CN109031229B - Probability hypothesis density method for target tracking in clutter environment - Google Patents
Probability hypothesis density method for target tracking in clutter environment Download PDFInfo
- Publication number
- CN109031229B CN109031229B CN201810692989.5A CN201810692989A CN109031229B CN 109031229 B CN109031229 B CN 109031229B CN 201810692989 A CN201810692989 A CN 201810692989A CN 109031229 B CN109031229 B CN 109031229B
- Authority
- CN
- China
- Prior art keywords
- clutter
- gaussian
- weight
- covariance matrix
- target
- 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.)
- Active
Links
Images
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
- 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
- G01S7/414—Discriminating targets with respect to background clutter
Abstract
The invention discloses a probability hypothesis density method for target tracking in a clutter environment, and belongs to the technical field of radars. Firstly, dividing the whole monitoring area into a plurality of clutter sub-areas by utilizing the difference of prior clutter densities according to the prior clutter map information of the monitoring area; in each tracking period, each clutter sub-region is respectively processed according to different parameters of different clutter rates; for each subarea, adopting GM-PHD assisted by echo amplitude to carry out filtering tracking; and an improved Gaussian component pruning scheme is adopted to improve the detection rate of the new target.
Description
Technical Field
The invention belongs to the technical field of radars, and particularly relates to a probability hypothesis density method for target tracking in a clutter environment.
Background
In recent years, target tracking technology based on random finite sets has gained wide attention. The target tracking theory based on the random finite set respectively models all state sets and observed values obtained by each observation into set value state and set value observation, thereby avoiding complex data association process and directly utilizing the optimal Bayes technology to realize target tracking under clutter background, association and uncertain detection. There are many possible theoretical approximation filters that can be used in engineering implementation, and the gaussian probability density hypothesis (GM-PHD) algorithm used herein is one of them. The standard GM-PHD algorithm still has some disadvantages. In practical situations, due to the complexity of environments such as terrain, the background of a monitored area is not uniform in space, so that the density of clutter in the monitored area is not uniform, and the GM-PHD algorithm assumes that the clutter is uniformly distributed in the whole monitored area and tracks the clutter, which inevitably causes instability of tracking performance. For the radar in the fixed monitoring area, on the premise that the clutter density is assumed to change only in space and is basically unchanged or slowly changed in time, the clutter distribution condition of the monitoring area is obtained and a prior clutter map is formed, and then the target tracking is assisted by utilizing the prior information of the clutter map. It is also feasible to use the echo amplitudes to assist target tracking, since the echo amplitudes from the target and the echo amplitudes from the clutter have different probability density distributions. In addition, in order to deal with the problem that the new target is not in the assumed target probability region, the strategy of cutting and combining the Gaussian components can be changed, and the initial probability of the new target is improved. Based on the above description, the invention provides a GM-PHD method combining a prior clutter map, a target echo amplitude and an improved pruning strategy.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a probability hypothesis density method for target tracking in a clutter environment.
The technical problem proposed by the invention is solved as follows:
a probability hypothesis density method for target tracking in clutter environment comprises the following steps:
A. establishing a priori clutter map
Modeling a clutter map of a monitoring area as a combination of a plurality of areas with different clutter densities, and constructing clutter prior information of the monitoring area on the assumption that the clutter density in each clutter area is uniformly distributed;
B. knowledge-aided tracking, for each simulation instant
According to the GM-PHD algorithm, the posterior intensity function v is determined from the state at the time k-1 k-1 (x) Calculating a state prediction strength function v at the moment k k|k-1 (x) (ii) a Predicting intensity function v from state at time k k|k-1 (x) Calculating the posterior intensity function v of the state at the moment k k (x) (ii) a On the basis of the calculation result of the standard GM-PHD algorithm, v is modified according to the echo amplitude and clutter distribution information k (x) The weight value of each Gaussian component;
step C: gaussian pruning and state extraction, to state posterior intensity function v k (x) Adopting a pruning strategy: deleting all Gaussian components with weight values smaller than a specified threshold, if all the Gaussian component weights derived from a certain measured value are smaller than the specified threshold, keeping the component with the maximum weight derived from the measured value, and modifying the weight of the component into the specified threshold; and normalizing the weight of the Gaussian components after the pruning strategy is adopted, calculating the merging distance of the Gaussian components, and merging the Gaussian components to obtain the final estimation result of the target state.
According to the method, the whole monitoring area is divided into a plurality of clutter sub-areas by utilizing the difference of prior clutter densities according to the prior clutter map information of the monitoring area; in each tracking period, each clutter sub-region is respectively processed according to different parameters of different clutter rates; and for each sub-area, adopting the GM-PHD assisted by the echo amplitude to carry out filtering tracking, and adopting an improved Gaussian component pruning scheme to improve the detection rate of the new target.
The invention has the beneficial effects that:
the algorithm of the invention utilizes priori knowledge to assist target tracking, utilizes priori clutter information of radar monitoring areas, and adopts different parameters to track and filter areas with different clutter densities; meanwhile, the amplitude information of the radar echo signal is utilized, and a filtering updating equation is modified according to different target echoes and clutter echo distributions; in addition, the pruning rule of the Gaussian mixture probability hypothesis density algorithm is modified, and the tracking starting probability of the target is improved.
Drawings
FIG. 1 is a prior knowledge assistance diagram;
fig. 2 is a flow chart of gaussian pruning and state extraction.
Detailed Description
The invention is further described below with reference to the figures and examples.
The present embodiment provides a method for probability hypothesis density of target tracking in a clutter environment, where an auxiliary schematic diagram of prior knowledge is shown in fig. 1, and the method includes the following steps:
step A, establishing a prior clutter map
Modeling a clutter map of a monitored region as a combination of regions of different clutter densities, assuming that the clutter density within each clutter region is subject to uniform distribution, the different clutter regions having different clutter densities, i.e. V = V 1 ∪V 2 ∪...∪V M Wherein V denotes the entire detection area, V 1 ~V M Representing different clutter areas divided by different clutter densities, wherein M is the total number of the clutter areas; recording clutter prior information for each clutter region including an average false alarm count for each clutter regionAnd false alarm location distribution density
Step B, knowledge-aided tracking
Step B-1, predicting:
the posterior intensity function of the target state vector x at time k-1 is assumed to be of the following Gaussian mixture:
wherein, J k-1 J =1, 2.. J. k-1 ,Andrespectively the weight, mean and covariance matrix of the jth Gaussian component at the time k-1,represents a mean value ofCovariance matrixA gaussian distribution of;
the state prediction strength function at the time k is as follows:
wherein p is S In order to target the probability of survival,andrespectively the mean and covariance matrices of the calculated jth predicted strength gaussian component at time k,the method comprises the following steps of (1) taking a motion prediction matrix of a target, Q a process noise covariance matrix, and superscript T as a transposition; j is a unit of γ,k For the number of new target gaussian components at time k, h =1,2 γ,k ; Andrespectively is the weight, the mean value and the covariance matrix of the h-th Gaussian component of the new target intensity function at the moment k,represents a mean value ofThe covariance matrix isA gaussian distribution of (d);
step B-2, updating:
predicting the state by a strength function v k|k-1 (x) The rewrite is:
wherein, J k|k-1 Predict the number of gaussian components of the intensity function for the state, l =1,2 k|k-1 ;Andrespectively the weight, mean and covariance matrix of the ith gaussian component,represents a mean value ofThe covariance matrix is(ii) a gaussian distribution of;
the gaussian posterior intensity formula updated at time k is:
wherein p is D Is a target detection probability, m z For efficient number of measurements, n =1,2 z ,z n Is a measure of the return of the radar,h is a target measurement matrix, R is a measurement noise covariance matrix,represents a mean value ofCovariance matrix ofA gaussian distribution of (d); weight ofThe calculation formula of (a) is as follows:
wherein λ is k As a measured value z n Average false alarm number of clutter area, c k As a measured value z n Distribution density of false alarm positions of the clutter region, a n As a measured value z n The amplitude of the echo of (a) is,a false alarm amplitude likelihood function representing a detection threshold of tau,representing a target amplitude likelihood function with a detection threshold of tau,represents a mean value ofThe covariance matrix is(ii) a gaussian distribution of;
step C, to Gaussian posterior intensity v k (x) Gaussian pruning and state extraction are performed, the flow chart of which is shown in fig. 2:
step C-1, deleting weighted valueLess than a predetermined threshold epsilon 1 All gaussian components of (a);
if the value z is measured n All derived Gaussian component weight values are less than epsilon 1 Then the component with the largest weight is retained and its weight value is modified to epsilon 1 ;
Step C-2, rewriting the Gaussian component after the step C-1 into the following form:
wherein, J k|k The number of gaussian components after gaussian pruning at the time k,andrespectively the weight, mean and covariance matrix of the u-th gaussian component,represents a mean value ofThe covariance matrix isA gaussian distribution of (d);
step C-3. Calculate v k|k (x) Combined distance of gaussian components:
wherein v =1,2 k|k ,w=1,2,...,J k|k And v ≠ w;
if d is vw Less than a merging threshold epsilon 2 The corresponding two gaussian components are combined into a single gaussian component, denoted as (w) vw ,m vw ,P vw ) And has:
step C-4, rewriting the Gaussian component after the step C-3 into the following form:
wherein, J k R =1,2.., J, the number of combined gaussian components k ,Andrespectively the weight, mean and covariance matrix of the r-th gaussian component,represents a mean value ofThe covariance matrix is(ii) a gaussian distribution of;
extraction of v k|k (x) The weight being greater than a threshold epsilon 3 As a final estimation result of the target state:
Claims (1)
1. a probability hypothesis density method for target tracking in clutter environment is characterized by comprising the following steps:
A. establishing a priori clutter map
Modeling a clutter map of a monitoring area as a combination of a plurality of areas with different clutter densities, and constructing clutter prior information of the monitoring area on the assumption that the clutter density in each clutter area is uniformly distributed;
the specific process of the step A is as follows:
modeling a clutter map of a monitored region as a combination of regions of different clutter densities, assuming that the clutter density within each clutter region is subject to uniform distribution, the different clutter regions having different clutter densities, i.e. V = V 1 ∪V 2 ∪...∪V M Wherein V represents the entire detection region, V 1 ~V M Representing different clutter areas divided by different clutter densities, wherein M is the total number of the clutter areas; recording clutter prior information for each clutter region, including the average of each clutter regionNumber of average false alarmsAnd false alarm location distribution density
B. Knowledge-aided tracking, for each simulation instant
According to the GM-PHD algorithm, from the state posterior intensity function v at the time k-1 k-1 (x) Calculating a state prediction strength function v at the moment k k|k-1 (x) (ii) a Predicting the intensity function v from the state at the time k k|k-1 (x) Calculating the posterior intensity function v of the state at the moment k k (x) (ii) a On the basis of the calculation result of the standard GM-PHD algorithm, v is modified according to the echo amplitude and clutter distribution information k (x) The weight value of each Gaussian component;
the specific process of the step B is as follows:
step B-1, predicting:
the posterior intensity function of the target state vector x at time k-1 is assumed to be of the following mixed gaussian form:
wherein, J k-1 J =1, 2.. J. k-1 ,Andrespectively the weight, mean and covariance matrix of the jth Gaussian component at the time k-1,represents a mean value ofCovariance matrixA gaussian distribution of;
the state prediction strength function at the time k is as follows:
wherein p is S In order to target the probability of survival,andrespectively the mean and covariance matrices of the calculated jth predicted strength gaussian component at time k,f is a motion prediction matrix of the target, Q is a process noise covariance matrix, and superscript T represents transposition; j. the design is a square γ,k For the number of new target gaussian components at time k, h =1,2 γ,k ; Andrespectively is the weight, the mean value and the covariance matrix of the h-th Gaussian component of the new target intensity function at the moment k,represents a mean value ofThe covariance matrix is(ii) a gaussian distribution of;
step B-2, updating:
predicting the state by a strength function v k|k-1 (x) The rewrite is:
wherein, J k|k-1 Predict the number of gaussian components of the intensity function for the state, l =1,2 k|k-1 ;Andrespectively the weight, mean and covariance matrix of the ith gaussian component,represents a mean value ofCovariance matrix ofA gaussian distribution of (d);
the gaussian posterior intensity formula updated at time k is:
wherein p is D Is the target detection probability, m z For an effective number of measurements, n =1,2,. Multidot.m z ,z n Is a measure of the return of the radar,h is a target measurement matrix, R is a measurement noise covariance matrix, represents a mean value ofThe covariance matrix isA gaussian distribution of (d); weight ofThe calculation formula of (a) is as follows:
wherein λ is k As a measured value z n Average false alarm number of clutter area, c k As a measured value z n Distribution density of false alarm positions of clutter areas, a n As a measured value z n The amplitude of the echo of (a) is,a false alarm amplitude likelihood function representing a detection threshold t,representing a target amplitude likelihood function with a detection threshold of tau,represents a mean value ofCovariance matrix of(ii) a gaussian distribution of;
step C: gaussian pruning and state extraction, to state posterior intensity function v k (x) Adopting a pruning strategy: deleting all Gaussian components with weight values smaller than a specified threshold, if all the Gaussian component weights derived from a certain measured value are smaller than the specified threshold, keeping the component with the maximum weight derived from the measured value, and modifying the weight of the component into the specified threshold; normalizing the weight of the Gaussian components after the pruning strategy is adopted, calculating the merging distance of the Gaussian components, and merging the Gaussian components to obtain a final estimation result of the target state;
the specific process of the step C is as follows:
step C-1, deleting weighted valueLess than a prescribed threshold value epsilon 1 All gaussian components of (a);
if the measured value z is n All derived Gaussian component weight values are less than epsilon 1 Then the component with the largest weight is retained and its weight value is modified to epsilon 1 ;
Step C-2, rewriting the Gaussian component after the step C-1 into the following form:
wherein, J k|k The number of gaussian components after gaussian pruning at the time k,andrespectively the weight, mean and covariance matrix of the u-th gaussian component,represents a mean value ofCovariance matrix ofA gaussian distribution of (d);
step C-3. Calculate v k|k (x) Combined distance of gaussian components:
wherein v =1,2 k|k ,w=1,2,...,J k|k And v ≠ w;
if d is vw Less than a merging threshold epsilon 2 The corresponding two gaussian components are combined into a single gaussian component, denoted as (w) vw ,m vw ,P vw ) And has the following:
step C-4, rewriting the Gaussian component after the step C-3 into the following form:
wherein, J k R =1, 2.., J, the number of combined gaussian components k ,Andrespectively the weight, mean and covariance matrix of the r-th gaussian component,represents a mean value ofCovariance matrix of(ii) a gaussian distribution of;
extraction of v k|k (x) The weight being greater than a threshold epsilon 3 As the final estimation result of the target state:
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711201677 | 2017-11-27 | ||
CN2017112016771 | 2017-11-27 |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109031229A CN109031229A (en) | 2018-12-18 |
CN109031229B true CN109031229B (en) | 2022-10-11 |
Family
ID=65520717
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810692989.5A Active CN109031229B (en) | 2017-11-27 | 2018-06-28 | Probability hypothesis density method for target tracking in clutter environment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109031229B (en) |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109917331B (en) * | 2019-01-11 | 2021-02-09 | 杭州电子科技大学 | Unknown clutter passive co-location method based on sparsity order optimization |
CN109886305B (en) * | 2019-01-23 | 2021-05-04 | 浙江大学 | Multi-sensor non-sequential measurement asynchronous fusion method based on GM-PHD filtering |
CN110308442B (en) * | 2019-06-10 | 2023-04-28 | 南京理工大学 | GM-PHD target tracking method of phased array radar in strong clutter environment |
CN110596693B (en) * | 2019-08-26 | 2021-10-22 | 杭州电子科技大学 | Multi-sensor GMPHD self-adaptive fusion method with iterative updating |
CN112328959B (en) * | 2020-10-14 | 2023-04-11 | 哈尔滨工程大学 | Multi-target tracking method based on adaptive extended Kalman probability hypothesis density filter |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2012149624A1 (en) * | 2011-05-04 | 2012-11-08 | Jacques Georgy | Two-stage filtering based method for multiple target tracking |
JP2014169865A (en) * | 2013-03-01 | 2014-09-18 | Hitachi Ltd | Target tracking device, target tracking program and target tracking method |
CN104156984A (en) * | 2014-05-30 | 2014-11-19 | 西北工业大学 | PHD (Probability Hypothesis Density) method for multi-target tracking in uneven clutter environment |
CN106772357A (en) * | 2016-11-23 | 2017-05-31 | 谭顺成 | AI PHD wave filters under signal to noise ratio unknown condition |
CN106896352A (en) * | 2017-04-17 | 2017-06-27 | 电子科技大学 | A kind of many radar asynchronous datas distribution fusion method theoretical based on random set |
CN113673565A (en) * | 2021-07-16 | 2021-11-19 | 杭州电子科技大学 | Multi-sensor GM-PHD self-adaptive sequential fusion multi-target tracking method |
-
2018
- 2018-06-28 CN CN201810692989.5A patent/CN109031229B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2012149624A1 (en) * | 2011-05-04 | 2012-11-08 | Jacques Georgy | Two-stage filtering based method for multiple target tracking |
CA2834916A1 (en) * | 2011-05-04 | 2012-11-08 | Jacques Georgy | Two-stage filtering based method for multiple target tracking |
JP2014169865A (en) * | 2013-03-01 | 2014-09-18 | Hitachi Ltd | Target tracking device, target tracking program and target tracking method |
CN104156984A (en) * | 2014-05-30 | 2014-11-19 | 西北工业大学 | PHD (Probability Hypothesis Density) method for multi-target tracking in uneven clutter environment |
CN106772357A (en) * | 2016-11-23 | 2017-05-31 | 谭顺成 | AI PHD wave filters under signal to noise ratio unknown condition |
CN106896352A (en) * | 2017-04-17 | 2017-06-27 | 电子科技大学 | A kind of many radar asynchronous datas distribution fusion method theoretical based on random set |
CN113673565A (en) * | 2021-07-16 | 2021-11-19 | 杭州电子科技大学 | Multi-sensor GM-PHD self-adaptive sequential fusion multi-target tracking method |
Non-Patent Citations (5)
Title |
---|
Jinglong Yang et al..Multiple extended target tracking algorithm based on GM-PHD filter and spectral clustering.《EURASIP Journal on Advances in Signal Processing 2014》.2014, * |
Weihua WU et al..Augmented state GM-PHD filter with registration errors for multi-target tracking by Doppler radars.《Signal Processing》.2015, * |
史玺.基于PHD滤波器的多目标跟踪方法研究.《中国优秀硕士学位论文全文数据库(电子期刊)信息科技辑》.2017,(第3期), * |
张鹤冰.概率假设密度滤波算法及其在多目标跟踪中的应用.《中国博士学位论文全文数据库(电子期刊)工程科技II辑》.2013,(第1期), * |
王芝.基于概率假设密度函数(PHD)的多目标跟踪方法研究.《中国优秀硕士学位论文全文数据库(电子期刊)信息科技辑》.2012,(第3期), * |
Also Published As
Publication number | Publication date |
---|---|
CN109031229A (en) | 2018-12-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109031229B (en) | Probability hypothesis density method for target tracking in clutter environment | |
CN104318059B (en) | Method for tracking target and tracking system for non-linear Gaussian Systems | |
CN105182291B (en) | The multi-object tracking method of the PHD smoothers of adaptive targets new life intensity | |
CN103885057B (en) | Adaptive strain sliding window multi-object tracking method | |
Houssineau et al. | PHD filter with diffuse spatial prior on the birth process with applications to GM-PHD filter | |
CN111722214B (en) | Method for realizing radar multi-target tracking PHD | |
CN104155650A (en) | Object tracking method based on trace point quality evaluation by entropy weight method | |
CN107064865A (en) | The passive co-located method of polar coordinates Dynamic Programming clustered based on depth | |
CN109633599B (en) | Multi-target tracking method for airborne early warning radar | |
CN105301584B (en) | The IPPHDF multiple-moving target tracking methods of fuzzy distance solution simultaneously | |
CN113673565B (en) | Multi-sensor GM-PHD self-adaptive sequential fusion multi-target tracking method | |
CN107167802A (en) | A kind of breath signal detection algorithm based on ULTRA-WIDEBAND RADAR | |
CN107436434B (en) | Track starting method based on bidirectional Doppler estimation | |
WO2023284698A1 (en) | Multi-target constant false alarm rate detection method based on deep neural network | |
CN105761276A (en) | Iteration RANSAC-based adaptive birth target intensity estimation GM-PHD multi-target tracking algorithm | |
CN104501812A (en) | Filtering algorithm based on self-adaptive new target strength | |
CN107037423A (en) | Multi-object tracking method is filtered with reference to the PHD of amplitude information | |
CN106168943A (en) | A kind of method for following the tracks of turning machine moving-target and system thereof | |
CN107102293B (en) | The passive co-located method of unknown clutter based on the estimation of sliding window integral density | |
CN103616680B (en) | Mobile dim target tracking-before-detecting method based on discrete variable rate sampling | |
CN106772357B (en) | AI-PHD filter multi-object tracking method under signal-to-noise ratio unknown condition | |
CN105353353B (en) | Multiple search particle probabilities assume the multi-object tracking method of density filtering | |
Luo et al. | Improved SMC-PHD filter for multi-target track-before-detect | |
Melzi et al. | Joint multiple target tracking and classification using the unscented Kalman particle PHD filter | |
Chen | An algorithm of mobile sensors data fusion tracking for wireless sensor networks |
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 |