CN101655561A - Federated Kalman filtering-based method for fusing multilateration data and radar data - Google Patents

Abstract

The invention relates to a Federated Kalman filtering-based method for fusing multilateration data and radar data, comprising seven processes: pretreatment, time and space registration, error correction, single-sensor line tracking or airborne trace interconnection and filtration, data association, data fusion as well as error estimation. On the basis of multiclass monitoring information, the technical scheme of the invention establishes data association mapping relations among all classes of targets, constructs a process method fusing the multilateration system data and the radar data, realizes deep fusion of all classes of information and comprehensive utilization on effective information, ensures high updating rate of a multilateration system and the whole precision of system monitoringwhen period is less than or equal to 1sec, reduces impact of radar measuring error on system monitoring precision, and significantly improves the tracking precision of the system.

Description

Multipoint positioning data and Radar Data Fusion method based on federated Kalman filtering

Technical field

The present invention relates to a kind of multisensor data fusion processing method, particularly a kind of multipoint positioning data and Radar Data Fusion method based on federated Kalman filtering.

Background technology

Aircraft moves or flight aloft on the ground, must carry out dynamic surveillance, implements overall process control, to guarantee flight safety.Therefore, various monitoring sensors have been installed all on airport and air route, to obtain the activity situation information of controlled object.

The multisensor data fusion processing technology is the gordian technique of airdrome scene movable guiding control system (A-SMGCS).It can be joined together the various sensor informations of connecting system, carries out fusion treatment, with unique, accurate, real-time, the effective information of acquisition target, and shows on terminal, implements scene monitoring and control task to help the controller.

Monitor message commonly used in the air traffic control system has occurred ADS-B information, multipoint location system (MLAT) information of satnav again in recent years based on air traffic control radar information.

Air traffic control radar comprises aviation management primary radar (PSR), aviation management secondary surveillance radar (SSR), the tradition that is air traffic control system monitors means, also be the equipment that is using on the airport, the monitoring range of SSR can reach 400 kilometers, and the monitoring range of PSR can be above 100 kilometers.Because the restriction of Radar Technology system, its coverage rate, data precision, Data Update cycle etc. have been limited the raising to target tracking accuracy and real-time.

Multipoint location system is the external a kind of novel surveillance technology that proposes, this technology makes full use of the A/C pattern and the S pattern of existing airborne answering machine, adopt multiple spot passive sensor receive mode, according to the mistiming of signal arrival Different Ground receiving station, determine the accurate position of aircraft or other moving target.Can discern target by SSR code in the A/C pattern or the unique address code of S pattern simultaneously.The bearing accuracy of multipoint location system is very high, can reach in 10 meters in the ground bearing accuracy; Its update cycle is very short, is generally less than 1 second.This system installs on the airport usually, and its coverage can not surpass 100 kilometers on airport on every side.

For nearly supervisory system and airdrome scene activity monitoring system are entered in five limits, they should the surveillance distance airport traffic in quite on a large scale, very accurate target position information is provided again.Traffic more complicated and variation are rapid on every side on the airport, and system must guarantee certain turnover rate.The access of air traffic control radar can guarantee the coverage of system, and the access of multipoint location system can guarantee coverage rate, renewal rate and the supervision precision of system.The present invention is exactly with two category informations while connecting system, carries out fusion treatment, respectively gets its length, obtains best information effect.

Present aviation management surveillance radar information and multipoint location system information are carried out fusion treatment and are also had many technological difficulties.

At first, air traffic control radar is different with the information updating cycle of multipoint location system, and falls far short, and in 4～10 seconds air traffic control radar information updating cycles, MLAT was less than 1 second; And multipoint location system does not have the stable cycle, sends repeatedly the information about same target in common 1 second; Secondly, the supervision precision of system requirements is very high, the supervision precision of tradition air traffic control radar on the air route can reach tens meters and can satisfy the control requirement, surveillance is relative with the supervision precision of scene activity monitoring system wants high a lot of and five limits are closely advanced, even several meters error also may cause great security incident on the ground sometimes, so it monitors that precision need guarantee in 10 meters.

Existing sensor Data Fusion is the Data Fusion at the multi-section air traffic control radar mostly, and whole implement process comprises that mainly space-time registration, flight path pairing and flight path merge this several links.

The space-time registration is because each sensor may not have the unified time benchmark, and measured target position data also all is in local coordinate system separately, just must be data-switching in unified space-time coordinates for each sensing data is merged.Time alignment technique commonly used at present mainly contains the timeslice technology and the two kinds of methods of starting point of sampling together.Spacial alignment is with the target location under the local coordinate system of different sensors, through conversion process such as rotation of coordinate, translation, projection, is scaled the data under the public rectangular coordinate system of system centre.

The flight path pairing is the track association that belongs to same target that will be detected by different sensors, each sensor flight path can be according to the position in the flight path data, highly, information such as the 3/A schema code of speed, flight path number and aircraft or unique identifier mates, and then be associated with corresponding flight planning, the realization of using for various control provides foundation.In real-time multiple target tracking process, the measurement that same target is set up on a plurality of sensors, must have certain similar features because of its physics source is identical, also must be because of the instability of noise jamming and sensor self performance, and the feature that causes these metric data to be set up is incomplete same.The purpose of data association is exactly the similar features with this measurement, judges whether the incomplete same metric data of these features comes from same target.The main method of data association has arest neighbors data association, probabilistic data association, JPDA, arest neighbors JPDA, flight path bifurcated method, maximum likelihood data association, Zero-one integer programming method and many subjunctives or the like.

It is that each single channel flight path is merged that the multisensor flight path merges, and obtains the system synthesis flight path.For data fusion, its structure difference will cause it that different system performances is arranged.From the angle of target following, multi-sensor fusion system has three kinds of typical structures: centralized fusion structure, distributed fusion structure and hybrid fusion structure.Centralized fusion structure can utilize the full detail of all the sensors to carry out state estimation, velocity estimation and predictor calculation.Its major advantage is to have utilized full detail, and the information loss of system is little, performance good, the estimation of the state of target, speed is optimum estimate.But contain much information, communication overhead is big, computing machine is had relatively high expectations.To be fusion center merge the local Filtering Estimation of each sensor to distributed fusion structure, obtains the overall situation and estimate.This structure communication overhead is little, fusion speed fast, it is low that computing power is required, but its performance is not as centralized fusion structure.Hybrid fusion structure is the applied in any combination of centralized fusion structure and distributed fusion structure, and it has the advantage of these two kinds of structures concurrently, but the calculated amount of complex structure, processor is very big.Referring to: Chang Le, " applied research of data fusion in the flight path data processing ",, Institutes Of Technology Of Nanjing's master thesis in 2006.And display, " error correction of radar intelligence (RADINT) data fusion system and track association technical research ",, Institutes Of Technology Of Nanjing's master thesis in 2007.

There is following shortcoming in the prior art scheme: the data fusion that carry out between air traffic control radar and the multipoint location system (1) can't make system's flight path reach higher precision.Air traffic control radar has bigger systematic error than multipoint location system.If merge according to the metric data of the fusion architecture in the existing scheme with the two, the systematic error that air traffic control radar is big can't obtain revising, and will directly influence the result of flight path pairing.If the threshold value of selecting when doing the flight path pairing is less, with should being correlated with, some can't be correlated with for the gauge point of same target, cause the flight path division.On promptly allowing to be correlated with, a part of systematic error of air traffic control radar is inherited by last fusion results.Even multipoint location system has very high measurement accuracy like this, can not assurance system output result have higher precision.(2) can't solve owing to certain sensor detection accuracy reduces the problem that causes the total system performance to descend.Sensor is along with the growth of tenure of use, and its detection accuracy will reduce.In addition, the variation of weather also can influence the detection accuracy of sensor to a certain extent.Yet, when the detection data that processes sensor is sent here, do not consider the detection accuracy of this sensor in the present actual use, treat and the data that all the sensors is sent are equal to.Like this, if there is certain sensor detection accuracy seriously to descend,, the decline of system accuracy will be caused directly according to existing technical scheme.(3) can't solve the problem that multipoint location system does not have stable period.Traditional air traffic control radar all has the fixing scan period, generally is 4 seconds.Existing technical scheme all is to carry out data fusion under such prerequisite.And multipoint location system does not have the stable update cycle, can send several measuring values of same target in general 1 second.Also can't handle such problem according to existing technical scheme.

Summary of the invention

Goal of the invention: technical matters to be solved by this invention is at the deficiencies in the prior art, a kind of multipoint positioning data and Radar Data Fusion method based on federated Kalman filtering is provided, carry out data fusion by the federated Kalman filtering device, guarantee that under the higher prerequisite of data updating rate, the overall precision of monitored object is higher than the precision of arbitrary single information source.

Technical scheme: the invention discloses a kind of multipoint positioning data and Radar Data Fusion method, may further comprise the steps based on federated Kalman filtering:

Pre-treatment step promptly receives the aim parameter measurement information that each sensor is sent, and according to analysiss of decoding of the data layout of each sensor definition, and the aim parameter measurement information of different-format is converted to consolidation form explains;

The space-time step of registration, soon the target position information in the aim parameter measurement information of each sensor acquisition is transformed under the same coordinate system, and is registered to synchronization; Target position information is the location entries that extracts from the aim parameter measurement information that sensor is received, described location entries is represented the target location with longitude and latitude, rectangular coordinate or polar form.

Point mark or the interconnected step of flight path are about to the different information processings constantly of same target that same sensor detect and become with a collection of flight path; When an aim parameter measurement information begins to carry out a mark or the interconnected process of flight path, at first be that the planimetric position that comprises at center is made as relevant ripple door with the certain limit of height with this target, this Bo Mennei has been deposited flight path as the affiliated partner hunting zone; Check the degree of agreement of this aim parameter measurement information and relevant each flight path of Bo Mennei then one by one, relevant factor comprises flight path number, the secondary code of target, the identity information and the velocity information of destination address; Set up factor of influence at every kind of key element, meet for just, be not inconsistent be combined into negative; For each factor of influence, if greater than preset threshold, expression is relevant, otherwise uncorrelated;

Adopt local filter to carry out each sensor flight path filter step, obtain the local optimum of target current location is estimated; Described flight path filtering may further comprise the steps: the equation of motion of target is considered as n dimensional linear dynamic system, during the sensor tracking target, the flight path that observes is considered as m dimensional linear recording geometry, and then the discrete description form of the state equation of i local filter and measurement equation is:

$\left\{\begin{array}{cc}{S}_i\left(k\right)=A\left(k\right)S_i(k-1)+w_i(k-1)& k=\mathrm{1,2},\·\·\·\\ X_i\left(k\right)=C_i\left(k\right)S_i\left(k\right)+v_i\left(k\right)& i=\mathrm{1,2},\·\·\·,q\end{array}\right.;$

S wherein _i(k) the expression system is at k state constantly; A (k) is the system state transition matrix; w _i(k-1) (k=1,2 ...) be illustrated in the random disturbance that k acts on system constantly, i.e. plant noise; Described plant noise is assumed to be the white Gaussian noise sequence, promptly has known zero-mean and covariance matrix Q _i(k); X _i(k) be observation vector; C _i(k) be observing matrix; v _i(k) (k=1,2 ...) be observation noise, be set at the white Gaussian noise sequence, promptly have known zero-mean and covariance matrix R _i(k);

The filtering recursion formula is:

ε′ _i(k)＝A(k)ε _i(k-1)A ^T(k)+Q(k-1)

$H_i\left(k\right)={\mathrm{\ϵ}}_i^{\′}\left(k\right)C_i^T\left(k\right){[C_i\left(k\right){\mathrm{\ϵ}}_i^{\′}\left(k\right)C_i^T\left(k\right)+R_i\left(k\right)]}^{-1}$

ε _i(k)＝[I-H _i(k)C _i(k)]ε′ _i(k)

Wherein, Q _i(k) be plant noise w _i(k-1) covariance matrix; R _i(k) be observation noise v _i(k) covariance matrix;

Be that k is constantly through filtered valuation; H _i(k) be gain matrix; ε _i(k) be the evaluated error covariance matrix; Through the filter process of decentralized concurrent operation, obtain local optimum and estimate (k=1,2 ...), and the filtering result in per step passed to senior filter;

The data association step judges that whether information that different sensors detects is about same target.When an aim parameter measurement information of certain sensor begins to carry out the data association process, at first be that the planimetric position that comprises at center is made as relevant ripple door with the certain limit of height with this target, this Bo Mennei has been deposited flight path as the affiliated partner hunting zone; Check the degree of agreement of this aim parameter measurement information and relevant each flight path of Bo Mennei then one by one, relevant factor comprises flight path number, the secondary code of target, the identity information and the velocity information of destination address; Set up factor of influence at every kind of key element, meet for just, be not inconsistent be combined into negative; For the summation of each factor of influence, if greater than preset threshold, expression is relevant, otherwise uncorrelated.Aim parameter measurement information for shutting mutually with flight path carries out flight path filtering and fusion.For the aim parameter measurement information of not shutting mutually with any flight path, set up new flight path.

Data fusion step, the information that the different sensors of same target is detected merges, and the local optimum that decentralized concurrent operation obtains according to the local filter of different sensors correspondence is estimated

(k=1,2 ...), in senior filter, merge; In senior filter, overall fusion results is:

S＝(ε ₁ ^-1+ε ₂ ^-1+…+ε _q ^-1) ^-1(ε ₁ ^-1S ₁+ε ₂ ^-1S ₂+…+ε _q ^-1S _q)

ε＝(ε ₁ ^-1+ε ₂ ^-1+…+ε _q ^-1) ^-1

Overall estimated result is fed back to each local filter, as the k estimated value of each local filter constantly:

S _i(k)＝S(k)

$Q_i^{-1}\left(k\right)={\mathrm{\α}}_i{Q}^{-1}\left(k\right)$

${\mathrm{\ϵ}}_i^{-1}\left(k\right)={\mathrm{\α}}_i{\mathrm{\ϵ}}^{-1}\left(k\right)$

α ₁+α ₂+…+α _q＝1

Wherein, i=1,2 ..., q, 0≤α _i≤ 1;

Senior filter is finished the optimal synthesis of information, forms the integrated information of global system; After each filtering stage is finished, estimate and the information distribution amount, carry out information feedback to each local filter, thereby adjust each parameter value of local filter in real time by the overall situation that senior filter will synthesize;

The average weighted weights of each sensor are determined according to the measurement position of the current test point of this sensor and the departure degree of physical location; The test point physical location of supposing i portion sensor is (x _i, y _i), and the position quantity measured value is

Alternate position spike between them is so:

${\mathrm{\Δr}}_i=|\sqrt{x_i^2+y_i^2}-\sqrt{{\hat{x}}_i^2+{\hat{y}}_i^2}|$

The gap of the physical location of test point and measurement position is that the current precision of position difference and sensor is inverse relation; Determine that by the position difference this sensor shared weights in the weighting fusion of all n portion sensors are:

${\mathrm{\&alpha;}}_i=\frac{\frac{1}{{\mathrm{\&Delta;r}}_i}}{\underset{i}{\overset{\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A2009100350310020H1.png"\; state="\{\{state\}\}">n</figure-callout>}}{\mathrm{\&Sigma;}}}\frac{1}{{\mathrm{\&Delta;r}}_i}}.$

Among the present invention, also comprise estimation of error step and error correction step after the data fusion step;

Described estimation of error step comprises, will ask difference to calculate through filtered each the target location vector measuring value corresponding with each air traffic control radar of flight path, obtains the deviate about each target; The deviate of all targets that an interior air traffic control radar of statistics current time in some cycles forward detects, and do on average to obtain the systematic error of the estimation of this air traffic control radar; The public flight path number that detects with this air traffic control radar in the comprehensive flight path is m, and n cycle added up; The position vector that is detected by air traffic control radar in k cycle j target is S _jFiltered comprehensive flight path position vector is M _jThe estimated value e of i portion air traffic control radar systematic error _iFor:

$e_i=\underset{j}{\overset{m}{\mathrm{\&Sigma;}}}\underset{k}{\overset{n}{\mathrm{\&Sigma;}}}(S_j-M_j);$

The systematic error e of the estimation of each air traffic control radar of statistics _i

Described error correction step comprises, with the estimating system error e of each air traffic control radar in the estimation of error step with respect to comprehensive flight path _iAim parameter measurement information behind the space-time registration is carried out error correction; If current time calculates the position vector estimating system error e of the i portion air traffic control radar that obtains from estimation of error _i, and current this road radar is m to the position vector measuring value of j target _j, the position vector after this target is corrected is

Then the pass between them is:

${\hat{m}}_j=m_j+e_i;$

This road radar circulated in these all targets that detect constantly revise the position quantity measured value of the systematic error that has been eliminated.

Among the present invention, in the space-time step of registration, be registered to synchronization and adopt the timeslice method, be about to a timing cycle as the temporal clustering point, in a temporal clustering, be reference point with this timing cycle, with the extrapolation of the data of other observation station or in be inserted into this reference point; With a plurality of observations in the temporal clustering as a virtual observation on temporal clustering point; The state transitions formula that adopts is with t ₁State extrapolation constantly or in be inserted to t ₂Constantly; Described state transitions formula is:

$\hat{X}\left(t_2\right)=\mathrm{\&Phi;}(t_2-t_1)\hat{X}\left(t_1\right)$

Wherein, Be state vector; Φ (t) is a state-transition matrix.

Among the present invention, in some mark or the interconnected step of flight path,, adopt minimal distance principle to determine the last flight path relevant with the aim parameter measurement information if search a plurality of flight paths relevant with the aim parameter measurement information at relevant Bo Mennei; If relevant Bo Mennei do not find can be relevant with this target flight path, just with this target as new flight path, add in the flight path chained list.

Among the present invention, comprise in the described error correction step that aim parameter measurement information report selects excellent, the aim parameter measurement information report that is used to multipoint location system is presented has the stable cycle, at first multipoint location system information of receiving and the multipoint positioning flight path of having set up are done pre-relevant, if shut mutually, just contrast both timestamp and position, whether judgement their alternate position spike in both mistimings is reasonable, if rationally, further compare timestamp, keep and to be separated by, otherwise give up near 1 second information input information as this update cycle.

Among the present invention, multipoint location system aim parameter measurement information report is selected and is excellently implemented in the error correction step.Behind the space-time registration, the aim parameter measurement information of multipoint location system is arrived under system coordinates and the time shaft by unified, owing to can send a plurality of information for same target multipoint location system in 1 second,, need select excellent to these a plurality of information in order to obtain 1 second stable period.Select excellent rule be select time stab from last one-period near 1 second and reasonably that information as the input in this cycle.Through choosing excellent after, just multipoint location system can be used as sensor and handle with stable period, normally enter subsequent processing steps, do the interconnected and filtering of a mark or flight path, and then carry out related, data fusion with the data of other sensors.

Beneficial effect: technical solution of the present invention is on the basis of multiclass monitor message, set up the data association mapping relations between all kinds of targets, make up a process approach with multipoint location system data and the processing of many Radar Data Fusion, the degree of depth that realizes multiclass information merges and to the comprehensive utilization of effective information, the overall precision of system monitoring when guaranteeing multipoint location system report turnover rate height (cycle＜=1 second), reduced of the influence of radar measurement error, the tracking accuracy of system is significantly improved the system monitoring precision.

The present invention has the following advantages: (1) according to the detection accuracy of each sensor test point, adjusts this sensor shared weight in fusion in real time, thereby has solved owing to certain sensor accuracy reduces the problem that causes the total system performance to descend.And each sensor of existing technology shared weight in fusion is fixed, and can't therefore when certain sensor performance decline, will directly cause system performance to descend according to the actual performance adjustment of sensor.

(2) adopt the Error Feedback correction technique, revise the systematic error of each sensor in real time, thereby improve the overall precision of system with respect to comprehensive flight path.Also do not utilize the high characteristics of multipoint location system precision in existing engineering is used, actually adopt such technology, when the systematic error of certain sensor was big, system can't perception, thereby the precision of the comprehensive flight path in back is merged in influence.

(3) several measurement informations about same target that multipoint location system is sent in 1 second select excellent processing, have both guaranteed 1 second cycle that multipoint location system is stable, and it is optimum at that time that the target that obtains from multipoint location system is measured.This type of disposal route at the multipoint location system characteristic does not also appear in the existing technical scheme.

Description of drawings

Below in conjunction with the drawings and specific embodiments the present invention is done further to specify, above-mentioned and/or otherwise advantage of the present invention will become apparent.

Fig. 1 is multipoint positioning data and the Radar Data Fusion method flow diagram that the present invention is based on federated Kalman filtering.

Fig. 2 is the interconnected process flow diagram of data among the present invention.

Fig. 3 is federated Kalman filtering structural drawing among the present invention.

Fig. 4 is the mapping table of internal system form and ASTERIX001 form in the embodiment of the invention.

Fig. 5 a, Fig. 5 b, Fig. 5 c are system of utilization the inventive method and the contrast of using the classic method system.

Embodiment:

As shown in Figure 1, multipoint positioning data and the Radar Data Fusion method based on federated Kalman filtering of the present invention comprises seven processes: pre-service, space-time registration, error correction, single-sensor point mark or flight path is interconnected and filtering, data association, data fusion and estimation of error.

Because single-sensor point mark or the interconnected related principle with multi-sensor data of flight path are similar, it is interconnected to be referred to as data, sets forth in the data interconnect portion.And therefore flight path filtering and data fusion close association is in logic also being described with a part.

(1) pre-treatment step

System is after receiving the aim parameter measurement information that each sensor is sent, at first will be according to the analysis of decoding of the data layout of each sensor definition, the error message that rejecting may exist, and the aim parameter measurement information of different-format is converted to the statement of internal system consolidation form, finish the information pre-service.

The information of the different sensors that system receives has different forms, they normally transmit according to the form that defines among the standard document ASTERIX of EUROCONTROL about the monitoring data exchange, it is to carry out information transmission according to the self-defining form of manufacturer that some sensors are also arranged, for example Alenia radar etc.The form of these forms and internal system definition there are differences, so system need resolve their content when receiving sensor information, the form of analysis result according to the internal system definition is assembled, and give the next procedure processing.With 001 format analysis in the ASTERIX document commonly used and be assembled into example.The ASTERIX001 form can be referring to EUROCONTROLSTANDARD DOCUMENT FOR Radar Data Exchange-Part 2a Transmission ofMonoradar Target Reports, SUR.ET1.ST05.2000-STD-02a-01, November 1997, ReleasedIssue.The corresponding relation of internal system form and ASTERIX001 form as shown in Figure 4.The internal system data layout has comprised by being drawn and has connect the needed data item of system in the sensing data form.ASTERIX001 is the form of air traffic control radar, and representation, destination address, catchword and these data item that provided by multipoint location system usually of height of target location longitude and latitude are not provided for it.Therefore, after obtaining the data of a frame ASTERIX001 form, each data item is extracted, store into then in the data item of corresponding system definition by the definition of form.In this process, because the problem or the transmission line failure of sensor itself, may exist some information be invalid, be not inconsistent with given form or do not meet logic, need do the rationality inspection to the content after resolving, unreasonable message rejected and take statistics.Next treatment step is assembled and delivered to information only with a grain of salt just according to the internal system form.

(2) space-time registration

After pre-service, the aim parameter measurement information that different sensors obtains is transformed under the consolidation form.But the position quantity measured value that comprises in the information obtains under each sensor coordinate system, and the coordinate system of sensor is common and system coordinate system is inconsistent.In addition, though sensor and system all be adopt gps clock to the time, may not necessarily be consistent but sensor obtains the moment of target with the system timing cycle, may early than or be later than regularly constantly, and must record at synchronization the target location of each sensor when relevant doing flight path.Therefore to the aim parameter measurement information do flight path relevant and merge before, must the coordinate system of target location is unified, time alignment, otherwise can cause relevant error and fusion results invalid.In space-time registration link, the target position information that each sensor is obtained is transformed under the system coordinate system, and is registered to synchronization, promptly finishes the information synchronization processing.

Spatial registration through conversion processes such as rotation of coordinate, translation, map projections, converts the aim parameter measured value under the different sensors coordinate system under the system coordinate system data.Owing to will be presented at the target location on the control seat man-machine interface, therefore adopting with system centre in the system usually is that initial point, direct north are that y axle, due east direction are that the plane right-angle coordinate of x axle is a reference frame.

Traditional sensors, for example SSR, the PSR target location that obtains is to be that initial point, direct north are the polar coordinate representation of angle benchmark with the sensing station.In addition, some novel sensors, for example MLAT, ADS-B etc., the target location of detecting is adopted terrestrial coordinate usually, and promptly longitude and latitude is represented, perhaps is that initial point, direct north are that y axle, due east direction are that the local rectangular coordinate system of x axle is represented with the sensing station.

With these data-switching under system coordinate system, different representations, flow path switch difference.If longitude and latitude form, at first adopt the Gauss projection formula just to calculate, be converted to the planimetric coordinates under the initial point coordinate system under the line, then true origin move to the system centre point (true origin that system centre point is set in system exactly, in the system all about the calculating of position all as with reference to); If polar form, at first being transformed into sensing station is under the part plan rectangular coordinate system of initial point, according to known sensor terrestrial coordinate, planimetric rectangular coordinates is become longitude and latitude by the Gauss projection inverse then, be transformed into again under system's rectangular coordinate system; If the part plan rectangular coordinate is scaled terrestrial coordinate equally earlier, be transformed into again under system's rectangular coordinate system.

Time alignment just is meant in sheet sometime, and the target observation data in this timeslice of each sensor acquisition are carried out interpolation or extrapolation, and it is registered on some unified time, just carries out time normalization and handles.The usually free chip technology and the two kinds of time alignment technique commonly used of starting point of sampling together.The timeslice technology is that whole observation process is divided into the some time sheet, and the data that collect on each timeslice are carried out temporal clustering, just can obtain several temporal clustering points like this on a timeslice.In a temporal clustering, be reference point with such temporal clustering point, with the interpolation of data of other observation station or be extrapolated to this reference point, a plurality of observations in such temporal clustering just can be regarded a virtual observation on temporal clustering point as.

The present invention adopts improved timeslice technology, omits cluster process, simply with the timing cycle of system as the temporal clustering point.In a temporal clustering, be reference point with this timing cycle, with the extrapolation of the data of other observation station or in be inserted into this reference point, a plurality of observations in such temporal clustering just can be regarded a virtual observation on temporal clustering point as.If with t ₁State extrapolation constantly or in be inserted to t ₂Constantly, the state transitions formula of employing is:

$\hat{x}\left(t_2\right)=\mathrm{\&Phi;}(t_2-t_1)\hat{X}\left(t_1\right)$

In the formula,

State vector for system; Φ (t) is a state-transition matrix.

Carry out the space-time registration by this technology, single channel is handled the flight path that obtains and can be guaranteed to be consistent with the sensor actual cycle.Comprehensive flight path for after merging can upgrade according to the minimum value in each cycle sensor.

(3) error correction

In various kinds of sensors information, because the bearing accuracy of multipoint location system exceeds a lot (reaching 7-10 rice) than general air traffic control radar, therefore the adding of this type of information will improve a lot as having introduced a reference data sources than air traffic control radar flight path precision through comprehensive flight path precision after the fusion treatment.Utilize the position of comprehensive flight path and the error mean between the correlation radar measuring value, can draw the air traffic control radar systematic error easily,, improve the precision of single portion radar measurement value by the systematic error of each radar of Error Feedback correction technique correction.

Each air traffic control radar of estimation of error stage statistics acquisition with respect to comprehensive flight path systematic error, is fed back to the error correction link.To the aim parameter measurement information behind the space-time registration in the correction of error correction link Real-time Error.

Suppose that the position vector estimating system error that current time calculates the i portion air traffic control radar that obtains from estimation of error is e _i, and current this road radar is m to the position vector measuring value of j target _j, the position vector after this target is corrected is

Then the pass between them is:

${\hat{m}}_j=m_j+e_i$

Like this, this road radar according to said method circulated in these all targets that detect constantly revise, the position quantity measured value of the systematic error that has been eliminated is given follow-up link and is handled.Other road air traffic control radar is also handled in this way.

The multipoint location system flight path can reach for several times in 1 second because its target reporting cycle is fixing, therefore need select excellently to its aim parameter measurement information report, and the aim parameter measurement information report that multipoint location system is presented has the stable cycle.At first multipoint location system information of receiving and the multipoint positioning flight path of having set up are done pre-relevant, if shut mutually, just contrast both timestamp and position, whether judgement their alternate position spike in both mistimings is reasonable, if rationally, further relatively timestamp keeps and is separated by near 1 second the information input information as this update cycle, otherwise gives up.The aim parameter measurement information report that multipoint location system is sent can approximately remain on 1 second through selecting the excellent back cycle.

(4) data are interconnected

Although interconnected single-sensor point mark or the interconnected related two kinds of situations of flight path of having comprised of data with multi-sensor data; but the principle of both of these case is similar; all be to set up certain sensor metric data and other relation of metric data constantly constantly, to determine that these metric data are whether from the processing procedure of same target.Just the former at be the metric data of a sensor, the latter at be the metric data of multi-section sensor.Their implementation process is also very similar.

The interconnected process of data is according to actual conditions, mainly consider three key elements: limit being not suitable for related right generation according to the data association thresholding of setting, realize direct association based on the related of code (for example: secondary code, destination address and flight path number etc.), and implement related according to factors such as minor increments.

As shown in Figure 2, when an aim parameter measurement information begins to carry out the interconnected process of data, at first be that the certain limit (comprising planimetric position and height) at center is made as relevant ripple door with this target, this Bo Mennei has been deposited flight path as the affiliated partner hunting zone; Check the degree of agreement of this aim parameter measurement information and relevant each flight path of Bo Mennei then one by one, main relevant factor comprises flight path number, secondary code, these identity informations of destination address and the velocity information of target.Set up factor of influence at every kind of key element, meet for just, be not inconsistent be combined into negative.The information of ordinary representation identity is higher to the influence of measurement information and track association possibility, therefore establish the absolute value of the factor of influence of the relevant factor of expression target identities such as secondary code greatlyyer, and speed is that the relevant factor of expression target travel characteristic is owing to change very fast, reliability is lower, so the factor of influence absolute value is also established lessly.Each key element is checked one by one, comprehensive then each factor of influence, if greater than certain threshold value, expression is shut mutually, otherwise uncorrelated.

In some cases, may search a plurality of flight paths relevant at relevant Bo Mennei with the aim parameter measurement information, at this moment adopt minimal distance principle to determine the last flight path relevant with the aim parameter measurement information, i.e. the position of which flight path and target measurement is the most approaching, and this target is just relevant with which flight path.

If relevant Bo Mennei do not find can be relevant with this target flight path, just with this target as new flight path, add in the flight path chained list.

(5) flight path filtering and data fusion

The flight path data fusion merges the information that the different sensors of same target detects, and the diverse location point of sending here according to different sensors, and filter forecasting obtains the optimum estimate to the target current location.

Among the present invention, Kalman filtering is the major technique means that multi-sensor information fusion is handled, and has designed the federated Kalman filtering device especially on this basis.This Design of Filter structure can guarantee local and overall wave filter all has higher precision, and has fault-tolerant preferably effect, and the calculated amount of senior filter is very little in addition, and synthetic method is simple, is easy to realize.

The basic thought of federated Kalman filtering device design is first dispersion treatment, the overall situation merges again, promptly in many non-similar subsystems, select a subsystem that information is comprehensive, output speed is high, reliability definitely guarantees as the common reference system, combine in twos with other subsystem, form the plurality of sub wave filter.Because the measurement accuracy of multipoint location system is significantly higher than air traffic control radar, therefore here with multipoint location system as the reference sensor.Each subfilter parallel running obtains to be based upon the local local optimum that measures on the basis of subfilter and estimates that these local optimums are estimated in senior filter synthetic by the weighted mean blending algorithm, estimates thereby obtain to be based upon all overall situations that measure on the basis.

According to the basic thought of above-mentioned federated Kalman filtering device design, in all the sensors of connecting system, select the fast sensor of a detection accuracy height, update cycle as the reference sensor, select multipoint location system as the reference sensor here.Reference sensor is combined respectively with other sensor, form local filter.In local filter, calculate the state estimation of local optimum by Kalman filtering, form the single channel flight path.Each single channel flight path is delivered to senior filter, in senior filter, merge calculating, obtain overall fusion results, form comprehensive flight path.Fusion parameters and result are fed back to each local filter, and local filter is further optimized filtering according to fusion parameters and the filtering parameter separately of adjustment as a result, thereby optimizes single channel flight path and fusion results, improves the precision of comprehensive flight path.Filter Structures figure as shown in Figure 3.

The equation of motion of considering target can be considered n dimensional linear dynamic system, and during the sensor tracking target, the flight path that observes can be considered m dimensional linear recording geometry, and then the discrete description form of the state equation of i local filter and measurement equation is:

$\left\{\begin{array}{cc}{S}_i\left(k\right)=A\left(k\right)S_i(k-1)+w_i(k-1)& k=\mathrm{1,2},\&CenterDot;\&CenterDot;\&CenterDot;\\ X_i\left(k\right)=C_i\left(k\right)S_i\left(k\right)+v_i\left(k\right)& i=\mathrm{1,2},\&CenterDot;\&CenterDot;\&CenterDot;,q\end{array}\right.$

S wherein _i(k) the expression system is at k state (can comprise position, speed and acceleration) constantly.A (k) is the system state transition matrix.w _i(k-1) (k=1,2 ...) expression acts on the random disturbance of system, promptly plant noise generally is assumed to be the white Gaussian noise sequence, has known zero-mean and covariance matrix Q _i(k).X _i(k) be observation vector.C _i(k) be observing matrix.v _i(k) (k=1,2 ...) be observation noise, be assumed to be the white Gaussian noise sequence, have known zero-mean and covariance matrix R _i(k).

The filtering recursion formula is:

ε′ _i(k)＝A(k)ε _i(k-1)A ^T(k)+Q(k-1)

$H_i\left(k\right)={\mathrm{\&epsiv;}}_i^{\&prime;}\left(k\right)C_i^T\left(k\right){[C_i\left(k\right){\mathrm{\&epsiv;}}_i^{\&prime;}\left(k\right)C_i^T\left(k\right)+R_i\left(k\right)]}^{-1}$

ε _i(k)＝[I-H _i(k)C _i(k)]ε′ _i(k)

Wherein, Q _i(k) be plant noise w _i(k-1) covariance matrix; R _i(k) be observation noise v _i(k) covariance matrix;

Be that k is constantly through filtered valuation; H _i(k) be gain matrix; ε _i(k) be the evaluated error covariance matrix.

The processing of the Kalman filter of the decentralized concurrent operation of process obtains local optimum and estimates

(k=1,2 ...), in senior filter, merge.In senior filter, overall fusion results is:

S＝(ε ₁ ^-1+ε ₂ ^-1+…+ε _q ^-1) ^-1(ε ₁ ^-1S ₁+ε ₂ ^-1S ₂+…+ε _q ^-1S _q)

ε＝(ε ₁ ^-1+ε ₂ ^-1+…+ε _q ^-1) ^-1

Overall estimated result is fed back to each local filter, as the k estimated value of each local filter constantly:

S _i(k)＝S(k)

$Q_i^{-1}\left(k\right)={\mathrm{\&alpha;}}_i{Q}^{-1}\left(k\right)$

${\mathrm{\&epsiv;}}_i^{-1}\left(k\right)={\mathrm{\&alpha;}}_i{\mathrm{\&epsiv;}}^{-1}\left(k\right)$

α ₁+α ₂+…+α _q＝1

Wherein, i=1,2 ..., q, 0≤α _i≤ 1.

Local filter is carried out filtering according to state equation and measurement equation, and the filtering result in per step is passed to senior filter.Senior filter is finished the optimal synthesis of information, forms the integrated information of global system.After each filtering stage was finished, the overall situation that will be synthesized by senior filter was estimated and according to the information distribution amount that " information distribution " principle forms, is carried out information feedback to each local filter, thereby can in real time suitable value be arrived in each parameter adjustment of local filter.

The average weighted weights of each sensor are determined according to the current performance of this sensor.And the current performance of sensor is to be determined by the measurement position of the current test point of this sensor and the departure degree of physical location.The test point physical location of supposing i portion sensor is (x _i, y _i), and the position quantity measured value is

Alternate position spike between them is so:

${\mathrm{\&Delta;r}}_i=|\sqrt{x_i^2+y_i^2}-\sqrt{{\hat{x}}_i^2+{\hat{y}}_i^2}|$

This position difference is big more to show that the physical location of test point is big more with the gap that measures the position, and the current performance of this sensor is poor more, otherwise performance is good more, promptly is oppositely relevant.Can determine that this sensor shared weights in the weighting fusion of all n portion sensors are by this alternate position spike:

${\mathrm{\&alpha;}}_i=\frac{\frac{1}{{\mathrm{\&Delta;r}}_i}}{\underset{i}{\overset{\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A2009100350310020H1.png"\; state="\{\{state\}\}">n</figure-callout>}}{\mathrm{\&Sigma;}}}\frac{1}{{\mathrm{\&Delta;r}}_i}}$

(6) estimation of error

After the data fusion link, will ask difference to calculate through filtered each the target location vector measuring value corresponding of flight path with each air traffic control radar, obtain deviate about each target.The deviate of all targets that an interior air traffic control radar of statistics current time in some cycles forward detects, and work is average, thus obtain the estimating system error of this air traffic control radar.

Suppose now to calculate the systematic error of i portion air traffic control radar.The public flight path number that detects with this air traffic control radar in the comprehensive flight path is m, and n cycle added up.The position vector that is detected by air traffic control radar in k cycle j target is S _jFiltered comprehensive flight path position vector is M _jThe estimating system error of i portion air traffic control radar is e _i, then there is following relation between them:

$e_i=\underset{j}{\overset{m}{\mathrm{\&Sigma;}}}\underset{k}{\overset{n}{\mathrm{\&Sigma;}}}(S_j-M_j)$

The systematic error of each air traffic control radar of herein adding up in the utilization of error correction link is revised position quantity measured value separately, thereby improves the measurement accuracy of each air traffic control radar, finally improves the overall precision of system.

The present invention has provided utilization the present invention front and back in the airport ground operational system, and the sectional drawing of landing aircraft display position on map is to show explanation beneficial effect of the present invention.Fig. 5 a is for adopting the full interface display figure of system of the present invention.Fig. 5 b monitors the aircraft displayed map that is about to landing for the preceding multi-section air traffic control radar that merges of utilization the present invention.Fig. 5 c merges air traffic control radar after for utilization the present invention and multipoint location system monitors the aircraft displayed map that is about to landing.The landing aircraft in the end closely advances when flight in stage should line up with runway, so its flight path should overlap with the runway extended line.Among Fig. 5 b, the result demonstration of adopting the generic card Kalman Filtering and only merging the multi-section air traffic control radar, comprehensive flight path upgrades drift off the runway extended line certain distance with 4 seconds cycle.Among Fig. 5 c, the result demonstration of adopting federated Kalman filtering and merging a multipoint location system and multi-section air traffic control radar, comprehensive flight path upgraded with 1 second cycle, and flight path can overlap well with the runway extended line.Contrast among above-mentioned two width of cloth figure to the aircraft surveillance effect as can be known, the flight path turnover rate significantly improved after utilization the present invention added multipoint location system, and the position of flight path is more realistic, more steady.

The invention belongs to the application software technology of computer science, relate to multisensor data fusion processing.By setting up polymorphic type sensing data Fusion Model, create federated Kalman filtering device and weighted mean blending algorithm, to the multi-section scene surveillance radar information of reversed return type not, advance nearly radar information and multipoint location system information is carried out fusion treatment, form dynamically flight path of continuous, level and smooth target (aircraft, vehicle).This technology is used for civil aviaton's Terminal Air Traffic Control five limits and advances the senior scene activity guiding of nearly supervisory system and airport control system (A-SMGCS).

The invention provides a kind of based on the multipoint positioning data of federated Kalman filtering and the thinking and the method for Radar Data Fusion method; the method and the approach of this technical scheme of specific implementation are a lot; the above only is a preferred implementation of the present invention; should be understood that; for those skilled in the art; under the prerequisite that does not break away from the principle of the invention, can also make some improvements and modifications, these improvements and modifications also should be considered as protection scope of the present invention.The all available prior art of each ingredient not clear and definite in the present embodiment is realized.

Claims (5)

1, a kind of multipoint positioning data and Radar Data Fusion method based on federated Kalman filtering is characterized in that, may further comprise the steps:

Pre-treatment step promptly receives the aim parameter measurement information that each sensor is sent, and according to analysiss of decoding of the data layout of each sensor definition, and the aim parameter measurement information of different-format is converted to consolidation form explains;

The space-time step of registration, soon the target position information in the aim parameter measurement information of each sensor acquisition is transformed under the same coordinate system, and is registered to synchronization; Target position information is the location entries that extracts from the aim parameter measurement information that sensor is received, described location entries is represented the target location with longitude and latitude, rectangular coordinate or polar form;

Point mark or the interconnected step of flight path are about to the different information processings constantly of same target that same sensor detect and become with a collection of flight path; When an aim parameter measurement information begins to carry out a mark or the interconnected process of flight path, at first be that the planimetric position that comprises at center is made as relevant ripple door with the scope of height with this target, this Bo Mennei has been deposited flight path as the affiliated partner hunting zone; Check the degree of agreement of this aim parameter measurement information and relevant each flight path of Bo Mennei then one by one, relevant factor comprises flight path number, the secondary code of target, the identity information and the velocity information of destination address; Set up factor of influence at every kind of key element, meet for just, be not inconsistent be combined into negative; For the summation of each factor of influence, if greater than preset threshold, expression is relevant, otherwise uncorrelated;

Adopt local filter to carry out each sensor flight path filter step, obtain the local optimum of target current location is estimated; Described flight path filtering may further comprise the steps: the equation of motion of target is considered as n dimensional linear dynamic system, during the sensor tracking target, the flight path that observes is considered as m dimensional linear recording geometry, and then the discrete description form of the state equation of i local filter and measurement equation is:

$\left\{\begin{array}{cc}{S}_i\left(k\right)=A\left(k\right)S_i(k-1)+w_i(k-1)& k=\mathrm{1,2},\&CenterDot;\&CenterDot;\&CenterDot;\\ X_i\left(k\right)=C_i\left(k\right)S_i\left(k\right)+v_i\left(k\right)& i=\mathrm{1,2},\&CenterDot;\&CenterDot;\&CenterDot;,q\end{array}\right.;$

S wherein _i(k) the expression system is at k state constantly; A (k) is the system state transition matrix; w _i(k-1) (k=1,2 ...) expression acts on the random disturbance of system, i.e. plant noise; Described plant noise is assumed to be the white Gaussian noise sequence, promptly has known zero-mean and covariance matrix Q _i(k); X _i(k) be observation vector; C _i(k) be observing matrix; v _i(k) (k=1,2 ...) be observation noise, be set at the white Gaussian noise sequence, promptly have known zero-mean and covariance matrix R _i(k);

The filtering recursion formula is:

ε′ _i(k)＝A(k)ε _i(k-1)A ^T(k)+Q(k-1)

$H_i\left(k\right)={\mathrm{\&epsiv;}}_i^{\&prime;}\left(k\right)C_i^T\left(k\right){[C_i\left(k\right){\mathrm{\&epsiv;}}_i^{\&prime;}\left(k\right)C_i^T\left(k\right)+R_i\left(k\right)]}^{-1}$

ε _i(k)＝[I-H _i(k)C _i(k)]ε′ _i(k)

Wherein, Q _i(k) be plant noise w _i(k-1) covariance matrix; R _i(k) be observation noise v _i(k) covariance matrix;

Be that k is constantly through filtered valuation; H _i(k) be gain matrix; ε _i(k) be the evaluated error covariance matrix; Through the filter process of decentralized concurrent operation, obtain local optimum and estimate

(k=1,2 ...), and the filtering result in per step passed to senior filter;

The data association step, judge that whether information that different sensors detects is about same target, when an aim parameter measurement information of certain sensor begins to carry out data association, the planimetric position that comprises that at first with this target is the center is made as relevant ripple door with the scope of height, and this Bo Mennei has been deposited flight path as the affiliated partner hunting zone; Check the degree of agreement of this aim parameter measurement information and relevant each flight path of Bo Mennei then one by one, relevant factor comprises flight path number, the secondary code of target, the identity information and the velocity information of destination address; Set up factor of influence at every kind of key element, meet for just, be not inconsistent be combined into negative; For the summation of each factor of influence, if greater than preset threshold, expression is relevant, otherwise uncorrelated; Aim parameter measurement information for shutting mutually with flight path carries out flight path filtering and fusion; For the aim parameter measurement information of not shutting mutually with any flight path, set up new flight path;

Data fusion step, the information that the different sensors of same target is detected merges, and the local optimum that decentralized concurrent operation obtains according to the local filter of different sensors correspondence is estimated (k=1,2 ...), in senior filter, merge; In senior filter, overall fusion results is:

S＝(ε ₁ ^-1+ε ₂ ^-1+…+ε _q ^-1) ^-1(ε ₁ ^-1S ₁+ε ₂ ^-1S ₂+…+ε _q ^-1S _q)

ε＝(ε ₁ ^-1+ε ₂ ^-1+…+ε _q ^-1) ^-1

Overall estimated result is fed back to each local filter, as the k estimated value of each local filter constantly:

S _i(k)＝S(k)

$Q_i^{-1}\left(k\right)={\mathrm{\&alpha;}}_i{Q}^{-1}\left(k\right)$

${\mathrm{\&epsiv;}}_i^{-1}\left(k\right)={\mathrm{\&alpha;}}_i{\mathrm{\&epsiv;}}^{-1}\left(k\right)$

α ₁+α ₂+…+α _q＝1

Wherein, i=1,2 ..., q, 0≤α _i≤ 1;

Senior filter is finished the optimal synthesis of information, forms the integrated information of global system; After each filtering stage is finished, estimate and the information distribution amount, carry out information feedback, adjust each parameter value of local filter in real time to each local filter by the overall situation that senior filter will synthesize;

The average weighted weights of each sensor are determined according to the measurement position of the current test point of this sensor and the departure degree of physical location; The test point physical location of supposing i portion sensor is (x _i, y _i), and the position quantity measured value is

Alternate position spike between them is so:

$\mathrm{\&Delta;}{r}_i=|\sqrt{x_i^2+y_i^2}-\sqrt{{\hat{x}}_i^2+{\hat{y}}_i^2}|$

The gap of the physical location of test point and measurement position is that the current precision of position difference and sensor is inverse relation; Determine that by the position difference this sensor shared weights in the weighting fusion of all n portion sensors are:

${\mathrm{\&alpha;}}_i=\frac{\frac{1}{\mathrm{\&Delta;}{r}_i}}{\underset{i}{\overset{n}{\mathrm{\&Sigma;}}}\frac{1}{\mathrm{\&Delta;}{r}_i}}.$

2, multipoint positioning data and Radar Data Fusion method based on federated Kalman filtering according to claim 1 is characterized in that, also comprise estimation of error step and error correction step after the data fusion step;

Described estimation of error step comprises, will ask difference to calculate through filtered each the target location vector measuring value corresponding with each air traffic control radar of flight path, obtains the deviate about each target; The deviate of all targets that an interior air traffic control radar of statistics current time in some cycles forward detects, and do on average to obtain the systematic error of the estimation of this air traffic control radar; The public flight path number that detects with this air traffic control radar in the comprehensive flight path is m, and n cycle added up; The position vector that is detected by air traffic control radar in k cycle j target is S _jFiltered comprehensive flight path position vector is M _jThe estimated value e of i portion air traffic control radar systematic error _iFor:

$e_i=\underset{j}{\overset{m}{\mathrm{\&Sigma;}}}\underset{k}{\overset{n}{\mathrm{\&Sigma;}}}(S_j-M_j);$

The systematic error e of the estimation of each air traffic control radar of statistics _i

Described error correction step comprises, with the estimating system error e of each air traffic control radar in the estimation of error step with respect to comprehensive flight path _iAim parameter measurement information behind the space-time registration is carried out error correction; If current time calculates the position vector estimating system error e of the i portion air traffic control radar that obtains from estimation of error _i, and current this road radar is m to the position vector measuring value of j target _j, the position vector after this target is corrected is

Then the pass between them is:

${\hat{m}}_j=m_j+e_i;$

This road radar circulated in these all targets that detect constantly revise the position quantity measured value of the systematic error that has been eliminated.

3, multipoint positioning data and Radar Data Fusion method based on federated Kalman filtering according to claim 1, it is characterized in that, in the space-time step of registration, be registered to synchronization and adopt the timeslice method, be about to a timing cycle as the temporal clustering point, in a temporal clustering, be reference point with this timing cycle, with the extrapolation of the data of other observation station or in be inserted into this reference point; With a plurality of observations in the temporal clustering as a virtual observation on temporal clustering point; The state transitions formula that adopts is with t ₁State extrapolation constantly or in be inserted to t ₂Constantly; Described state transitions formula is:

$\hat{X}\left(t_2\right)=\mathrm{\&Phi;}(t_2-t_1)\hat{X}\left(t_1\right)$

Wherein,

Be state vector; Φ (t) is a state-transition matrix.

4, multipoint positioning data and Radar Data Fusion method based on federated Kalman filtering according to claim 1, it is characterized in that, in some mark or the interconnected step of flight path, if search a plurality of flight paths relevant, adopt minimal distance principle to determine the last flight path relevant with the aim parameter measurement information with the aim parameter measurement information at relevant Bo Mennei; If relevant Bo Mennei do not find can be relevant with this target flight path, just with this target as new flight path, add in the flight path chained list.

5, multipoint positioning data and Radar Data Fusion method based on federated Kalman filtering according to claim 2, it is characterized in that, comprise in the described error correction step that aim parameter measurement information report selects excellent, at first multipoint location system information of receiving and the multipoint positioning flight path of having set up are done pre-relevant, if shut mutually, just according to both speed of timestamp and position calculation flight path, judge the maximal rate whether this speed can reach greater than flight path, if be not more than, further compare timestamp, keep and to be separated by, otherwise give up near 1 second information input information as this update cycle.

Images