A multi-sensor air target track data fusion method based on comprehensive criterion

Abstract

The application belongs to the field of data processing and aviation technology, and discloses a multi-sensor air target track data fusion method based on comprehensive criteria, which solves the problem of multi-air target track data fusion in multi-air platform multi-element sensors through track target grouping, track target fusion, time track association and target track fitting. The application fully considers factors such as distributed multi-sensor type, detection accuracy, detection persistence and detection interference, can stably maintain the situation information of multiple air targets, and improves target positioning accuracy.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of data processing and aviation technology, and particularly relates to a multi-sensor air target track data fusion method based on comprehensive criteria. BACKGROUND

[0002] In the process of air platform cooperative combat, the sensor system as an important means of target detection provides basic target information input for combat decision-making, and is an indispensable part in the combat process. Due to the great gap in detection performance such as detection range, detection accuracy, anti-interference ability, etc. of different sensors, through the comprehensive use of multi-type sensor data on multiple combat platforms, the sustained and accurate positioning of targets is realized, which plays a very key role in air combat victory.

[0003] The air platform is equipped with radar, electronic countermeasure system, data link, photoelectric radar and other types of sensors. The air target detection results obtained by these sensors are mostly track data, that is, the current time coordinate point of the target is output. Due to different detection methods, the radar outputs the target azimuth angle, elevation angle, distance and other format information, the electronic countermeasure system outputs the target azimuth angle and elevation angle, and the data link outputs the target longitude, latitude, height and other format information. Moreover, due to the limitation of detection principle, the radar and electronic countermeasure system cannot guarantee the sustained detection of the target; due to the limitation of detection accuracy, the data provided by the data link has large error and is discontinuous, which cannot meet the real-time countermeasure demand.

[0004] At present, air combat has developed from the original single aircraft combat to cooperative combat between multiple combat platforms, and the comprehensive use of multiple air platforms realizes the complementary advantages and information integration capability. The multi-platform heterogeneous sensor data fusion helps to maintain the enemy target information with high quality and provides strong data support for air combat decision-making, which is the key to realize cooperative combat. At present, there are mainly two problems in the air platform data fusion in China. One is that the theoretical research is divorced from engineering application. Some colleges have carried out data fusion algorithm research based on simulation, but due to the low fidelity of the simulation model, the actual combat situation of the air platform cannot be simulated, and the related theories and methods have low engineering application value. The other is that the cooperative combat system is not perfect. Some research institutions have carried out some exploration in single-platform sensor data fusion and have produced some application value, but the multi-aircraft cooperative sensor data fusion is still not perfect. SUMMARY

[0005] To solve the above problems, the application provides a multi-sensor air target track data fusion method based on comprehensive criteria. By establishing a track target correlation comprehensive criterion algorithm, the target position, target speed, target number and other track target information are used to comprehensively analyze the detection results of multiple sensors, and finally the high-precision track information of all targets is given, which effectively improves the air platform sensor track data fusion capability.

[0006] A multi-sensor air target track data fusion method based on comprehensive criteria, comprising the following steps:

[0007] (1) Track target grouping

[0008] For the target track information obtained at each time, the target position, speed, height and other information are comprehensively considered, a target grouping comprehensive criterion is established, the same entity target detected by different sensors is classified and grouped, and a plurality of entity target groups are formed.

[0009] The target minimum position error dD = |P1-P2|, the target minimum speed error dV = |V1-V2|, and the target minimum height error dH = H1-H2 are set, a plurality of targets of a sensor of a certain platform are selected as a reference, and each target is compared with other sensors: the target whose position error is less than dD and whose speed error is less than dV and whose height error is less than dH is identified as the same group target; otherwise, it is considered as a new reference target, and recursive grouping is performed.

[0010] (2) Track target fusion

[0011] Based on the target grouping result of step (1), the target tracks in the same group are fused, the fusion weight values of each sensor are set, and the fusion results of each group of tracks are calculated by weighted average.

[0012] The target track fusion weights of different types of sensors S1, S2…S n are set as α1, α2…α n The data of different sensors of the same group target are fused by weighting Where X j is the fusion result of the jth group target, k j is the number of targets of the jth group target, is the fusion weight of the ith target in the jth group target (consistent with the sensor type), is the coordinate data (including position and speed) of the ith target in the jth group target.

[0013] (3) Time track association

[0014] For the target track fused in step (2), a time track association comprehensive criterion is established through target number, position, target speed and other information, the current track is matched and associated with the existing track, and the consistency and continuity of the target are maintained. The specific process is as follows:

[0015] (3a) For the first frame of target track data, the target grouping result of step (2) is used as the initial track point of each target;

[0016] (3b) The point association method is as follows: Set the minimum error for position association dD′, the minimum error for velocity association dV′, and the minimum error for height association dH′. For the target point set P′={p1′,p′2…p′} of the previous frame, u}, and the current frame target point set P = {p1, p2, ..., p v}, where p and p′ are 6-dimensional vectors containing position and velocity information (global North-Sky-East coordinate system), i.e. (x n ,x u ,x e ,v n ,v u ,v e For each point p in set P, compare it with all points in set P′. Among the points that simultaneously satisfy the conditions that the position error is less than dD′, the velocity error is less than dV′, and the height error is less than dH′, select the point p with the smallest distance error. j ′ as p i Related points;

[0017] (3c) For target point data that is not in the first frame, after grouping it in step (2), it is associated with the previous frame data of the existing target track. If the association with the existing track is successful, it is used as the latest point of the track. If it cannot be associated with the existing track, it is used as a new target for track management.

[0018] (3d) After the point traces are associated, the historical point traces of different targets are managed. Newly emerging dots are treated as new targets and managed accordingly.

[0019] (4) Target trajectory fitting

[0020] For the target coarse track generated in step (3), in order to address the problems of discontinuity, non-smoothness and large error of the existing target flight track, the tracks of multiple targets are fitted by methods such as track association error detection and track smoothing, so as to generate the track fitting result that is closest to the target flight trajectory.

[0021] For each target trajectory, the UnivariateSpline curve fitting method is used. The fitting method parameters k and s are set, and the number of fitting points N is used. A sliding window method is adopted, that is, the N points after the reciprocal of the target are taken for fitting each time. The fitting function is generated in real time, and the value of the function at the current time point is calculated as the fitting result of the current point of the target.

[0022] Beneficial effects: The application is aimed at air target track data, and a target grouping and time track association judgment basis considering target position, speed, height and other information is designed. Through four main steps of target grouping, track fusion, track association and track fitting, effective fusion of multi-sensor multi-target track data is realized, the same target is filtered from complex observation information, and target observation information error is also reduced. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 is the implementation flowchart of the application.

[0024] Figure 2 is a data track data fusion result case diagram of the application. DETAILED DESCRIPTION

[0025] As shown in Figure 1 , a multi-sensor air target track data fusion method based on comprehensive criteria, the implementation process includes the following steps:

[0026] (1) Track target grouping

[0027] Set the target minimum position error dD, the target minimum speed error dV, the target minimum height error dH, select a plurality of targets of a sensor of a platform as a reference, and compare each target with other sensors comprehensively: the target whose position error is less than dD and whose speed error is less than dV and whose height error is less than dH is identified as the same group of targets, otherwise, it is considered as a new reference target, and recursive grouping is performed.

[0028] In this embodiment, the target minimum position error dD is set to 3000 m, the target minimum speed error dV is set to 100 m / s, and the target minimum height error dH is set to 1000 m. The track data of three types of sensor targets of two aircrafts is grouped, and a total of two groups G1 and G2 are formed.

[0029] (2) Track target fusion

[0030] Set the target track fusion weights of different types of sensors S1, S2…S n α1, α2…α n , weight and fuse the data of different sensors of the same group of targets wherein X j is the fusion result of the jth group of targets, k j is the target quantity of the jth group of targets, is the fusion weight corresponding to the ith target in the jth group of targets (consistent with the sensor type), is the coordinate data (including position and speed) of the ith target in the jth group of targets.

[0031] In this embodiment, for radar S1, data link S2, electronic warfare S3 and other three types of sensors, the target track fusion weight is set as a1=90, a2=20, a3=10, the track fusion is performed for the target in G1 group to obtain the fusion track X1, and the track fusion is performed for the target in G2 group to obtain the fusion track X2.

[0032] (3) Time track association

[0033] For each time point target track fusion result generated in step (2), the target time sequence track is associated. The specific process is as follows:

[0034] (3a) For the first frame target track fusion data and as the starting point of the track of the two targets;

[0035] (3b) Set the position association minimum error dD'=1000m, the speed association minimum error dV'=100m / s, and the height association minimum error dH'=100m. For the first frame target and the second frame target and satisfy the association minimum error limit, and satisfy the association minimum error limit, then and two time tracks of the same target, and two time tracks of another target are considered; (3c) For the track

[0036] and the last frame track of the existing track and are associated, and so on; (3d) Management of target track

[0037]

[0038] (4) Target track fitting

[0039] For each target track, the UnivariateSpline curve fitting method is adopted, the fitting method parameters k and s are set, the number of fitting tracks N is set, the sliding window is adopted, that is, the last N tracks of the target are taken each time for fitting, the fitting function is generated in real time, and the value of the function at the current time point is calculated as the fitting result of the current track of the target.

[0040] ​In this embodiment, the fitting parameter k is set to the default value, N = 20, s n = 800000000, s u = 500000000, s e = 800000000, s vn = 80000, s vu = 50000, s ve = 80000, the north-sky-east position coordinates and the speed of the target are fitted respectively, a UnivariateSpline curve function is generated, and the fitting result of the latest point is calculated as a fitting point track for track management.

[0041] In this embodiment, the experimental environment parameters are as follows: CPU: Intel(R) Xeon(R) Platinum 8163 @ 2.5 GHz, memory: 64 G, operating system: Ubuntu 18, based on x64 processor. The processing of the data set and the simulation experiment of data fitting are both based on Python 3.6 and programmed using numpy and scipy framework.

[0042] The following table is a test result statistical table of the point track data fusion of the present application.

[0043]

[0044]

Claims

1. A multi-sensor air target track data fusion method based on comprehensive criterion, characterized in that, Comprising the following steps: (1) Target grouping For each time point of the target track information obtained, a target grouping comprehensive criterion is established, the same entity target detected by different sensors is classified and grouped, and a plurality of entity target groups are formed; A target minimum position error dD = |P1-P2|, a target minimum speed error dV = |V1-V2|, and a target minimum height error dH = H1-H2 are set, a plurality of targets of a sensor of a certain platform are selected as a reference, and each target is compared with other sensors: the target whose position error is less than dD, and whose speed error is less than dV, and whose height error is less than dH is identified as the same group target; otherwise, it is considered to be a new reference target, and recursive grouping is performed; (2) Target track fusion Based on the target grouping result of step (1), the target tracks in the same group are fused, the fusion weight values of each sensor are set, and the fusion results of each group of tracks are calculated by weighted average; Set different types of sensors S1, S2…S n Target point trace fusion weight α1, α2…α n , the same group of targets for different sensor data weighted fusion Where X j is the fusion result of the jth group of targets, k j is the number of targets in the jth group of targets, α ji is the fusion weight corresponding to the ith target in the jth group of targets, is the coordinate data of the ith target in the jth group of targets; (3) Time track association For the target track fused in step (2), a time track association comprehensive criterion is established, and the current track is matched and associated with the existing track; the specific process is as follows: (3a) For the first frame of target track data, the target grouping result of step (2) is used as the initial track point of each target; (3b) The point association method is as follows: Set the minimum error for position association dD′, the minimum error for velocity association dV′, and the minimum error for height association dH′. For the target point set P′={p′1,p′2…p′} in the previous frame... u }, and the current frame target point set P = {p1, p2, ..., p v }, where p and p′ are 6-dimensional vectors, i.e. (x n ,x u ,x e ,v n ,v u ,v e For each point p in set P, compare it with all points in set P′. Among the points that simultaneously satisfy the conditions that the position error is less than dD′, the velocity error is less than dV′, and the height error is less than dH′, select the point p′ with the smallest distance error. j As p i Related points; (3c) For the target track data other than the first frame, after grouping by step (2), the last frame data of the existing target track is associated, if the association with the existing track is successful, it is used as the latest track of the track, if the association with the existing track fails, it is used as a new target for track management; (3d) After the association of the tracks is completed, the historical tracks of different targets are managed The newly appearing tracks are managed as new targets; (4) Target track fitting For the target rough track generated in step (3), the track association error detection and track smoothing processing method are used to fit the tracks of a plurality of targets respectively, and the track fitting result closest to the target flight trajectory is generated; For each target track, the UnivariateSpline curve fitting method is used, the fitting method parameters k and s are set, the track quantity N is set, the sliding window method is used, the last N tracks of the target are taken each time for fitting, the fitting function is generated in real time, the value of the function at the current time point is calculated as the fitting result of the current track of the target.

Images