CN108734218A - A kind of information fusion method and device of multisensor syste - Google Patents

A kind of information fusion method and device of multisensor syste Download PDF

Info

Publication number
CN108734218A
CN108734218A CN201810497192.XA CN201810497192A CN108734218A CN 108734218 A CN108734218 A CN 108734218A CN 201810497192 A CN201810497192 A CN 201810497192A CN 108734218 A CN108734218 A CN 108734218A
Authority
CN
China
Prior art keywords
sensor
interest
value
current time
observation
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201810497192.XA
Other languages
Chinese (zh)
Other versions
CN108734218B (en
Inventor
张亚男
金正具
印思琪
姜永强
孙宏轩
许梦兴
郑财
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
BOE Technology Group Co Ltd
Beijing BOE Optoelectronics Technology Co Ltd
Original Assignee
BOE Technology Group Co Ltd
Beijing BOE Optoelectronics Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by BOE Technology Group Co Ltd, Beijing BOE Optoelectronics Technology Co Ltd filed Critical BOE Technology Group Co Ltd
Priority to CN201810497192.XA priority Critical patent/CN108734218B/en
Publication of CN108734218A publication Critical patent/CN108734218A/en
Application granted granted Critical
Publication of CN108734218B publication Critical patent/CN108734218B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/251—Fusion techniques of input or preprocessed data
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques

Abstract

The embodiment of the present invention provides a kind of information fusion method and device of multisensor syste, is related to sensor information process field, can significantly promote the adaptive faculty and arithmetic accuracy in complex situations of information fusion algorithm.This method includes：Observational equation is established according to default Chance Failure Rate；Determine the sensor of interest that can be merged into row information at current time in multisensor syste；According to observational equation calculate sensor of interest the observation at current time theoretical value；Value of the calculating sensor of interest in the observation residual error statistical property at current time；Judge that sensor of interest whether there is mistake in the observation at current time in the value of the observation residual error statistical property at current time and the magnitude relationship of default chi square distribution value according to sensor of interest；Observation according to sensor of interest at current time obtains the result that sensor of interest merges at current time into row information with the presence or absence of the situation of mistake.

Description

A kind of information fusion method and device of multisensor syste
Technical field
The present invention relates to sensor information process field more particularly to a kind of information fusion method of multisensor syste and Device.
Background technology
AR/VR (Augmented Reality/Virtual Reality, augmented reality/virtual reality) system, medical treatment are strong Multisensor syste is belonged in the complication systems such as health sensor wisdom system, target tracking system, image-signal processing system, And all kinds of multisensor systes need to use distinctive algorithm when the collected information of each sensor is carried out fusion treatment It completes.But existing multiple sensor information amalgamation method does not consider to observe random unreliable the case where observing random fault The observation of any sensor all may be because that the situation of mistake occur in all kinds of random faults i.e. in information fusion process, simultaneously In existing multiple sensor information amalgamation method also do not consider obtain multisensor syste state value when system noise and Observation noise of each sensor when obtaining observation whether there is the possibility of correlation；And actual multisensor syste There is the observation noise of observation random fault while its system noise and each sensor mostly in information fusion process All there is certain correlation, so introducing observation random fault is more in line with reality, so existing multisensor is believed Fusion method is ceased less to be applicable in for actual conditions.
Invention content
The embodiment of the present invention provides a kind of information fusion method and device of multisensor syste, can be more realistic Optimal state estimation is carried out to multisensor syste, significantly promoted information fusion algorithm adaptive faculty and in complicated feelings Arithmetic accuracy under condition.
In order to achieve the above objectives, the embodiment of the present invention adopts the following technical scheme that：
In a first aspect, a kind of information fusion method and device of multisensor syste are provided, including：
Observational equation is established according to default Chance Failure Rate；
Determine the sensor of interest that can be merged into row information at current time in multisensor syste；
According to observational equation calculate sensor of interest the observation at current time theoretical value；
Value of the calculating sensor of interest in the observation residual error statistical property at current time；
According to sensor of interest the value and default chi square distribution value of the observation residual error statistical property at current time size Relationship judges that sensor of interest whether there is mistake in the observation at current time；
Observation according to sensor of interest at current time obtains sensor of interest with the presence or absence of the situation of mistake and is working as The result that the preceding moment merges into row information.
Optionally, there are correlations between the system noise and observation noise in multisensor syste, meet following public Formula：
νi(k)=βiw(k-1)+ηi(k)；
Wherein, νi(k) be multisensor syste in i-th of sensor in the observation noise at k moment, βiFor multisensor system In the relevance parameter of the system noise at k moment, w (k-1) is for the observation noise of i-th sensor and multisensor syste in system Multisensor syste is in the system noise at k-1 moment, ηi(k) it is that i-th of sensor is random at the k moment in multisensor syste Zero mean Gaussian white noise；I is positive integer.
Optionally, the state value of multisensor syste at any one time includes system noise, any in multisensor syste The observation of sensor at any one time includes observation noise.
Optionally, determine that the sensor of interest that can be merged into row information at current time in multisensor syste includes：
It is determined at current time according to the sampling ratio of each sensor in value at the time of current time and multisensor syste The sensor of interest that can be merged into row information.
Optionally, calculation formula calculating sensor of interest includes in the value of the observation residual error statistical property at current time：
Compared according to observation in the theoretical value of the observation at current time and the sampling of sensor of interest according to sensor of interest Value of the residual error statistical property calculation formula calculating sensor of interest in the observation residual error statistical property at current time.
Optionally, the value according to sensor of interest in the observation residual error statistical property at current time and default chi square distribution value Magnitude relationship judge that sensor of interest includes with the presence or absence of mistake in the observation at current time：
When sensor of interest is more than default chi square distribution in the absolute value of the value of the observation residual error statistical property at current time When value, determine that there are mistakes for observation of the sensor of interest at current time；
When sensor of interest is less than or equal to default card side in the absolute value of the value of the observation residual error statistical property at current time When Distribution Value, determine that mistake is not present in observation of the sensor of interest at current time.
Optionally, the observation according to sensor of interest at current time obtains target sensing with the presence or absence of the situation of mistake The result that device merges at current time into row information includes：
Observation according to sensor of interest at current time obtains sensor of interest with the presence or absence of the situation of mistake and is working as The kalman gain value at preceding moment；
Kalman gain value according to sensor of interest at current time calculates sensor of interest and carries out letter at current time State Estimation of Mean value after breath fusion and target covariance.
Optionally, the observation according to sensor of interest at current time obtains target sensing with the presence or absence of the situation of mistake Kalman gain value of the device at current time include：
When sensor of interest current time observation exist mistake when, determine sensor of interest current time card Germania yield value is zero；
When sensor of interest current time observation there is no mistake when, then according to kalman gain calculation formula meter Kalman gain value of the calculation sensor of interest at current time.
Optionally, the kalman gain value according to sensor of interest at current time calculates sensor of interest at current time State Estimation of Mean value and target covariance after being merged into row information include：
According to sensor of interest current time kalman gain value with the descending sequence of sampling rate by target Sensor calculates the updated value of the state mean prediction value of sensor of interest according to state mean prediction value more new formula successively；Shape State Estimation of Mean value is the updated value of the state mean prediction value of the sensor of interest of sampling rate minimum in sensor of interest；
According to sensor of interest current time kalman gain value with the descending sequence of sampling rate by target Sensor calculates the updated value of the target covariance of sensor of interest according to target covariance more new formula successively；Target covariance It is the updated value of the target covariance of the sensor of interest of sampling rate minimum in sensor of interest.
Second aspect provides a kind of information fuse device of multisensor syste, including：
Establishing equation module, for establishing observational equation according to default Chance Failure Rate；
Judgment module is merged, can be passed into the target that row information merges at current time for determining in multisensor syste Sensor；
Computing module is observed, the observational equation for being established according to establishing equation module calculates what fusion judgment module determined Theoretical value of the sensor of interest in the observation at current time；
Residual error statistical property computing module, for calculating the sensor of interest of fusion judgment module determination at current time Observe the value of residual error statistical property；
Random fault judgment module, the sensor of interest for being calculated according to residual error statistical property computing module is when current The value of the observation residual error statistical property at quarter and the magnitude relationship of default chi square distribution value judge sensor of interest at current time Observation whether there is mistake；
Fusion calculation module, for obtaining sensor of interest when current according to the judging result of random fault judgment module Carve the result merged into row information.
Optionally, which further includes state computation module；
State computation module is used to calculate target according to the observational equation that establishing equation module is established in observation computing module Sensor is before the theoretical value of the observation at current time, state value of the calculating sensor of interest at current time.
Optionally, fusion judgment module is specifically used for：
It is determined at current time according to the sampling ratio of each sensor in value at the time of current time and multisensor syste The sensor of interest that can be merged into row information.
Optionally, residual error statistical property computing module is specifically used for：
Compared according to observation in the theoretical value of the observation at current time and the sampling of sensor of interest according to sensor of interest Value of the residual error statistical property calculation formula calculating sensor of interest in the observation residual error statistical property at current time.
Optionally, random fault judgment module is specifically used for：
When sensor of interest is more than default chi square distribution in the absolute value of the value of the observation residual error statistical property at current time When value, determine that there are mistakes for observation of the sensor of interest at current time；
When sensor of interest is less than or equal to default card side in the absolute value of the value of the observation residual error statistical property at current time When Distribution Value, determine that mistake is not present in observation of the sensor of interest at current time.
Optionally, fusion calculation module includes kalman gain computing unit, fusion treatment unit；
Kalman gain computing unit, which is used to obtain sensor of interest according to the judging result of random fault judgment module, to exist The kalman gain value at current time；
Card of the sensor of interest that fusion treatment unit is used to be calculated according to kalman gain computing unit at current time Germania yield value calculates state Estimation of Mean value and target covariance of the sensor of interest after current time merges into row information.
Optionally, kalman gain computing unit is specifically used for：
When random fault judgment module determines that sensor of interest when the observation at current time has mistake, determines target Kalman gain value of the sensor at current time is zero；
When random fault judgment module determine sensor of interest current time observation there is no when mistake, then basis Kalman gain value of the kalman gain calculation formula calculating sensor of interest at current time.
Optionally, fusion treatment unit includes that state Estimation of Mean value computation subunit and target covariance calculate son list Member；
State Estimation of Mean value computation subunit, the sensor of interest for being calculated according to kalman gain computing unit exist The kalman gain value at current time, it is with sampling rate descending sequence that sensor of interest is pre- according to state mean value successively Measured value more new formula calculates the updated value of the state mean prediction value of sensor of interest；State Estimation of Mean value is sensor of interest The updated value of the state mean prediction value of the sensor of interest of middle sampling rate minimum；
Target covariance computation subunit, the sensor of interest for being calculated according to kalman gain computing unit is current The kalman gain value at moment is updated sensor of interest with the descending sequence of sampling rate according to target covariance successively Formula calculates the updated value of the target covariance of sensor of interest；Target covariance is that sampling rate is minimum in sensor of interest The updated value of the target covariance of sensor of interest.
The third aspect, the embodiment of the present invention provide a kind of computer equipment, including memory, processor；It is deposited on memory The computer program that can be run on a processor is contained, processor realizes the side provided such as first aspect when executing computer program Method.
Fourth aspect, the embodiment of the present invention provide a kind of computer storage media, are stored with computer program, the calculating The method provided such as first aspect is provided when machine program is executed by processor.
The information fusion method and device of multisensor syste provided in an embodiment of the present invention, this method include：According to pre- If Chance Failure Rate establishes observational equation；It determines in multisensor syste and can be passed into the target that row information merges at current time Sensor；According to the state value of state equation and dynamic sliding average equation calculation sensor of interest current time；According to observation side Theoretical value of the journey calculating sensor of interest in the observation at current time；Observation residual error of the calculating sensor of interest at current time The value of statistical property；The value in the observation residual error statistical property at current time and default chi square distribution value according to sensor of interest Magnitude relationship judges that sensor of interest whether there is mistake in the observation at current time；According to sensor of interest at current time Observation obtain the result that is merged into row information at current time of sensor of interest with the presence or absence of the situation of mistake.The present invention is real The technical solution of example offer is provided, existing CKF algorithms are based on, is introduced in observational equation and observes random random fault, then sentenced The disconnected sensor of interest that can be merged then judges observation random fault, is finally obtained according to judging result more The result of sensing system information fusion.Because technical solution provided in an embodiment of the present invention when establishing observational equation at the beginning Random observation failure is just introduced, and corresponding judgement then has been carried out to random fault, finally gets multisensor letter It is after breath fusion as a result, so the more realistic situation of technical solution provided in an embodiment of the present invention compared with prior art, for The fitness of environment is also with regard to higher.
Description of the drawings
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below There is attached drawing needed in technology description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this Some embodiments of invention for those of ordinary skill in the art without creative efforts, can be with Obtain other attached drawings according to these attached drawings.
Fig. 1 is a kind of information fusion method flow diagram of multisensor syste provided in an embodiment of the present invention；
Fig. 2 is the information fusion method flow diagram of another multisensor syste provided in an embodiment of the present invention；
Fig. 3 is that multisensor syste provided in an embodiment of the present invention samples schematic diagram；
Fig. 4 is a kind of information fusion method flow diagram for multisensor syste that another embodiment of the present invention provides；
Fig. 5 is a kind of information fuse device structural schematic diagram of multisensor syste provided in an embodiment of the present invention；
Four category information fusion methods are for one-dimensional position in Fig. 6 Monte-Carlo Simulation experiments provided in an embodiment of the present invention Estimation effect comparison diagram (band 10% observes random fault but the CKF methods that only merge can detect failure and debugging)；
Four category information fusion methods are for two-dimensional position in Fig. 7 Monte-Carlo Simulation experiments provided in an embodiment of the present invention Estimation effect comparison diagram (band 10% observes random fault but the CKF methods that only merge can detect failure and debugging)；
Four category information fusion methods are for third dimension position in Fig. 8 Monte-Carlo Simulation experiments provided in an embodiment of the present invention Estimation effect comparison diagram (band 10% observes random fault but the CKF methods that only merge can detect failure and debugging)；
Fig. 9 is that three classes information fusion method ties up position for first in Monte-Carlo Simulation provided in an embodiment of the present invention experiment The comparison diagram (not band observation random fault) for the estimation effect set；
Figure 10 is that three classes information fusion method is tieed up for second in Monte-Carlo Simulation provided in an embodiment of the present invention experiment The comparison diagram (not band observation random fault) of the estimation effect of speed；
Figure 11 is three classes information fusion method in Monte-Carlo Simulation provided in an embodiment of the present invention experiment for the third dimension The comparison diagram (not band observation random fault) of the estimation effect of acceleration；
Figure 12 is that three classes information fusion method is tieed up for first in Monte-Carlo Simulation provided in an embodiment of the present invention experiment The comparison diagram of the estimation effect of position (band 10% observes random fault but can detect failure and debugging)；
Figure 13 is that three classes information fusion method is tieed up for second in Monte-Carlo Simulation provided in an embodiment of the present invention experiment The comparison diagram of the estimation effect of speed (band 10% observes random fault but can detect failure and debugging)；
Figure 14 is three classes information fusion method in Monte-Carlo Simulation provided in an embodiment of the present invention experiment for the third dimension The comparison diagram of the estimation effect of acceleration (band 10% observes random fault but can detect failure and debugging).
Specific implementation mode
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete Site preparation describes, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other Embodiment shall fall within the protection scope of the present invention.
It should be noted that in the embodiment of the present invention, " illustrative " or " such as " etc. words make example, example for indicating Card or explanation.Be described as in the embodiment of the present invention " illustrative " or " such as " any embodiment or design scheme do not answer It is interpreted than other embodiments or design scheme more preferably or more advantage.Specifically, " illustrative " or " example are used Such as " word is intended to that related notion is presented in specific ways.
It should also be noted that, the embodiment of the present invention in, " (English：Of) ", " corresponding (English： Corresponding, relevant) " and " corresponding (English：Corresponding it) " can use with sometimes, it should be pointed out that It is that, when not emphasizing its difference, meaning to be expressed is consistent.
For the ease of clearly describing the technical solution of the embodiment of the present invention, in an embodiment of the present invention, use " the One ", the printed words such as " second " distinguish function and the essentially identical identical entry of effect or similar item, and those skilled in the art can To understand that the printed words such as " first ", " second " are not to be defined to quantity and execution order.
The fitness of existing multiple sensor information amalgamation method and actual conditions is relatively low, does not consider occur in practice generally Observation random fault situation the case where causing the observation of sensor to be not present, so it is directed to actual multisensor syste For precision it is inadequate, it is less practical.
In view of the above-mentioned problems, shown in referring to Fig.1, the embodiment of the present invention provides a kind of information fusion side of multisensor syste Method, including：
101, observational equation is established according to default Chance Failure Rate.
In practice, state equation can be also established while establishing observational equation.
102, the sensor of interest that can be merged into row information at current time in multisensor syste is determined.
Illustratively, 102 steps are specially：According to each sensor in value at the time of current time and multisensor syste Sampling than determine current time can into row information merge sensor of interest.
103, according to observational equation calculate sensor of interest the observation at current time theoretical value.
Specifically, because introducing observation random fault in observational equation, according to the observation calculated in equation For theoretical value, and in practice observation occur the when of observing random fault be not present it is different with theoretical value, but believing It must be used again to accurate observation when not breaking down when breath fusion, so the theoretical value of observation is introduced herein, after For the influence one more breakdown judge process of observation random fault in continuous method flow；
In addition, it is necessary to explanation, the theoretical value of observation in practice generally will use be calculated to state value, institute With optional, further include before the progress of 103 steps：State value of the calculating sensor of interest at current time.
104, value of the calculating sensor of interest in the observation residual error statistical property at current time.
Optionally, 104 steps include：It is adopted according to each sensor in value at the time of current time and multisensor syste Sample is than determining the sensor of interest that can be merged into row information at current time；
Wherein sample the sampling rate than referring to the sensor that sampling rate is most fast in multisensor syste and each sensor Sampling rate sum ratio；The value of the sampling ratio of each sensor is obtained before 101 steps；Illustratively, such as institute in Fig. 3 Show, the sampling of sensor 1 is than being 1, and than being 3, it is 4 that the sampling of sensor 3, which is compared, for the sampling of sensor 2.
Illustratively, it is set as 0 at the time of multisensor syste being brought into operation each time, value is then at the time of when operation one second It is 1, and so on, value is k at the time of at the time of when running ks.
105, the value according to sensor of interest in the observation residual error statistical property at current time and default chi square distribution value Magnitude relationship judges that sensor of interest whether there is mistake in the observation at current time.
Optionally, 105 steps specifically include：When sensor of interest is in the value of the observation residual error statistical property at current time When absolute value is more than default chi square distribution value, determine that there are mistakes for observation of the sensor of interest at current time；
When sensor of interest is less than or equal to default card side in the absolute value of the value of the observation residual error statistical property at current time When Distribution Value, determine that mistake is not present in observation of the sensor of interest at current time.
106, the observation according to sensor of interest at current time obtains sensor of interest with the presence or absence of the situation of mistake In the result that current time merges into row information.
Illustratively, 106 steps specifically include：
1061, the observation according to sensor of interest at current time obtains sensor of interest with the presence or absence of the situation of mistake In the kalman gain value at current time.
Optionally, 1061 steps include：When sensor of interest current time observation exist mistake when, determine target Kalman gain value of the sensor at current time is zero；When mistake is not present in observation of the sensor of interest at current time When, then according to kalman gain calculation formula calculate sensor of interest current time kalman gain value.
1062, according to sensor of interest current time kalman gain value calculate sensor of interest current time into State Estimation of Mean value after row information fusion and target covariance.
Optionally, 1062 steps include：According to sensor of interest current time kalman gain value with sampling rate Sensor of interest is calculated the state of sensor of interest by descending sequence according to state mean prediction value more new formula successively The updated value of mean prediction value；State Estimation of Mean value is the state of the sensor of interest of sampling rate minimum in sensor of interest The updated value of mean prediction value；
According to sensor of interest current time kalman gain value with the descending sequence of sampling rate by target Sensor calculates the updated value of the target covariance of sensor of interest according to target covariance more new formula successively；Target covariance It is the updated value of the target covariance of the sensor of interest of sampling rate minimum in sensor of interest.
It should be noted that the information fusion of multisensor is generally to calculate do well Estimation of Mean value and mesh in practice Covariance is marked as final result, so including 1061 and 1062 steps in above-mentioned 106 step；In addition, above-mentioned kalman gain Required parameter value can be by existing in calculation formula, state mean prediction value more new formula and target covariance more new formula Some CKF algorithms etc. obtain, and the present invention is not particularly limited this.
The information fusion method of multisensor syste provided in an embodiment of the present invention, this method include：According to default random Failure rate establishes observational equation；Determine the sensor of interest that can be merged into row information at current time in multisensor syste； According to the state value of state equation and dynamic sliding average equation calculation sensor of interest current time；It is calculated according to observational equation Theoretical value of the sensor of interest in the observation at current time；It calculates observation residual error of the sensor of interest at current time and counts special The value of property；It is closed in the value of the observation residual error statistical property at current time and the size of default chi square distribution value according to sensor of interest System judges that sensor of interest whether there is mistake in the observation at current time；According to sensor of interest current time observation Value obtains the result that sensor of interest merges at current time into row information with the presence or absence of the situation of mistake.The embodiment of the present invention carries The technical solution of confession is based on existing CKF algorithms, is introduced in observational equation and observes random random fault, and then judging can be with The sensor of interest merged then judges observation random fault, finally obtains multisensor according to judging result The result of system information fusion.Because technical solution provided in an embodiment of the present invention is just introduced when establishing observational equation at the beginning Random observation failure, and then corresponding judgement has been carried out to random fault, finally get multi-sensor information fusion Afterwards as a result, so the more realistic situation of technical solution provided in an embodiment of the present invention compared with prior art, for environment Fitness is also with regard to higher.
For the detailed process of clearer explanation multiple sensor information amalgamation method provided in an embodiment of the present invention, reference Shown in Fig. 2, the embodiment of the present invention provides the information fusion method of another multisensor syste, including：
201, it establishes the state equation of multisensor syste and observational equation is established according to default Chance Failure Rate.
Specifically, state equation is：
X (k+1)=A (k) x (k)+w (k)；
Observational equation is：
yi(ki)=γi(ki)[Ci(ki)xi(ki)]+νi(ki)；
x(k)∈RnState value for multisensor syste at the k moment, x (k+1) are multisensor syste at the k+1 moment State value, A (k) ∈ Rn*nSystematic state transfer matrix for multisensor syste at the k moment, w (k) are multisensor syste in k The system noise at moment；Wherein n refers to the dimension of multisensor syste, for example, the detection of certain multisensor syste parameter have speed, The dimension of acceleration and the displacement then multisensor syste is 3；
yi(ki)∈RmFor the reason of the observation at k-th moment of i-th of the sensor in multisensor syste belonging to itself By value because assuming that observation random fault is loss of data in the present embodiment, in observational equation for indicate observation with The γ of machine failurei(ki) ∈ R be value be 0 or 1 and meet Bernoulli Jacob distribution random sequence (in practice can according to observation with The Type Change of machine failure), it is mainly used for showing the observed failure rate of multisensor syste, γ in observational equationi(ki) observation Value shows that observation is broken down when being 0 when, γi(ki) observe showing to observe when value is 1 and not break down, 0 in all values In the ratio that accounts for be observed failure rate；Ci(ki)∈Rm*nFor kth of i-th of sensor belonging to itself in multisensor syste The observing matrix at a moment, xi(ki) be multisensor syste in k-th moment of i-th of sensor belonging to itself state Value, νi(ki) it is k-th moment observation noise of i-th of sensor belonging to itself in multisensor syste；M refers to sensor Dimension, m are less than or equal to n；
Because in multisensor syste in practice, existing between the system noise and observation noise in multisensor syste Certain correlation, so in the present embodiment：
νi(k)=βiw(k-1)+ηi(k)；
νi(k) be multisensor syste in i-th of sensor in the observation noise at k moment, βiIt is in multisensor syste For the observation noise and multisensor syste of i sensor in the relevance parameter of the system noise at k moment, w (k-1) is more sensings Device system is in the system noise at k-1 moment, ηi(k) be multisensor syste in i-th of sensor the k moment random zero mean White Gaussian noise；I is positive integer；Experimenter is rule of thumb depending on relevance parameter concrete foundation reality or when emulation experiment Depending on historical data；
Optionally, the state value of multisensor syste at any one time includes system noise, any in multisensor syste The observation of sensor at any one time includes observation noise.
Illustratively, it is set as 0 at the time of multisensor syste being brought into operation each time, value is then at the time of when operation one second It is 1, and so on, value at the time of when running ks is k.
202, the sensor of interest that can be merged into row information at current time in multisensor syste is determined.
With reference to shown in Fig. 3, by taking some is there are the multisensor syste of three different sensors as an example, in figure, identical height The vertical bands point straight line of degree belongs to same sensor；Because the sampling rate of three sensors is different, it is impossible to every A moment can merge fixed several sensor informations, so needing the information to which sensor of current time can at this time Interpretation is carried out with fusion, and judgment mode in practice is there are a variety of, and in embodiments of the present invention, at the time of according to current time The sampling of value and each sensor according to remainder function than determining the sensor of interest that can be merged into row information at current time；
The remainder function is：
For the sampling ratio of p-th of sensor of interest in sensor of interest, k is moment value, and p is positive integer；Because of sampling The most fast sensor of rate necessarily participate in information each time fusion, if so the sampling ratio of sensor can by it is current when Value divides exactly at the time of quarter, then shows that the sensor and the most fast sensor of speed rates rate are sampled at current time, It can be merged into row information.
Wherein sample the sampling rate than referring to the sensor that sampling rate is most fast in multisensor syste and each sensor Sampling rate sum ratio；The value of the sampling ratio of each sensor is obtained before 101 steps；Illustratively, such as institute in Fig. 3 Show, the sampling of sensor 1 is than being 1, and than being 3, it is 4 that the sampling of sensor 3, which is compared, for the sampling of sensor 2.
203, according to the state value of state equation and dynamic sliding average equation calculation sensor of interest current time.
Specifically, each sensor in multisensor syste is asynchronous multi tate, multi-sensor information is being carried out When fusion, because primary information fused data only exists primary transfer, and the biography most fast by transmission rate is embodied in specifically is shifted In the data update of sensor, so in the information fusion process at each moment, it is only directed to the most fast sensor of transmission rate State value calculates ability use state equation and is calculated, remaining sensor is then calculated using dynamic sliding average equation； So with reference to shown in Fig. 3,203 steps include：
2031, the state value according to first object sensor in previous moment calculates first object sensing according to state equation State value of the device at current time.
Wherein, first object sensor is the sensor of interest that sampling rate is most fast in sensor of interest.
2032, the state-transition matrix according to first object sensor in the state value and state equation at current time, According to the sensor of interest in dynamic sliding average equation calculation sensor of interest in addition to first object sensor when current The state value at quarter.
The dynamic sliding average equation is：
Wherein, niFor the sampling ratio of i-th of sensor in multisensor syste, x1(niki) it is that first object sensor exists The n-th of i-th of sensorikiThe state value at a moment.
204, the state value and observational equation according to sensor of interest at current time calculate sensor of interest when current The theoretical value of the observation at quarter.
Specifically, because introducing observation random fault in observational equation, according to the observation calculated in equation For theoretical value, and in practice observation occur the when of observing random fault be not present it is different with theoretical value, but believing It must be used again to accurate observation when not breaking down when breath fusion, so the theoretical value of observation is introduced herein, after For the influence one more breakdown judge process of observation random fault in continuous method flow.
205, according to sensor of interest in the state Estimation of Mean value and target covariance of previous moment, foundation volume karr Graceful filtering CKF algorithms calculate state mean prediction value, observation predicted value, observational variance of the sensor of interest at current time and predict Value and target covariance predicted value.
Wherein, state Estimation of Mean value is the estimated value of the mean value of state value；Target covariance be state value mean value with The covariance of observation；State mean prediction value is the predicted value of the mean value of state value；Observe the prediction that predicted value is observation Value；Observational variance predicted value is the predicted value of the variance of observation；Target covariance predicted value is the predicted value of target covariance；
It should be noted that in the state mean prediction value, observation predicted value, observation that calculate the first moment of sensor of interest State Estimation of Mean value and mesh of the sensor of interest of foundation in previous moment when variance predicted value and target covariance predicted value It is preset value to mark covariance, is obtained before 101 steps.
206, observation residual error of the sensor of interest at current time is calculated according to observation residual error statistical property calculation formula to unite Count the value of characteristic.
Since multi-sensor information fusion process summarizes and introduces observation random fault in the embodiment of the present invention, so to every One sensor carry out being required for before the update of subsequent state Estimation of Mean value and target covariance carrying out the judgement of failure with The case where determining kalman gain, and specifically judge whether to deposit by the way of calculating residual error statistical property in the embodiment of the present invention In observation random fault, so optionally, 106 steps are specially in embodiments of the present invention：
According to sensor of interest in the theoretical value of the observation at current time and observation predicted value and sensor of interest Sample the observation residual error statistical property at current time than foundation observation residual error statistical property calculation formula calculating sensor of interest Value；
Observing residual error statistical property calculation formula is：
Wherein,For p-th of sensor of interest the residual error statistical property at k moment value,It is P sensor of interest the observation at k moment theoretical value,It is pre- in the observation at k moment for p-th of sensor of interest Measured value,It is p-th of sensor of interest in the covariance at k moment, wlFor the weights of volume point, l is of volume point Number,For the theoretical value of p-th of sensor of interest observation of first of volume point at the k moment,For P-th of sensor of interest observation noise variance at the k moment.
207, the value according to sensor of interest in the observation residual error statistical property at current time and default chi square distribution value Magnitude relationship judges that sensor of interest whether there is mistake in the observation at current time.
Illustratively, observation mistake that may be present includes but not limited to following any：Caused by loss of data Observation is not present, sensor fault causes the theoretical value of the observation actually obtained and observation to differ, to sensor The incorrect theoretical value for leading to the observation actually obtained and observation of observation process when observation differs.
Optionally, with reference to shown in Fig. 4,207 steps include：
2071, when sensor of interest is more than default card side in the absolute value of the value of the observation residual error statistical property at current time When Distribution Value, determine that there are mistakes for observation of the sensor of interest at current time.
Specifically, the value of the observation residual error meter characteristic of sensor of interestThe chi square distribution that degree of freedom is m is obeyed, M is the observation dimension of the sensor of interest, so can determine default card side point from chi-square distribution table after default confidence level ImplantationWhenThen illustrate that observation is broken down, observation of the sensor of interest at current time There are mistakes for value.
2072, it is less than or equal in the absolute value of the value of the observation residual error statistical property at current time when sensor of interest default When chi square distribution value, determine that mistake is not present in observation of the sensor of interest at current time.
WhenThen illustrate that observation is not broken down, observation of the sensor of interest at current time Mistake is not present in value.
208, sensor of interest is obtained at current time with the presence or absence of the situation of mistake according to the observation of sensor of interest Kalman gain value.
Optionally, with reference to shown in Fig. 4,208 steps include：
2081, mistake in the presence of observation at current time when sensor of interest, determines sensor of interest when current The kalman gain value at quarter is zero.
2082, when sensor of interest current time observation there is no mistake when, then worked as according to sensor of interest The observational variance predicted value and target covariance predicted value at preceding moment calculate sensor of interest according to kalman gain calculation formula In the kalman gain value at current time；
The kalman gain calculation formula is：
Kalman gain for p-th of sensor of interest at the k moment,It is passed for p-th of target Target covariance predicted value of the sensor at the k moment,Observational variance for p-th of sensor of interest at the k moment is predicted Value.
Illustratively, 1081 and 1,082 two steps can specifically be indicated with following formula：
209, according to sensor of interest current time state mean prediction value, kalman gain value, the reason of observation By value, observation predicted value, state mean prediction value, target covariance predicted value and observational variance predicted value, according to state mean value Predicted value more new formula and target covariance more new formula calculate shape of the sensor of interest after current time merges into row information State Estimation of Mean value and target covariance.
Optionally, with reference to shown in Fig. 4,209 steps include：
2091, according to sensor of interest current time state mean prediction value, kalman gain value, the reason of observation By being worth and observation predicted value, with the descending sequence of sampling rate by sensor of interest successively according to state mean prediction value more New formula calculates the updated value of the state mean prediction value of sensor of interest；By the target of sampling rate minimum in sensor of interest The updated value of the state mean prediction value of sensor is the state Estimation of Mean value after the fusion of sensor of interest information.
Illustratively, state mean prediction value more new formula is：
Wherein,For p-th of sensor of interest the state mean prediction value at k moment updated value,Kalman gain value for p-th of sensor of interest at the k moment,It is p-th of sensor of interest in k The state mean prediction value at moment.
2092, according to sensor of interest in target covariance predicted value, kalman gain value and the observation side at current time Sensor of interest is calculated mesh by poor predicted value according to target covariance more new formula successively with the descending sequence of sampling rate Mark the updated value of the target covariance of sensor；By the target association side of the sensor of interest of sampling rate minimum in sensor of interest The updated value of difference is the target covariance after the fusion of sensor of interest information.
Illustratively, target covariance more new formula is：
Wherein,For p-th of sensor of interest the target covariance at k moment updated value,For Target covariance predicted value of p-th of sensor of interest at the k moment,It is p-th of sensor of interest at the k moment Observational variance predicted value.
It should be noted that in aforesaid state mean prediction value more new formula and target covariance more two formula of new formula Parameter needs acquired by CKF algorithms, so there are following iterative relations：P-th of sensor of interest exists in sensor of interest The state Estimation of Mean value of previous moment is that -1 sensor of interest of pth is pre- in the state mean value at current time in sensor of interest The updated value of measured value；
P-th of sensor of interest pth -1 in the target covariance of previous moment is sensor of interest in sensor of interest Updated value of a sensor of interest in the target covariance at current time；
The sampling rate of p-th of sensor of interest is merely greater than p-1 target sensing in sensor of interest in sensor of interest The sensor of interest of the sampling rate of device；The sampling rate of -1 sensor of interest of pth is only smaller than target biography in sensor of interest The sampling rate of p-th of sensor of interest in sensor；P is positive integer.
The state mean value after the information fusion of any moment multisensor syste can be obtained according to above method flow to estimate Evaluation and target covariance complete the information fusion estimation of multisensor in multisensor syste.
The information fusion method for the multisensor syste that above-mentioned specific embodiment provides includes：Establish multisensor syste State equation simultaneously establishes observational equation according to default Chance Failure Rate；Determining in multisensor syste can carry out at current time The sensor of interest of information fusion；According to the shape of state equation and dynamic sliding average equation calculation sensor of interest current time State value；According to sensor of interest current time state value and observational equation calculate sensor of interest current time observation The theoretical value of value；According to sensor of interest in the state Estimation of Mean value and target covariance of previous moment, foundation volume karr Graceful filtering CKF algorithms calculate state mean prediction value, observation predicted value, observational variance of the sensor of interest at current time and predict Value and target covariance predicted value；Target covariance is the covariance of state mean value and observation；It is special according to observation residual error statistics Property calculation formula calculate sensor of interest observation residual error statistical property value；Count special according to the observation residual error of sensor of interest Property value and default chi square distribution value magnitude relationship judge sensor of interest observation whether there is random fault；According to target The observation of sensor whether occur the case where random fault obtain sensor of interest current time kalman gain value；According to Sensor of interest is in the state mean prediction value at current time, kalman gain value, observation, observation predicted value, state mean value Predicted value, target covariance predicted value and observational variance predicted value, according to state mean prediction value more new formula and target association side Poor more new formula calculates state Estimation of Mean value and target covariance of the sensor of interest after current time merges into row information. As can be seen that the technical solution that above-mentioned specific embodiment provides, is based on existing CKF algorithms, observation is introduced in observational equation Then random random fault judges the sensor of interest that can be merged, then judged observation random fault with true Determine kalman gain value, the sequential blending algorithm for reusing the information fusion applied to asynchronous multirate sensor passes target Sensor is merged according to the ascending sequence of sampling rate, finally calculate fusion after the completion of state Estimation of Mean value and Target covariance.Because the technical solution just introduces random observation failure when establishing observational equation at the beginning, and then Corresponding judgement has been carried out to random fault, has finally calculated the knot after multi-sensor information fusion using sequential blending algorithm Fruit, so technical solution provided in an embodiment of the present invention is stronger to the adaptive faculty of environment compared with prior art, more realistic feelings Condition, precision higher.
In order to which vivider illustrates that the information fusion algorithm that above-described embodiment provides is compared and existing information blending algorithm Superiority is illustrated with actual illiteracy Quattro the simulation experiment result below：
For the estimation effect of evaluation status mean value, the root-mean-square error of root-mean-square error RMSE and average time are introduced TARMSE, for scaling information blending algorithm to the index of state Estimation of Mean value effect, value is smaller, illustrates when the two parameters Estimation effect is better,
Wherein, T is the length of the emulation signal of Monte-Carlo Simulation, and M is the number of Monte-Carlo Simulation, when x (k) is k The state value of multisensor syste is carved,It is l*The state Estimation of Mean that the k moment of secondary Monte-Carlo Simulation obtains Value.
UKF (Unscented Kalman Filter, lossless Kalman filtering) method, CKF are used in emulation experiment (Cubage Kalman Filter, volume Kalman filtering) method and the UKF methods of fusion with it is provided in an embodiment of the present invention For method (the CKF methods of fusion) into the comparison of row index, it is 600 to be provided with emulation signal length, and state dimension is three-dimensional, the One-dimensional is position, and the second dimension is speed, and the third dimension is acceleration, and to three-dimensional state into line trace, i.e. observation is also three-dimensional for observation；
First, it is merged with provided in an embodiment of the present invention with original UKF methods, CKF methods and fusion UKF methods CKF methods do emulation experiment, all introduce 10% observation random fault in four kinds of methods, wherein UKF methods, CKF methods and Fusion UKF methods itself can't exclude observation random fault, and the knot of table 1, table 2 and Fig. 6-Fig. 8 can be obtained in experiment Fruit：
Table 1
Table 2
As it can be seen from table 1 the estimation effect that the CKF methods of fusion provided in an embodiment of the present invention finally obtain exists In the case of observing random fault, it to be substantially better than UKF methods, CKF methods and fusion UKF methods, as can be seen from Table 2 conduct The UKF methods of the same fusion for multisensor syste are compared with fusion CKF methods provided in an embodiment of the present invention, the present invention The method run time that embodiment provides is shorter, more efficient；
It can be seen that with reference to the RMSE index curves of Fig. 6-emulation experiments shown in Fig. 8 and have fault detect and exclusion The index curve of the CKF methods of fusion is more stablized, and shows compared with existing UKF methods, CKF methods and fusion UKF methods, The influence provided in an embodiment of the present invention that can exclude observation random fault and raising system stability, are more suitable for actual multiple Heterocycle border.
In addition, in order to more further show the advantage of technical solution provided in an embodiment of the present invention, by the embodiment of the present invention In fault detection and exclusion method be introduced into existing UKF methods, CKF methods and fusion UKF methods in, it can be deduced that it is as follows The experimental result of table 3, table 4, table 5, table 6 and Fig. 9-Figure 14；
With reference to the following table 3 and table 4 as can be seen that either observing random fault either with or without setting, the embodiment of the present invention proposes The estimated accuracy of state mean value of information fusion method be all substantially better than existing UKF methods and CKF methods, referring concurrently to figure The RMSE index curves of the emulation experiment of 9- Figure 14 can be seen that in the 3 d case, information provided in an embodiment of the present invention The RMSE curves of blending algorithm are optimal compared to the estimation effect of existing UKF methods and CKF methods also minimum i.e. state mean value, with The UKF methods of existing fusion are suitable compared to the estimation effect of then state mean value；
Table 3
Table 4
And with reference to table 5 and table 6 as can be seen that with or without random fault, information provided in an embodiment of the present invention is observed The emulation experiment of fusion method, which goes out the time needed for result, will be significantly less than the UKF methods of existing fusion；
Table 5
UKF methods and CKF methods are single-sensor algorithm, so run time is than the fusion for multisensor syste UKF and fusion CKF run time it is short.
Table 6
So in general, the information fusion method of multisensor syste provided in an embodiment of the present invention is than existing side Method effect is all good.
Referring to Figure 5, the embodiment of the present invention also provides a kind of information fuse device 01 of multisensor syste, including：
Establishing equation module 501, for establishing observational equation according to default Chance Failure Rate；
Judgment module 502 is merged, for determining the mesh that can be merged into row information at current time in multisensor syste Mark sensor；
Computing module 503 is observed, the observational equation for being established according to establishing equation module 501 calculates fusion judgment module Theoretical value of 502 sensor of interest determined in the observation at current time；
Residual error statistical property computing module 504, for calculating the sensor of interest of the fusion determination of judgment module 502 current The value of the observation residual error statistical property at moment；
Random fault judgment module 505, the sensor of interest for being calculated according to residual error statistical property computing module 504 exist The value of the observation residual error statistical property at current time and the magnitude relationship of default chi square distribution value judge sensor of interest current The observation at moment whether there is mistake；
Fusion calculation module 506 exists for obtaining sensor of interest according to the judging result of random fault judgment module 505 The result that current time merges into row information.
Optionally, because the observation theoretical value of sensor calculates the state value for needing to use sensor in practice, The device further includes state computation module 507；
The observational equation that state computation module 507 is used to be established according to establishing equation module 501 in observation computing module 503 Sensor of interest is calculated before the theoretical value of the observation at current time, state value of the calculating sensor of interest at current time.
Optionally, fusion judgment module 502 is specifically used for：
It is determined at current time according to the sampling ratio of each sensor in value at the time of current time and multisensor syste The sensor of interest that can be merged into row information.
Optionally, residual error statistical property computing module 504 is specifically used for：
Compared according to observation in the theoretical value of the observation at current time and the sampling of sensor of interest according to sensor of interest Value of the residual error statistical property calculation formula calculating sensor of interest in the observation residual error statistical property at current time.
Optionally, random fault judgment module 505 is specifically used for：
When sensor of interest is more than default chi square distribution in the absolute value of the value of the observation residual error statistical property at current time When value, determine that there are mistakes for observation of the sensor of interest at current time；
When sensor of interest is less than or equal to default card side in the absolute value of the value of the observation residual error statistical property at current time When Distribution Value, determine that mistake is not present in observation of the sensor of interest at current time.
Optionally, fusion calculation module 506 includes kalman gain computing unit 5061, fusion treatment unit 5062；
Kalman gain computing unit 5061 is used to obtain target according to the judging result of random fault judgment module 505 and pass Kalman gain value of the sensor at current time；
The sensor of interest that fusion treatment unit 5062 is used to be calculated according to kalman gain computing unit is at current time Kalman gain value calculate sensor of interest current time into row information merge after state Estimation of Mean value and target association Variance.
Optionally, kalman gain computing unit 5061 is specifically used for：
When random fault judgment module 505 determines that sensor of interest when the observation at current time has mistake, determines Kalman gain value of the sensor of interest at current time is zero；
When random fault judgment module 505 determine sensor of interest current time observation there is no mistake when, then According to kalman gain calculation formula calculate sensor of interest current time kalman gain value.
Optionally, fusion treatment unit 5062 includes state Estimation of Mean value computation subunit 50621 and target covariance Computation subunit 50622；
State Estimation of Mean value computation subunit 50621, the target for being calculated according to kalman gain computing unit pass Kalman gain value of the sensor at current time, with the descending sequence of sampling rate by sensor of interest successively according to state Mean prediction value more new formula calculates the updated value of the state mean prediction value of sensor of interest；State Estimation of Mean value is target The updated value of the state mean prediction value of the sensor of interest of sampling rate minimum in sensor；
Target covariance computation subunit 50622, the target for being calculated according to kalman gain computing unit 5061 pass Kalman gain value of the sensor at current time, with the descending sequence of sampling rate by sensor of interest successively according to target Covariance more new formula calculates the updated value of the target covariance of sensor of interest；Target covariance is sampled in sensor of interest The updated value of the target covariance of the sensor of interest of rate minimum.
It should be noted that in the information fuse device of multisensor syste provided in an embodiment of the present invention, can also wrap It is used to obtain the sampling ratio of parameter such as sensor when multisensor syste operation, the sampling rate of sensor containing acquisition module Deng directly can also directly acquiring the parameter needed for oneself by above-mentioned modules, be not particularly limited herein.
The information fuse device of multisensor syste provided in an embodiment of the present invention, because the device includes：Establishing equation Module, for establishing observational equation according to default Chance Failure Rate；Merge judgment module, for determine in multisensor syste The sensor of interest that current time can merge into row information；Computing module is observed, for what is established according to establishing equation module Theoretical value of the observational equation calculating sensor of interest in the observation at current time；Residual error statistical property computing module is used for root It is residual in the observation at current time that the sensor of interest that fusion judgment module determines is calculated according to observation residual error statistical property calculation formula The value of poor statistical property；Random fault judgment module, the sensor of interest for being calculated according to residual error statistical property computing module Judge that sensor of interest is being worked as in the value of the observation residual error statistical property at current time and the magnitude relationship of default chi square distribution value The observation at preceding moment whether there is mistake；Fusion calculation module, for being obtained according to the judging result of random fault judgment module The result for taking sensor of interest to be merged into row information at current time.It, can be with so technical solution provided in an embodiment of the present invention It on the basis of based on existing CKF algorithms, is introduced in observational equation and observes random random fault, then judge to carry out The sensor of interest of fusion then judges observation random fault, finally obtains multisensor syste according to judging result The result of information fusion.Because technical solution provided in an embodiment of the present invention just introduced when establishing observational equation at the beginning with Machine observes failure, and has then carried out corresponding judgement to random fault, after finally getting multi-sensor information fusion As a result, so the more realistic situation of technical solution provided in an embodiment of the present invention compared with prior art, the adaptation for environment Degree is also with regard to higher.
The embodiment of the present application provides a kind of computer equipment, including memory, processor.Being stored on the memory can be The computer program run on processor, the processor realize multisensor syste above-mentioned when executing above computer program Information incorporates method.Wherein, above-mentioned storage medium includes：ROM, RAM, magnetic disc or CD etc. are various can to store program code Medium.
The embodiment of the present application provides a kind of computer-readable medium, is stored with computer program, the computer program quilt Processor realizes that the information of multisensor syste above-mentioned incorporates method when executing.
The above description is merely a specific embodiment, but scope of protection of the present invention is not limited thereto, any Those familiar with the art in the technical scope disclosed by the present invention, all answer by the change or replacement that can be readily occurred in It is included within the scope of the present invention.Therefore, protection scope of the present invention should be subject to the protection scope in claims.

Claims (18)

1. a kind of information fusion method of multisensor syste, which is characterized in that including：
Observational equation is established according to default Chance Failure Rate；
Determine the sensor of interest that can be merged into row information at current time in the multisensor syste；
According to the observational equation calculate the sensor of interest the observation at current time theoretical value；
Calculate the sensor of interest the observation residual error statistical property at current time value；
According to the sensor of interest the value and default chi square distribution value of the observation residual error statistical property at current time size Relationship judges that the observation at current time of the sensor of interest whether there is mistake；
Observation according to the sensor of interest at current time obtains the sensor of interest with the presence or absence of the situation of mistake In the result that current time merges into row information.
2. according to the method described in claim 1, it is characterized in that, system noise and observation in the multisensor syste are made an uproar There are correlations between sound, meet following formula：
νi(k)=βiw(k-1)+ηi(k)；
Wherein, νi(k) be the multisensor syste in i-th of sensor in the observation noise at k moment, βiFor more sensings In device system the observation noise of i-th sensor and the multisensor syste the system noise at k moment relevance parameter, W (k-1) is the multisensor syste in the system noise at k-1 moment, ηi(k) it is i-th of biography in the multisensor syste Random zero mean white Gaussian noise of the sensor at the k moment；I is positive integer.
3. according to the method described in claim 2, it is characterized in that, the state value packet of the multisensor syste at any one time Containing the system noise, the observation of any sensor at any one time includes observation noise in the multisensor syste.
4. according to the method described in claim 1, it is characterized in that, at current time in the determination multisensor syste Can include into the sensor of interest of row information fusion：
It is determined at current time according to the sampling ratio of each sensor in value at the time of current time and the multisensor syste The sensor of interest that can be merged into row information.
5. according to the method described in claim 1, it is characterized in that, it is described calculate the sensor of interest current time sight Survey residual error statistical property value include：
Compare foundation in the theoretical value of the observation at current time and the sampling of the sensor of interest according to the sensor of interest Observation residual error statistical property calculation formula calculate the sensor of interest the observation residual error statistical property at current time value.
6. according to the method described in claim 1, it is characterized in that, it is described according to the sensor of interest current time sight Survey residual error statistical property value and default chi square distribution value magnitude relationship judge the sensor of interest current time sight Measured value includes with the presence or absence of mistake：
When the sensor of interest is more than the default card side in the absolute value of the value of the observation residual error statistical property at current time When Distribution Value, determine that there are mistakes for observation of the sensor of interest at current time；
When the sensor of interest is less than or equal to described preset in the absolute value of the value of the observation residual error statistical property at current time When chi square distribution value, determine that mistake is not present in observation of the sensor of interest at current time.
7. according to the method described in claim 1, it is characterized in that, it is described according to the sensor of interest current time sight Measured value obtains the result that the sensor of interest merges at current time into row information with the presence or absence of the situation of mistake：
Observation according to the sensor of interest at current time obtains the sensor of interest with the presence or absence of the situation of mistake In the kalman gain value at current time；
Kalman gain value according to the sensor of interest at current time calculate the sensor of interest current time into State Estimation of Mean value after row information fusion and target covariance.
8. the method according to the description of claim 7 is characterized in that it is described according to the sensor of interest current time sight Measured value obtains kalman gain value of the sensor of interest at current time with the presence or absence of the situation of mistake：
When the sensor of interest current time observation exist mistake when, determine the sensor of interest at current time Kalman gain value be zero；
When the sensor of interest current time observation there is no mistake when, then according to kalman gain calculation formula meter Calculate the sensor of interest current time kalman gain value.
9. the method according to the description of claim 7 is characterized in that it is described according to the sensor of interest current time card Germania yield value calculates state Estimation of Mean value and target association of the sensor of interest after current time merges into row information Variance includes：
According to the sensor of interest the kalman gain value at current time will be described with the descending sequence of sampling rate Sensor of interest calculates the state mean prediction value of the sensor of interest according to state mean prediction value more new formula successively Updated value；The state Estimation of Mean value is the state mean value of the sensor of interest of sampling rate minimum in the sensor of interest The updated value of predicted value；
According to the sensor of interest the kalman gain value at current time will be described with the descending sequence of sampling rate Sensor of interest calculates the updated value of the target covariance of the sensor of interest according to target covariance more new formula successively；Institute State the updated value that target covariance is the target covariance of the sensor of interest of sampling rate minimum in the sensor of interest.
10. a kind of information fuse device of multisensor syste, which is characterized in that including：Establishing equation module, fusion judge mould Block, observation computing module, observation residual error statistical property computing module, random fault judgment module and fusion calculation module；
The establishing equation module, for establishing observational equation according to default Chance Failure Rate；
The fusion judgment module, for determining the mesh that can be merged into row information at current time in the multisensor syste Mark sensor；
The observation computing module, the observational equation for being established according to the establishing equation module calculate the fusion and sentence The theoretical value of observation of the sensor of interest that disconnected module determines at current time；
The residual error statistical property computing module is being worked as calculating the sensor of interest that the fusion judgment module determines The value of the observation residual error statistical property at preceding moment；
The random fault judgment module, the sensor of interest for being calculated according to the residual error statistical property computing module Judge the sensor of interest in the value of the observation residual error statistical property at current time and the magnitude relationship of default chi square distribution value It whether there is mistake in the observation at current time；
The fusion calculation module, for obtaining the sensor of interest according to the judging result of the random fault judgment module In the result that current time merges into row information.
11. device according to claim 10, which is characterized in that the fusion judgment module is specifically used for：
It is determined at current time according to the sampling ratio of each sensor in value at the time of current time and the multisensor syste The sensor of interest that can be merged into row information.
12. device according to claim 10, which is characterized in that the residual error statistical property computing module is specifically used for：
Compare foundation in the theoretical value of the observation at current time and the sampling of the sensor of interest according to the sensor of interest Observation residual error statistical property calculation formula calculate the sensor of interest the observation residual error statistical property at current time value.
13. device according to claim 10, which is characterized in that the random fault judgment module is specifically used for：
When the sensor of interest is more than the default card side in the absolute value of the value of the observation residual error statistical property at current time When Distribution Value, determine that there are mistakes for observation of the sensor of interest at current time；
When the sensor of interest is less than or equal to described preset in the absolute value of the value of the observation residual error statistical property at current time When chi square distribution value, determine that mistake is not present in observation of the sensor of interest at current time.
14. device according to claim 10, which is characterized in that the fusion calculation module includes that kalman gain calculates Unit, fusion treatment unit；
The kalman gain computing unit is used to obtain the target according to the judging result of the random fault judgment module Kalman gain value of the sensor at current time；
The sensor of interest that the fusion treatment unit is used to be calculated according to the kalman gain computing unit is current The kalman gain value at moment calculates state Estimation of Mean value of the sensor of interest after current time merges into row information With target covariance.
15. device according to claim 14, which is characterized in that the kalman gain computing unit is specifically used for：
When the random fault judgment module determines that the sensor of interest when the observation at current time has mistake, determines Kalman gain value of the sensor of interest at current time is zero；
When the random fault judgment module determine the sensor of interest current time observation there is no mistake when, then According to kalman gain calculation formula calculate the sensor of interest current time kalman gain value.
16. device according to claim 14, which is characterized in that the fusion treatment unit includes state Estimation of Mean value Computation subunit and target covariance computation subunit；
The state Estimation of Mean value computation subunit, the target for being calculated according to the kalman gain computing unit Kalman gain value of the sensor at current time, with the descending sequence of sampling rate by the sensor of interest successively according to The updated value of the state mean prediction value of the sensor of interest is calculated according to state mean prediction value more new formula；The state is equal Value estimated value is the updated value of the state mean prediction value of the sensor of interest of sampling rate minimum in the sensor of interest；
The target covariance computation subunit, the target for being calculated according to the kalman gain computing unit sense Kalman gain value of the device at current time, with the descending sequence of sampling rate by the sensor of interest successively according to mesh Mark covariance more new formula calculates the updated value of the target covariance of the sensor of interest；The target covariance is the mesh Mark the updated value of the target covariance of the sensor of interest of sampling rate minimum in sensor.
17. a kind of computer equipment, which is characterized in that including memory, processor；Being stored on the memory can be described The computer program run on processor, the processor are realized when executing the computer program as claim 1-9 is any Method described in.
18. a kind of computer-readable medium, is stored with computer program, which is characterized in that the computer program is handled Such as claim 1-9 any one of them methods are realized when device executes.
CN201810497192.XA 2018-05-22 2018-05-22 Information fusion method and device of multi-sensor system Active CN108734218B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810497192.XA CN108734218B (en) 2018-05-22 2018-05-22 Information fusion method and device of multi-sensor system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810497192.XA CN108734218B (en) 2018-05-22 2018-05-22 Information fusion method and device of multi-sensor system

Publications (2)

Publication Number Publication Date
CN108734218A true CN108734218A (en) 2018-11-02
CN108734218B CN108734218B (en) 2021-01-15

Family

ID=63934951

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810497192.XA Active CN108734218B (en) 2018-05-22 2018-05-22 Information fusion method and device of multi-sensor system

Country Status (1)

Country Link
CN (1) CN108734218B (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109848988A (en) * 2019-01-24 2019-06-07 深圳市普森斯科技有限公司 A kind of scan matching method and system based on the fusion of history multiframe point cloud information
CN109978055A (en) * 2019-03-26 2019-07-05 京东方科技集团股份有限公司 The information fusion method and system of multisensor syste, computer equipment and medium
CN111062359A (en) * 2019-12-27 2020-04-24 广东海洋大学深圳研究院 Two-stage Kalman filtering fusion method based on noise sequential decorrelation
CN112134544A (en) * 2020-09-25 2020-12-25 郑州佛光发电设备有限公司 Filtering method, device, equipment and computer readable storage medium

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101630445A (en) * 2009-06-10 2010-01-20 民航数据通信有限责任公司 Method for processing data of asynchronous multiple sensors based on noise error and system thereof
CN101696906A (en) * 2009-10-28 2010-04-21 上海电机学院 Multi-source information fusion based state monitoring system and method of wind-driven generator
US20130246006A1 (en) * 2012-03-13 2013-09-19 King Fahd University Of Petroleum And Minerals Method for kalman filter state estimation in bilinear systems
CN103575278A (en) * 2013-10-08 2014-02-12 北京理工大学 Robust filtering method aiming at parameter uncertainty and observation loss
CN103714045A (en) * 2014-01-09 2014-04-09 北京理工大学 Information fusion estimation method for asynchronous multi-rate non-uniform sampled observation data
CN104297557A (en) * 2014-10-08 2015-01-21 北京航空航天大学 United navigation autonomous integrity monitoring method applicable to free flight of plurality of aircraft

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101630445A (en) * 2009-06-10 2010-01-20 民航数据通信有限责任公司 Method for processing data of asynchronous multiple sensors based on noise error and system thereof
CN101696906A (en) * 2009-10-28 2010-04-21 上海电机学院 Multi-source information fusion based state monitoring system and method of wind-driven generator
US20130246006A1 (en) * 2012-03-13 2013-09-19 King Fahd University Of Petroleum And Minerals Method for kalman filter state estimation in bilinear systems
CN103575278A (en) * 2013-10-08 2014-02-12 北京理工大学 Robust filtering method aiming at parameter uncertainty and observation loss
CN103714045A (en) * 2014-01-09 2014-04-09 北京理工大学 Information fusion estimation method for asynchronous multi-rate non-uniform sampled observation data
CN104297557A (en) * 2014-10-08 2015-01-21 北京航空航天大学 United navigation autonomous integrity monitoring method applicable to free flight of plurality of aircraft

Non-Patent Citations (7)

* Cited by examiner, † Cited by third party
Title
吴昊 等: "基于广义M估计的鲁棒容积卡尔曼滤波目标跟踪算法", 《物理学报》 *
宁涛 等: "一类噪声相关多传感器系统的新型序贯式融合滤波", 《大连理工大学学报》 *
徐树生: "船舶动力定位系统多传感器信息融合方法研究", 《万方数据知识服务平台》 *
敖志刚 等: "多传感器自适应容积卡尔曼滤波融合算法", 《计算机应用研究》 *
李刚 等: "抗差与自适应组合的卡尔曼滤波算法在动态导航中的研究", 《重庆邮电大学学报（自然科学版）》 *
甘林海 等: "基于异步多速率观测的噪声去相关融合算法", 《传感器与微系统》 *
葛泉波 等: "基于状态转换的顺序式异步数据融合算法", 《系统工程与电子技术》 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109848988A (en) * 2019-01-24 2019-06-07 深圳市普森斯科技有限公司 A kind of scan matching method and system based on the fusion of history multiframe point cloud information
CN109978055A (en) * 2019-03-26 2019-07-05 京东方科技集团股份有限公司 The information fusion method and system of multisensor syste, computer equipment and medium
CN111062359A (en) * 2019-12-27 2020-04-24 广东海洋大学深圳研究院 Two-stage Kalman filtering fusion method based on noise sequential decorrelation
CN111062359B (en) * 2019-12-27 2023-03-31 广东海洋大学深圳研究院 Two-stage Kalman filtering fusion method based on noise sequential decorrelation
CN112134544A (en) * 2020-09-25 2020-12-25 郑州佛光发电设备有限公司 Filtering method, device, equipment and computer readable storage medium
CN112134544B (en) * 2020-09-25 2024-03-19 郑州佛光发电设备股份有限公司 Filtering method, device, equipment and computer readable storage medium

Also Published As

Publication number Publication date
CN108734218B (en) 2021-01-15

Similar Documents

Publication Publication Date Title
CN108734218A (en) A kind of information fusion method and device of multisensor syste
CN107948166A (en) Traffic anomaly detection method and device based on deep learning
CN107248039B (en) Real-time quality control method and device based on medical specimen detection project result
CN108900546A (en) The method and apparatus of time series Network anomaly detection based on LSTM
TWI660277B (en) Information processing device and information processing method
CN110084292A (en) Object detection method based on DenseNet and multi-scale feature fusion
CN110263652A (en) Laser point cloud data recognition methods and device
CN109587001A (en) A kind of performance indicator method for detecting abnormality and device
CN107958230A (en) Facial expression recognizing method and device
CN104679818A (en) Video keyframe extracting method and video keyframe extracting system
CN107368820B (en) Refined gesture recognition method, device and equipment
CN106909886A (en) A kind of high accuracy method for traffic sign detection and system based on deep learning
CN108474867A (en) High resolution precipitation amount data recovery system and its method
CN110046212B (en) Road condition change information determining method and device, computer equipment and storage medium
CN110111885B (en) Attribute prediction method, attribute prediction device, computer equipment and computer readable storage medium
CN103218533A (en) Estimation method of series system reliability degree lower confidence limit of exponential distribution
CN109740609A (en) A kind of gauge detection method and device
CN109858526A (en) Sensor-based multi-target track fusion method in a kind of target following
CN106289243A (en) A kind of magnetometer automatic calibrating method and system
CN107564125A (en) Drainage pipeline networks inspection treating method and apparatus
CN109918718A (en) Unmanned plane during flying method for evaluating quality, device, storage medium and electronic equipment
CN112313702B (en) Display control device, display control method, and display control program
CN112699950B (en) Medical image classification method, image classification network processing method, device and equipment
CN104992428A (en) Image falling rook volume determination method based on K mean value cluster analysis
CN106294485B (en) Determine the method and device in significant place

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant