CN109991597A - Weak-expansion-target-oriented tracking-before-detection method - Google Patents
Weak-expansion-target-oriented tracking-before-detection method Download PDFInfo
- Publication number
- CN109991597A CN109991597A CN201910273270.2A CN201910273270A CN109991597A CN 109991597 A CN109991597 A CN 109991597A CN 201910273270 A CN201910273270 A CN 201910273270A CN 109991597 A CN109991597 A CN 109991597A
- Authority
- CN
- China
- Prior art keywords
- target
- particle
- particles
- state
- probability
- 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.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 25
- 239000002245 particle Substances 0.000 claims abstract description 141
- 238000005259 measurement Methods 0.000 claims abstract description 63
- 238000000034 method Methods 0.000 claims abstract description 33
- 238000001914 filtration Methods 0.000 claims abstract description 16
- 230000008569 process Effects 0.000 claims abstract description 7
- 238000012544 monitoring process Methods 0.000 claims abstract description 5
- 230000004083 survival effect Effects 0.000 claims description 53
- 238000009826 distribution Methods 0.000 claims description 12
- 238000012937 correction Methods 0.000 claims description 11
- 238000012952 Resampling Methods 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 6
- 238000005315 distribution function Methods 0.000 claims description 5
- 238000012546 transfer Methods 0.000 claims description 3
- 238000009827 uniform distribution Methods 0.000 claims description 3
- 238000009825 accumulation Methods 0.000 abstract description 6
- 238000003672 processing method Methods 0.000 description 3
- 230000009897 systematic effect Effects 0.000 description 3
- 238000007476 Maximum Likelihood Methods 0.000 description 1
- 240000007594 Oryza sativa Species 0.000 description 1
- 235000007164 Oryza sativa Nutrition 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 235000009566 rice Nutrition 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/66—Radar-tracking systems; Analogous systems
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
Landscapes
- Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention belongs to the technical field of target detection and tracking, and particularly discloses a tracking-before-detection method for a weakly extended target. In the invention, under a Bayes filtering framework, a particle approximation means is used for recursively predicting and updating the target state, thereby realizing the detection and tracking of the weakly expanded target, wherein in the updating step, the intensity measurement data accumulation of the resolution units occupied by the target is completed through the particles, and when the accumulation length reaches a set value, the power of each resolution unit occupied by the target is estimated, and the power is used as a parameter of a likelihood function to update the particle weight. The method can realize the detection and tracking of the weakly extended target without target prior information, avoids the complicated nonlinear parameter estimation process of the traditional method, has no specific requirement on the type of the intensity measurement data, can be widely applied to various civil and military systems such as a video monitoring system, a robot navigation system, a military target detection and tracking system and the like, and has wide market prospect and application value.
Description
Technical Field
The invention relates to the technical field of target detection and tracking, in particular to a tracking-before-detection method for a weakly extended target.
Background
Under the background that the demands of civil monitoring and military reconnaissance are deepening, the sensor is required to have the capability of quickly detecting and stably tracking a weak extended target. Under the condition of low signal-to-noise ratio, the track-before-detect (TBD) technology can more effectively realize the joint detection and tracking of the target compared with the traditional track-after-detect (TAD) technology. The TBD technology was first applied to target detection of infrared image sequences by Maybeck in 1983, and thereafter widely applied to the field of radar detection. The basic idea of the TBD technology is that before detection judgment, the signal-to-noise ratio of target detection is improved through the energy accumulation of echo signals on a target motion track, and the combined detection and tracking of a weak target are realized. The TBD technology directly applies the intensity measurement data of each resolution unit output by the sensor, so that the loss of the echo information caused by threshold processing is avoided.
The existing implementation methods for TBD technology can be roughly divided into two categories: one is a batch processing method represented by methods such as three-dimensional matched filtering, dynamic programming, Harvard transformation, multi-order hypothesis testing and the like, and real-time performance is difficult to ensure; the other type is a recursive processing method represented by recursive Bayesian filtering, wherein the success of particle filtering approximation in nonlinear application undoubtedly improves the application value of the method. For the study of single target TBD technology, most of the efforts are still based on point target assumptions. With the widespread use of high resolution sensors, the point target assumption has been difficult to meet engineering application requirements. Aiming at the weak extension target, the existing research still takes TAD technical thought as the main point, and TBD technical achievement for realizing joint detection and tracking based on intensity measurement data is not much. The invention focuses on a tracking-before-detection method for a weakly extended target.
The existing few weakly-extended target TBD algorithms are updated under a Bayesian filtering framework, and target power parameters are assumed to be known a priori, or signal-to-noise ratios of measurement sets are known. This type of assumption is often difficult to satisfy since the detection target tends to be non-cooperative. Therefore, the scholars introduce the idea of parameter estimation to estimate the target and noise power by the method of Maximum Likelihood Estimation (MLE), but the estimation algorithm has at least the following two problems: firstly, under the condition that the target intensity obeys Rice distribution, a group of height nonlinear equations need to be solved for parameter estimation; secondly, under the condition that the target intensity obeys exponential distribution, the estimation algorithm completely fails.
Disclosure of Invention
The invention aims to solve the technical problem of weak extension target joint detection and tracking under the condition that target prior information is unknown. In order to overcome the defects and shortcomings of the existing method, the invention provides a tracking method before detection for a weakly extended target.
The invention is realized by the following technical scheme:
a tracking-before-detection method for a weak extension target comprises the following steps:
step 1, filter initialization: the method comprises the following steps of determining a model and parameters of a system, and setting prior information of the system, wherein the method is based on the following steps:
(1.1) defining a target dynamic model xk=f(xk-1)+wkAnd a sensor intensity measurement model zk=h(xk)+vkWherein x iskIs the state vector of the target at time k, f (-) is the target state transfer function, wkIs subject to pwDistributed process noise, assumed to be independently and identically distributed; z is a radical ofkIs the set of intensity measurements of the sensor at time k, h (-) describes the target state vector xkAnd intensity measurement zkRelation between vkIs subject to pvDistributed and independent of wkThe measurement noise of (2) is also assumed to be independently and identically distributed.
And (1.2) setting system prior information. Since the filter is recursively processing the data, each iteration requires the particle state and the survival probability at the previous time. Therefore, the probability of survival of the target at the initial moment and the distribution function for generating the new particle state must both be set a priori. The setting of the prior information comprises the following three aspects:
a. setting survival probability of target at initial moment as p0|0;
b. The existence m of the targetkModeling as a two-state Markov chain and setting pbAnd psIs a constant number, where pb=prob{mk=1|mk-10 represents the probability that the target does not exist at time k-1 and "grows" at time k, ps=prob{mk=1|mk-11 represents the probability that the target is present at time k-1 and "survives" at time k;
c. setting two distribution functions for generating new particle states, i.e. corresponding to pbIs suggested probability distribution bk|k-1(x) And corresponds to psIs suggested probability distributionWherein the target prior information is unknown, bk|k-1(x) A uniform distribution can be assumed;derived from the target dynamic model as pw(xk-f(xk-1))。
(1.3) determining the particle number N + B of the particle filter and the maximum occupied resolution unit number L of the targetmaxThe filter initialization is completed, where N represents the number of "surviving" particles at each time and B represents the number of "new" particles at each time.
Step 2, predicting the target survival probability and the particle state: since the filter is recursive in time, the estimated target survival probability p at time k-1 is usedk-1|k-1And particle stateAs it is passed to the time point k,representing the motion state of the nth particle at time k-1. From the particle filter it can be determined:
(2.1) the survival probability of the target at the k moment is predicted to be as follows according to the target 'newborn' probability, 'survival' probability and the survival probability at the k-1 moment set in the step 1:
pk|k-1=pb(1-pk-1|k-1)+pspk-1|k-1
wherein p isk|k-1Is the predicted value of the survival probability of the target at the moment k.
(2.2) predicting the states of the N + B particles at the k time by the two proposed probability distributions set in step 1Comprises the following steps:
wherein,representing the predicted state of the nth particle by a "pseudo-weight" proportional to the weight of the particle "The prediction weight of the nth particle is represented.
From this, a set of particle prediction states is determined, including "new" particles and "surviving" particlesAnd the survival probability of the target at the time kMeasured value pk|k-1。
Step 3, inputting the intensity measurement data generated by the sensor into a filter: and (3) under a Bayesian filtering framework, after the prediction operation mentioned in the step (2) is completed, the filter corrects and updates the state of the particle set by acquiring the measurement data of the sensor at the moment k. The intensity measurement data of each resolution unit provided by the sensor at the moment k is used as a measurement set to be input into a filter, and the value of each unit of the measurement set corresponds to the echo power of the corresponding resolution unit in the monitoring area and is recorded as the echo powerWhere I is the total number of sensor resolution cells,representing the echo power of the i-th resolution element at time k.
Step 4, target power parameter estimation: for intensity measurement data, the resolution cell affected by the target contains both noise and power information of the target. Under the condition that the noise power is assumed to be known a priori, in order to obtain the parameters of the likelihood function, target power information needs to be estimated from the measurement data input in the step 3. Different from the traditional method, the method utilizes the characteristic that the particle state gradually converges to the target real state after multiple iterations to set the length W of the accumulated frame numberlAnd then, accumulating the intensity measured value of each time the particles occupy the resolution unit, and subtracting the noise power to obtain an expectation value as an estimated value of the target power parameter after the accumulated frame number reaches an accumulated length. At time k, since the intensity measurement information of all resolution cells is input in step 3, the respective particles accumulate the intensity measurements of the respective occupied resolution cells into the corresponding set of particle measurements. When the length of the measurement set of particles reaches WlThen, the power parameter P of each resolution unit occupied by the corresponding particlen(r) can be estimated according to the following formula:
where t represents the number of frames in which the measurement data is located, r is the r-th distance resolving unit occupied by the particle n,representing the echo power of the r-th range resolution element in the t-th frame of measurement data,representing the set of distance-resolving elements, σ, occupied by the particle n2Is to measure the noise power, LmaxFor the maximum number of occupied resolution cells of the target, i.e.In order to ensure that the estimated value has definite physical meaning, namely that the power value is constantly greater than zero, the method adds the operation of taking the maximum value between the estimated value and zero.
Step 5, correcting the target survival probability and the particle state: and in a Bayes filtering framework, correcting and updating the target survival probability and the particle state on the basis of the measurement data input into the filter in the step 3 according to the target survival probability predicted value and the particle prediction state obtained in the step 2. First, for each particle state in measurement set zkCalculates corresponding likelihood ratios on all the resolution cellsMultiplying the ratio by the corresponding particle weight to serve as the weight after particle correction, wherein the target power parameter is given by the step 4 when the likelihood ratio is calculated; secondly, the weights of all particles after the joint correction and the target predicted survival probability p obtained in step 2.1 are combinedklk-1For the target survival probability p at time kk|kUpdating is carried out; finally, the weights of the particles are normalized and the state of the target k moment is estimatedThe specific process is as follows:
(5.1) calculating likelihood ratios of the respective particles:
wherein, g1(. is a likelihood function of the presence of the target, g0(. is a likelihood function for noise only;for the state of beingThe likelihood ratio of the particles over the entire measurement set, Pn(r) particles corresponding to the particles estimated in step 4The target power of (1).
(5.2) updating the particle weights by using the likelihood ratios obtained in step 5.1 and the "pseudo weights" correction in step 2:
at the same time, order
(5.3) correction of the probability of target survival: updating the target survival probability with the predicted survival probability of step 2.1 and the corrected weights of step 5.2:
pk|k=s/(1-pk|k-1+s)
wherein,
(5.4) normalizing the weights and estimating the target state:
wherein,representing the normalized particle weight, and estimating the target state according to the following formula under the minimum mean square error criterion:
thus, the target survival probability p corrected at the moment k can be obtainedk|kAnd the state of all particlesObtaining the estimated state of the target k moment
Step 6, resampling particles: in order to avoid the problem that particle exhaustion is generated after multiple iterations of particle filtering, a systematic resampling step is introduced. And resampling the corrected particles according to the weight value to obtain N particles, and transferring the particles to the next iterative filtering to finish the recursive processing of the filter.
And 7, repeating the iteration steps 2 to 6 along with the continuous extension of the time k.
Compared with the prior art, the invention has the following technical effects:
firstly, the method can exert the advantages of joint detection and tracking by utilizing the TBD technology, and filtering is carried out on the strength measurement data of each resolution unit output by the sensor, so that the weak extension target is quickly detected and stably tracked.
Secondly, the method approximately estimates the target power parameter through the accumulation of the particles to multi-frame measurement, thereby avoiding the problem of difficult solution of a nonlinear equation set faced by the MLE estimation method; meanwhile, the estimation algorithm has no specific requirements on the distribution type met by the measurement data, and has universality.
Drawings
FIG. 1 is a general block diagram of the processing method of the present invention.
Detailed Description
In order to better explain the technical solution of the present invention, the following describes the embodiments of the present invention with reference to the accompanying drawings. Fig. 1 is a tracking method before detection for a weakly extended target according to the present invention, wherein the method is a particle filtering approximation implementation scheme, and includes the following steps:
step 1, filter initialization: the initialization step involves determining the system model and parameters, and setting a priori information for the system.
(1.1) defining a target dynamic model x by the kinematic characteristics (such as uniform motion or uniform acceleration motion) of the target and the data acquisition mode of the sensork=f(xk-1)+wkAnd a sensor intensity measurement model zk=h(xk)+vkWherein x iskIs the state vector of the target at time k, f (-) is the target state transfer function, wkIs subject to pwDistributed process noise, assumed to be independently and identically distributed; z is a radical ofkIs the set of intensity measurements of the sensor at time k, h (-) describes the target state xkAnd intensity measurement zkRelation between vkIs subject to pvDistributed and independent of wkThe measurement noise of (2) is also assumed to be independently and identically distributed.
And (1.2) setting system prior information. Since the filter is recursively processing the data, each iteration requires the particle state and the survival probability at the previous time. Therefore, the probability of survival of the target at the initial moment and the distribution function for generating the new particle state must both be set a priori. The setting of the prior information is divided into the following three aspects:
a. setting survival probability of target at initial moment as p0|0Usually, the probability takes a smaller value, preferably 0.01;
b. the existence m of the targetkModeling as a two-state Markov chain and setting pbAnd psIs a constant number, where pb=prob{mk=1|mk-10 represents the probability that the target does not exist at time k-1 and "grows" at time k, ps=prob{mk=1|mk-11 represents the probability that the target is present at time k-1 and "survives" at time k. Since the probability of the target "surviving" is much greater than the probability of the target "newborn", pbUsually smaller in value and psThe value is large, and preferably 0.01 and 0.99 respectively;
c. setting two distribution functions for generating new particle states, i.e. corresponding to pbIs suggested probability distribution bk|k-1(x) And corresponds to psIs suggested probability distributionWherein the target prior information is unknown, bk|k-1(x) A uniform distribution can be assumed;derived from the target dynamic model as pw(xk-f(xk-1))。
(1.3) determining the number of particles N + B of the particle filter and the maximum occupancy of the targetNumber of resolution cells LmaxThe filter initialization is completed, where N represents the number of "surviving" particles at each time and B represents the number of "new" particles at each time.
Step 2, predicting the target survival probability and the particle state: since the filter is recursive in time, the estimated target survival probability p at time k-1 is usedk-1|k-1And particle stateIs transmitted to the moment k, whereinRepresenting the motion state of the nth particle. From the particle filter it can be determined:
(2.1) the survival probability of the target at the k moment is predicted to be as follows according to the target 'newborn' probability, 'survival' probability and the survival probability at the k-1 moment set in the step 1:
pk|k-1=pb(1-pk-1|k-1)+pspk-1|k-1
wherein p isk|k-1Is a predicted value of survival probability at the moment k.
(2.2) predicting the states of the N + B particles at the k time by the two proposed probability distributions set in step 1Comprises the following steps:
wherein,represents the prediction weight of the nth particle,representing a predicted state of the nth particle; using "pseudo-weights" in the calculation "To simplify and reduce the computational complexity, i.e.:
from this, a set of particle prediction states comprising "new" particles and "surviving" particles can be derivedAnd a predicted value p of the survival probability of the target at the moment kk|k-1。
Step 3, inputting the intensity measurement data generated by the sensor into a filter: under a Bayesian filtering framework, after the prediction operation mentioned in the step 2 is completed, the filter needs to acquire measurement data of the sensor at the time k to correct and update the state of the particle set. The intensity measurement data of each resolution unit provided by the sensor at the moment k is used as a measurement set to be input into a filter, and the value of each unit of the measurement set corresponds to the echo power of the corresponding resolution unit in the monitoring area and is recorded as the echo powerWhere I is the total number of sensor resolution cells,representing the echo power of the i-th resolution element at time k.
Step 4, target power parameter estimation: for intensity measurement data, the resolution cell affected by the target contains both noise and power information of the target.Under the condition that the noise power is assumed to be known a priori, in order to obtain the parameters of the likelihood function, target power information needs to be estimated from the measurement data input in the step 3. Different from the traditional method, the method utilizes the characteristic that the particle state gradually converges to the target real state after multiple iterations to set the length W of the accumulated frame numberlAnd then, measuring the intensity of the accumulated particles occupying the resolution unit every time, and subtracting the noise power to obtain an expectation value as an estimated value of the target power parameter after the accumulated frame number reaches an accumulated length. At time k, since the intensity measurement information of all resolution cells is input in step 3, each particle accumulates the intensity measurements each occupying a resolution cell into a corresponding particle measurement accumulation set. When the length of the measurement accumulation set of particles reaches WlThen, the target power parameter P of each resolution unit occupied by the corresponding particlen(r) can be estimated by:
wherein r is the r-th distance resolution cell occupied by the particle n,representing the echo power of the r-th range resolution element in the t-th frame of measurement data,representing the set of distance-resolving elements, σ, occupied by the particle n2Is to measure the noise power, LmaxFor the maximum number of occupied resolution cells of the target, i.e.In order to ensure that the estimated value has definite physical significance, namely the power value is constantly larger than zero, the method adds the operation of taking the maximum value between the estimated value and zero.
Step 5, correcting the target survival probability and the particle state: in a Bayesian filtering framework, the target survival probability obtained according to the step 2And (3) correcting and updating the target survival probability and the particle state on the basis of the measurement data input into the filter in the step 3. First, for each particle state in measurement set zkCalculates corresponding likelihood ratios on all the resolution cellsMultiplying the ratio by the corresponding particle weight to serve as the weight after particle correction, wherein the target power parameter is given by the step 4 when the likelihood ratio is calculated; secondly, the weights of all particles after the joint correction and the target predicted survival probability p obtained in step 2.1 are combinedk|k-1For the target survival probability p at time kk|kUpdating is carried out; finally, the weights of the particles are normalized and the state of the target k moment is estimatedThe specific process is as follows:
(5.1) calculating likelihood ratios of the respective particles:
wherein, g1(. is a likelihood function of the presence of the target, g0(. is a likelihood function for noise only;for the state of beingThe likelihood ratio of the particles over the entire measurement set, Pn(r) particles corresponding to the particles estimated in step 4The target power of (1).
(5.2) correcting the particle weight using the likelihood ratio obtained in step 5.1 and the "pseudo weight" in step 2:
at the same time, order
(5.3) correction of the probability of target survival: updating the target survival probability with the predicted survival probability of step 2.1 and the corrected weights of step 5.2:
pk|k=s/(1-pk|k-1+s)
wherein,
(5.4) normalizing the weights and estimating the target state:
wherein,representing the normalized particle weight, and estimating the target state according to the following formula under the minimum mean square error criterion:
thus, the target survival probability p corrected at the moment k can be obtainedk|kAnd the state of all particlesObtaining the estimated state of the target k moment
Step 6, resampling particles: in order to avoid the problem that particle exhaustion is generated after multiple iterations of particle filtering, a systematic resampling step is introduced. And (3) resampling the corrected particles according to the weight value by using a systematic resampling method to obtain N particles, and transmitting the particles to iterative filtering at the next moment to finish the recursive processing of the filter.
And 7, repeating the iteration steps 2 to 6 along with the continuous extension of the time k.
The above is only one embodiment of the present invention, and the protection scope of the present invention is not limited to the above examples, and all technical solutions belonging to the idea of the present invention belong to the protection scope of the present invention. It should be noted that various modifications and adaptations to those skilled in the art without departing from the principles of the present invention should be considered as within the scope of the present invention.
Claims (2)
1. A tracking method before detection for a weak extension target is characterized by comprising the following steps:
step 1, filter initialization: the method comprises the following steps of determining a model and parameters of a system, and setting prior information of the system, wherein the method is based on the following steps:
(1.1) defining a target dynamic model xk=f(xk-1)+wkAnd a sensor intensity measurement model zk=h(xk)+vkWherein x iskIs the state vector of the target at time k, f (-) is the target state transfer function, wkIs subject to pwDistributed process noise, assumed to be independently and identically distributed; z is a radical ofkIs the set of intensity measurements of the sensor at time k, h (-) describes the target state vector xkAnd intensity measurement zkRelation between vkIs subject to pvDistributed and independent of wkThe measurement noise of (2) is also assumed to be independently and identically distributed;
(1.2) setting system prior information, and dividing the system prior information into the following three aspects:
a. setting survival probability of target at initial moment as p0|0;
b. The existence m of the targetkModeling as a two-state Markov chain and setting pbAnd psIs a constant number, where pb=prob{mk=1|mk-10 represents the probability that the target does not exist at time k-1 and "grows" at time k, ps=prob{mk=1|mk-11 represents the probability that the target is present at time k-1 and "survives" at time k;
c. setting two distribution functions for generating new particle states, i.e. corresponding to pbIs suggested probability distribution bk|k-1(x) And corresponds to psIs suggested probability distributionWherein the target prior information is unknown, bk|k-1(x) A uniform distribution can be assumed;derived from the target dynamic model as pw(xk-f(xk-1));
(1.3) determining the particle number N + B of the particle filter and the maximum occupied resolution unit number L of the targetmaxCompleting the initialization of the filter, wherein N represents the number of 'surviving' particles at each moment, and B represents the number of 'newborn' particles at each moment;
step 2, predicting the target survival probability and the particle state: target survival probability p obtained by estimating k-1 timek-1|k-1And particle stateAs it is passed to the time point k,representing the motion state of the nth particle at the k-1 moment; from the particle filter it can be determined:
(2.1) the survival probability of the target at the k moment is predicted to be as follows according to the target 'newborn' probability, 'survival' probability and the survival probability at the k-1 moment set in the step 1:
pk|k-1=pb(1-pk-1|k-1)+pspk-1|k-1
wherein p isk|k-1The predicted value of the target survival probability at the moment k is obtained;
(2.2) predicting the states of the N + B particles at the k time by the two proposed probability distributions set in step 1Comprises the following steps:
wherein,representing the predicted state of the nth particle by a "pseudo-weight" proportional to the weight of the particle "Represents a prediction weight of the nth particle;
from this, a set of particle prediction states is determined, including "new" particles and "surviving" particlesAnd a predicted value p of the survival probability of the target at the moment kk|k-1;
Step 3, inputting the intensity measurement data generated by the sensor into a filter: the intensity measurement data of each resolution unit provided by the sensor at the moment k is used as a measurement set to be input into a filter, and the value of each unit of the measurement set corresponds to the echo power of the corresponding resolution unit in the monitoring area and is recorded as the echo powerWhere I is the total number of sensor resolution cells,representing the echo power of the ith resolution cell at the moment k;
step 4, target power parameter estimation: setting the length W of the accumulated frame number by using the characteristic that the particle state gradually converges to the target real state after multiple iterationslThen, accumulating the intensity measured value of each time the particles occupy the resolution unit, and subtracting the noise power to obtain an expectation value as an estimated value of a target power parameter after the accumulated frame number reaches an accumulated length; at the moment k, because the intensity measurement information of all the resolution cells is input in the step 3, the intensity measurement of each particle occupying the resolution cell is accumulated into a corresponding particle measurement set; when the length of the measurement set of particles reaches WlThen, the power parameter P of each resolution unit occupied by the corresponding particlen(r) can be estimated according to the following formula:
where t represents the number of frames in which the measurement data is located, r is the r-th distance resolving unit occupied by the particle n,representing the echo power of the r-th range resolution element in the t-th frame of measurement data,representing the set of distance-resolving elements, σ, occupied by the particle n2Is to measure the noise power, LmaxFor the maximum number of occupied resolution cells of the target, i.e.The maximum value operation between the estimated value and zero is adopted to ensure that the power value is constantly larger than zero;
step 5, correcting the target survival probability and the particle state: first, for each particle state in measurement set zkCalculates corresponding likelihood ratios on all the resolution cellsMultiplying the ratio by the corresponding particle weight to serve as the weight after particle correction, wherein the target power parameter is given by the step 4 when the likelihood ratio is calculated; secondly, the weights of all particles after the joint correction and the target predicted survival probability p obtained in step 2.1 are combinedk|k-1For the target survival probability p at time kk|kUpdating is carried out; finally, the weights of the particles are normalized and the state of the target k moment is estimatedThe specific process is as follows:
(5.1) calculating likelihood ratios of the respective particles:
wherein, g1(. is a likelihood function of the presence of the target, g0(. is a likelihood function for noise only;for the state of beingThe likelihood ratio of the particles over the entire measurement set, Pn(r) particles corresponding to the particles estimated in step 4The target power of (1);
(5.2) updating the particle weights by using the likelihood ratios obtained in step 5.1 and the "pseudo weights" correction in step 2:
at the same time, order
(5.3) correction of the probability of target survival: updating the target survival probability with the predicted survival probability of step 2.1 and the corrected weights of step 5.2:
pk|k=s/(1-pk|k-1+s)
wherein,
(5.4) normalizing the weights and estimating the target state:
wherein,representing the normalized particle weight, and estimating the target state according to the following formula under the minimum mean square error criterion:
thus, the target survival probability p corrected at the moment k can be obtainedk|kAnd the state of all particlesObtaining the estimated state of the target k moment
Step 6, resampling particles: resampling the corrected particles according to the weight value to obtain N particles, and transferring the particles to the next iterative filtering to complete the recursive processing of the filter;
and 7, repeating the iteration steps 2 to 6 along with the continuous extension of the time k.
2. The method for tracking weak extended target before detection as claimed in claim 1, wherein p in step (1.2)0|0Preferably, the value of (A) is 0.01, pbAnd psPreferably values of 0.01 and 0.99.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910273270.2A CN109991597A (en) | 2019-04-04 | 2019-04-04 | Weak-expansion-target-oriented tracking-before-detection method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910273270.2A CN109991597A (en) | 2019-04-04 | 2019-04-04 | Weak-expansion-target-oriented tracking-before-detection method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109991597A true CN109991597A (en) | 2019-07-09 |
Family
ID=67130933
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910273270.2A Pending CN109991597A (en) | 2019-04-04 | 2019-04-04 | Weak-expansion-target-oriented tracking-before-detection method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109991597A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111580084A (en) * | 2020-05-13 | 2020-08-25 | 中国人民解放军国防科技大学 | Multi-distance extended target-oriented multi-Bernoulli pre-detection tracking method |
CN113408422A (en) * | 2021-06-21 | 2021-09-17 | 电子科技大学 | Multi-frame joint detection tracking and classification method suitable for weak targets |
WO2022116375A1 (en) * | 2020-12-01 | 2022-06-09 | 中国人民解放军海军航空大学 | Method for performing tracking-before-detecting on multiple weak targets by high-resolution sensor |
CN115097437A (en) * | 2022-06-06 | 2022-09-23 | 哈尔滨工程大学 | Underwater target tracking track approaching and crossing solution method based on label multi-Bernoulli pre-detection tracking algorithm |
CN115407299A (en) * | 2022-09-15 | 2022-11-29 | 中国人民解放军国防科技大学 | Weak and small target detection method and device based on Bernoulli filter algorithm |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2014106133A (en) * | 2012-11-28 | 2014-06-09 | Toshiba Corp | Target detection device, and target detection method |
CN106569193A (en) * | 2016-11-01 | 2017-04-19 | 西安电子科技大学 | Sea-surface small target detection method based on front-back revenue reference particle filter |
CN107340514A (en) * | 2017-07-10 | 2017-11-10 | 中国人民解放军海军航空工程学院 | Hypersonic weak signal target RAE HT TBD integration detection methods in three dimensions |
CN108919225A (en) * | 2018-07-26 | 2018-11-30 | 中国人民解放军海军航空大学 | Range extension target Multichannel fusion detection method under the uniform environment in part |
CN109471081A (en) * | 2018-11-07 | 2019-03-15 | 中国人民解放军国防科技大学 | Single pulse radar weak and small target combined detection and state estimation method |
-
2019
- 2019-04-04 CN CN201910273270.2A patent/CN109991597A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2014106133A (en) * | 2012-11-28 | 2014-06-09 | Toshiba Corp | Target detection device, and target detection method |
CN106569193A (en) * | 2016-11-01 | 2017-04-19 | 西安电子科技大学 | Sea-surface small target detection method based on front-back revenue reference particle filter |
CN107340514A (en) * | 2017-07-10 | 2017-11-10 | 中国人民解放军海军航空工程学院 | Hypersonic weak signal target RAE HT TBD integration detection methods in three dimensions |
CN108919225A (en) * | 2018-07-26 | 2018-11-30 | 中国人民解放军海军航空大学 | Range extension target Multichannel fusion detection method under the uniform environment in part |
CN109471081A (en) * | 2018-11-07 | 2019-03-15 | 中国人民解放军国防科技大学 | Single pulse radar weak and small target combined detection and state estimation method |
Non-Patent Citations (1)
Title |
---|
蔡飞: ""雷达弱小目标检测与跟踪技术研究"", 《中国博士学位论文全文数据库信息科技辑》 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111580084A (en) * | 2020-05-13 | 2020-08-25 | 中国人民解放军国防科技大学 | Multi-distance extended target-oriented multi-Bernoulli pre-detection tracking method |
WO2022116375A1 (en) * | 2020-12-01 | 2022-06-09 | 中国人民解放军海军航空大学 | Method for performing tracking-before-detecting on multiple weak targets by high-resolution sensor |
CN113408422A (en) * | 2021-06-21 | 2021-09-17 | 电子科技大学 | Multi-frame joint detection tracking and classification method suitable for weak targets |
CN115097437A (en) * | 2022-06-06 | 2022-09-23 | 哈尔滨工程大学 | Underwater target tracking track approaching and crossing solution method based on label multi-Bernoulli pre-detection tracking algorithm |
CN115407299A (en) * | 2022-09-15 | 2022-11-29 | 中国人民解放军国防科技大学 | Weak and small target detection method and device based on Bernoulli filter algorithm |
CN115407299B (en) * | 2022-09-15 | 2024-06-18 | 中国人民解放军国防科技大学 | Weak and small target detection method and device based on Bernoulli filter algorithm |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109991597A (en) | Weak-expansion-target-oriented tracking-before-detection method | |
CN108255791B (en) | Maneuvering target tracking method based on distributed sensor consistency | |
CN107703496B (en) | Interactive multimode Bernoulli filtering maneuvering weak target tracking-before-detection method | |
CN110689576A (en) | Automatic ware-based dynamic 3D point cloud normal distribution AGV positioning method | |
CN106526585B (en) | Tracking before target detection based on the filtering of Gaussian particle gesture probability hypothesis density | |
CN103115631B (en) | System and method for correcting imaging parameter error of remote sensing camera | |
CN111127523B (en) | Multi-sensor GMPHD self-adaptive fusion method based on measurement iteration update | |
CN110187335B (en) | Particle filter tracking-before-detection method for target with discontinuous characteristic | |
CN101719278B (en) | Automatic tracking method for video microimage cells based on KHM algorithm | |
CN110138698B (en) | Initial phase joint estimation method and device for frequency offset of high-order modulation linear mixed signal | |
CN116500575B (en) | Extended target tracking method and device based on variable decibel leaf theory | |
CN113466890A (en) | Lightweight laser radar inertial combination positioning method and system based on key feature extraction | |
CN112346010B (en) | Dual-computer passive positioning method based on scale difference and time difference | |
CN108333571B (en) | Multi-sensor multi-frame joint detection tracking method based on trace point sequence fusion | |
CN111145115A (en) | Non-uniform correction method based on scene adaptation and hardware implementation device thereof | |
CN107247257B (en) | Distributed multi-sensor track-before-detect method based on likelihood function approximation | |
CN114296087B (en) | Online Bayesian compression underwater imaging method, system, equipment and medium | |
CN104331087B (en) | Robust underwater sensor network target tracking method | |
CN114626307A (en) | Distributed consistent target state estimation method based on variational Bayes | |
CN117433538A (en) | Multi-source heterogeneous sensor track fusion method | |
CN108834043B (en) | Priori knowledge-based compressed sensing multi-target passive positioning method | |
CN111883265A (en) | Target state estimation method applied to fire control system | |
CN109188424B (en) | Distributed multi-sensor multi-target tracking method based on measurement consistency | |
CN108845299B (en) | Multi-sensor multi-frame joint detection algorithm based on posterior information fusion | |
CN110986946A (en) | Dynamic pose estimation method and device |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190709 |
|
WD01 | Invention patent application deemed withdrawn after publication |