A global sensor system error partition registration algorithm based on similarity principle

Abstract

The invention discloses a global sensor system error partition registration algorithm based on the similarity principle, which includes the following steps: S1, collecting all output tracks of multiple sensors; S2, using the track space similarity principle, calculating the correlation between tracks, and automatically selecting a track set for error registration; S3, partitioning the track set for error registration; S4, using a genetic algorithm to calculate the system errors of different sectors of each sensor for registration; S5, calculating the global matching variance; S6, selecting the sensor system error with the smallest global variance for sensor error registration, and completing the global sensor system error partition registration algorithm based on the similarity principle. This method aims at the error registration problem in multi-sensor fusion, uses the spatial similarity principle of the target track, uses a genetic algorithm to solve the system error of each sensor, and uses partition registration for the inconsistency of the system error of each direction of the sensor, and adopts the global matching variance minimum criterion to achieve accurate registration of the system error in each region. This algorithm can be widely used in multi-sensor data fusion systems, solves the problems of multiple batches and missed batches of fused tracks caused by target-related errors caused by sensor system errors, and improves the accuracy of the situation after data fusion.

Description

Technical Field

[0001] The patent of this invention relates to the field of electronic software, and specifically to an automatic registration algorithm for multi-sensor system errors. Background Art

[0002] The prerequisite for successful data fusion is to transform the measurement data of each sensor into a common reference coordinate system without error after registering the system error. If the system error is not compensated, the system error will increase the track tracking error, and even phantom targets will appear, which will seriously reduce the monitoring performance of the entire system. For data fusion, the system error causes a large deviation between the tracks of the same target from different sensors, which brings ambiguity and difficulty to track association and integration, reduces the performance of the integrated system track, and loses the characteristics that the multi-sensor itself should have. Therefore, it is necessary to seek an effective system error correction algorithm when processing multi-radar data.

[0003] The difficulty of systematic error registration lies in:

[0004] (1) It is difficult to calculate the sensor system error based on the sensor detection data. There are multiple fuzzy solutions, which makes it difficult to accurately estimate the sensor system error.

[0005] (2) The errors of sensors in all directions are usually inconsistent. Using a unified error estimation method for registration will make it difficult to meet the requirements of all-round accurate registration after registration.

[0006] There are three main existing error registration methods:

[0007] (1) Using external devices such as ADS_B to calculate sensor system errors. This method requires external ADS_B equipment and has low accuracy because ADS_B does not have accurate time registration.

[0008] (2) Using drones or other aircraft as cooperative targets for error calibration. This method has high accuracy, but is difficult to implement and has high costs.

[0009] (3) Using sensor data to estimate system errors, commonly used methods include least squares estimation, maximum likelihood estimation, real-time precision control, etc. These methods are greatly affected by random errors and are closely related to the target position. The results are not ideal in practical applications and the problem of inconsistent sensor errors in all directions is not solved. Summary of the invention

[0010] In response to the above-mentioned deficiencies in the prior art, the global sensor system error partitioning registration algorithm based on the similarity principle provided by the present invention divides the sensor into different sectors, and calculates the sensor system error in each sector separately, thereby solving the inconsistency of the sensor errors in all directions; the similarity principle is used to automatically select the registration track set, and there is no need for manual selection of the registration track; a genetic algorithm is used to optimize the solution of the system error, thereby solving the fuzzy multi-solution problem of the system error and improving the solution speed.

[0011] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is:

[0012] A method for comprehensive suppression of false flight paths of low-altitude radar based on feature statistics is provided, which comprises the following steps:

[0013] S1, collect all output tracks of multiple sensors;

[0014] S2. Using the principle of track spatial similarity, the correlation between tracks is calculated, and the track set used for error registration is automatically selected to avoid manual participation in selecting the registration track set;

[0015] S3, error-registered track set partitioning;

[0016] S4, using genetic algorithm to calculate the systematic error of all sectors of each sensor;

[0017] S4, calculating the global matching variance;

[0018] S6. Select the sensor system error with the smallest global matching variance to perform sensor error registration, and complete the global sensor system error partition registration algorithm based on the similarity principle.

[0019] In the common detection area of ​​multiple sensors, the detection estimates of the same target by different sensors are similar, and the detection error of the sensor is independent of the target position. The registration track combination is automatically selected by measuring the similarity of the tracks.

[0020] For the i-th track and the j-th track, the similarity factor can be expressed as

[0021]

[0022] Δ ij (k)=(R i (k)cos(θ i (k))-R j (k)cos(θ j (k))) 2 +(R i (k)sin(θ i (k))-R j(k)sin(θ j (k))) 2 When k

[0023] The Euclidean distance between the two tracks, R i (k) is the distance of the ith track, θ i (k) is the bearing of the ith track.

[0024] ε ij The smaller it is, the higher the similarity between track i and track j.

[0025] Calculate ε for all the tracks in the common detection area of ​​the two sensors ij , select a set of track combinations that satisfy the following formula {(i1,j1),(i2,j2),K,(i n ,j n )} is the registration track group.

[0026] min(∑ε ij ), ε ij ≤δ

[0027] Furthermore, the specific method of step S3 is:

[0028] The sensor detection area is divided into 16 sectors according to the azimuth, each sector is 22.5 degrees, and the registration track set {(i1,j1),(i2,j2),K,(i n ,j n )} are assigned to the corresponding sectors of the two sensors. The track of the Nth sector of sensor A can be expressed as

[0029]

[0030] If there is no track that can be paired in the sector, then

[0031] A N =[0]

[0032] Furthermore, the specific method of step S4 is:

[0033] Track combination after partition pairing and Build an error registration model.

[0034] represents the measurement of sensor A and sensor B at time k; λ=[ΔR A ,Δθ A ,ΔR B ,Δθ B ] T represents the systematic error measured by radar A and radar B. The coordinates after error registration can be expressed as:

[0035]

[0036]

[0037] After registration at the same time, the observation points of radar A and radar B on the same target should coincide, that is,

[0038]

[0039] Construct an objective function and use nonlinear optimization methods to estimate the system error of each radar. The goal of error registration is to make the target tracks observed by the two radars close to the target's true position. If the observation points of the two radars are registered to the target's true position, the distance between the two measurement points after registration should be 0. Therefore, the goal of the two radar registration is to find a set of system error values ​​to minimize the distance between the track points after registration. Therefore, the registration goal can be expressed as follows:

[0040]

[0041] G mn (k) represents the Euclidean distance of the same target measured by radar sector A m and radar sector B n at the same time after system error registration, indicating the degree of difference between the two measurement points after registration. mn (k) = 0 means that after registration, the measurement points of the two radars coincide with the real position of the target. However, due to the existence of random errors in the measurement, G mn (k)≠0, the purpose of error registration is to find a suitable set of λ to make min(G mn (k)). For N times of measurement data, we need to find a set of λ within a certain range so that The solution process uses genetic algorithm.

[0042] Furthermore, the specific method of step S5 is:

[0043] Through the above four steps, the estimated values ​​of the system errors of several regions of the two sensors A and B can be obtained. For the sectors without paired tracks, the estimated values ​​of the two sectors before and after are used for least squares interpolation calculation. The system error of sensor A can be expressed as

[0044]

[0045] The corresponding variance can be expressed as

[0046]

[0047] When there are multiple systematic error solutions, a set of systematic error solutions that satisfy min(S) is selected as the registration systematic error.

[0048] The beneficial effects of the present invention are as follows: the algorithm can automatically select a set of tracks for system error registration by using the large-scale similarity of multi-sensor detection tracks, reduce the human involvement process, and improve the automation of error registration; adopt an error registration algorithm based on a genetic algorithm to improve the error registration solution speed and error estimation accuracy; adopt a partition registration method to estimate the sensor system error of different sectors, and solve the problem of low system error accuracy caused by the inconsistency of sensor directional errors. The algorithm can be applied to a multi-sensor data fusion system to solve the core problem of system error registration and effectively improve the accuracy and adaptability of the data fusion system. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a schematic diagram of the process of the present invention;

[0050] Figure 2 This is a comparison chart of the detection tracks of 7 radar actual data before and after alignment by this algorithm. DETAILED DESCRIPTION

[0051] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0052] like Figure 1 As shown in FIG. 1 , the global sensor system error partition registration algorithm based on the similarity principle includes the following steps:

[0053] S1, collect all output tracks of multiple sensors;

[0054] S2. Perform spatial coordinate conversion to convert the detection tracks of all sensors into a unified coordinate system. Utilize the track similarity principle to calculate the correlation between tracks and automatically select a track set for error registration, avoiding manual participation in selecting the registration track set.

[0055] S3, error-registered track set partitioning;

[0056] S4, all sector errors of each sensor system using genetic algorithm;

[0057] S4, calculating the global matching variance;

[0058] S6. Select the sensor system error with the smallest global matching variance to perform sensor error registration, and complete the global sensor system error partition registration algorithm based on the similarity principle.

[0059] The specific method of step S2 is:

[0060] In the common detection area of ​​multiple sensors, the detection estimates of the same target by different sensors are similar, and the detection error of the sensor is independent of the target position. The registration track combination is automatically selected by measuring the similarity of the tracks.

[0061] For the i-th track and the j-th track, the similarity factor can be expressed as

[0062]

[0063] Δ ij (k)=(R i (k)cos(θ i (k))-R j (k)cos(θ j (k))) 2 +(R i (k)sin(θ i (k))-R j (k)sin(θ j (k))) 2 When k

[0064] The Euclidean distance between the two tracks, R i (k) is the distance of the ith track, θ i (k) is the bearing of the ith track.

[0065] ε ij The smaller it is, the higher the similarity between track i and track j.

[0066] Calculate ε for all the tracks in the common detection area of ​​the two sensors ij , select a set of track combinations that satisfy the following formula {(i1,j1),(i2,j2),K,(i n ,j n )} is the registration track group.

[0067] min(∑ε ij ), ε ij ≤δ

[0068] The specific method of step S3 is:

[0069] The sensor detection area is divided into 16 sectors according to the azimuth, each sector is 22.5 degrees, and the registration track set {(i1,j1),(i2,j2),K,(in ,j n )} are assigned to the corresponding sectors of the two sensors. The track of the Nth sector of sensor A can be expressed as

[0070]

[0071] If there is no track that can be paired in the sector, then

[0072] A N =[0]

[0073] The specific method of step S4 is:

[0074] Track combination after partition pairing and Build an error registration model.

[0075] represents the measurement of sensor A and sensor B at time k; λ=[ΔR A ,Δθ A ,ΔR B ,Δθ B ] T represents the systematic error measured by radar A and radar B. The coordinates after error registration can be expressed as:

[0076]

[0077]

[0078] After registration at the same time, the observation points of radar A and radar B on the same target should coincide, that is,

[0079]

[0080] Construct an objective function and use nonlinear optimization methods to estimate the system error of each radar. The goal of error registration is to make the target tracks observed by the two radars close to the target's true position. If the observation points of the two radars are registered to the target's true position, the distance between the two measurement points after registration should be 0. Therefore, the goal of the two radar registration is to find a set of system error values ​​to minimize the distance between the track points after registration. Therefore, the registration goal can be expressed as follows:

[0081]

[0082] G mn (k) represents the Euclidean distance of the same target measured by radar sector A m and radar sector B n at the same time after system error registration, indicating the degree of difference between the two measurement points after registration. mn(k) = 0 means that after registration, the measurement points of the two radars coincide with the real position of the target. However, due to the existence of random errors in the measurement, G mn (k)≠0, the purpose of error registration is to find a suitable set of λ to make min(G mn (k)). For N times of measurement data, we need to find a set of λ within a certain range so that The solution process uses genetic algorithm.

[0083] The specific method of step S5 is:

[0084] Through the above four steps, the estimated values ​​of the system errors of several regions of the two sensors A and B can be obtained. For the sectors without paired tracks, the estimated values ​​of the two sectors before and after are used for least squares interpolation calculation. The system error of sensor A can be expressed as

[0085]

[0086] The corresponding variance can be expressed as

[0087]

[0088] When there are multiple systematic error solutions, a set of systematic error solutions that satisfy min(S) is selected as the registration systematic error.

[0089] In the specific implementation process, this method uses a Windows 10 system and the programming language used is C++. In this implementation process, the simulation data and measured data used by this method are used to verify the accuracy of the system error registration algorithm.

[0090] The system error registration algorithm of the present invention is used for simulation calculation and compared with the traditional method for verification. The verification results of the consistency of the errors in all directions are shown in Table 1.

[0091] Table 1 Simulation results of registration algorithm when all directions errors are consistent

[0092]

[0093]

[0094] The verification results of inconsistent directional errors are shown in Table 2.

[0095] Table 2 Simulation results of error registration algorithm when the errors in all directions are inconsistent

[0096]

[0097] As shown in Table 1, when the errors in all directions are consistent, the distance systematic error estimation error of the algorithm of the present invention is 7.854, and the best of the existing methods is the RTOC method, whose distance systematic error estimation error is 76.596, and the algorithm of the present invention is improved by 9.75 times; the azimuth systematic error estimation error of the algorithm of the present invention is 0.079, and the best of the other methods is the maximum likelihood estimation method, whose azimuth systematic error estimation error is 0.096, and the algorithm of the present invention is improved by 1.21 times, but the distance systematic estimation value of the method of the present invention is 13.16 times higher than that of the maximum likelihood method. This implementation case shows that the algorithm of the present invention is more suitable for the joint estimation of the systematic errors of distance and azimuth than the traditional method, and the estimation accuracy is improved by one order of magnitude.

[0098] As shown in Table 2, when the errors in all directions are inconsistent, the mean error variance of the distance system error estimation of the algorithm of the present invention is 10.0578, and the maximum likelihood estimation method has the smallest distance error estimation variance among other methods, with an estimated variance of 55.64; the error variance of the azimuth system error estimation of the algorithm of the present invention is 0.028, and the best of the other methods is the RTOC method, whose azimuth system error estimation error variance is 0.245, which is 8.135 times higher than the algorithm of the present invention. This implementation case shows that when the errors in all directions are inconsistent, the error estimation results of the traditional error registration algorithm diverge, and basically do not have the effect of error registration. The estimated error variance of the algorithm of the present invention is still kept within 15%, and has a good error registration effect.

[0099] In summary, the algorithm of the present invention utilizes the large-scale similarity of multi-sensor detection tracks to automatically select a track set for system error registration, reduce the human participation process, and improve the automation of error registration; adopts an error registration algorithm based on a genetic algorithm to improve the error registration solution speed and error estimation accuracy; adopts a partition registration method to estimate the sensor system error in different sectors, solving the problem of low system error accuracy caused by the inconsistency of sensor errors in all directions. The algorithm has been verified by simulation and actual data. Under the premise of unknown sensor system errors, it can directly estimate the distance and azimuth system errors of the sensor through the jointly detected target track data. It has few restrictions in engineering applications, high error estimation accuracy, and can effectively improve the positioning and tracking accuracy after fusion. It is a universal and effective multi-sensor system error registration algorithm that can be widely used in multi-sensor data fusion systems to improve the accuracy and adaptability of data fusion systems.

Claims

1. A global sensor system error partitioning registration algorithm based on similarity principle, characterized in that: The following steps are involved: S1, collect all output tracks of multiple sensors; S2, using the principle of track spatial similarity, calculate the correlation between tracks, automatically select the track set for error registration, and avoid manual participation in selecting the registration track set; S3, error-registered track set partitioning; S4, using genetic algorithm to calculate the system error of different sectors of each sensor; S5, calculating the global matching variance; S6, selecting the sensor system error with the smallest global matching variance for sensor error registration, and completing the global sensor system error partition registration algorithm based on the similarity principle; The specific method of step S3 is: The sensor detection area is divided into 16 sectors according to the azimuth, each sector is 22.5 degrees, and the registration track set {(i1, j1), (i2, j2), …, (i n , j n )} are assigned to the corresponding sectors of the two sensors; the track of the Nth sector of sensor A is expressed as If there is no paired track in the sector, then A N ＝[0]。 2. The global sensor system error partition registration algorithm based on similarity principle according to claim 1 is characterized in that: The specific method of step S2 is: Definition: For the i-th track and the j-th track, the similarity factor is expressed as represents the Euclidean distance between two tracks at time k, R i (k) is the distance of the ith track, θ i (k) is the bearing of the ith track; The smaller it is, the higher the similarity between track i and track j is; All tracks in the common detection area of ​​the two sensors are calculated separately , select a set of track combinations that satisfy the following formula {(i1,j1),(i2,j2),…,(i n ,j n )} is the registration track group; 。 3. The global sensor system error partition registration algorithm based on similarity principle according to claim 1 is characterized in that: The specific method of step S4 is: Track combination after partition pairing and Establish error registration model; represents the measurement of sensor A and sensor B at time k; λ = [ΔR A ,Δθ A ,ΔR B ,Δθ B ] T represents the systematic error measured by radar A and radar B. The coordinates after error registration are expressed as: After registration at the same time, the observation points of radar A and radar B on the same target coincide, which satisfies Construct an objective function and use nonlinear optimization method to estimate the system error of each radar. The goal of two radar registration is to find a set of system error values ​​to minimize the distance between the registered track points. Therefore, the registration goal is expressed as follows: Gmn(k) represents the Euclidean distance of the same target measured by radar sector m and radar sector n at the same time after system error registration, indicating the degree of difference between the two measurement points after registration; for the data of M measurements, find a set of λ such that Take the minimum value; the genetic algorithm is used in the solution process.

Images