CN109683150A - Multigroup/extension Target ellipse shape estimates evaluation method - Google Patents
Multigroup/extension Target ellipse shape estimates evaluation method Download PDFInfo
- Publication number
- CN109683150A CN109683150A CN201811640647.5A CN201811640647A CN109683150A CN 109683150 A CN109683150 A CN 109683150A CN 201811640647 A CN201811640647 A CN 201811640647A CN 109683150 A CN109683150 A CN 109683150A
- Authority
- CN
- China
- Prior art keywords
- estimation
- extension
- multigroup
- target
- extension 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.)
- Granted
Links
- 238000011156 evaluation Methods 0.000 title claims abstract description 97
- 238000001914 filtration Methods 0.000 claims description 32
- 238000000034 method Methods 0.000 claims description 30
- 238000004422 calculation algorithm Methods 0.000 claims description 15
- 230000001133 acceleration Effects 0.000 claims description 4
- 235000013399 edible fruits Nutrition 0.000 claims 1
- 238000005259 measurement Methods 0.000 abstract description 6
- 230000035945 sensitivity Effects 0.000 abstract description 2
- 230000000875 corresponding effect Effects 0.000 description 12
- 238000012545 processing Methods 0.000 description 11
- 230000006870 function Effects 0.000 description 9
- 101001086191 Borrelia burgdorferi Outer surface protein A Proteins 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 238000009792 diffusion process Methods 0.000 description 5
- 238000001514 detection method Methods 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 4
- 241000208340 Araliaceae Species 0.000 description 3
- 238000000342 Monte Carlo simulation Methods 0.000 description 3
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 3
- 235000003140 Panax quinquefolius Nutrition 0.000 description 3
- 230000003247 decreasing effect Effects 0.000 description 3
- 235000008434 ginseng Nutrition 0.000 description 3
- 239000000203 mixture Substances 0.000 description 3
- 238000012544 monitoring process Methods 0.000 description 3
- 239000002245 particle Substances 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 241001123248 Arma Species 0.000 description 2
- 230000002596 correlated effect Effects 0.000 description 2
- 238000009472 formulation Methods 0.000 description 2
- 238000003709 image segmentation Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 240000007594 Oryza sativa Species 0.000 description 1
- 235000007164 Oryza sativa Nutrition 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 238000012552 review Methods 0.000 description 1
- 235000009566 rice Nutrition 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000012360 testing method Methods 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
- 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
- G01S7/415—Identification of targets based on measurements of movement associated with the target
-
- 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
- G01S13/72—Radar-tracking systems; Analogous systems for two-dimensional tracking, e.g. combination of angle and range tracking, track-while-scan radar
- G01S13/723—Radar-tracking systems; Analogous systems for two-dimensional tracking, e.g. combination of angle and range tracking, track-while-scan radar by using numerical data
Landscapes
- Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
本发明公开了一种多群/扩展目标椭圆形状估计评价方法，解决了现有性能评价体系不能反应目标的过估与漏估及椭圆长轴朝向估计误差的问题，实现步骤有：获取量测数据；对量测数据滤波得到椭圆形状估计；椭圆形状估计匹配；计算IOU值；得到椭圆长轴朝向误差及势估计错误惩罚后的性能评价指标数值；依NIS值评价滤波器性能；判断是否接收新量测数据，是则更新时刻，继续性能评价，否则结束。本发明通过椭圆形状估计匹配，选择惩罚函数惩罚椭圆长轴朝向估计误差，并惩罚过估或漏估情况获得性能评价结果。本发明响应快、灵敏度高、精度高，能用于目标识别、战场监视、视频监控、空中交通管制等领域。
The invention discloses a multi-group/expanded target ellipse shape estimation and evaluation method, which solves the problems that the existing performance evaluation system cannot reflect the over-estimation and omission of the target and the estimation error of the orientation of the long axis of the ellipse. data; filter the measurement data to obtain the ellipse shape estimation; the ellipse shape estimation matches; calculate the IOU value; obtain the ellipse long axis orientation error and the performance evaluation index value after the potential estimation error penalty; evaluate the filter performance according to the NIS value; judge whether to accept or not New measurement data, if yes, update the time and continue performance evaluation, otherwise end. The present invention obtains performance evaluation results by estimating and matching the ellipse shape, selecting a penalty function to penalize the estimation error of the long axis of the ellipse, and penalizing over-estimation or underestimation. The invention has quick response, high sensitivity and high precision, and can be used in the fields of target identification, battlefield surveillance, video surveillance, air traffic control and the like.
Description
Technical field
The invention belongs to radar target tracking technical fields, are related to multigroup/extension Target ellipse shape estimation shape and estimate
Matching treatment is counted, specifically a kind of multigroup/extension Target ellipse shape estimation filter evaluation method is used for radar target tracking
And group/extension target shape tracking performance evaluation etc..
Background technique
In traditional target tracking domain, it will usually ignore target shape and target is regarded as a little.And with radar, infrared
Equal sensor resolutions are higher and higher, and the target relevant information of acquisition is more and more, when sensor single target is detected it is more
When a measurement, target must be regarded as group/extension target just to handle.Therefore only target cannot be tracked, together
When also to estimate its shape.The shape tracking estimation of group/extension target can be not only used for target identification, missile defence, battlefield
The military fields such as monitoring, may be also used in the civil fields such as video monitoring, air traffic control.In recent years, estimate group/extension mesh
The algorithm of mark shape mainly has random matrix (random matrix) method, random hyperplane (random hypersurface)
3 kinds of method of method and Gaussian process (gaussian process).It is dry since there are noises in detection environment in actual conditions
Disturb, thus from sensor collection to metric data in extract detailed shape information be often extremely challenging, so
In many actual groups/extension target following application, people often pay close attention to simple shape, for this purpose, the group of Most current/
Extend method for tracking target be all to be modeled using ellipse to target shape, ellipse can also provide about target direction of motion,
The relevant supplementary information of range of scatter and size.It is objective as group/extension Target Tracking Filter algorithm continuously improves innovation
On need a kind of effective group/extension target shape tracking filter algorithm assessment indicator system.
Article " a kind of multiple targets Gaussian Mixture PHD filtering based on oval random hypersurface model that Zhang Hui et al. is delivered
In device ", because not there is also effectively evaluating index that can be used to measure algorithm to the estimation performance of target shape at that time, divide
Not Ji Suan OSPA distance between centroid position, oval long semi-minor axis and Class area estimation value and true value, realize to performance of filter
Evaluation.
In the article " the extension target SMC-PHD filtering based on the convex RHM of star " that Wang Xue et al. is delivered, use
Performance Evaluating Indexes of Intersection-over-union (IOU) value as its shape algorithm for estimating, IOU performance evaluation refer to
Mark is a kind of relevant Performance Evaluating Indexes of shape, by intersect between feedback estimation shape and true shape and phase simultaneously area it
Than illustrating superiority-inferiority of the shape algorithm for estimating in terms of shape estimation.
However for multigroup/extension target elliptical shape estimation filter, OSPA Distance evaluation index can not provide this
Filter estimates the upper relevant Performance Evaluating Indexes of shape in elliptical shape so that can not evaluate elliptical shape estimation tracking filter
The superiority-inferiority that device algorithm estimates shape is not properly suited for the evaluation index of elliptical shape estimation filter.By IOU performance
Evaluation index applies in multigroup/extension target tracking domain, due to tracking scene and tracking purpose needs,
The mistake that IOU Performance Evaluating Indexes do not react in the gesture estimation mistake occurred during tracking, which is estimated, estimates situation with leakage, also without anti-
It answers elliptical shape estimation ellipse towards error, therefore cannot directly apply to the estimation evaluation of multigroup/extension target elliptical shape.
Summary of the invention
Estimate and leak the purpose of the present invention is in view of the above shortcomings of the prior art, proposition can react to cross in gesture estimation mistake
Estimate and oval multigroup/extension Target ellipse shape estimation filter evaluation method towards error.
The present invention is a kind of multigroup/extension Target ellipse shape estimation filter evaluation method, which is characterized in that including such as
Lower step:
(1) radar sensor metric data is obtained in real time；
(2) radar sensor metric data is filtered, obtains multigroup/extension Target ellipse shape estimation ginseng
Number: the metric data that radar sensor obtains is inputted into multigroup to be evaluated/extension Target ellipse shape estimation filter and is carried out
Multigroup/extension target shape estimates tracking filter, obtains multigroup/extension Target ellipse shape estimation parameter, mainly includes: each
A elliptical shape central coordinate of circle, length shaft length, transverse direction and movement velocity size, acceleration magnitude；
(3) multigroup/extension Target ellipse shape estimation parameter matching: use estimates elliptical shape apart from division methods
Parameter is divided into multiple units, there is corresponding elliptical shape estimation matching result, in each unit in each unit
It include a true group/extension Target ellipse form parameter and corresponding all multigroups/extension Target ellipse shape with result
Estimate parameter；
(4) all multigroups/extension Target ellipse shape estimation and the true group/extension ellipse target in each unit are calculated
Between IOU value: use matching result, calculate all multigroups/extension mesh in the matched unit of true group/extension target
The elliptical shape intersection and phase and area ratio between elliptical shape estimation are marked, shape estimated result in each unit is obtained
IOU Performance Evaluating Indexes numerical value；
(5) estimation transverse is calculated towards error, punish towards error, is obtained transverse and is punished towards error
Performance Evaluating Indexes numerical value (Aep) afterwards: the transverse court first in calculating multigroup/extension Target ellipse shape estimation parameter
To difference of the transverse in parameter and true group/extension Target ellipse form parameter between；Then punishment letter is chosen
Number does corresponding numerical value punishment, carries out estimation transverse towards error and punishes (Angle error penalty (Aep)), most
The Performance Evaluating Indexes numerical value (Aep) obtained afterwards；
(6) according to judgement, Performance Evaluating Indexes numerical value after the gesture estimation mistake punishment in each unit is obtained
(cardinalized error penalty (Cep)): judging, which whether there is in each unit, estimates or leaks the case where estimating, and is examining
Consider to be obtained simultaneously on the basis of shape is estimated and estimates or leak the Performance Evaluating Indexes numerical value (Cep) estimated；
(7) multigroup/extension Target ellipse shape estimation filter is evaluated with the moment Evaluation on Total Performance index value (NIS)
The superiority-inferiority of tracking: extracting the Performance Evaluating Indexes value (New index system (NIS)) of the moment all unit totality, uses
In doing superiority-inferiority explanation to multigroup/extension target shape estimation filter tracking performance；
(8) judge whether radar sensor also receives new metric data, if so, renewable time, returns and execute
Step (2) recycles real-time tracking, otherwise, executes step (9)；
(9) terminate multigroup/extension target following, complete filter tracks performance evaluation.
The present invention mainly on the basis of traditional I/O U Performance Evaluating Indexes, is estimated by multigroup/extension Target ellipse shape
It counts matching process and carries out division unit, solve the transverse existing for multigroup/extension Target ellipse shape estimation filter
It towards error, crosses to estimate or leak and estimates situation the problem of existing assessment indicator system can not provide definite evaluation, carrying out accordingly
After the punishment of Performance Evaluating Indexes numerical value and combination, the Evaluation on Total Performance index value for finally obtaining the moment is excellent for filter
Pessimum explanation.
Compared with the prior art, the present invention has the following advantages:
First, optimal secondary mode distribution (Optimal Sub-pattern Assignment) the most widely used at present
OSPA Distance evaluation index, measuring the error size between real goal and estimated result is by calculating the distance between mass center
It is calculated, has ignored the comparison of the evaluation to vpg connection most important in shape estimation filter；The perfect mesh of the present invention
Evaluation index problem relevant for shape estimation aspect, is not only realized in preceding multigroup/extension target shape tracking estimation field
Shape estimates the foundation of the assessment indicator system of vpg connection, also achieves the evaluation to other parameters evaluated error size.
Second, IOU Performance Evaluating Indexes are widely used in the appraisement system of object detection and image segmentation field, at present
Also have in multigroup/extension target shape tracking estimation field, do shape estimation evaluation index, but be used for multigroup/expansion
It opens up in target shape tracking estimation evaluation index, error, the multigroup/extension target gesture estimation mistake of oval direction is estimated and estimated with leakage
Situation cannot all embody, and therefore, for the present invention on the basis of IOU Performance Evaluating Indexes, perfect shape estimates performance evaluation
While On Index, also solves common mistake in multigroup/extension target shape estimation tracking field and estimate and estimate problem with leakage
The problem of elliptical shape filter evaluation index is influenced, and joined elliptical shape common in elliptical shape estimation for the first time
Influence towards error to multigroup/extension target shape tracking filter estimation method evaluation index.
Detailed description of the invention
Fig. 1 is flow chart of the invention；
Fig. 2 is multigroup/true elliptical shape of extension target and motion diagram in simulator service scene of the present invention；
Fig. 3 is that the present invention uses the filtering estimated result figure after SMC-RHM-PHD filter；
Fig. 4 is Performance Evaluating Indexes NIS proposed by the present invention and IOU Performance Evaluating Indexes comparison diagram.
Specific embodiment
The present invention is described in detail with reference to the accompanying drawing.
Embodiment 1
Recently as the proposition of a large amount of multigroup/extension target shape estimation filtering algorithm, numerous scholars are to multigroup/expansion
Exhibition Target ellipse shape estimation tracking filter carries out continuously improving innovation, but in terms of its superiority-inferiority evaluation index, at present
OSPA Distance evaluation index is widely applied and does superiority-inferiority evaluation, estimates that upper shape is related in shape since the filter can not be provided
Performance Evaluating Indexes, can not evaluate the filter practice to shape estimation direction superiority-inferiority, therefore not exclusively be applicable in
In multigroup/extension Target ellipse shape estimation tracking filter evaluation index；Also small part scholar refers to IOU performance evaluation
Mark apply in multigroup/extension target tracking domain, due to track scene and track purpose needs, to target institute
The accuracy requirement for locating the acquisition of information such as position and number is higher and higher, and IOU Performance Evaluating Indexes can not react during tracking
Mistake in the gesture estimation mistake of appearance is estimated and estimates situation with leakage, elliptical shape estimates ellipse towards error.
A kind of more effective group/extension target shape track algorithm assessment indicator system is objectively needed, the present invention is suitable
This actual needs is answered, a kind of multigroup/extension Target ellipse shape estimation filter evaluation method is proposed.Referring to Fig. 1, this hair
It is bright that superiority-inferiority evaluation is carried out to multigroup/extension Target ellipse shape estimation tracking filter, pass through the multigroup/expansion obtained after filtering
It opens up Target ellipse shape estimation parameter calculated performance evaluation index value to illustrate filter tracking performance evaluation with this, includes as follows
Step:
(1) radar sensor metric data is obtained in real time, these metric data include two-dimensional spatial location coordinate information, are
Initial time data acquired.True group/extension ellipse target elliptical shape is obtained, and really group/extension ellipse target is total
Number.
(2) radar sensor metric data is filtered, obtains multigroup/extension Target ellipse shape estimation ginseng
Number: the metric data that radar sensor obtains is inputted into multigroup to be evaluated/extension Target ellipse shape estimation filter and is carried out
Multigroup/extension Target ellipse shape estimates tracking filter, obtains multigroup/extension Target ellipse shape estimation parameter, estimation ginseng
Number includes: the central coordinate of circle of each elliptical shape for being obtained after filtering processing estimation, length shaft length, transverse towards with
And movement velocity size, acceleration magnitude etc..Here each parameter obtained is multigroup to be evaluated/extension Target ellipse shape
It is obtained after estimation filter filtering processing.
If it is desired to being evaluated using the present invention certain multigroup/extension Target ellipse shape estimation filter, seek to
Metric data is input in this kind of filter, which is exactly multigroup to be evaluated/extension Target ellipse shape estimation filter
Wave device.The multigroup referred to later/extension Target ellipse shape estimation is multigroup to be evaluated/extension Target ellipse shape estimation
It is obtained after filter filtering processing.
Wherein multigroup/extension target refers to, multiple to be detected by a sensor the movement mesh of multiple measurements in synchronization
Mark.Multigroup/extension Target ellipse shape estimation refers to, is filtered by multigroup/extension Target ellipse shape estimation filter
After obtain multiple elliptical estimations.
(3) multigroup/extension Target ellipse shape estimation parameter matching: use is apart from division methods, with reference to true group/extension
The elliptical shape estimation parameter obtained after step (2) filtering is divided into multiple units by ellipse target, and the unit being divided into is always a
Number is equal to true group/extension ellipse target total number, has corresponding multigroup/extension Target ellipse shape to estimate in each unit
Count parameter Estimation matching result, the matching result in each unit include a true group/extension Target ellipse form parameter and
Corresponding all multigroups/extension Target ellipse shape estimates parameter.
Wherein, true group/extension Target ellipse form parameter includes the elliptical circle of true group/extension target diffusion
Heart coordinate, length shaft length and transverse direction.
(4) all multigroups/extension Target ellipse shape estimation and the true group/extension Target ellipse in each unit are calculated
IOU value between shape: the matching result being mentioned to using step (3) calculates the matched list of true group/extension target
Elliptical shape intersection and phase and area ratio in member between all multigroups/extension Target ellipse shape estimation, obtain each
Multiple IOU Performance Evaluating Indexes numerical value in unit, the multigroup/expansion for including in the number of IOU value and the unit in each unit
It is equal to open up Target ellipse shape estimation number.
(5) estimation transverse is calculated towards error, punish towards error, is obtained transverse and is punished towards error
Performance Evaluating Indexes numerical value (Aep) afterwards: traversing all units, firstly for each unit, calculates each more in the unit
True group/extension target of the transverse for including in group/extension Target ellipse shape estimation parameter towards parameter and the unit
The absolute value of difference of the transverse between in elliptical shape parameter；Then selection penalty is done corresponding numerical value and is punished
It penalizes, that is, carries out estimation transverse towards error and punish (Angle error penalty (Aep)), according to ellipse towards error
The characteristic of value, penalty must in [0 pi/2] value interval monotone decreasing, the slope of penalty is also as oval direction
Error amount increases and reduces, i.e. penalty curve is steeper.The estimation transverse for finally exporting each unit is punished towards error
Penalize Performance Evaluating Indexes numerical value (Aep).
(6) according to judgement, Performance Evaluating Indexes numerical value after the gesture estimation mistake punishment in each unit is obtained
(cardinalized error penalty (Cep)): judging, which whether there is in each unit, estimates or leaks the case where estimating, according to
Different judging results estimates situation if there is no crossing to estimate or leak, then handles in this step without punishment；If it exists, then into
Row is crossed to estimate or leak and estimates corresponding punishment processing, no matter which kind of judging result, being all finally will be in the base for considering that shape is estimated
Performance Evaluating Indexes numerical value (Cep) after gesture estimation mistake is punished is obtained on plinth simultaneously.
(7) with current time Evaluation on Total Performance index value (NIS) evaluation multigroup/extension Target ellipse shape estimation filtering
The superiority-inferiority of device tracking: Performance Evaluating Indexes value (the New index system of the current time all unit totality is extracted
(NIS)), for doing superiority-inferiority explanation to multigroup to be evaluated/extension target shape estimation filter tracking performance.In the present invention
Current time described herein is initial time when first time, is hereafter the updated moment every time.
(8) judge whether radar sensor also receives new metric data, if so, renewable time, returns and execute
Step (2) recycles real-time tracking, otherwise, executes step (9): when radar sensor is also when receiving metric data, illustrating radar
Still there is multigroup/extension target Continuous movement in sensor observation area, updates the tracking moment at this time, returns to circulation and executes step (2)-
(8), performance evaluation update constantly is carried out to multigroup/extension Target ellipse shape estimation tracking filter.Otherwise, no reception measures
When data, illustrates that observation area is moved without target, execute step (9).
(9) terminate multigroup/extension target following, complete filter shape tracking estimation performance evaluation.
Thinking of the invention is to be sensed using multigroup to be evaluated/extension Target ellipse shape estimation filter to radar
The metric data that device obtains does multigroup/extension Target ellipse shape estimation, obtains elliptical shape and estimates parameter, passes through estimation first
True group/extension target and filtering estimated result are divided into multiple units, then based on each unit, calculated by matching
Elliptical shape estimation parameter and true group/extension Target ellipse shape between intersect and phase simultaneously area ratio, i.e. IOU, followed by
It is poor with the angle of real goal elliptical shape long axis direction to calculate elliptical shape estimation long axis direction, does shape using penalty
Estimation punishes that, followed by gesture estimation mistake is judged whether there is, the main foundation of judgement estimates matching towards evaluated error
Problem is estimated as a result, mainly judging whether there is to estimate or leak, and corresponding operation is carried out by judging result again, finally obtains this
Moment Evaluation on Total Performance index value (NIS), for evaluating multigroup to be evaluated/extension Target ellipse shape estimation filter
Tracking performance.
Embodiment 2
Multigroup/extension Target ellipse shape estimation filter evaluation method is with embodiment 1, multigroup/expansion described in step (3)
The estimation parameter matching of Target ellipse shape is opened up to comprise the following steps that mainly according to apart from division methods
3.1 according to all multigroups/extension Target ellipse shape estimation parameter and all true groups/extension Target ellipse shape
Parameter calculates Euclidean distance of the above-mentioned elliptical shape two-by-two between the center of circle.
The smallest that group of multigroup of Euclidean distance/extension Target ellipse shape is estimated parameter and true group/extension target by 3.2
Elliptical shape parameter matches together, that is, shows that this group of multigroup/extension Target ellipse shape estimation parameter is the true group/extension
The filtering of Target ellipse form parameter is estimated.
3.3 a true group/extension Target ellipse form parameter and matched all elliptical shape estimations are returned
Into identical unit.How many organizes true group/extension Target ellipse form parameter in this example, is just divided into how many a units,
And it is independent from each other between each unit.
Multigroup of the present invention/extension Target ellipse shape estimates characteristic parameter matching method, applies to multigroup/extension Target ellipse shape
Shape is estimated dividing elements, accurate dividing elements can be more accurately carried out, so that the performance obtained is commented later in parameter matching
Valence index is more authentic and valid.
Embodiment 3
Multigroup/extension Target ellipse shape estimation filter evaluation method is with embodiment 1-2, meter described in step (4)
Calculate the IOU value of all multigroups in each unit/between the estimation of extension Target ellipse shape and true group/extension Target ellipse shape
Method be monte carlo method, be described in detail as follows:
Use [ae be] respectively indicate long axis and minor axis length in multigroup/extension Target ellipse shape estimation parameter, [ar
br] respectively indicate long axis and minor axis length in the unit in true group/extension Target ellipse form parameter；With multigroup/extension
Target ellipse shape estimates that the midpoint of circle center line connecting between elliptical shape and true group/extension Target ellipse shape is mass center, this
Example is using the long axis in 4 times of true group/extension Target ellipse form parameters as side length, i.e. 4ar, a rectangular area is done, in the area
By 10,000 points of generation are uniformly distributed in domain, belong to multigroup/extension Target ellipse shape in this ten thousand points by counting
Estimation is oval, and belongs to true group/extension Target ellipse shape and share N number of point；True group/extension target is more with this at this time
Elliptical shape intersecting area S between the estimation of group/extension Target ellipse shape∩It is obtained by following formula:
Then corresponding phase and area S∪It can be obtained by following formula:
S∪=π ar·br+π·ae·be-S∩
Then IOU value can be obtained by following formula:
It seeks IOU value above and why selects monte carlo method to be since this method being capable of express statistic acquisition dependency number
According to, and IOU value is calculated by what above data was capable of efficiency；Certainly selection is generated to obey in the rectangular area and is uniformly distributed
Points it is more, the IOU value sought also will increase computation complexity closer to true value, extend calculate the time, influence reality
Shi Xing.The present invention passes through many experiments, and when 10,000 points of use, and rectangular area side length value is 4arWhen can quickly
IOU value is accurately acquired again, therefore the above value is the best value of this example, in the specific implementation, according to specific needs, Ke Yijin
The different value of row.Compared to the method for seeking IOU value by image processing method after drawing two correlation ellipse mappings, originally
Inventing the monte carlo method used has the characteristics that easier, efficiency.
Embodiment 4
Multigroup/extension Target ellipse shape estimation filter evaluation method is with embodiment 1-3, meter described in step (5)
Estimation transverse is calculated towards error, punished towards error, obtains transverse towards the performance evaluation after error punishment
Index value (Aep) method, steps are as follows:
5.1 according to the matching result in step (3) in each unit, passes through each group of multigroup/extension Target ellipse respectively
Shape estimate parameter in transverse towards parameter value, be calculated the transverse towards with the true group in the unit/
The absolute value for extending difference between the angle of Target ellipse long axis direction, is denoted as | Δ φ |, and | Δ φ | ∈ [0 pi/2].
5.2 selection penalty f (| Δ φ |), it punishes towards angular error.It must satisfy according to penalty
Angular error is bigger, punishes that more serious and angle error value is bigger, penalty curve is steeper, i.e. the bigger rule of punishment dynamics
Then.The penalty that the present invention chooses is cosine function in trigonometric function as penalty, i.e. f (| Δ φ |)=cos (|
Δφ|)。
5.3 combine referring to as a result, obtaining the evaluation after punishing towards angular error for the Iou value being calculated in step (4)
Scale value (Angle error penalty (Aep)), is calculated by following formula:
Aep=Iouf (| Δ φ |)=IouCos (| Δ φ |)
The present invention is regular according to the punishment of formulation after having understood the oval characteristic towards angular error of estimation why
It selects the cosine function in trigonometric function as penalty of the invention, is because cosine function is empty in the value of [0 pi/2]
Interior is the function of a monotone decreasing, and its two-dimensional curve meets the feature steeper with the bigger curve of angle error value,
Two above feature agrees with the punishment rule of formulation completely；And it can be seen that working as | Δ φ |=0, f (| Δ φ |)=1, at this time
There are Aep=Iou, i.e., when estimation it is oval towards error be zero, that is to say, that there is no estimation it is oval towards error when, without
Punishment.When | Δ φ |=pi/2, f (| Δ φ |)=0, there are Aep=0, and estimation ellipse at this time has reached maximum towards error,
That is to say, the estimation group/extension target direction of motion and complete mistake in actual direction is illustrated, in cross-shaped, therefore by property
Can evaluation index value zero setting to react the estimated result of the mistake.Either existing IOU Performance Evaluating Indexes and or OSPA
Distance evaluation index, always all not to the estimated result after multigroup/extension Target ellipse shape estimation filter filtering processing
The oval evaluation towards error is carried out, but in multigroup/extension target tracking domain, oval direction is often exactly multigroup/expansion
Target direction of motion is opened up, the present invention is doing above combination, referring to according to the performance evaluation after combination according to the characteristic of cosine function
Scale value provides a strong reference for multigroup/extension target motion intention judgement confidence level, has very high reality
Meaning.
Embodiment 5
Multigroup/extension Target ellipse shape estimation filter evaluation method is with embodiment 1-4, root described in step (6)
It is judged that obtaining Performance Evaluating Indexes numerical value Cep after the gesture estimation mistake punishment in each unit, specifically comprise the following steps:
Matching result in (6a) each unit according to step (3), judged whether there is each unit and estimated
Or situation is estimated in leakage: judged whether there is to estimate or leak and estimate situation method and be: true group/extension target in each unit only matches
When estimating parameter to 1 group of multigroup/extension Target ellipse shape, then it is assumed that there is no crossing to estimate or leak to estimate situation, execute step
(6e) is directly exported.Otherwise it is assumed that multigroup/extension Target ellipse shape estimation filter is to the true group/extension target state estimator
Filtering processing occurs crossing to estimate or leak estimating situation, thens follow the steps (6b), does further judgement.
(6b) estimates the processing respectively of situation to existing to estimate or leak: judgement determination was to estimate or leak to estimate situation: when the list
When true group/extension object matching in member estimates parameter to 2 or more group multigroups/extension Target ellipse shape, then it is assumed that
The filter moment occurs crossing to the true group/extension target state estimator filtering processing estimates situation, then executes (6c)；When one
When true group/extension target is not matched to any elliptical shape estimated result, then it is assumed that the filter moment is to the group/extension
There is leakage and estimates situation in target state estimator filtering processing, then executes (6d)；
(6c) carried out estimating punishment, exported Cep value: all Aep evaluation index values in the unit obtained using step (5)
Do a mean operation, use the result as the multigroup/extension target shape estimation filter under the scene in the unit
True group/extension target Cep value；Why use mean value computation result as true group/extension target moment here
The Performance Evaluating Indexes value after estimating punishment is crossed, is because true group/extension target in each unit is always when estimating excessively
There are 2 or more group shape estimated results, therefore in the calculating by step (5), can obtain and multigroup/expansion for including in the unit
Open up the same Aep Performance Evaluating Indexes value of Target ellipse shape estimation number, thus using the Performance Evaluating Indexes value of preference or
Be deviation Performance Evaluating Indexes value be unsuitable for being used as filter to the true group in the unit/extension target state estimator filtering with
Track Performance Evaluating Indexes value, and the numerical value that mean operation obtains is a good measure of central tendency, has and is quick on the draw, determines sternly
Close, concise legibility calculates simple, suitable further calculation and smaller the advantages that being influenced by sampling variation and mean operation reaction
Sensitive, the either large or small variation of each data can influence final result, therefore the present invention uses owning in the unit
The mean value of the Aep value property for being described filter group/extension target state estimator filter tracking true in the unit being crossed in the case of estimating
Energy evaluation index value is preferably to select.
(6d) carries out leakage and estimates punishment, exports Cep=0: directly will be to single group/extension target Performance Evaluating Indexes value
Zero setting processing, i.e. Cep=0；Leakage is estimated, in traditional IOU Performance Evaluating Indexes, estimates situation in the event of leakage, often
It is ignored, the present invention solves the problems, such as that group/extension target capabilities evaluation index value that meeting is ignored, directly estimates leakage is estimated in leakage
Assign zero.
(6e) does not do any punishment, the Cep=Aep exported at this time.
The mistake that the above method of the present invention is intended to react in the estimation mistake of the gesture in multigroup/extension target shape tracking field is estimated
The influence different degrees of to Performance Evaluating Indexes is estimated with leakage.Compared in traditional I/O U Performance Evaluating Indexes value, the present invention is solved
How to embody estimate with leakage estimate to shape estimation performance evaluation influence, by above method, realized in the case of estimating with leakage
In the case of estimating the payment method of different forms with obtained estimate estimate with leakage in the case of respective Performance Evaluating Indexes value, and can
To make practical evaluation to multigroup/extension Target ellipse shape estimation filter by the value.
Embodiment 6
Multigroup/extension Target ellipse shape estimation filter evaluation method is with embodiment 1-5, use described in step (7)
The moment Evaluation on Total Performance index value (NIS) evaluates multigroup/extension Target ellipse shape estimation filter tracking superiority-inferiority,
It specifically describes are as follows: the Cep index value in all units acquired to step (6) is averaged calculating, and obtained result is exactly this
The Evaluation on Total Performance index value of the moment multigroup/extension Target ellipse shape estimation track algorithm.In order to embody the moment
Evaluation on Total Performance index value, the present invention by Evaluation on Total Performance index NIS value, illustrate multigroup under the moment scene/
Target shape estimation filter tracking performance is extended, multigroup of the same race/extension target shape estimation filter can be compared in difference
The size description of the NIS value filter obtained under the multigroup at moment/extension target difference tracking scene be more suitable for it is any with
Track scene can also compare different multigroups/extension target shape estimation filter under identical multigroup/extension target following scene
The size of obtained NIS value judges which multigroup/extension target shape estimation filter is more suitable for tracking the scene.The present invention
Mean operation is done to the whole Cep values obtained in all units, using mean operation can embody all groups/extension target this when
Carve the Performance Evaluating Indexes value trend after carrying out shape estimation filter tracking, moreover it is possible to whether unexpected according to Performance Evaluating Indexes curve
There is sharp fall and judge whether the moment filter leakage occurs and estimate, this is because estimating in step of the present invention (6d) for leakage
The processing of punishment is the Cep value zero setting that will be leaked in the case of estimating, therefore occurs leaking to estimate will lead to the relatively low situation of mean value.
A more detailed example is given below, the present invention is further described
Embodiment 7
Multigroup/extension Target ellipse shape estimation filter evaluation method is with embodiment 1-6, in conjunction in attached drawing 1, to this hair
Bright specific steps are further described.
Step 1, radar sensor Correlated Case with ARMA Measurement data are obtained.
When multigroup/extension target setting in motion is when in radar sensor observation area, radar sensor receive multigroup/
Target Correlated Case with ARMA Measurement is extended, initial time is recorded as, clocks and carve k=1.
Step 2, multigroup/extension Target ellipse shape estimation: the metric data obtained according to radar sensor, selection is constructed
Elliptical shape estimation filter to be evaluated carries out multigroup/extension target shape and estimates tracking filter, and output filters estimated result,
That is multigroup/extension Target ellipse shape estimation.
Multigroup/extension target shape estimation tracking filter selected to use in this example, based on oval random hypersurface
(RHM) particle probabilities assume density (SMC-PHD) filter.The filter describes group/extension by constructing a major and minor axis
Target sizes, the ellipse of long axis direction description extension target direction of motion, using Gaussian Profile, the random hypersurface ruler of approximate ellipse
The distribution of the factor is spent, and tracking is filtered to multigroup/extension target using SMC-PHD filtering method, uses the filter
The metric data that radar sensor receives is filtered, multigroup/extension under the k moment tracking scene can be obtained
The elliptical shape of target estimates parameter.Elliptical shape estimation parameter mainly contains, each group/extension target state estimator ellipse
The central coordinate of circle of shape, length shaft length, transverse towards and group/extension target speed size, acceleration magnitude
Estimation.
Step 3, true multigroup/extension ellipse target is matched with multigroup/extension Target ellipse shape estimation parameter, output
With result.
Using apart from division methods, realize that true multigroup/extension ellipse target and multigroup/extension Target ellipse shape are estimated
Matching the multigroup obtained after filtering/extension Target ellipse shape estimation is divided into N number of unit, the size etc. of N referring to Fig. 1
The number in radar observation region is moved in true group/extension ellipse target.Calculate multigroup/expansion obtained in all steps 2
Opening up each elliptical shape in the estimation of Target ellipse shape estimates the center of circle between each true multigroup/extension ellipse target center of circle
The smallest elliptical shape estimation of Euclidean distance is matched in same unit, together by Euclidean distance with true group/extension ellipse target
The different estimated results that one true group/extension ellipse target is matched to all are integrated into same unit, that is, show these ellipses
Shape estimation is all to belong to true group/extension ellipse target filtering estimation；Different true group/extension ellipse targets and its
The different estimated results being matched to are integrated into different units, and are independent from each other between these units, are not influenced each other.
Described here, true group/extension ellipse target refers to, using ellipse to single group/extension target diffusion shape
The real motion target modeled.True multigroup/extension ellipse target refers to that multiple true groups/extension ellipse target is constituted
Mass motion target complex.
Step 4, calculate the estimation of all multigroups in same unit/extension Target ellipse shape and the true group in the unit/
Extend the IOU value between ellipse target.
By all matching results in the division unit described in step 3, multigroup/extension mesh in each unit is calculated
Mark elliptical shape estimate the elliptical shape between matched true group/extension ellipse target intersection and mutually and area it
Than obtaining the IOU Performance Evaluating Indexes numerical value of the shape estimation tracking filter of each unit.In k moment each unit how many
A filtering estimated result can then obtain respective numbers IOU value.
Step 5, estimation transverse in same unit is calculated to carry out transverse towards error and punish towards error, it is defeated
Performance Evaluating Indexes value (Aep) after k moment transverse is punished towards error out.
According to the matching result in step 3, by elliptical shape estimate in ellipse towards parameter value, ellipse is calculated
For the transverse of shape estimation towards the absolute value with true group/extension ellipse target direction angle difference, i.e. estimation is oval
Long axis is denoted as towards error | Δ φ |, and | Δ φ | ∈ [0 pi/2].Then penalty is chosen, to transverse towards error
It is punished.According to penalty in value interval [0 pi/2] must monotone decreasing, and it is bigger to meet angular error, punishment
When more serious and angle error value is bigger, slope is smaller, the bigger feature of penalty value change rate.The penalty that the present invention chooses
It is the cosine function in trigonometric function as penalty, i.e. f (| Δ φ |)=cos (| Δ φ |).It falls into a trap then in conjunction with step 4
Obtained IOU value as a result, obtain transverse towards error punishment after evaluation index value (Angle error
penalty(Aep)).Then Aep is calculated by following formula:
Aep=Iouf (| Δ φ |)=Ioucos (| Δ φ |)
The present invention considers the estimation transverse in the ignorance of traditional I/O U Performance Evaluating Indexes towards error problem, because
The transverse of true multigroup/extension ellipse target is towards the direction of motion for being often exactly target, to judging multigroup/extension target
The accuracy of motion intention provides important confidence level reference, for example, in military field to judge target attack, cruise or
Equal strategic intents are withdrawn to have important practical significance.
Step 6, it judged whether there is and estimates or leak the case where estimating, if so, carrying out estimating or leaking the corresponding punishment estimated；It is no
Then, any punishment is not done, and output considers that the gesture of vpg connection is estimated to cross in mistake and estimates or leak the Performance Evaluating Indexes in the case of estimating
Numerical value (cardinalized error penalty (Cep)):
According to the matching result in the division unit in step 3, judged whether there is to estimate or leak and estimates the method for situation and be,
When true group/extension ellipse target in the unit has only been matched to 1 elliptical shape estimated result, then it is assumed that there is no cross to estimate
Or situation is estimated in leakage；Otherwise it is assumed that the k moment, there is true group/extension target estimation tracking filter in the unit in filter
It crosses to estimate or leak and estimates situation；Judgement was to estimate or leak the method for estimating situation and corresponding penalty method is as follows: when one true
Group/extension object matching to 2 or more estimated results when, then it is assumed that there is the group/extension target in the moment
It crosses and estimates situation, the mistake that the present invention uses estimates punishment for for a group/extension target, all estimated results that will match to are passed through
It crosses the Aep evaluation index value that step 5 obtains and does a mean operation, use the result as the multigroup/extension target shape estimation
Track algorithm under the scene for single group/extension target Performance Evaluating Indexes value.When a true group/extension
When the none of matching of target is estimated, then it is assumed that the moment leakage occurs to the group/extension target and estimates situation, and the present invention adopts
It is exactly that will handle single group/extension target Performance Evaluating Indexes value zero setting that punishment is estimated in leakage.The gesture that step 6 obtains is estimated
Performance Evaluating Indexes value after meter mistake punishment, is denoted as Cep.
Step 7, Performance Evaluating Indexes value overall in k moment all units is extracted, to multigroup/extension Target ellipse shape
Estimation filter carries out performance evaluation.
For multigroup/extension target, group/extension target in each unit can obtain a gesture after step 6
Performance Evaluating Indexes value Cep after estimation mistake punishment.And finally wants reaction multigroup/extension target shape and estimate track algorithm
Filtering performance the moment superiority-inferiority, it is desirable that the moment totality filtering performance evaluation index value is illustrated.It mentions
It takes the Performance Evaluating Indexes value of final k moment totality, Cep value exactly is acquired to each group/extension target and is averaged place
Reason, obtained result are exactly the Evaluation on Total Performance index value of the moment multigroup/extension target shape estimation track algorithm, note
It is NIS；By the NIS numerical value of output, for comparing multigroup/extension target shape estimation tracking filter superiority-inferiority explanation.
The present invention can be used for unmanned aerial vehicle group tracking, aircraft carrier warship group tracking etc., passing under the tracking scene according to filter
NIS numerical value can determine under the tracking scene, the confidence level of filter tracks performance as reference.
Step 8, judge whether radar sensor also receives new metric data, if so, following after moment k is added 1
Ring executes step 2-8, continues performance evaluation to multigroup/extension Target ellipse shape estimation tracking filter.Otherwise, that is, it passes
Sensor is not received by new metric data, illustrates to move in observation area without target, executes step 9.
Step 9, terminate multigroup/extension target following, complete filter shape tracking estimation performance evaluation.
IOU Performance Evaluating Indexes are widely used in the appraisement system of object detection and image segmentation field, also useful at present
In multigroup/extension target shape tracking estimation field, shape estimation evaluation index is done, but is being used for multigroup/extension target
Elliptical shape tracks in estimation filter evaluation index, to the error of oval direction, multigroup/extension target gesture estimation mistake estimate with
Situation is estimated in leakage cannot all embody, and therefore, the present invention solves elliptical shape estimation on the basis of IOU Performance Evaluating Indexes and refers to
While marking not perfect problem, also solves common mistake in multigroup/extension target shape estimation tracking field and estimate and estimate with leakage
Problem joined common in elliptical shape estimation the influence problem of elliptical shape estimation filter evaluation index for the first time
Influence of the elliptical shape towards error to multigroup/extension target shape tracking filter estimation method evaluation index.
Effect of the invention is described further below with reference to emulation experiment.
Embodiment 8
Multigroup/extension Target ellipse shape estimation filter evaluation method with embodiment 1-7,
Emulation experiment condition:
The hardware test platform of emulation experiment of the present invention is: processor Intel Core i3-7100 CPU, dominant frequency are
3.90GHz, memory 4GB；Software platform are as follows: 7 Ultimate of Windows, 64 bit manipulation systems, MATLAB R2010b.
Emulation content:
Simulating scenes of the invention are to track number changed multigroup/extension mesh at any time in two-dimensional tracking scene
Mark, the observation area of emulation are [- 5 × 104,8×104]×[-5×104,5×104], unit is rice (m), sampling period T=1
(s), noise v is measuredkStandard deviation sigmax=σy=500m.It is up to 8 groups/extension target in observation area, and all groups/
It extends target and has moved 30s altogether in observation area.The number of tracing area clutter locating for target obeys the Poisson that mean value is 3
Distribution.Starting tracking moment multigroup/extension target sum is 1, stochastical sampling total number of particles N=1000.Wherein in 11s, hair
Raw group's division.23s occurs group and merges, and overall running track and true oval diffusion shape are as shown in Figure 2.Use base
It is filtered in particle PHD (SMC-PHD) filter of oval random hypersurface (RHM), filter result is as shown in Figure 3.
Analysis of simulation result:
Fig. 2 is multigroup/true elliptical shape of extension target and motion diagram in simulator service scene of the present invention, all one in figure
Each and every one self-existent elliptical shape is that true multigroup/extension target diffusion shape moves in radar sensor observation area
The position really passed through in 30s.True multigroup/inference rule namely of the present invention, abscissa and ordinate in figure
The rectangular area collectively formed is exactly the observation area of the radar detection.
Fig. 3 is to estimate tracking result figure, that is, multigroup/extension using the filtering of SMC-RHM-PHD filter in this example
Target ellipse shape estimation figure, Fig. 3 (a) are that overall filtering estimates that tracking result figure, Fig. 3 (b) are that the filtering of partial enlargement is estimated
Tracking result figure is counted, in order to more intuitively embody between true multigroup/extension Target ellipse diffusion shape and elliptical shape estimation
Existing error, solid oval is true multigroup/extension ellipse target present position and really spreads elliptical shape in figure, in figure
Dotted ellipse is to do the elliptical shape estimation that elliptical shape estimates that filter tracking obtains by SMC-RHM-PHD filter to correspond to
Elliptical shape in observation area estimates present position, in figure × amount that is received for multigroup/extension target by radar sensor
The corresponding position of measured data 2-d spatial coordinate, these metric data not only from multigroup/extension target, also come from radar
Existing noise in sensor itself and observation area.Filtering estimated result can be reflected in figure is constantly present error,
But macroscopic error is all smaller and the error at each moment is all random inconsistent, therefore is not available naked eyes and distinguishes
The superiority-inferiority of filter tracks performance, the present invention try hard to the tracking performance superiority and inferiority that filter is specifically evaluated from the angle of amount sum number
Property, specific Performance Evaluating Indexes are given, are illustrated in conjunction with figure (4).
NIS Performance Evaluating Indexes of the present invention under the tracking scene are as shown in figure 4, Fig. 4 is performance proposed by the present invention
Evaluation index NIS and IOU Performance Evaluating Indexes comparison diagram.
Two curves in comparison diagram 4 are as it can be seen that this two curves are filtered by Fig. 2, tracking scene shown in Fig. 3 and filter
Wave result obtains: when unexpected sharp fall occurs in NIS performance indicator of the invention and a certain moment afterwards is fast
When quick-recovery, this can judge that, at the moment, filter leakage occurs to the shape estimation filter tracking of target and estimates situation；?
3rd, 5,8 moment, filter often will appear the leakage at a moment to newborn target due to its characteristic due to the new life of target
Estimate, and traditional I/O U Performance Evaluating Indexes, abscissa the 3rd, 5,8 moment, estimating situation to leakage did not made any reaction, this be because
Mistake of the shape estimation filter in gesture estimation mistake cannot be estimated for traditional I/O U Performance Evaluating Indexes and estimate situation with leakage and make instead
It answers, and reviews NIS Performance Evaluating Indexes value of the invention, the sharp fall at a moment occur to reflect that feelings are estimated in the leakage
Condition.Period between the 9 to 14th moment of abscissa, NIS performance index value proposed by the invention and IOU coincide substantially,
It is because multigroup/extension target numbers are less in the period, filter performance under the tracking scene is more excellent.But the 14th
Into the period between the 19th moment, due to the quick increasing of the multigroup in the period in observation area/extension target numbers
Long that elliptical shape estimation filter is caused the biggish evaluated error of meeting occur, error is mainly derived from oval direction estimation and misses
Difference, but the IOU value tendency in the period in the period between IOU Performance Evaluating Indexes value and the 9 to 14th moment is substantially
Equal, this is because IOU Performance Evaluating Indexes can not embody the transverse in elliptical shape estimation towards error, comparison is originally
The NIS Performance Evaluating Indexes proposed are invented, due to having the oval technical characterstic punished towards evaluated error, so index value exists
On the basis of original, certain punishment joined, it is proposed by the present invention to elliptical shape estimation ellipse towards error rdativery sensitive
NIS Performance Evaluating Indexes value compares the period between the 14th to the 19th moment in the period between the 9th to the 14th moment
Interior numerical value tendency illustrates that the present invention is higher towards error-detecting precision to ellipse.It can therefore be appreciated that multigroup/extension target
For SMC-RHM-PHD filter when group/extension target numbers are excessive, filtering estimated result will appear more serious oval court
To error.
In brief, multigroup disclosed by the invention/extension Target ellipse shape estimates evaluation method, solves current performance
The mistake that assessment indicator system cannot react in target state estimator, which is estimated, to be estimated with leakage and the technical issues of transverse is towards evaluated error,
Realize that step has (1) to obtain metric data；(2) filter is filtered metric data to obtain elliptical shape estimation；(3) oval
Shape estimation matching；(4) IOU value is calculated；(5) transverse is obtained towards the Performance Evaluating Indexes numerical value after error punishment
(Aep)；(6) according to judgement, Performance Evaluating Indexes numerical value (Cep) after gesture estimation mistake punishment is obtained；(7) according to NIS value, evaluation
Performance of filter；(8) judge whether sensor receives metric data, if so, renewable time, continues performance of filter evaluation,
Otherwise, terminate performance of filter evaluation.The present invention considers that transverse is estimated or leaked towards evaluated error and mistake and estimates situation, proposes
Estimate to match by elliptical shape, selects penalty to punish transverse towards evaluated error and then passes through judgement
It crosses and estimates or leak the solution estimated and then punished to obtain performance evaluation index value, advantages of the present invention is to transverse
It can also embody to estimate or leak in real time towards evaluated error sensitivity and estimate situation, application field has the military affairs such as target identification, battlefield surveillance
The civil fields such as field and video monitoring, air traffic control.
Claims (5)
1. a kind of multigroup/extension Target ellipse shape estimation filter evaluation method, which comprises the steps of:
(1) radar sensor metric data is obtained in real time；
(2) radar sensor metric data is filtered, obtains multigroup/extension Target ellipse shape estimation parameter: will
The metric data that radar sensor obtains input multigroup to be evaluated/extension Target ellipse shape estimation filter carry out multigroup/
It extends target shape and estimates tracking filter, obtain multigroup/extension Target ellipse shape estimation parameter, mainly include: each ellipse
Round central coordinate of circle, length shaft length, transverse direction and movement velocity size, acceleration magnitude；
(3) multigroup/extension Target ellipse shape estimation parameter matching: use estimates parameter apart from division methods, by elliptical shape
Multiple units are divided into, there is corresponding elliptical shape estimation matching result, the matching knot in each unit in each unit
Fruit includes that a true group/extension Target ellipse form parameter is estimated with corresponding all multigroups/extension Target ellipse shape
Parameter；
(4) all multigroups/extension Target ellipse shape in each unit is calculated to estimate between true group/extension ellipse target
IOU value: matching result is used, it is ellipse to calculate all multigroups/extension target in the matched unit of true group/extension target
Elliptical shape intersection and phase and area ratio between round estimation, obtain the IOU of shape estimated result in each unit
It can evaluation index numerical value；
(5) estimation transverse is calculated towards error, punish towards error, obtains transverse towards after error punishment
Performance Evaluating Indexes numerical value: transverse first in calculating multigroup/extension Target ellipse shape estimation parameter towards parameter with
Difference of the transverse between in true group/extension Target ellipse form parameter；Then penalty is chosen to do accordingly
Numerical value punishment, carry out estimation transverse towards error punish, the Performance Evaluating Indexes numerical value finally obtained；
(6) it according to judgement, obtains Performance Evaluating Indexes numerical value after the gesture estimation mistake punishment in each unit: judging each unit
Interior whether there is is estimated or leaks the case where estimating, and obtains on the basis of considering shape estimation or cross to estimate or leak the performance estimated to comment simultaneously
Valence index value；
(7) with moment Evaluation on Total Performance index value NIS evaluation multigroup/extension Target ellipse shape estimation filter tracking
Superiority-inferiority: extracting the Performance Evaluating Indexes value of the moment all unit totality, for multigroup/extension target shape estimation filtering
Device tracking performance does superiority-inferiority explanation；
(8) judge whether radar sensor also receives new metric data, if so, renewable time, returns to step
(2)-(8) recycle real-time tracking, otherwise, execute step (9)；
(9) terminate multigroup/extension target following, complete filter tracks performance evaluation.
2. multigroup according to claim 1/extension Target ellipse shape estimates evaluation method, which is characterized in that step (3)
Described in the estimation parameter matching of multigroup/extension Target ellipse shape, use apart from division methods, comprise the following steps that
3.1, which calculate all multigroups/extension Target ellipse shape, estimates between all true groups/extension Target ellipse shape center of circle
Euclidean distance；
3.2 match the smallest estimation elliptical shape estimation of Euclidean distance together with true group/extension ellipse target, that is, show
The elliptical shape is estimated as true group/extension ellipse target filtering estimation；
Each true group/extension ellipse target and matched all elliptical shape estimations are grouped into difference by 3.3 respectively
Unit in.
3. multigroup according to claim 1/extension Target ellipse shape estimates evaluation method, which is characterized in that step (5)
Described in calculating estimate transverse towards error, punished towards error, obtain transverse towards after error punishment
Performance Evaluating Indexes numerical value, comprise the following steps that
5.1 estimate the transverse in parameter according to the matching result in step (3), by multigroup/extension Target ellipse shape
Towards parameter value, the oval absolute value towards with true group/extension ellipse target direction angle error value of estimation is calculated,
It is denoted as | Δ φ |, and | Δ φ | ∈ [0 pi/2]；
5.2 selection penalty f (| Δ φ |), it punishes towards angular error.It must satisfy direction according to penalty
Angular error is bigger, punishes rule more serious and bigger towards angle error value, that penalty curve is steeper.The present invention according to
It is the cosine function in trigonometric function that the above punishment rule, which chooses penalty, i.e. f (| Δ φ |)=cos (| Δ φ |)；
5.3 combine step (4) in be calculated Iou value as a result, obtain towards angular error punish after evaluation index value,
It is calculated by following formula:
Aep=Iouf (| Δ φ |)
According to the characteristic of the penalty, it will thus be seen that when | Δ φ |=0, f (| Δ φ |)=1, there is Aep=Iou at this time,
I.e. when estimation it is oval towards error be zero, that is to say, that it is bright when there is no towards error, without punishment.When | Δ φ |=π/
2, f (| Δ φ |)=0, there are Aep=0, and estimation ellipse at this time has reached maximum towards error, illustrate estimation group/extension
The direction of motion of target and the complete mistake in actual direction, in cross-shaped, by index value zero setting to react estimating for the mistake
Count filter result.
4. multigroup according to claim 1/extension Target ellipse shape estimates evaluation method, which is characterized in that step (6)
Described according to judgement, determine in each unit gesture estimation mistake punishment after Performance Evaluating Indexes numerical value, specifically include as
Lower step:
(6a) was judged whether there is to estimate or leak and is estimated situation according to the matching result in each unit；If so, thening follow the steps
(6b) does further judgement, otherwise, executes step (6e): judging whether there is to estimate or leak and estimates situation method and be: in the unit
True group/extension ellipse target shape when being only matched to 1 group of elliptical shape estimation parameter, then it is assumed that estimate or leak there is no crossing
Estimate situation；Otherwise it is assumed that this true group/extension target, which existed to estimate or leak, estimates situation；
(6b) was determined by judgement to be estimated or leakage is estimated: when in the unit a true group/extension object matching to 2 or more
When group elliptical shape estimation parameter, then it is assumed that the moment occurs crossing to the group/extension target state estimator filtering estimates situation；If one
A true group/extension target is not matched to any elliptical shape estimation parameter, then it is assumed that the moment is to the true group/extension mesh
There is leakage and estimates situation in mark estimation filtering；When occurred that estimate situation when, then follow the steps (6c), leakage occur when estimating situation,
Then follow the steps (6d):
(6c) carried out estimating punishment: do a mean operation using Aep evaluation index values all in the unit, use the result as
The multigroup/extension target shape estimation track algorithm under the scene for single group/extension target Cep value；
(6d) carries out leakage and estimates punishment: will directly handle single group/extension target Performance Evaluating Indexes value zero setting, i.e. Cep
=0；
(6e) does not do any punishment: the Cep=Aep exported at this time.
5. multigroup according to claim 1/extension Target ellipse shape estimates evaluation method, which is characterized in that step (7)
Described in the superiority and inferiority of moment Evaluation on Total Performance index value evaluation multigroup/extension Target ellipse shape estimation filter tracking
Property, is described in detail below: the Cep index value in all units acquired to step (6) is averaged calculating, and obtained result is just
It is the Evaluation on Total Performance index value of the moment multigroup/extension Target ellipse shape estimation track algorithm；By comparing NIS
The size of value and multiple moment are linked to be the smoothness of curve, analyze the tracking performance of the filter, obtain the filter and exist
Superiority-inferiority explanation under different situations.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811640647.5A CN109683150B (en) | 2018-12-29 | 2018-12-29 | Multi-group/expansion target elliptical shape estimation evaluation method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811640647.5A CN109683150B (en) | 2018-12-29 | 2018-12-29 | Multi-group/expansion target elliptical shape estimation evaluation method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109683150A true CN109683150A (en) | 2019-04-26 |
| CN109683150B CN109683150B (en) | 2023-04-07 |
Family
ID=66191361
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811640647.5A Active CN109683150B (en) | 2018-12-29 | 2018-12-29 | Multi-group/expansion target elliptical shape estimation evaluation method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109683150B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113485385A (en) * | 2021-07-13 | 2021-10-08 | 中国人民解放军战略支援部队信息工程大学 | UUV cluster formation configuration design method based on error ellipse |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2000075023A (en) * | 1998-08-27 | 2000-03-14 | Mitsubishi Electric Corp | Sensor group management device |
| CN105373667A (en) * | 2015-11-27 | 2016-03-02 | 西安交通大学 | Multi-group section perturbation method for uncertainty analysis of reactor physics calculation |
| CN107797106A (en) * | 2017-05-08 | 2018-03-13 | 南京航空航天大学 | A kind of PHD multiple target tracking smooth filtering methods of the unknown clutter estimations of acceleration EM |
-
2018
- 2018-12-29 CN CN201811640647.5A patent/CN109683150B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2000075023A (en) * | 1998-08-27 | 2000-03-14 | Mitsubishi Electric Corp | Sensor group management device |
| CN105373667A (en) * | 2015-11-27 | 2016-03-02 | 西安交通大学 | Multi-group section perturbation method for uncertainty analysis of reactor physics calculation |
| CN107797106A (en) * | 2017-05-08 | 2018-03-13 | 南京航空航天大学 | A kind of PHD multiple target tracking smooth filtering methods of the unknown clutter estimations of acceleration EM |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113485385A (en) * | 2021-07-13 | 2021-10-08 | 中国人民解放军战略支援部队信息工程大学 | UUV cluster formation configuration design method based on error ellipse |
| CN113485385B (en) * | 2021-07-13 | 2023-11-07 | 中国人民解放军战略支援部队信息工程大学 | UUV cluster formation configuration design method based on error ellipse |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109683150B (en) | 2023-04-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Stein et al. | Director's cut: Analysis and annotation of soccer matches | |
| CN112819303B (en) | Aircraft tracking efficiency evaluation method and system based on PCE proxy model | |
| CN103941233B (en) | The radar interval alternately radiation control method of tracking worked in coordination with by a kind of multi-platform main passive sensor | |
| CN107677997A (en) | Extension method for tracking target based on GLMB filtering and Gibbs samplings | |
| Venanzi et al. | Trust-based fusion of untrustworthy information in crowdsourcing applications | |
| CN104730537A (en) | Infrared/laser radar data fusion target tracking method based on multi-scale model | |
| CN109858526A (en) | Sensor-based multi-target track fusion method in a kind of target following | |
| CN115128597B (en) | Maneuvering target tracking method under non-Gaussian noise based on IMM-STEKF | |
| CN115877376A (en) | Millimeter-wave radar gesture recognition method and recognition system based on multi-head self-attention mechanism | |
| CN118330632B (en) | An integrated estimation method for the external shape and centroid state of radar cluster targets | |
| CN114445456B (en) | Data-driven intelligent maneuvering target tracking method and device based on partial model | |
| US7020046B1 (en) | System and method for target motion analysis with intelligent parameter evaluation plot | |
| CN101308206B (en) | Circumferential track mobile target tracking method under white noise background | |
| CN109683150A (en) | Multigroup/extension Target ellipse shape estimates evaluation method | |
| CN106443618A (en) | High-resolution radar track association method aided by standard deviation ellipse parameters | |
| CN107391289B (en) | Usability evaluation method for three-dimensional pen type interactive interface | |
| CN106772357B (en) | AI-PHD filter multi-object tracking method under signal-to-noise ratio unknown condition | |
| CN116738364B (en) | Multi-source sensor measuring fusion method based on angle correlation | |
| CN110031803B (en) | Fusion localization method of dual infrared sensors with random measurement noise | |
| CN118965475A (en) | A Maneuvering Target Cluster Tracking Method Based on Gaussian Process | |
| CN104330772B (en) | The bistatic location method of comprehensive trace formula UKF filtering algorithm based on multidirectional optimizing | |
| Xu et al. | Low scattering performance and scattering center characteristics of streamlined space target | |
| CN118962662B (en) | A 3D multi-extended target tracking method based on inter-target reference deviation state | |
| Yu et al. | A New Quantitative Evaluation Strategy for Deep Generated Sar Images | |
| Qiu et al. | Anti-Occlusion Visual Tracking Algorithm for UAVs with Multi-Feature Adaptive Fusion |
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 |