of the target, and effectively improves the probability of track target identification through the score comparison method, solves the defects that it is difficult to accurately and quickly identify and distinguish the person and the vehicle when the target moves at low speed and the amplitude fluctuation is relatively small, is easier to understand and more accurate and fast compared with traditional identification algorithms, and provides a new reference direction for radar classification and identification.

Description

TECHNICAL FIELD

[0001] The application relates to a battlefield detection radar person-vehicle classification and identification method. BACKGROUND

[0002] With the development of radar technology, the target detection capability thereof is widely applied to various fields, such as a battlefield reconnaissance field, and usually, a radar or the like is required to quickly distinguish a moving vehicle and a pedestrian, so as to facilitate accurate attack on a target.

[0003] A radar antenna transmits millimeter waves to irradiate a target, and then a receiving device receives a signal reflected by the target, and after processing of the signal, a plurality of characteristic values of the target are obtained, and whether each characteristic value represents a vehicle or a pedestrian is judged according to the characteristic value, and a conventional way of target identification by the radar is to acquire speed and amplitude information of the target, and the target is distinguished and identified.

[0004] When the target moves at a low speed, it is difficult to distinguish the target by the speed, and the radar can basically only distinguish the person and the vehicle by amplitude information of the target, but the difference between amplitude fluctuations of the person and the vehicle is small, which causes the identification result to be unstable, if other methods are used for target identification, although the identification probability is high, the operation amount is large, and a long time is required, and fast identification cannot be achieved, and therefore, accurate and fast identification and distinction between the person and the vehicle are urgent problems to be solved. SUMMARY

[0005] The main purpose of the application is to provide a battlefield detection radar person-vehicle classification and identification method, which can accurately and quickly identify and distinguish between the person and the vehicle.

[0006] The purpose of the application can be achieved by adopting the following technical scheme.

[0007] A battlefield detection radar person-vehicle classification and identification method comprises the following steps:

[0008] Step 1, acquiring a radar receiving signal, performing moving target detection analysis on the receiving signal, and extracting a speed and an RCS value of a moving target;

[0009] Step 2, establishing a Gaussian probability distribution function according to the target speed and the target RCS, setting a mean value and a standard deviation corresponding to the person and the vehicle, performing target classification probability statistics by the Gaussian distribution function, and calculating a target category score according to the target classification probability and a target characteristic value proportion; the calculation of the target category score according to the target classification probability and the target characteristic value proportion is specifically as follows:

[0010] The probabilities of the target being a person and a car are calculated by the Gaussian distribution functions respectively, and four parameters P1, P2, P3 and P4 are obtained, wherein P1 is the probability of determining the target as a person at the current speed, P2 is the probability of determining the target as a car at the current speed, P3 is the probability of determining the target as a person at the current RCS, and P4 is the probability of determining the target as a car at the current RCS;

[0011] Based on the above parameters, the probability scores of the target being a person and a car are:

[0012] P(person) = A*P1 + B*P3, P(car) = A*P2 + B*P4, the score value is between 0 and 100, wherein A represents the proportion of the speed feature calculation score, B represents the proportion of the RCS feature calculation score, the sum of A and B is 1, if the P(person) score is greater than the P(car) score, the target is determined as a person, and if the P(person) score is less than the P(car) score, the target is determined as a car.

[0013] Preferably, the target classification probability calculation process is:

[0014] Step 2.1, the RCS and speed information of the default target conforms to the Gaussian distribution, and the target is a person or a car, and the mean and standard deviation corresponding to the person and the car are preset;

[0015] Step 2.2, the Gaussian probability distribution function based on the target speed and the target RCS is obtained:

[0016] In the formula, F(X) represents the probability distribution of a specific continuous random variable X, X is the target speed or the target RCS, σ is the standard deviation, σ 2 is the standard deviation, u is the mean value of the Gaussian distribution, e is a constant, x represents the specific value of the random variable at a specific time, and dx represents the difference between two adjacent data points. In the formula, 1 standard deviation on the left and right of X is taken as the probability of the value, the target speed or the target RCS is substituted into F(X), and then F(X) is normalized to make the value range 0-100, that is, the probability distribution of determining whether the target is a person according to the target speed is obtained.

[0017] Preferably, when the target speed is greater than 4.1 m / s, A = 1 and B = 0, at this time P(person) = 1*P1 + 0*P3 and P(car) = 1*P2 + 0*P4;

[0018] When the target speed is less than 2 m / s, A = 0.8 and B = 0.2, at this time P(person) = 0.8*P1 + 0.2*P3 and P(car) = 0.8*P2 + 0.2*P4;

[0019] When the target speed is not greater than 4.1 m / s and not less than 2 m / s, A=0.6, B=0.4, at this time, P (person) =0.6*P1+0.4*P3, P (car) =0.6*P2+0.4*P4.

[0020] Preferably, in the process of calculating the target category score, if the distance width and the speed width have obvious characteristic values, the corresponding target category score is added.

[0021] Preferably, in the process of target classification probability statistics by the Gaussian distribution function, if the distance width and the speed width have obvious characteristic values, the target category is directly judged.

[0022] Preferably, for the target forming a stable track, the all single-point track recognition scores p (person) and p (car) contained in the track are averaged and comprehensively judged, and the judgment mode is that the scores of the last n points of the track are averaged, n>1, to obtain a target recognition total score for comparison.

[0023] The beneficial technical effects of the present application are: the present application establishes a Gaussian probability distribution function through the radar cross section (RCS) and the speed of the target, and effectively improves the probability of track target recognition through score comparison, solves the defects that it is difficult to accurately and quickly identify and distinguish people and cars when the target moves at low speed and the amplitude fluctuation is small, is easier to understand and more accurate and fast compared with the traditional identification algorithm, and provides a new reference direction for radar classification and identification. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 It is a working flow structure schematic diagram according to the present application;

[0025] Figure 2 It is a probability distribution structure schematic diagram according to the present application;

[0026] Figure 3 It is a target recognition score structure schematic diagram according to the present application;

[0027] Figure 4 It is a feature proportion calculation structure schematic diagram according to the present application;

[0028] Figure 5 It is a track recognition flow structure schematic diagram according to the present application;

[0029] Figure 6 It is a feature proportion calculation structure schematic diagram according to the present application;

[0030] Figure 7 It is a comparison schematic diagram of the traditional person-car distinguishing probability and the target feature-based person-car distinguishing probability at 500 m-1000 m according to the present application;

[0031] Figure 8 A comparison diagram of a conventional vehicle-person distinction probability and a target feature-based vehicle-person distinction probability based on a target speed of 2 m / s according to the present application. DETAILED DESCRIPTION

[0032] In order to make the technical solution of the present application more clear and explicit to those skilled in the art, the present application will be further described in detail below in conjunction with embodiments and drawings, but the embodiments of the present application are not limited thereto.

[0033] As shown in the drawings, the battlefield detection radar vehicle-person classification and identification method provided by the present embodiment comprises the following steps: Figures 1-4

[0034] Step 1: The radar antenna transmits millimeter wave to irradiate a target, and then a receiving device receives the signal reflected by the target. After processing, the speed of the moving target and the radar cross section (RCS) of the moving target are obtained.

[0035] Step 2: According to the target speed and target RCS, a Gaussian probability distribution function is established, the mean and standard deviation corresponding to a person and a vehicle are set, and the target category probability is calculated. When the distance width and speed width have obvious characteristic values, direct judgment is made (for example, the radial speed or vector speed of two consecutive points of the target > 4 m / s, which is determined as a vehicle; the radial speed of two consecutive points < 2.4 m / s, the distance width < 6 m, the RCS < 1 m 2 , which is determined as a person).

[0036] The target classification probability calculation process is as follows: the RCS and speed information of the target (person and vehicle) are assumed to conform to Gaussian distribution.

[0037] Table 1: Track identification parameter table

[0038]

[0039] According to the definition of Gaussian distribution, the sum of probabilities is 1, that is, the area enclosed by the probability density function and the X axis (the X axis is the target speed and the Y axis is the target speed probability) is equal to 1.

[0040] Gaussian probability distribution function established based on target speed and target RCS: In the formula, F(X) represents the probability distribution of a specific continuous random variable X, X can be target speed or target RCS, σ is the standard deviation, σ 2 is the standard variance, u is the mean of Gaussian distribution, e is a constant, x represents the specific value of the random variable (target speed or target RCS) at a specific time, and dx represents the difference between two adjacent data points (data points refer to one value of target speed or target RCS). In the formula, 1 times the standard deviation on the left and right of X is taken as the probability of the value.

[0041] From​Figure 2 It can be seen that, in Figure 2 The x-axis represents the speed, and the y-axis represents the probability that the current speed is a person, Figure 2 It can be seen that, in Figure Three It can be seen that, in Figure Two The area represents the probability that the target is a person when the target speed is 1.5 m / s,

[0042]

[0043] The probability that the target is a person when the target speed is 4 m / s,

[0044]

[0045] The probability distribution of whether the target is a person according to the target speed is obtained by normalizing F(x) to a value range of 0-100, i.e. Figure 3 , where the x-axis is the speed, and the y-axis is the probability of a person's speed;

[0046] The calculation process of the probability score of the target being a person and a car is as follows: the target RCS information and speed information are extracted, and the probabilities of the target being a person and a car are calculated through the above Gaussian distribution functions, and four parameters (P1-P4) are obtained

[0047] Table 2: Track recognition probability table

[0048]

[0049] In this table: P1 is the probability of determining that the target is a person at the current speed, P2 is the probability of determining that the target is a car at the current speed, P3 is the probability of determining that the target is a person at the current RCS, P4 is the probability of determining that the target is a car at the current RCS, and the current speed and RCS refer to the speed and RCS value of the current extracted moving target,

[0050] Based on the above parameters, the probability score of the target being a person and a car is:

[0051] P(person) = A*P1 + B*P3, P(car) = A*P2 + B*P4, the score is between 0 and 100, the sum of A and B is 1, when the speed is less than 2 m / s, the proportion of the score calculated by the speed feature determines the values of A and B, the speed of the target affects the values of A and B, when one of A and B is 0, the target can be directly determined, the values of A and B affect the scores of P(person) and P(car), the score is obtained, which affects the determination of the target, where P(person) refers to the score of the probability of the target being a person, and P(car) refers to the score of the probability of the target being a car, where A represents the proportion of the score calculated by the speed feature, and B represents the proportion of the score calculated by the RCS feature;

[0052] When the distance width and the speed width have obvious characteristic values, the scores are added, as shown in Figure 5

[0053] The specific calculation method is shown in Figure 4

[0054] When the target speed is greater than 4.1 m / s, A=1 and B=0, at this time, P(person)=1*P1+0*P3 and P(car)=1*P2+0*P4, that is, whether the fast-moving target is judged as a car or a person completely depends on the speed, and has nothing to do with the RCS feature, so there is no need to calculate the RCS related operation, thereby achieving the effect of quickly judging the target type;

[0055] When the target speed is less than 2 m / s, A=0.8 and B=0.2, at this time, P(person)=0.8*P1+0.2*P3 and P(car)=0.8*P2+0.2*P4;

[0056] When the target speed is not greater than 4.1 m / s and not less than 2 m / s, A=0.6 and B=0.4, at this time, P(person)=0.6*P1+0.4*P3 and P(car)=0.6*P2+0.4*P4;

[0057] When the RCS is 5, the mean value and the standard deviation are known, and P3 and P4=1-P3 can be obtained by substituting the Gaussian function;

[0058] For plot recognition, specifically:

[0059] If the target RCS>15+4*5, the person score is assigned 0 and the car score is assigned 100;

[0060] If the target RCS<15+4*5, the RCS and speed probability weighted score is

[0061] If the distance width>24 m, the car score is added by 40;

[0062] If the distance width<25 m and the speed width>0.6 m / s, the person score is added by 40;

[0063] If the speed width is not greater than 0.6 m / s and the track speed>4.1 m / s, the person score is assigned 0 and the car score is assigned 100;

[0064] The P(person) and P(car) are recorded and compared;

[0065] Through score comparison, if P(person) and P(car) are both less than 20 points, the target type is judged as an unknown type, if P(person) is greater than P(car), it is judged as a person, and if P(person) is less than P(car), it is judged as a car, which can effectively improve the probability of plot target recognition;

[0066] ​​Step 3, for the target forming stable track, average all single point track recognition scores p(man) and p(car) included in the track, and make a comprehensive judgment, as shown in (track recognition process); Figure 6

[0067] The judgment method is to average the scores of the last 10 points of the track to obtain the total score of target recognition, and then compare them;

[0068] If the target distance is greater than 8500m, the car score is added by 20 and the man score is reduced by 20;

[0069] If the track speed of two consecutive points is greater than 4.1m / s and the track point number is greater than 5, the car score is added by 1000;

[0070] If the radial distance of two consecutive points is greater than 10km, the car score is added by 1000;

[0071] If the RCS of two consecutive points is less than 35 and the distance width is less than 10 and the speed width is greater than 0.6, the man score is added by 1000;

[0072] If the average RCS of two consecutive times is less than 5 and the average distance width is less than 15 and the average speed width is greater than 0.4, the man score is added by 1000;

[0073] If the average speed of two points is less than 2.4 and the distance width is less than 6 and the average RCS is less than 1, the man score is added by 1000;

[0074] The man and car scores are calculated and compared according to the above;

[0075] In addition, there are some special judgments: if there are obvious car and man features for two consecutive times, the target recognition is confirmed, and the corresponding class total score is added by 1000;

[0076] The obvious features are as follows:

[0077] 1) The radial velocity or vector velocity of two consecutive points is greater than 4m / s, which is determined as a car;

[0078] 2) The radial distance of two consecutive points is greater than 10km, which is determined as a car;

[0079] 3) The radial velocity of two consecutive points is less than 2.4m / s, the distance width is less than 6m, and the RCS is less than 1m 2 , which is determined as a man;

[0080] 4) The distance width of two consecutive points is less than 10m, the RCS is less than 20m^2, and the speed width is greater than 0.6m / s, which is determined as a man;

[0081] 5) The average distance width of two consecutive times is less than 15m, the average RCS of the track is less than 5m^2, and the average speed width of the track is greater than 0.4m / s, which is determined as a man;​

[0082] The score comparison mode is:

[0083] If the scores of the person and the vehicle are both less than 20, the track speed is less than 4.1 m / s, and the number of track points is less than 5, the target type is determined as an unknown type;

[0084] If the score of the person is greater than the score of the vehicle, the target is determined as a person; if the score of the vehicle is greater than the score of the person, the target is determined as a vehicle, which can further improve the target recognition probability.

[0085] In the embodiment, the track recognition process is as shown in Figure 5

[0086] In summary, in the embodiment, the Gaussian probability distribution function is established through the radar reflection cross section (RCS) and the speed of the target, the score comparison method is used, the probability of track target recognition is effectively improved, the defect that the person and the vehicle are difficult to accurately and quickly recognize and distinguish when the target moves at low speed and the amplitude fluctuation is small is solved, compared with the traditional recognition algorithm, it is easier to understand, more accurate and fast, and a new reference direction is provided for radar classification and recognition.

[0087] The above is only further embodiments of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can make equivalent replacement or change according to the technical scheme and concept of the present application within the scope disclosed by the present application, which belongs to the protection scope of the present application.​

Claims

1. A method for classifying and identifying people and vehicles using battlefield detection radar, characterized in that: The method comprises the following steps Step 1, obtaining radar receiving signals, performing moving target detection analysis on the receiving signals, and extracting the speed and RCS value of the moving target; Step 2, establishing a Gaussian probability distribution function according to the target speed and target RCS, setting the mean value and standard deviation corresponding to the person and vehicle, performing target classification probability statistics through the Gaussian distribution function, and calculating the target category score according to the target classification probability and target characteristic value proportion; The target category score calculated according to the target classification probability and target characteristic value proportion is specifically: The probabilities that the target is a person and a vehicle are calculated through the Gaussian distribution function respectively, and four parameters P1, P2, P3 and P4 are obtained, wherein P1 is the probability that the target is a person determined by the current speed, P2 is the probability that the target is a vehicle determined by the current speed, P3 is the probability that the target is a person determined by the current RCS, and P4 is the probability that the target is a vehicle determined by the current RCS; Based on the above parameters, the probability scores of the target being a person and a vehicle are: P(person) = A*P1 + B*P3, P(vehicle) = A*P2 + B*P4, the score value is between 0 and 100, wherein A represents the proportion of the speed characteristic calculation score, B represents the proportion of the RCS characteristic calculation score, the sum of A and B is 1, if the P(person) score is greater than the P(vehicle) score, the target is determined to be a person, and if the P(person) score is less than the P(vehicle) score, the target is determined to be a vehicle.

2. The battlefield detection radar method for classifying and identifying people and vehicles according to claim 1, characterized in that: The target classification probability calculation process is: Step 2.1, the RCS and speed information of the target conform to Gaussian distribution by default, the target is a person or a vehicle, and the mean value and standard deviation corresponding to the person and vehicle are preset; Step 2.

2. The Gaussian probability distribution function is established based on the target speed and target RCS: In the formula, F(X) represents the probability distribution of a specific continuous random variable X, X is the target speed or target RCS, σ is the standard deviation, σ 2 is the standard variance, u is the average value of the Gaussian distribution, e is a constant, x represents the specific value of the random variable at a specific time, and dx represents the difference between adjacent data points. In the formula, the value of X is taken as 1 standard deviation on the left and right, the target speed or target RCS is substituted into F(X), and F(X) is normalized to make its value range 0-100, thereby obtaining the probability distribution of determining whether the target is a person according to the target speed.

3. The battlefield detection radar person-vehicle classification and identification method according to claim 2, characterized in that: When the target speed is greater than 4.1 m / s, A = 1, B = 0, at this time P(person) = 1*P1 + 0*P3, P(vehicle) = 1*P2 + 0*P4; When the target speed is less than 2 m / s, A = 0.8, B = 0.2, at this time P(person) = 0.8*P1 + 0.2*P3, P(vehicle) = 0.8*P2 + 0.2*P4; When the target speed is not greater than 4.1 m / s and not less than 2 m / s, A = 0.6, B = 0.4, at this time P(person) = 0.6*P1 + 0.4*P3, P(vehicle) = 0.6*P2 + 0.4*P4.

4. The battlefield detection radar method for classifying and identifying people and vehicles according to claim 3, characterized in that: In the process of calculating the target category score, if obvious characteristic values appear in the distance width and speed width, the corresponding target category score is added.

5. The method of claim 3, wherein the method further comprises: determining a distance between the radar and the vehicle; and determining a height of the vehicle based on the distance and the radar return. In the process of performing target classification probability statistics through the Gaussian distribution function, if obvious characteristic values appear in the distance width and speed width, the target category is directly determined.

6. The method of claim 1, wherein the method further comprises: determining a distance between the radar and the vehicle; and determining a speed of the vehicle based on the distance and the time. For a target forming a stable track, the scores p(person) and p(vehicle) of all single-point tracks contained in the track are averaged and comprehensively judged, and the judgment mode is: the scores of the last n points of the track are averaged, n > 1, to obtain a target recognition total score for comparison.

Images