CN105761276B - Based on the iteration RANSAC GM-PHD multi-object tracking methods that adaptively newborn target strength is estimated - Google Patents
Based on the iteration RANSAC GM-PHD multi-object tracking methods that adaptively newborn target strength is estimated Download PDFInfo
- Publication number
- CN105761276B CN105761276B CN201510943173.1A CN201510943173A CN105761276B CN 105761276 B CN105761276 B CN 105761276B CN 201510943173 A CN201510943173 A CN 201510943173A CN 105761276 B CN105761276 B CN 105761276B
- Authority
- CN
- China
- Prior art keywords
- submodule
- phd
- target
- ransac
- newborn
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20004—Adaptive image processing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
Landscapes
- Radar Systems Or Details Thereof (AREA)
Abstract
A kind of GM-PHD multi-object tracking methods of the adaptive newborn target strength estimation based on iteration RANSAC, including I-RANSAC new lives module of target detection and PHD filter modules.Wherein:The I-RANSAC new lives module of target detection includes:It effectively measures and generates submodule, assume to generate submodule, hypothesis testing submodule, wherein:Effectively measurement generation submodule is connected with current measurement and current state estimation respectively transmits effective measurement of currently " sliding window ", assumed by the flight path that stochastical sampling generates from effective measure assuming that generating submodule and being connected to transmit with effective measurement submodule, hypothesis testing submodule with assume that generating submodule is connected the empirical results transmitted to currently assuming flight path, that is, the location information of the newborn target confirmed.The method of the present invention has the advantages that effective, robust, can be widely applied to the multiple target trackings such as radar, robot, video monitoring field.
Description
Technical field
It is specifically a kind of that density is assumed based on Gaussian-mixture probability the present invention relates to a kind of video multi-target tracking
(GM-PHD) the dynamic video multi-object tracking method filtered.
Background technology
Multiple target tracking (MTT) technology is in intelligent monitoring, robot self-navigation, and the different fields such as biology obtain extensively
General application.The purpose of MTT is from including to estimate the number and state of multiple target in the measurement set of clutter.Due in measurement
There are the reasons such as clutter and the uncertainty of detection, the accurate multiple target for tracking number of variations is still a problem.Tradition
Multiple target tracking apply " measurement-flight path " data association technique, it is intended to multiple target tracking problem is decoupled as multiple single goals
Tracking problem.The joint probability data association side of the propositions such as more hypothesis methods (MHT) of the propositions such as Reid and Bar-Shalom
Method (JPDA) is most representative and validity two kinds of correlating methods.Due to the combined characteristic of method, data correlation method
Common drawback is that calculation amount is huge, thus limits the practical application of such algorithm.
In recent years, the multi-object tracking method for being based on stochastic finite collection (RFS) achieves quantum jump, and has obtained extensively
Concern.Uncertain measurement is expressed as random collection with multiple target state based on RFS theoretical frames method, and then by more mesh
Mark tracking problem is expressed as multidimensional Bayesian filter, so as to avoid the data correlation of " measurement-flight path ".Wherein, Mahler
The PHD implementation methods that the probability hypothesis density (PHD) of proposition filters and scholar proposes later are most representative.PHD uses more mesh
The approximate posteriority for replacing multiple target of first order statistic (probability hypothesis density, i.e. PHD) for marking the posterior probability density of random set is general
Rate density simplifies the integral operation in multidimensional Bayesian filter recurrence formula so that being implemented as multidimensional Bayesian filter can
Energy.But there is number of targets estimation inaccuracy in PHD filters.In order to solve this problem, Mahler has also been proposed base
Number probability hypothesis density (Cardinalized Probability Hypothesis Density, CPHD) filter.CPHD is logical
It crosses while the intensity function of spread state random set and radix is distributed to obtain accurate number of targets estimation.PHD and CPHD filters
Although the new life of multiple target, hatching and problem of death can be handled, there are still a disadvantage, i.e. standard PHD and CPHD filtering
Device assumes that the intensity function of newborn target is known.However, the intensity function of newborn target is not readily available under normal circumstances.
Therefore, which constrains the extensive use of PHD filters.
Through the literature search of existing technologies, IEEE Transactions ons of the B.RISTIC et al. in 2012
Aerospace and Electronic Systems (International Electrical and Electronic Engineering Association's avionics system journal)
" the Adaptive target birth intensity for PHD and CPHD filters delivered on the phase of volume 48 the 2nd
It is proposed in (the newborn objective self-adapting intensity function in PHD and CPHD filters) " a kind of based on the adaptive of particle PHD filters
Answer newborn target strength method.This method surveys the newborn target strength function of design by each frame amount, and measurement space is divided into measurement can
Subspace v can not be surveyed by surveying subspace p and measurement, construct newborn target strength by likelihood function for p, but v is still needed
Known prior information, without solving the construction problem of newborn target strength function at all.Michael Beard et al. propose newborn
Gaussian Mixture PHD filters when target strength is distributed as being uniformly distributed, but still it is known to assume that newborn target strength is distributed as
's.Ou Yangcheng et al. is directed to document, proposes a kind of normalization factor modification method, and the flight path to improve former algorithm, which normalizes, to be lost
Weighing apparatus problem, but this method still needs to known portions prior information, and also this method still includes after the filtering a large amount of clutters, is weakened
PHD filters go the major advantage of noise wave removing.Wang et al. is proposed using probability likelihood ratio test (SPRT) side all measured
Method carries out the detection of newborn target, and then constructs birth target strength function.But since it is excellent using the combination all measured
Change, inevitably brings huge computation burden.
Invention content
The present invention in view of the above-mentioned deficiencies in the prior art, provides a kind of based on iteration stochastical sampling consistency
(RANSAC) the GM-PHD multi-object tracking methods of the adaptive newborn target strength estimation of algorithm.The present invention proposes that one kind can be with
Iteration RANSAC (I-RANSAC) algorithm for going out newborn target location from measurement collective estimation, based on the newborn target position estimated
Confidence ceases, and constructs the intensity function of newborn target, needs known target intensity function to solve original PHD filters
It is insufficient.The present invention gives the adaptive GM-PHD multiple target trackings frames of the I-RANSAC algorithms based on proposition, solve original
Beginning GM-PHD multi-object tracking method needs the problem of known newborn target strength function, simple and strong robustness with realizing
Advantage.
The present invention is achieved by the following technical solutions:
The present invention includes:I-RANSAC new lives module of target detection and PHD filter modules, wherein:I-RANSAC new life mesh
Mark detection module in " sliding window " metric data and current state estimation be connected transmit be likely to occur in current sliding window it is new
The location information of raw target, so-called " sliding window " refer to the continuous measurement set of fixed frame number, and the set is over time
And move forward, i.e., after a new frame amount, which measures, to be come, which is surveyed, " sliding window " set is added backmost, and give up " sliding
A frame amount of foremost is surveyed in window " set, to keep " sliding window " set to measure the fixation of frame number.Specifically, " sliding window " is
One metrology queue moved forward with the time.PHD filter modules and I-RANSAC new lives module of target detection and current amount
The state estimation random set and number of targets for surveying the transmission target complete that is connected estimate random set.
The I-RANSAC new lives module of target detection includes:It effectively measures and generates submodule, assume to generate submodule, vacation
If examining submodule, wherein:It effectively measures generation submodule and is connected transmission currently with current measurement and current state estimation respectively
Effective measurement of " sliding window ", it is assumed that generation submodule is connected with effective measurement submodule to be transmitted from effective measurement by adopting at random
Sample and the flight path that generates it is assumed that hypothesis testing submodule with assume to generate submodule and be connected the experience transmitted to currently assuming flight path
As a result, the location information of the newborn target confirmed.
The PHD filter modules include:Newborn target strength function structure submodule, prediction submodule, update submodule,
Gauss member trims submodule and state extracts submodule, wherein:Newborn target strength function structure submodule and I-RANSAC are new
Hypothesis testing submodule in raw module of target detection, which is connected, transmits newborn target strength function, prediction submodule and newborn target
Intensity function builds submodule and is connected the prediction Gauss member parameter of transmission objectives state random set PHD, update submodule respectively with
Prediction submodule and the updated Gauss member parameter for measuring the transmission objectives state random set PHD that is connected, Gauss member trim submodule
Block is connected with update submodule transmits the Gauss member of the dbjective state random set PHD after updated Gauss member is merged and deleted
Parameter, state extract submodule and Gauss member trimming submodule be connected transmission objectives state estimation random set and number of targets estimate with
Machine collection.
Compared with prior art, present invention has the advantages that:Newborn module of target detection based on I-RANSAC is PHD
Filter provides newborn target strength information;It solves the problems, such as that PHD filters need to know newborn target strength function, protects
The robustness and reliability of PHD filters are demonstrate,proved.This method is simple and effective, easy to implement, and robustness is good, the army of can be widely applied to
Thing and civilian multiple target tracking field.
Description of the drawings
Fig. 1 is the adaptive GM-PHD multiple-target systems block diagram based on I-RANSAC in the embodiment of the present invention;
Fig. 2 is tracking scene and target plot in the embodiment of the present invention;
Fig. 3 is the measurement figure including clutter in the directions each frame x and the directions y in the embodiment of the present invention;
Fig. 4 is in the embodiment of the present invention in the x and y direction to the location estimation result of time;
Fig. 5 is that the embodiment of the present invention gives the result that same multiple target scene is estimated by standard GM-PHD filters;
Fig. 6 is to provide the average real goal number in 100 Monte Carlos, standard GM-PHD filters in the embodiment of the present invention
The comparison result for the number of targets estimated with the method for the present invention,;
Fig. 7 is that the embodiment of the present invention has suffered the average standard GM-PHD filters in 100 Monte Carlos and the method for the present invention
The comparison figure of the OSPA distances of estimation;
Specific implementation mode
It elaborates below in conjunction with the accompanying drawings to the embodiment of the present invention:The present embodiment before being with technical solution of the present invention
It puts and is implemented, give detailed embodiment and specific operating process, but protection scope of the present invention is not limited to down
The embodiment stated.
As shown in Figure 1, the present embodiment includes:I-RANSAC new lives module of target detection and PHD filter modules, wherein:I-
RANSAC new life module of target detection includes:It effectively measures and generates submodule, assume to generate submodule, hypothesis testing submodule,
Wherein:Effectively measurement generation submodule is connected with current measurement and current state estimation respectively transmits the effective quantity of currently " sliding window "
It surveys, it is assumed that generate submodule and effectively measure submodule and be connected transmission flight path for being generated by stochastical sampling from effectively measurement
The empirical results transmitted to currently assuming flight path it is assumed that hypothesis testing submodule is connected with hypothesis generation submodule, that is, confirm
The location information of newborn target.
Effective effective measurement for measuring generation submodule realization for detecting newborn target.
The hypothesis generates submodule and realizes various possible newborn goal hypothesis.
The hypothesis testing submodule realizes the inspection of newborn goal hypothesis, to confirm possible newborn target.
PHD filter modules include:Newborn target strength function structure submodule, prediction submodule, update submodule, Gauss
Member trimming submodule and state extract submodule.Wherein:Newborn target strength function structure submodule and I-RANSAC new life mesh
Mark the newborn target strength function of the connected transmission of hypothesis testing submodule in detection module, prediction submodule and new life target strength
Function builds submodule and is connected the prediction Gauss member parameter of transmission objectives state random set PHD, update submodule respectively with prediction
Submodule and measurement are connected the updated Gauss member parameter of transmission objectives state random set PHD, Gauss member trim submodule and
Update the Gauss radix scrophulariae of the dbjective state random set PHD after the connected transmission of submodule merges and deletes to updated Gauss member
Number, state extract that submodule and Gauss member trimming submodule are connected transmission objectives state estimation random set and number of targets is estimated at random
Collection.
The new life target strength function submodule realizes the structure of the intensity function of newborn target.
The prediction submodule realizes the prediction of the Gauss member parameter of the PHD of dbjective state random set.
The update submodule realizes the Gauss member parameter of update target prediction state random set PHD.
The Gauss member trimming submodule realizes the close Gauss member of combined distance, and the Gauss of the minimum weights of removal
Member, for reducing calculation amount and clutter.
The state extracts the desired value corresponding to Gauss member of the submodule extraction weights more than threshold value (generally taking 0.5) i.e.
For state random set, Gauss member number is number of targets random set.
The specific work process of effect measurement generation submodule is in the present embodiment:
1) input new frame measures, and is put into the end of " sliding window " queue, and the first frame amount for removing " sliding window " queue is surveyed, and realizes " sliding
The movement of window " queue.
2) it sets current state and is estimated as xk, calculate the measurement residuals d that frame is corresponded in " sliding window "k(i)=zk(i)-HkFkxk,
Middle Hk, FkIt is the measurement matrix and state-transition matrix of system respectively;And the norm of residual error:
S in formulakIt is the covariance matrix of residual vector.If gk(i) distance be less than setting thresholding, then it is corresponding measurement be considered as with
There are track associations, which is removed.So processing then obtains effectively measuring set.
Assume that the specific work process for generating submodule is in the present embodiment:It is measured before collection in the survey of two frame amount from effective
The subset for including a sample is respectively randomly selected, its model M is calculated by least square method according to the two sample points;Effective quantity
The set surveyed after concentrating two samples for removing and extracting is called the complementary set that commercial weight surveys collection.In complementary set with the small Mr. Yu of the error of model M
The sample of one given threshold is called interior point, they constitute set and are called consistent collection.The hypothesized model of model M, that is, current.
The specific work process of hypothesis testing submodule is in the present embodiment:If the radix unanimously collected is more than given threshold
Value, then it is assumed that correct model parameter has been obtained, and model is recalculated by least square method using the sample unanimously concentrated, it should
New model is a newborn object module;It after determining a fresh target, is replaced by complementary set and effectively measures collection, repeat above-mentioned vacation
If generating and hypothesis testing process, you can confirm multiple newborn targets.
Object module is in the present embodiment:
1) state equation of target is non-linear asymptotic turning (CT) model:
xk=F (ω) xk-1+Gvk-1 (3)
Wherein, ω is rate of turn, when with target centroid coordinate and its speed come when describing target, state xkIt can indicate
For:
xk=(locX, k, locY, k, velX, k, velY, k, ω)T (4)
F (ω) is state-transition matrix,
State-noise vkIt is the zero-mean Gaussian noise using Q as covariance, σvFor system noise standard deviation.
2) systematic observation model is zk=Hxk+wk, wherein observing matrix H is:
Observation noise wkIt is zero-mean Gaussian noise, covariance isσwFor observation noise standard deviation.
Some parameter settings of the present embodiment PHD filter modules are as follows:Take the survival probability P of each targetS=0.99,
Target detection probability PD=0.99.State-noise standard deviation sigmav=2m/s2, observation noise standard deviation sigmaw=8.Sampling time interval
It is taken as T=1s.Clutter is obeyed in tracing area [0,20000] × [0,20000] and is uniformly distributed, and the number of each moment clutter
It is λ that mesh, which obeys parameter,c=10-7m-2Poisson distribution (i.e. average each frame have 40 clutters).
The specific work process of newborn target strength function submodule is in the present embodiment:
Assuming that the k moment detects m newborn target by I-RANSAC new life module of target detection, then with each newborn mesh
Cursor position is as location mean value, using given value as speed mean value, and using given larger positive definite numerical matrix as variance matrix
Intensity function of the Multi-dimensional Gaussian distribution function as each newborn target is constructed, and its weights is taken as 0.03.
Predict that the specific work process of submodule is in the present embodiment:
Assuming that the Gauss member parameter that the k moment describes PHD isWherein, JkFor the k moment Gauss member
Number,For the mean value of i-th of Gauss member of k moment,It is its weights,It is corresponding covariance.
1) it predicts newly generating target, if it is J that the k+1 moment, which newly generates target number,γ, k+1, then to j=1 ...,
Jγ, k+1HaveWhereinRespectively newborn mesh
The weights of this yuan of absolute altitude, state it is expected (mean value) and covariance;
2) to hatching target prediction, if hatching target number is Jβ, k+1, then to j=1 ..., Jβ, k+1, l=1 ..., JkHaveWhereinTo hatch mesh
The weights of this meta-model of absolute altitude,Respectively (5) formula and the state-transition matrix in (6) formula and state-noise association side
Difference.
3) prediction calculating is carried out to the target that exists, if its survival probability is pS, then to j=1 ..., Jk, according to following
Formula updates weights, mean value and covariance:
The specific work process of update submodule is in the present embodiment:
1) the measurement random set obtained using moving object detection module is denoted as Zk+1, and (7) formula observing matrix H and amount
It surveys noise covariance R and updates weights, mean value and covariance.pDDescribed in detection probability embodiment as above, then to being not detected
Target with formula (8)~(10) update:To j=1 ..., Jk+1|k, wherein Jk+1|k=Jγ, k+1+lJβ, k+1+Jk+1, have:
2) to the target update detected, i.e., using the center-of-mass coordinate of moving object detection module as measure random set into
Row update calculates, to each z ∈ Zk+1, calculate:
If the probability of the clutter RFS of Poisson distribution is κk(z), to j=1 ..., Jk+1|kHave:
Implementation result
The tracking scene of the present embodiment is two-dimentional multiple target scene, has 8 targets in monitor area.The new target that generates is obeyed
Poisson distribution.In order to be compared with original PHD filters, 5 target initial positions of setting are it is known that fresh target is obeyed
Poisson distribution, intensity are:Wherein ωγ=0.03,
The initial position of remaining 3 target and moment are unknown.
Fig. 2 gives the targetpath in simulating scenes.
Fig. 3 is the measurement figure including clutter in the directions each frame x and the directions y.
Fig. 4 is the method for the present invention in the x and y direction to the location estimation result of time.From fig. 4, it can be seen that of the invention
Method not only can accurately estimate the newborn target of 5 known initial positions, but also can accurately estimate 3 unknown startings
The newborn target of position.
Fig. 5 gives the result that same multiple target scene is estimated by standard GM-PHD filters.Wherein PHD filtering parts
Parameter setting is identical with the method for the present invention.As can be seen that GM-PHD filters are only capable of accurately estimating 5 known start bits
The newborn target set, and the newborn target of 3 unknown initial positions cannot be estimated.This is because GM-PHD filters do not have
The unknown fresh target ability in initial position is determined.
Fig. 6 provides the average real goal number in 100 Monte Carlos, standard GM-PHD filters and the method for the present invention estimation
Number of targets comparison result, wherein red solid line indicate real goal number, blue dotted line indicate the method for the present invention estimation mesh
Number is marked, the dotted line of black bands point indicates the number of targets of standard GM-PHD filters estimation.What can be removed finds out, standard GM-
The estimation of PHD filters gives mistake target numbers in the period after the appearance of newborn target are estimated, and the method for the present invention
It is then whole to provide correct target numbers estimation.
Fig. 7 give the average standard GM-PHD filters in 100 Monte Carlos and the method for the present invention estimation OSPA away from
From comparison figure.OSPA distances are a kind of measurements of " distance " between two random collections, it can be used for evaluating track algorithm
Performance, it is better to be worth the smaller tracking performance for illustrating algorithm.
Therefore, the method for the present invention has well solved standard GM-PHD filters and needs known target original intensity function
Problem has very strong practicability, validity and robustness.
Claims (7)
1. a kind of GM-PHD multi-object tracking methods of the adaptive newborn target strength estimation based on iteration RANSAC, feature
It is, the method is to be based on I-RANSAC new lives module of target detection and PHD filter modules;Wherein:I-RANSAC new life mesh
Mark detection module in " sliding window " metric data and current state estimation be connected transmit be likely to occur in current sliding window it is new
The location information of raw target, PHD filter modules are connected transmission all with I-RANSAC new lives module of target detection and current measure
The state estimation random set and number of targets of target estimate random set;" sliding window " is a measurement moved forward with the time
Queue;
The I-RANSAC new lives module of target detection includes:It effectively measures and generates submodule, assume to generate submodule, assume inspection
Submodule is tested, wherein:Effectively measurement generation submodule is connected to transmit and currently " slide with currently measurement and current state estimation respectively
Effective measurement of window ", it is assumed that generation submodule is connected with effective measurement generation submodule to be transmitted from effectively measuring by random
Sampling and the flight path that generates are it is assumed that hypothesis testing submodule is connected the warp transmitted to currently assuming flight path with generation submodule is assumed
It tests as a result, confirming the location information of newborn target;
The PHD filter modules include:Newborn target strength function structure submodule, prediction submodule, update submodule, Gauss
Member trimming submodule and state extract submodule, wherein:Newborn target strength function structure submodule and I-RANSAC new life mesh
Mark the newborn target strength function of the connected transmission of hypothesis testing submodule in detection module, prediction submodule and new life target strength
Function builds submodule and is connected the prediction Gauss member parameter of transmission objectives state random set PHD, update submodule respectively with prediction
Submodule trims submodule with the current updated Gauss member parameter for measuring the transmission objectives state random set PHD that is connected, Gauss member
Block is connected with update submodule transmits the Gauss member of the dbjective state random set PHD after updated Gauss member is merged and deleted
Parameter, state extract submodule and Gauss member trimming submodule be connected transmission objectives state estimation random set and number of targets estimate with
Machine collection.
2. the GM-PHD multiple targets of the adaptive newborn target strength estimation according to claim 1 based on iteration RANSAC
Tracking, characterized in that effective measure generates the end that submodule enters new input quantity measuring " sliding window " queue, together
When remove " sliding window " queue first frame amount survey, realize " sliding window " queue movement, then according to current state estimation and " sliding window "
The measurement of middle corresponding frame calculates residual error and its norm, according to the going or staying that the size of the norm determines to measure, to form effective amount
Survey set.
3. the GM-PHD multiple targets of the adaptive newborn target strength estimation according to claim 1 based on iteration RANSAC
Tracking, characterized in that the hypothesis generates submodule and respectively randomly selected from effective measure during two frame amount are surveyed before collection
The subset for including a sample calculates its model M according to the two sample points by least square method, the vacation of model M, that is, current
If model.
4. the GM-PHD multiple targets of the adaptive newborn target strength estimation according to claim 1 based on iteration RANSAC
Tracking, characterized in that the hypothesis testing submodule is compared the radix unanimously collected with given threshold value, if one
The radix of collection is caused to be more than given threshold value, then it is assumed that obtain correct model parameter, and using the sample unanimously concentrated by most
Small square law recalculates model, which is a newborn object module;After determining a fresh target, replaced by complementary set
It changes and effectively measures collection, repeat above-mentioned hypothesis generation and hypothesis testing process, you can confirm multiple newborn targets;Collection is measured from effective
Before two frame amount survey in respectively randomly select the subset for including a sample, calculated by least square method according to the two sample points
Go out its model M;It effectively measures the set after concentrating two samples for removing and extracting and is called the complementary set that commercial weight surveys collection;In complementary set with mould
The sample that the error of type M is less than a certain given threshold is called interior point, and the set that they are constituted is called consistent collection;Model M, that is, current
Hypothesized model.
5. the GM-PHD multiple targets of the adaptive newborn target strength estimation according to claim 1 based on iteration RANSAC
Tracking, characterized in that the described newborn target strength function structure submodule then using each newborn target location as
Location mean value using given value as speed mean value, and constructs one using given larger positive definite numerical matrix as variance matrix
Intensity function of the Multi-dimensional Gaussian distribution function as each newborn target, and its weights is taken as 0.03.
6. the GM-PHD multiple targets of the adaptive newborn target strength estimation according to claim 1 based on iteration RANSAC
Tracking, characterized in that the prediction submodule includes to hatch dbjective state mesh to newborn dbjective state random set PHD
The Gauss member parameter of mark random set PHD and existing dbjective state random set PHD is predicted.
7. the GM-PHD multiple targets of the adaptive newborn target strength estimation according to claim 1 based on iteration RANSAC
Tracking, characterized in that the state extracts submodule and refers to after update submodule and Gauss member trimming submodule
The Gauss member of dbjective state random set PHD handle, extract Gauss member of its weights more than 0.5 as Target state estimator
The number of random set, Gauss member of the weights more than 0.5 estimates random set as number of targets.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510943173.1A CN105761276B (en) | 2015-12-15 | 2015-12-15 | Based on the iteration RANSAC GM-PHD multi-object tracking methods that adaptively newborn target strength is estimated |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510943173.1A CN105761276B (en) | 2015-12-15 | 2015-12-15 | Based on the iteration RANSAC GM-PHD multi-object tracking methods that adaptively newborn target strength is estimated |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105761276A CN105761276A (en) | 2016-07-13 |
CN105761276B true CN105761276B (en) | 2018-11-09 |
Family
ID=56342125
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510943173.1A Active CN105761276B (en) | 2015-12-15 | 2015-12-15 | Based on the iteration RANSAC GM-PHD multi-object tracking methods that adaptively newborn target strength is estimated |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105761276B (en) |
Families Citing this family (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107220658B (en) * | 2017-05-10 | 2019-05-31 | 中国人民解放军军械工程学院 | A kind of images match point is to screening technique |
CN107273659B (en) * | 2017-05-17 | 2021-01-05 | 中国科学院光电技术研究所 | RANSAC algorithm-based improved track prediction method for space debris photoelectric tracking |
CN108333569B (en) * | 2018-01-19 | 2021-01-12 | 杭州电子科技大学 | Asynchronous multi-sensor fusion multi-target tracking method based on PHD filtering |
CN109298725B (en) * | 2018-11-29 | 2021-06-15 | 重庆大学 | Distributed multi-target tracking method for group robots based on PHD filtering |
CN110780269B (en) * | 2019-10-08 | 2022-03-25 | 河海大学 | Explicit multi-target tracking method based on GM-PHD filter under self-adaptive new growth intensity |
CN111340853B (en) * | 2019-12-04 | 2023-06-02 | 杭州电子科技大学 | Multi-sensor GMPHD self-adaptive fusion method based on OSPA iteration |
CN111562571B (en) * | 2020-05-28 | 2022-04-29 | 江南大学 | Maneuvering multi-target tracking and track maintaining method for unknown new-born strength |
CN112364292B (en) * | 2020-09-24 | 2024-05-03 | 北京电子工程总体研究所 | Ransac-based dense target tracking method, ransac-based dense target tracking device, ransac-based dense target tracking equipment and medium |
CN112198503A (en) * | 2020-10-16 | 2021-01-08 | 无锡威孚高科技集团股份有限公司 | Target track prediction optimization method and device and radar system |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101894381A (en) * | 2010-08-05 | 2010-11-24 | 上海交通大学 | Multi-target tracking system in dynamic video sequence |
CN104501812A (en) * | 2014-12-05 | 2015-04-08 | 江南大学 | Filtering algorithm based on self-adaptive new target strength |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CA2834916C (en) * | 2011-05-04 | 2017-10-17 | Jacques Georgy | Two-stage filtering based method for multiple target tracking |
-
2015
- 2015-12-15 CN CN201510943173.1A patent/CN105761276B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101894381A (en) * | 2010-08-05 | 2010-11-24 | 上海交通大学 | Multi-target tracking system in dynamic video sequence |
CN104501812A (en) * | 2014-12-05 | 2015-04-08 | 江南大学 | Filtering algorithm based on self-adaptive new target strength |
Non-Patent Citations (5)
Title |
---|
A Probabilistic Hypothesis Density Filter for Traffic Flow Estimation in the Presence of Clutter;Matthieu Canaud et al;《2012 Workshop on Sensor Data Fusion: Trends, Solutions, Applications (SDF)》;20120904;31-36 * |
Multiple Target Tracking using Recursive RANSAC;Peter C. Niedfeldt et al;《2014 American Control Conference (ACC)》;20140604;3393-3398 * |
基于概率假设密度的目标数变化视频跟踪算法;吴静静 等;《江苏大学学报(自然科学版)》;20151130;第36卷(第6期);697-704 * |
基于自适应新生目标强度的概率假设密度滤波;吴静静 等;《系统仿真学报》;20151130;第27卷(第11期);2741-2747 * |
自适应GM-PHD滤波器在多目标追踪的应用;苍岩 等;《哈尔滨工程大学学报》;20151130;第36卷(第11期);1526-1531 * |
Also Published As
Publication number | Publication date |
---|---|
CN105761276A (en) | 2016-07-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105761276B (en) | Based on the iteration RANSAC GM-PHD multi-object tracking methods that adaptively newborn target strength is estimated | |
CN103729637B (en) | Extended target probability hypothesis density filtering method based on cubature Kalman filtering | |
CN105182291B (en) | The multi-object tracking method of the PHD smoothers of adaptive targets new life intensity | |
CN104237879B (en) | A kind of multi-object tracking method in radar system | |
CN106526585B (en) | Tracking before target detection based on the filtering of Gaussian particle gesture probability hypothesis density | |
CN104155650A (en) | Object tracking method based on trace point quality evaluation by entropy weight method | |
CN101894381A (en) | Multi-target tracking system in dynamic video sequence | |
CN106054169B (en) | Multistation Radar Signal Fusion detection method based on tracking information | |
WO2012009947A1 (en) | Device and method of signature-driven multi-target tracking | |
CN106772353B (en) | A kind of multi-object tracking method and system suitable for flicker noise | |
CN108333569A (en) | A kind of asynchronous multiple sensors fusion multi-object tracking method based on PHD filtering | |
CN104849702B (en) | Radar system error combined estimation method is filtered using the GM EPHD of ADS B datas | |
CN109002835A (en) | A kind of particle filter data correlation method based on maximum entropy fuzzy clustering | |
CN107066806B (en) | Data Association and device | |
CN106291498B (en) | A kind of detecting and tracking combined optimization method based on particle filter | |
CN108802722A (en) | It is a kind of based on tracking before the Faint target detection virtually composed | |
CN107526070A (en) | The multipath fusion multiple target tracking algorithm of sky-wave OTH radar | |
CN104133983B (en) | Multiple-target Bernoulli distribution type fusion method based on weighting negative first-order RD (Renyi Divergence) | |
CN102915545A (en) | OpenCV(open source computer vision library)-based video target tracking algorithm | |
CN108344981A (en) | Multisensor Asynchronous towards clutter detects TSBF multi-object tracking methods | |
CN109239702B (en) | Airport low-altitude flying bird number statistical method based on target state set | |
CN106054167B (en) | More extension method for tracking target based on density filter | |
Chavali et al. | Hierarchical particle filtering for multi-modal data fusion with application to multiple-target tracking | |
CN111259332B (en) | Fuzzy data association method and multi-target tracking method in clutter environment | |
CN107037423A (en) | Multi-object tracking method is filtered with reference to the PHD of amplitude information |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |