CN109683150B - Multi-group/expansion target elliptical shape estimation evaluation method - Google Patents

Multi-group/expansion target elliptical shape estimation evaluation method Download PDF

Info

Publication number
CN109683150B
CN109683150B CN201811640647.5A CN201811640647A CN109683150B CN 109683150 B CN109683150 B CN 109683150B CN 201811640647 A CN201811640647 A CN 201811640647A CN 109683150 B CN109683150 B CN 109683150B
Authority
CN
China
Prior art keywords
estimation
group
target
extended
elliptical shape
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
Application number
CN201811640647.5A
Other languages
Chinese (zh)
Other versions
CN109683150A (en
Inventor
宋骊平
岑汉
潘燕鹏
杨平
邹志彬
王菲菲
宋飞宇
柴嘉波
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xidian University
Original Assignee
Xidian University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Xidian University filed Critical Xidian University
Priority to CN201811640647.5A priority Critical patent/CN109683150B/en
Publication of CN109683150A publication Critical patent/CN109683150A/en
Application granted granted Critical
Publication of CN109683150B publication Critical patent/CN109683150B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
    • G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
    • G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
    • G01S7/415—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

The invention discloses a multi-group/extended target elliptical estimation evaluation method, which solves the problems that the existing performance evaluation system can not reflect the overestimation and underestimation of targets and the elliptical major axis orientation estimation error, and comprises the following implementation steps: acquiring measurement data; filtering the measured data to obtain an oval estimation; matching elliptical shape estimation; calculating an IOU value; obtaining performance evaluation index values after the orientation error of the long axis of the ellipse and the potential estimation error punishment; evaluating the performance of the filter according to the NIS value; and judging whether to receive new measurement data, if so, updating the moment, and continuing the performance evaluation, otherwise, ending the performance evaluation. The method comprises the steps of matching ellipse estimation, selecting a penalty function to penalize an ellipse long axis orientation estimation error, and penalizing an overestimation or underestimation condition to obtain a performance evaluation result. The invention has the advantages of quick response, high sensitivity and high precision, and can be used in the fields of target identification, battlefield monitoring, video monitoring, air traffic control and the like.

Description

Multi-group/expansion target elliptical shape estimation evaluation method
Technical Field
The invention belongs to the technical field of radar target tracking, relates to shape estimation matching processing of multi-group/extended target elliptical shape estimation, and particularly relates to a multi-group/extended target elliptical shape estimation filter evaluation method which is used for performance evaluation of radar target tracking, group/extended target shape tracking and the like.
Background
In the conventional field of target tracking, the target is usually regarded as a point regardless of the target shape. As the resolution of sensors such as radar and infrared sensors is higher and higher, the acquired information related to the target is more and more, and when the sensors detect a plurality of quantities on a single target, the target must be treated as a group/expanded target. It is not possible to track only the target while estimating its shape. The shape tracking estimation of the group/extended target can be used in military fields such as target identification, missile defense and battlefield surveillance, and can also be used in civil fields such as video monitoring and air traffic control. In recent years, algorithms for estimating a group/spread target shape mainly include 3 random matrix (random matrix) methods, random hyper plane (random hyper plane) methods, and gaussian process (gaussian process) methods. In practice, it is often very challenging to extract detailed shape information from metrology data collected from sensors due to noise interference in the detection environment, so in many practical crowd/extended target tracking applications, one tends to focus on simple shapes, for which most current crowd/extended target tracking methods model the target shape using an ellipse, which also provides relevant additional information about the direction of target motion, diffusion range, and size. With the continuous improvement and innovation of the group/extended target tracking filter algorithm, an effective group/extended target shape tracking filter algorithm evaluation index system is objectively needed.
Zhang Hui et al, in a "group target gaussian mixture PHD filter based on an elliptical random hyper-curved model", because no effective evaluation index exists at that time to measure the estimation performance of the algorithm on the target shape, the OSPA distances between the centroid position, the ellipse major and minor half axes, and the area estimation value and the true value are calculated respectively, thereby realizing the evaluation on the filter performance.
Wang Xue et al, in the "extended target SMC-PHD filtering based on star-convex RHM", an Intersection-over-Intersection (IOU) value is used as a performance evaluation index of a shape estimation algorithm, the IOU performance evaluation index is a shape-related performance evaluation index, and the superiority and inferiority of the shape estimation algorithm in shape estimation are described by the ratio of Intersection and phase-to-phase areas between a feedback estimation shape and a real shape.
However, for the multi-group/extended-target elliptical shape estimation filter, the OSPA distance evaluation index cannot provide a performance evaluation index of the filter on shape correlation in elliptical shape estimation, so that the superiority and inferiority of the elliptical shape estimation tracking filter algorithm on shape estimation cannot be evaluated, and the evaluation index is not fully applicable to the elliptical shape estimation filter. The IOU performance evaluation index is applied to the field of multi-cluster/extended target tracking, and due to the reasons of tracking scenes and the requirement of tracking purposes, the IOU performance evaluation index does not reflect the over-estimation and under-estimation conditions in potential estimation errors occurring in the tracking process, and does not reflect the elliptical shape estimation elliptical orientation errors, so that the IOU performance evaluation index cannot be directly applied to elliptical shape estimation evaluation of multi-cluster/extended targets.
Disclosure of Invention
In view of the above-mentioned deficiencies of the prior art, the present invention proposes a multi-group/extended target elliptical shape estimation filter evaluation method capable of reflecting over-estimation and under-estimation in potential estimation errors and elliptical orientation errors.
The invention discloses a multi-group/expansion target elliptical estimation filter evaluation method, which is characterized by comprising the following steps of:
(1) Acquiring measurement data of a radar sensor in real time;
(2) Filtering the measured data of the radar sensor to obtain multi-group/extended target elliptical estimation parameters: inputting measured data obtained by a radar sensor into a multi-cluster/extended target elliptical shape estimation filter to be evaluated to perform multi-cluster/extended target shape estimation tracking filtering to obtain multi-cluster/extended target elliptical shape estimation parameters, and mainly comprising the following steps of: the center coordinates of the ellipses, the lengths of the long and short shafts, the orientations of the long shafts of the ellipses, the motion speed and the acceleration are measured;
(3) Multi-cluster/extended target elliptical shape estimation parameter matching: dividing the elliptical shape estimation parameters into a plurality of units by using a distance division method, wherein each unit has a corresponding elliptical shape estimation matching result, and the matching result in each unit comprises a real group/extended target elliptical shape parameter and all corresponding multi-group/extended target elliptical shape estimation parameters;
(4) Calculating the IOU values between all multi-cluster/extended target elliptical shape estimates and the true cluster/extended elliptical targets within each cell: using the matching result to calculate the ratio of elliptical intersection and parallel area between the real group/expanded target and all the elliptical estimation of the multi-group/expanded target in the unit matched with the real group/expanded target, and obtaining the IOU performance evaluation index value of the shape estimation result in each unit;
(5) Calculating and estimating the orientation error of the long axis of the ellipse, and performing orientation error punishment to obtain a performance evaluation index value (Aep) after the orientation error punishment of the long axis of the ellipse: firstly, calculating the difference between the orientation parameters of the major axes of the ellipses in the multi-group/expansion target elliptical shape estimation parameters and the orientation of the major axes of the ellipses in the real group/expansion target elliptical shape parameters; then selecting a penalty function to make a corresponding numerical penalty, and estimating an ellipse long axis orientation error penalty (Aep)), and finally obtaining a performance evaluation index numerical value (Aep);
(6) According to the judgment, obtaining a potential estimation error punishment post-performance evaluation index value (Cep) in each unit: judging whether the over-estimation or the underestimation exists in each unit, and simultaneously obtaining the performance evaluation index value (Cep) of the over-estimation or the underestimation on the basis of considering the shape estimation;
(7) And (3) evaluating the advantages and disadvantages of the multi-group/extended target elliptical shape estimation filter tracking by using the total performance evaluation index value (NIS) at the moment: extracting a performance evaluation index value (New index system (NIS)) of all units at the moment, and using the performance evaluation index value to explain the tracking performance of the multi-cluster/extended target shape estimation filter;
(8) Judging whether the radar sensor receives new measurement data, if so, updating the time, returning to the step (2), circularly tracking in real time, and otherwise, executing the step (9);
(9) And finishing the multi-group/extended target tracking and finishing the evaluation of the tracking performance of the filter.
The invention mainly divides units by a multi-group/extended target elliptical shape estimation matching method on the basis of the traditional IOU performance evaluation index, solves the problem that the orientation error, over-estimation or under-estimation of the elliptical long axis of a multi-group/extended target elliptical shape estimation filter cannot give exact evaluation in the existing evaluation index system, and finally obtains the total performance evaluation index value at the moment for explaining the superiority and inferiority of the filter after corresponding performance evaluation index value punishment and combination.
Compared with the prior art, the invention has the following advantages:
first, the most widely used Optimal Sub-pattern Assignment (OSPA) distance evaluation index is currently used, which measures the magnitude of the error between the real target and the estimation result calculated by calculating the distance between centroids, ignoring the evaluation contrast on the most important shape in the shape estimation filter; the invention improves the problem of the evaluation index related to the shape estimation in the current multi-group/extended target shape tracking estimation field, not only realizes the establishment of the evaluation index system in the shape estimation shape aspect, but also realizes the evaluation of the estimation error of other parameters.
Secondly, the IOU performance evaluation index is widely applied to an evaluation system in the field of object detection and image segmentation, and is also used in the field of multi-cluster/extended target shape tracking estimation for shape estimation evaluation at present, but in the field of multi-cluster/extended target shape tracking estimation, the error of the ellipse orientation and the overestimation and underestimation conditions of multi-cluster/extended target potential estimation cannot be reflected, so that the invention perfects the problem of the shape estimation performance evaluation index on the basis of the IOU performance evaluation index, solves the problem of the influence of the overestimation and underestimation problems common in the field of multi-cluster/extended target shape estimation and tracking on the evaluation index of an elliptical filter, and adds the influence of the common ellipse orientation error in the elliptical shape estimation on the evaluation index of the multi-cluster/extended target shape tracking filter estimation method for the first time.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a diagram of true elliptical shapes and motions of multiple clusters/extended targets in a simulation application scenario of the present invention;
FIG. 3 is a graph of the filtering estimation result after the SMC-RHM-PHD filter is used in the present invention;
fig. 4 is a comparison graph of the performance evaluation indexes NIS and IOU according to the present invention.
Detailed Description
The present invention is described in detail below with reference to the attached drawings.
Example 1
In recent years, with the proposal of a large number of multi-group/extended target shape estimation filtering algorithms, broad learners continuously improve and innovate a multi-group/extended target elliptical shape estimation tracking filter, but in the aspect of the evaluation indexes of superiority and inferiority, OSPA distance evaluation indexes are widely applied to superiority and inferiority evaluation at present, and because the performance evaluation indexes of the filter on shape estimation cannot be given, the superiority and inferiority of the filter actually applied to the shape estimation direction cannot be evaluated, the evaluation indexes are not completely suitable for the evaluation indexes of the multi-group/extended target elliptical shape estimation tracking filter; the IOU performance evaluation index cannot reflect the overestimation and underestimation conditions in the potential estimation error and the elliptical shape estimation ellipse orientation error in the tracking process due to the fact that the IOU performance evaluation index is higher and higher in accuracy requirement on information acquisition such as the position and the number of the target due to the tracking scene and the requirement of the tracking purpose.
The invention provides a multi-group/extended target elliptical estimation filter evaluation method to meet the actual requirement. Referring to fig. 1, the present invention performs a performance evaluation on a multi-group/extended target elliptical shape estimation tracking filter, and calculates a performance evaluation index value by using a multi-group/extended target elliptical shape estimation parameter obtained after filtering, so as to explain the filtering tracking performance evaluation, including the following steps:
(1) And acquiring measurement data of the radar sensor in real time, wherein the measurement data comprise two-dimensional space position coordinate information and are data acquired at the initial moment. The ellipse shape of the real group/expanded ellipse target and the total number of the real group/expanded ellipse targets are obtained.
(2) Filtering the measured data of the radar sensor to obtain multi-group/extended target elliptical estimation parameters: inputting measured data obtained by a radar sensor into a multi-cluster/extended target elliptical estimation filter to be evaluated to perform multi-cluster/extended target elliptical estimation tracking filtering to obtain multi-cluster/extended target elliptical estimation parameters, wherein the estimation parameters comprise: and the circle center coordinates, the length of the long and short axes, the orientation of the long axis of the ellipse, the motion speed, the acceleration and the like of each elliptical estimation obtained after filtering processing. All the obtained parameters are obtained after filtering processing of a multi-group/expansion target elliptical shape estimation filter to be evaluated.
If one wants to use the present invention to evaluate a multi-constellation/extended target elliptical shape estimation filter, it is input with the measured data, and the filter is the multi-constellation/extended target elliptical shape estimation filter to be evaluated. The later mentioned multi-group/expansion target elliptical shape estimation is obtained after the filtering processing of the multi-group/expansion target elliptical shape estimation filter to be evaluated.
Wherein a multi-cluster/extended target refers to multiple moving targets that can be detected by sensors at the same time. The multi-group/extended target elliptical shape estimation is an estimation of a plurality of elliptical shapes obtained after filtering processing by a multi-group/extended target elliptical shape estimation filter.
(3) Multi-cluster/extended target elliptical shape estimation parameter matching: dividing the elliptical shape estimation parameters obtained after filtering in the step (2) into a plurality of units by using a distance division method and referring to the real group/extended elliptical target, wherein the total number of the divided units is equal to the total number of the real group/extended elliptical target, each unit is internally provided with a corresponding multi-group/extended elliptical shape estimation parameter estimation matching result, and the matching result in each unit comprises a real group/extended target elliptical shape parameter and all corresponding multi-group/extended target elliptical shape estimation parameters.
The real group/expanded target elliptical shape parameters include the center coordinates, the length of the long axis and the short axis, and the orientation of the long axis of the ellipse of the real group/expanded target diffusion.
(4) Calculate the IOU values between all the multi-cluster/extended target elliptical shape estimates and the true cluster/extended target elliptical shape within each cell: and (4) calculating the elliptical shape intersection and phase-to-area ratio between the real group/extended target and all the multiple group/extended target elliptical shape estimates in the unit matched with the real group/extended target by using the matching result mentioned in the step (3) to obtain multiple IOU performance evaluation index values in each unit, wherein the number of the IOU values in each unit is equal to the number of the multiple group/extended target elliptical shape estimates contained in the unit.
(5) Calculating and estimating the orientation error of the long axis of the ellipse, and performing orientation error punishment to obtain a performance evaluation index value (Aep) after the orientation error punishment of the long axis of the ellipse: traversing all the units, firstly, calculating the absolute value of the difference between the orientation parameter of the long axis of the ellipse contained in each multi-group/expansion target elliptical shape estimation parameter in each unit and the orientation of the long axis of the ellipse in the real group/expansion target elliptical shape parameter of each unit; then selecting a penalty function to make corresponding numerical penalty, namely, estimating the ellipse long axis orientation error penalty (Aep)), according to the property of the ellipse orientation error value, the penalty function must be monotonously decreased in a value interval of [0 pi/2 ], the slope of the penalty function is also decreased along with the increase of the ellipse orientation error value, namely, the penalty function curve is steeper. And finally, outputting the estimated ellipse major axis orientation error punishment performance evaluation index value (Aep) of each unit.
(6) According to the judgment, obtaining a potential estimation error punishment post-performance evaluation index value (Cep) in each unit: judging whether the over-estimation or the under-estimation exists in each unit, and according to different judgment results, if the over-estimation or the under-estimation does not exist, performing no punishment processing in the step; and if so, carrying out punishment processing corresponding to over-estimation or under-estimation, and finally obtaining a potential estimation error punishment post-performance evaluation index value (Cep) on the basis of considering the shape estimation no matter what judgment result is.
(7) Evaluating the advantages and disadvantages of the multi-cluster/extended target elliptical shape estimation filter tracking by using the current time overall performance evaluation index value (NIS): and extracting a performance evaluation index value (New index system (NIS)) of all the unit totalities at the current moment for performing a superior-inferior explanation on the tracking performance of the multi-cluster/extended target shape estimation filter to be evaluated. The current time described here in the present invention is the initial time for the first time, and is the updated time every time thereafter.
(8) Judging whether the radar sensor receives new measurement data, if so, updating the time, returning to the step (2), circularly tracking in real time, and otherwise, executing the step (9): when the radar sensor receives the measured data, the observation area of the radar sensor still keeps moving continuously, the tracking time is updated, the steps (2) - (8) are circularly executed, and the performance evaluation updating is continuously carried out on the multi-group/extended target elliptical estimation tracking filter. Otherwise, when the measured data is not received, it indicates that the observation area has no target motion, and step (9) is executed.
(9) And finishing multi-group/extended target tracking and finishing evaluation of the shape tracking estimation performance of the filter.
The method comprises the steps of performing multi-group/extended target oval estimation on measurement data acquired by a radar sensor by using a multi-group/extended target oval estimation filter to be evaluated to acquire oval estimation parameters, dividing a real group/extended target and a filter estimation result into a plurality of units by estimation matching, calculating the ratio of intersection and phase area between the oval estimation parameters and the oval shape of the real group/extended target, namely IOU (input/output), calculating the included angle difference between the orientation of the oval estimation long axis and the orientation of the oval long axis of the real target, performing shape estimation orientation estimation error penalty by using a penalty function, judging whether potential estimation errors exist or not according to the estimation matching result, mainly judging whether the problems of over-estimation or under-estimation exist or not according to the estimation matching result, performing corresponding operation according to the judgment result, and finally obtaining a time overall performance evaluation index value (NIS) for evaluating the tracking performance of the multi-group/extended target oval estimation filter to be evaluated.
Example 2
The multi-group/extended target elliptical shape estimation filter evaluation method is the same as that in embodiment 1, and the multi-group/extended target elliptical shape estimation parameter matching in step (3) is mainly based on a distance division method, and includes the following steps:
3.1 calculating Euclidean distance between every two circle centers of the elliptical shape according to all the elliptical shape estimation parameters of the multi-group/extended target and all the elliptical shape parameters of the real group/extended target.
3.2 matching the set of multi-group/extended target elliptical shape estimation parameters with the minimum euclidean distance with the true group/extended target elliptical shape parameters, i.e. indicating that the set of multi-group/extended target elliptical shape estimation parameters is the filtered estimation of the true group/extended target elliptical shape parameters.
3.3 one true group/expanded target elliptical shape parameter and all elliptical shape estimates matching it are grouped into the same cell. In this example, how many sets of the real group/extended target elliptical shape parameters are divided into how many cells, and each cell is independent of each other.
The multi-group/extended target elliptical shape estimation parameter matching method is applied to multi-group/extended target elliptical shape estimation parameter matching, and can perform unit division more accurately and accurately, so that performance evaluation indexes obtained later are more real and effective.
Example 3
The method for evaluating the multi-group/extended target elliptical shape estimation filter is the same as that of embodiment 1-2, and the method for calculating the IOU values between all the multi-group/extended target elliptical shape estimates and the real group/extended target elliptical shape in each cell described in step (4) is a monte carlo method, and the detailed description is as follows:
using a e b e ]Respectively represent the lengths of the major axis and the minor axis in the multi-group/extended target elliptical shape estimation parameters, [ a ] r b r ]Respectively representing the lengths of the long axis and the short axis in the real group/extended target elliptical shape parameter in the unit; the midpoint of the circle center connecting line between the multi-group/expanded target elliptical shape estimated elliptical shape and the real group/expanded target elliptical shape is taken as the centroid,in this example, the long axis of the 4 times real group/extended target elliptical shape parameter is used as the side length, i.e. 4a r Making a rectangular area, generating ten thousand points in the area according to uniform distribution, and counting N total points which belong to both a multi-group/extended target elliptical shape estimation ellipse and a real group/extended target elliptical shape in the ten thousand points; at this time, the elliptical shape intersection area S between the real group/expanded target and the multiple group/expanded target elliptical shape estimation ∩ The following formula is obtained:
Figure GDA0003874994870000081
then the corresponding phase and area S ∪ Can be obtained by the following formula:
S ∪ ＝π·a r ·b r +π·a e ·b e -S ∩
the IOU value can be obtained by the following equation:
Figure GDA0003874994870000082
the Monte Carlo method is selected for solving the IOU value, because the method can quickly count and obtain related data, and the IOU value can be efficiently calculated through the data; certainly, the more the number of points uniformly distributed in the rectangular area is selected, the closer the obtained IOU value is to the true value, but the calculation complexity is increased, the calculation time is prolonged, and the real-time performance is affected. After a large number of experiments, when ten thousand points are used, the side length of the rectangular area is 4a r The IOU value can be rapidly and accurately obtained, so that the value is the optimal value of the embodiment, and different values can be carried out according to specific requirements in specific implementation. Compared with a method for calculating the IOU value through an image processing method after two related ellipses are plotted and drawn, the Monte Carlo method used by the invention has the characteristics of simplicity and convenience and efficiency.
Example 4
The multi-cluster/extended target elliptical shape estimation filter evaluation method is the same as that in the embodiment 1-3, the method for calculating and estimating the elliptical long axis orientation error and performing orientation error punishment in the step (5) to obtain a performance evaluation index value (Aep) after the elliptical long axis orientation error punishment comprises the following steps:
5.1 according to the matching result in each unit in the step (3), respectively estimating the parameter value of the long axis orientation of the ellipse in the parameter through each group of multi-group/extended target elliptical shapes, and calculating to obtain the absolute value of the difference between the angle of the long axis orientation of the ellipse and the angle of the long axis orientation of the true group/extended target elliptical shape in the unit, which is marked as | Δ φ |, and | Δ φ ∈ [0 π/2].
And 5.2, selecting a penalty function f (| delta phi |) to penalty the orientation angle error. According to the rule that the penalty function must satisfy the condition that the angle error is larger, the penalty is more serious, the angle error value is larger, the penalty function curve is steeper, and the penalty strength is larger. The penalty function selected by the invention is a cosine function in a trigonometric function as the penalty function, namely f (| delta phi |) = cos (| delta phi |).
5.3 combining the result of Iou calculated in step (4), obtaining an evaluation index value (Angle error penalty (Aep)) after an Angle error penalty, and calculating by the following formula:
Aep＝Iou·f(|Δφ|)＝Iou·cos(|Δφ|)
after the characteristics of the estimated ellipse orientation angle error are known, the cosine function in the trigonometric function is selected as the penalty function according to the established penalty rule, because the cosine function is a monotone decreasing function in the value space of [0 pi/2 ], and the two-dimensional curve meets the characteristic that the curve is steeper along with the larger angle error value, and the two characteristics are completely matched with the established penalty rule; and it can be seen that: when | Δ Φ | =0,f (| Δ Φ |) =1, there is Aep = Iou, i.e., no penalty is made when the estimated ellipse orientation error is zero, that is, there is no estimated ellipse orientation error. When | Δ Φ | = pi/2,f (| Δ Φ |) =0, aep =0 exists, the estimated ellipse orientation error at this time is the maximum, which means that the motion direction of the estimated group/extended target is completely wrong with the actual direction and is crossed, therefore, the performance evaluation index value is set to zero to reflect the wrong estimation result. No matter the prior IOU performance evaluation index or OSPA distance evaluation index is adopted, the evaluation of the ellipse orientation error of the estimation result after the filtering processing of the multi-group/extended target ellipse shape estimation filter is never carried out, but in the field of multi-group/extended target tracking, the ellipse orientation is often the motion direction of the multi-group/extended target.
Example 5
The multi-group/extended target elliptical shape estimation filter evaluation method is the same as that in the embodiment 1-4, and the performance evaluation index value Cep after the potential estimation error penalty in each unit is obtained according to the judgment in the step (6), and the method specifically comprises the following steps:
(6a) Judging whether the overestimation or underestimation condition exists for each unit according to the matching result in each unit in the step (3): the method for judging whether the overestimation or underestimation condition exists comprises the following steps: and (5) when the real group/expansion target in each unit is only matched with 1 group of multi-group/expansion target elliptical shape estimation parameters, determining that no over-estimation or under-estimation condition exists, executing the step (6 e), and directly outputting. Otherwise, if the situation of overestimation or underestimation occurs in the filtering process of the multi-group/extended target elliptical estimation filter on the real group/extended target estimation, the step (6 b) is executed to make further judgment.
(6b) Separate handling of the presence of overestimation or underestimation: judging whether the over-estimation or the under-estimation is determined: when the real group/expansion target in the unit is matched with 2 or more groups of multi-group/expansion target elliptical shape estimation parameters, considering that the filter carries out the overestimation on the real group/expansion target estimation filtering processing at the moment, and executing (6 c); when a real group/expansion target is not matched with any elliptical estimation result, considering that the filtering processing of the group/expansion target estimation at the moment of the filter has a missing estimation condition, and executing (6 d);
(6c) And (4) performing over-estimation punishment, and outputting a Cep value: performing an average operation on all the Aep evaluation index values in the unit obtained in the step (5), and using the result as a Cep value of the multi-cluster/extended target shape estimation filter to the real cluster/extended target in the unit under the scene; the reason why the mean calculation result is used as the performance evaluation index value after the overestimation penalty of the real group/extended target at the time is that under the overestimation condition, the real group/extended target in each unit always has 2 or more group shape estimation results, so that through the calculation of the step (5), the same Aep performance evaluation index value as the number of the multi-group/extended target elliptical shape estimation values contained in the unit can be obtained, therefore, the use of a better performance evaluation index value or a biased performance evaluation index value is not suitable for being used as a filter for estimating the filtering tracking performance evaluation index value of the real group/extended target in the unit, while the value obtained by the mean calculation is a good concentration number, and has the advantages of sensitive reaction, determination, simplicity and easy solution, simple calculation, suitability for further calculation, small influence of sampling change and the like, and the mean calculation reaction is sensitive, and large or small change of each data can influence the final result, so that the mean of all Aep values in the unit is used for describing the selection of the filter for the overestimation performance evaluation index value under the better condition of the real group/extended target in the overestimation condition.
(6d) Performing underestimation punishment, and outputting the Cep =0, namely directly setting the performance evaluation index value of the single group/extension target to zero, namely the Cep =0; for the underestimation, in the traditional IOU performance evaluation index, if the underestimation condition occurs, the underestimation is often ignored, the invention solves the problem that the underestimation is ignored, and directly assigns zero to the underestimated group/extended target performance evaluation index value.
(6e) Without any penalty, the output at this time, cep = Aep.
The method aims to reflect the influence of overestimation and underestimation in potential estimation errors in the field of multi-group/extended target shape tracking on performance evaluation indexes in different degrees. Compared with the traditional IOU performance evaluation index value, the method solves the problem how to reflect the influence of overestimation and underestimation on the shape estimation performance evaluation, realizes different form punishment modes under the overestimation condition and the underestimation condition through the method so as to obtain the respective performance evaluation index values under the overestimation condition and the underestimation condition, and can make practical evaluation on the multi-group/extended target elliptical shape estimation filter through the value.
Example 6
The method for evaluating the multi-group/extended target elliptical shape estimation filter is the same as that in embodiments 1-5, and the step (7) of evaluating the tracking advantages and disadvantages of the multi-group/extended target elliptical shape estimation filter by using the total performance evaluation index value (NIS) at that time is specifically described as follows: and (4) carrying out averaging calculation on the Cep index values in all the cells obtained in the step (6), wherein the obtained result is the overall performance evaluation index value of the multi-group/extended target elliptical shape estimation tracking algorithm at the moment. In order to embody the overall performance evaluation index value at the moment, the invention explains the tracking performance of the multi-cluster/extended target shape estimation filter under the scene at the moment through the overall performance evaluation index NIS value, can explain which tracking scene the filter is more suitable for by comparing the sizes of the NIS values obtained by the same multi-cluster/extended target shape estimation filter under different tracking scenes of the multi-cluster/extended target at different moments, and can judge which multi-cluster/extended target shape estimation filter is more suitable for tracking the scene by comparing the sizes of the NIS values obtained by different multi-cluster/extended target shape estimation filters under the same multi-cluster/extended target tracking scene. The method performs mean operation on all the Cep values obtained in all the units, can reflect the performance evaluation index value trend of all the groups/extended targets after shape estimation filtering tracking at the moment by using the mean operation, and can judge whether the filter has underestimation at the moment according to whether the performance evaluation index curve suddenly drops greatly, because the Cep value under the underestimation condition is set to be zero in the underestimation punishment processing in the step (6 d) of the method, the condition that the mean value is low due to underestimation occurs.
A more detailed example is given below to further illustrate the invention
Example 7
The method for evaluating the multi-group/extended target elliptical shape estimation filter is the same as that of embodiments 1-6, and the specific steps of the present invention are further described with reference to fig. 1.
Step 1, obtaining relevant measurement data of a radar sensor.
When the multiple swarm/extended targets start to move in the observation area of the radar sensor, the radar sensor receives the multiple swarm/extended target related measurement and records the initial time, and the time k =1.
Step 2, constructing multi-group/extended target elliptical shape estimation: according to the measured data obtained by the radar sensor, an elliptical estimation filter to be evaluated is selected to perform multi-cluster/extended target shape estimation tracking filtering, and a filtering estimation result, namely multi-cluster/extended target elliptical estimation, is output.
The multi-cluster/extended target shape estimation tracking filter selected for use in this example is an elliptical stochastic hyper-curved (RHM) based particle probability hypothesis density (SMC-PHD) filter. The filter describes the size of a group/an extended target through a long axis and a short axis, describes an ellipse of the extended target in the motion direction through the long axis, approximates the distribution of scale factors of an ellipse random hypersurface by Gaussian distribution, carries out filtering tracking on a plurality of groups/extended targets by using an SMC-PHD filtering method, and carries out filtering processing on measured data received by a radar sensor through the filter, so that the elliptical estimation parameters of the plurality of groups/extended targets in the tracking scene at the moment k can be obtained. The ellipse estimation parameters mainly comprise the circle center coordinates, the length of the long axis and the short axis, the ellipse long axis direction, the motion speed and the acceleration of each group/expansion target.
And 3, matching the real multi-group/expanded elliptical target with the multi-group/expanded elliptical target elliptical estimation parameters, and outputting a matching result.
Matching of the real multi-group/extended elliptical target and the multi-group/extended elliptical target elliptical shape estimation is achieved by using a distance division method, referring to fig. 1, the multi-group/extended elliptical target elliptical shape estimation obtained after filtering is divided into N units, and the size of N is equal to the number of real group/extended elliptical target motions in a radar observation area. Calculating Euclidean distances from the circle centers of all the elliptical shape estimation results in the multi-group/extended target elliptical shape estimation obtained in the step 2 to the circle centers of all the real multi-group/extended elliptical targets, matching the elliptical shape estimation with the minimum Euclidean distance and the real group/extended elliptical targets into the same unit, and merging different estimation results matched with the same real group/extended elliptical target into the same unit, namely, indicating that the elliptical shape estimation results are filtering estimation of the real group/extended elliptical target; different real group/expanded ellipse targets and different estimation results matched with the real group/expanded ellipse targets are merged into different units, and the units are independent from each other and do not influence each other.
As used herein, a true group/extended ellipse target refers to a true moving target that uses an ellipse to model the diffusion shape of a single group/extended target. The true multi-cluster/extended elliptical target refers to an overall moving target cluster formed by a plurality of true clusters/extended elliptical targets.
And 4, calculating the IOU value between all the multi-group/extended target elliptical shape estimation in the same unit and the real group/extended elliptical target in the unit.
And 3, calculating the ratio of the elliptic shape intersection and the parallel area between the elliptic shape estimation of the multiple groups/extended targets in each unit and the real groups/extended elliptic targets matched with the elliptic shape estimation of the multiple groups/extended targets by using all matching results in the divided units to obtain the IOU performance evaluation index value of the shape estimation tracking filtering of each unit. And obtaining a corresponding number of IOU values according to the number of filtering estimation results in each unit at the moment k.
And 5, calculating the orientation error of the estimated ellipse long axis in the same unit, punishing the orientation error of the ellipse long axis, and outputting a performance evaluation index value (Aep) obtained after punishing the orientation error of the ellipse long axis at the moment k.
And (4) according to the matching result in the step (3), calculating to obtain an absolute value of an angle difference value between the orientation of the long axis of the ellipse estimated by the ellipse and the orientation of the real group/extended ellipse target through the parameter value of the orientation of the ellipse in the ellipse estimation, namely estimating the orientation error of the long axis of the ellipse, and marking as | delta phi | which belongs to [0 pi/2 ]. And selecting a penalty function to penalize the orientation error of the long axis of the ellipse. According to the characteristics that the penalty function must be monotonously decreased within a value interval [0 pi/2 ], and the larger the angle error is, the more serious the penalty is, and the larger the angle error value is, the smaller the slope is, and the larger the penalty value change rate is. The selected penalty function is a cosine function in a trigonometric function as a penalty function, namely f (| delta phi |) = cos (| delta phi |). Then, the result of the IOU value calculated in step 4 is combined to obtain an evaluation index value (Angle error penalty (Aep)) with the penalty of the error of the ellipse major axis orientation. Aep is calculated as follows:
Aep＝Iou·f(|Δφ|)＝Iou·cos(|Δφ|)
the method considers the problem of error of the orientation of the long axis of the ellipse estimated in the conventional IOU performance evaluation index, and because the orientation of the long axis of the ellipse of the real multi-group/extended elliptical target is often the moving direction of the target, the method provides an important credibility reference for judging the accuracy of the moving intention of the multi-group/extended target, and has important practical significance for judging strategic intentions such as target attack, cruise or retreat and the like in the military field.
Step 6, judging whether the situation of overestimation or underestimation exists, and if so, carrying out corresponding punishment of overestimation or underestimation; otherwise, without any penalty, outputting a performance evaluation index value (Cep) under the condition of overestimation or underestimation in the potential estimation error considering the aspect of the shape:
judging whether the overestimation or underestimation exists according to the matching result in the dividing unit in the step 3, wherein when the real group/expanded elliptical target in the unit is only matched with 1 elliptical estimation result, the overestimation or underestimation does not exist; otherwise, considering the k moment, the filter tracks and filters the estimation of the real group/extended target in the unit to generate the over-estimation or under-estimation condition; the method for judging the overestimation or underestimation condition and the corresponding punishment method are as follows: when a real group/expansion target is matched with more than 2 estimation results, the situation that the group/expansion target is over-estimated at the moment is considered to occur, the over-estimation penalty adopted by the invention is that for a group/expansion target, the Aep evaluation index values obtained by the matched estimation results through the step 5 are subjected to mean operation, and the result is used as the performance evaluation index value of the multi-group/expansion target shape estimation tracking algorithm for the single group/expansion target in the scene. When a real group/extension target does not have any matching estimation, the situation of missing estimation of the group/extension target at the moment is considered to occur, and the missing estimation punishment adopted by the invention is to zero-set the performance evaluation index value of the single group/extension target. And 6, recording the performance evaluation index value obtained in the step 6 after the potential estimation error punishment as Cep.
And 7, extracting the overall performance evaluation index values in all the units at the time k, and performing performance evaluation on the multi-group/expansion target elliptical shape estimation filter.
For multiple groups/extended targets, the group/extended targets in each unit can obtain a performance evaluation index value Cep subjected to potential estimation error punishment after the step 6. Finally, it is required to explain the overall filter performance evaluation index value at the time to reflect the superiority and inferiority of the filter performance of the multi-cluster/extended target shape estimation and tracking algorithm at the time. Extracting the final performance evaluation index value of the k-time totality, namely averaging the Cep value obtained by each group/extended target, and recording the obtained result as the totality performance evaluation index value of the multi-group/extended target shape estimation tracking algorithm at the time as NIS; the output NIS value is used for comparing the advantages and disadvantages of the multi-cluster/extended target shape estimation tracking filter.
The method can be used for unmanned aerial vehicle group tracking, aircraft carrier group tracking and the like, and can determine the reliability of the tracking performance of the filter in the tracking scene according to the passing NIS value of the filter in the tracking scene as a reference.
And 8, judging whether the radar sensor receives new measurement data, if so, adding 1 to the moment k, and then circularly executing the steps 2-8 to continuously evaluate the performance of the multi-group/extended target elliptical estimation tracking filter. Otherwise, that is, the sensor does not receive the new measurement data, which indicates that there is no target motion in the observation area, and step 9 is executed.
And 9, finishing multi-group/extended target tracking and finishing evaluation of the shape tracking estimation performance of the filter.
The IOU performance evaluation index is widely applied to an evaluation system in the field of object detection and image segmentation, is also used in the field of multi-cluster/extended target shape tracking estimation to serve as a shape estimation evaluation index at present, but in the evaluation index of a multi-cluster/extended target elliptical shape tracking estimation filter, the error of the elliptical orientation and the overestimation and underestimation conditions of multi-cluster/extended target potential estimation cannot be reflected, so that the invention solves the problem of imperfect elliptical shape estimation index on the basis of the IOU performance evaluation index, solves the problem of influence of common overestimation and underestimation problems in the field of multi-cluster/extended target shape estimation tracking on the elliptical shape estimation filter evaluation index, and adds the influence of common elliptical shape orientation error in elliptical shape estimation on the evaluation index of the multi-cluster/extended target shape tracking filter estimation method for the first time.
The effect of the present invention will be further explained with the simulation experiment.
Example 8
The multi-group/extended target elliptical shape estimation filter evaluation method is the same as in examples 1 to 7,
simulation experiment conditions are as follows:
the hardware test platform of the simulation experiment of the invention is as follows: the Intel Core i3-7100 CPU of the processor has a main frequency of 3.90GHz and a memory of 4GB; the software platform is as follows: windows 7 flagship edition, 64-bit operating system, MATLAB R2010b.
Simulation content:
the simulation scene of the invention is a multi-group/extended target with the tracking number changing with time in a two-dimensional tracking scene, and the observation area of the simulation is [ -5 × 10 [) 4 ,8×10 4 ]×[-5×10 4 ,5×10 4 ]In meters (m), sampling period T = 1(s), measurement noise v k Standard deviation of (a) x ＝σ y =500m. There are a maximum of 8 cluster/extended objects within the observation area and all cluster/extended objects move a total of 30s within the observation area. Eyes of a userThe number of clutter in the tracking area where the target is located follows Poisson distribution with the average value of 3. The total number of start tracking time multi-cluster/extended targets is 1, and the total number of randomly sampled particles N =1000. Where at 11s, cluster splitting occurs. At 23s, cluster merging occurs, and the overall trajectory and true elliptical diffusion shape are shown in fig. 2. The filtering was performed using an elliptic Random Hypersurface (RHM) based particle PHD (SMC-PHD) filter, the filtering results are shown in fig. 3.
And (3) simulation result analysis:
fig. 2 is a diagram of true elliptical shapes and motion of multiple cluster/extended targets in a simulation application scene, where all the elliptical shapes independently exist are positions where the true multiple cluster/extended target diffusion shapes actually pass through within 30s of motion in an observation area of a radar sensor. Namely, the real multi-cluster/extended target map of the invention, the rectangular area formed by the abscissa and the ordinate in the map is the observation area detected by the radar.
Fig. 3 is a diagram of a result of filtering estimation and tracking using an SMC-RHM-PHD filter in this example, that is, a diagram of an ellipse estimation of multiple clusters/extended targets, fig. 3 (a) is a diagram of a result of filtering estimation and tracking of the overall filtering estimation and fig. 3 (b) is a diagram of a result of filtering estimation and tracking of a local enlargement, in order to more intuitively reflect an error existing between an ellipse diffusion shape of a true multiple cluster/extended target and an ellipse estimation of an ellipse, in the diagram, a solid line ellipse is a position where the true multiple cluster/extended target is located and a true diffusion ellipse, in the diagram, a dotted line ellipse is a position where an ellipse estimation of an ellipse obtained by performing ellipse estimation, filtering and tracking through an SMC-RHM-PHD filter corresponds to an ellipse estimation in an observation area, and x is a position where measurement data received by a radar sensor from the multiple clusters/extended target, and also from the radar sensor itself and noise existing in the observation area are two-dimensional space coordinates. In the figure, although errors always exist in the filter estimation result, the macroscopic errors are small and the errors at each moment are random and inconsistent, so that the advantages and the disadvantages of the tracking performance of the filter cannot be distinguished by naked eyes.
The NIS performance evaluation index of the present invention in the tracking scenario is shown in fig. 4, and fig. 4 is a comparison diagram of the performance evaluation index NIS and the IOU performance evaluation index provided by the present invention.
Comparing the two curves in fig. 4, the two curves are obtained from the tracking scenarios and the filter filtering results shown in fig. 2 and 3: when the NIS performance index of the invention is suddenly and greatly reduced and is quickly recovered at a certain later moment, the situation that the filter performs filtering tracking on the shape estimation of the target at the moment to avoid estimation missing can be judged; due to the new generation of the target at the 3 rd, 5 th and 8 th moments, the filter tends to generate a moment of missing estimation on the new target due to the characteristics of the filter, and the traditional IOU performance evaluation index does not react to the missing estimation situation at the 3 rd, 5 th and 8 th moments of the abscissa, because the traditional IOU performance evaluation index cannot react to the overestimation and the missing estimation situation of the shape estimation filter in the potential estimation error, and the NIS performance evaluation index value of the invention is inversely observed, and the NIS performance evaluation index value is greatly reduced at a moment to reflect the missing estimation situation. In the time interval between the 9 th and 14 th time points on the abscissa, the NIS performance index value proposed by the present invention is basically consistent with the IOU, because the number of the multi-cluster/extended targets in the time interval is small, and the performance of the filter is better in the tracking scene. However, in the time period between the 14 th time and the 19 th time, a large estimation error occurs in the elliptical shape estimation filter due to the rapid increase of the number of the multiple groups/expansion targets in the observation area in the time period, and the error mainly comes from the elliptical orientation estimation error, but the IOU performance evaluation index value in the time period is approximately equal to the IOU value trend in the time period between the 9 th time and the 14 th time, because the IOU performance evaluation index cannot reflect the elliptical major axis orientation error in the elliptical shape estimation. Therefore, it can be known that the filtering estimation result of the SMC-RHM-PHD filter with multiple clusters/extended targets has a relatively serious ellipse orientation error when the number of clusters/extended targets is too large.
In short, the multi-cluster/extended target elliptical shape estimation evaluation method disclosed by the invention solves the technical problems that the existing performance evaluation index system cannot reflect overestimation and underestimation in target estimation and ellipse long axis orientation estimation errors, and the implementation steps comprise (1) measurement data acquisition; (2) The filter filters the measured data to obtain an elliptical estimation; (3) elliptical shape estimate matching; (4) calculating an IOU value; (5) Obtaining a performance evaluation index value (Aep) after the ellipse major axis orientation error punishment; (6) Obtaining a performance evaluation index value (Cep) after potential estimation error punishment according to judgment; (7) evaluating the performance of the filter according to the NIS value; (8) And judging whether the sensor receives the measured data, if so, updating the moment, and continuing the performance evaluation of the filter, otherwise, ending the performance evaluation of the filter. The invention considers the ellipse long axis direction estimation error and the overestimation or underestimation condition, provides a solution method for punishing the ellipse long axis direction estimation error by elliptical shape estimation matching and selecting a punishment function, and then punishment is carried out by judging the overestimation or underestimation, so as to obtain a performance evaluation index value.

Claims (5)

1. A multi-group/expansion target elliptical shape estimation evaluation method is characterized by comprising the following steps:
(1) Acquiring measurement data of a radar sensor in real time;
(2) Filtering the measured data of the radar sensor to obtain multi-group/extended target elliptical estimation parameters: inputting measured data obtained by a radar sensor into a multi-cluster/extended target elliptical shape estimation filter to be evaluated to perform multi-cluster/extended target shape estimation tracking filtering to obtain multi-cluster/extended target elliptical shape estimation parameters, and mainly comprising the following steps of: the center coordinates of the ellipses, the lengths of the long and short shafts, the orientations of the long shafts of the ellipses, the motion speed and the acceleration are measured;
(3) Multi-cluster/extended target elliptical shape estimation parameter matching: dividing the elliptical shape estimation parameters into a plurality of units by using a distance division method, wherein each unit has a corresponding elliptical shape estimation matching result, and the matching result in each unit comprises a real group/extended target elliptical shape parameter and all corresponding multi-group/extended target elliptical shape estimation parameters;
(4) Calculating the IOU values between all multi-cluster/extended-target elliptical shape estimates and the true cluster/extended elliptical targets within each cell: using the matching result to calculate the ratio of elliptical intersection and parallel area between the real group/expanded target and all the elliptical estimation of the multi-group/expanded target in the unit matched with the real group/expanded target, and obtaining the IOU performance evaluation index value of the shape estimation result in each unit;
(5) Calculating and estimating the orientation error of the long axis of the ellipse, and punishing the orientation error to obtain a performance evaluation index value after punishing the orientation error of the long axis of the ellipse: firstly, calculating the difference between the orientation parameters of the long axes of the ellipses in the multi-group/extended target elliptical shape estimation parameters and the orientation parameters of the long axes of the ellipses in the real group/extended target elliptical shape parameters; then selecting a penalty function to carry out corresponding numerical penalty, carrying out estimation on the penalty of the orientation error of the long axis of the ellipse, and finally obtaining a performance evaluation index value;
(6) According to the judgment, obtaining the performance evaluation index value after the potential estimation error punishment in each unit: judging whether the over-estimation or the underestimation exists in each unit, and simultaneously obtaining the performance evaluation index value of the over-estimation or the underestimation on the basis of considering the shape estimation;
(7) And (3) evaluating the advantages and disadvantages of the multi-group/extended target elliptical shape estimation filter tracking by using the total performance evaluation index value NIS at the moment: extracting performance evaluation index values of all units at the moment, and using the performance evaluation index values to explain the tracking performance of the multi-group/extended target shape estimation filter;
(8) Judging whether the radar sensor receives new measurement data, if so, updating the time, returning to execute the steps (2) - (8), and circularly tracking in real time, otherwise, executing the step (9);
(9) And finishing multi-group/extended target tracking and finishing the evaluation of the tracking performance of the filter.
2. The multi-group/extended-target elliptical shape estimation evaluation method according to claim 1, wherein the multi-group/extended-target elliptical shape estimation parameter matching in step (3) uses a distance division method, comprising the steps of:
3.1 calculating Euclidean distances between all multi-group/extended target elliptical shape estimates and the centers of all real group/extended target elliptical shapes;
3.2 matching the estimated elliptical shape estimate with the minimum Euclidean distance with the real group/expanded elliptical target, namely indicating that the elliptical shape estimate is the filtering estimate of the real group/expanded elliptical target;
3.3 attributing each true group/expanded ellipse target and all ellipse shape estimates matching it to a different cell respectively.
3. The multi-group/extended-target elliptical shape estimation and evaluation method of claim 1, wherein the step (5) of calculating and estimating an elliptical long axis orientation error, and performing orientation error penalty to obtain a performance evaluation index value after ellipse long axis orientation error penalty comprises the following steps:
5.1 according to the matching result in the step (3), calculating an absolute value of an angle error value between the orientation of the estimated ellipse and the orientation of the real group/extended ellipse target by using the parameter value of the orientation of the major axis of the ellipse in the multi-group/extended target ellipse shape estimation parameter, and marking the absolute value as | delta phi |, wherein | delta phi | ∈ [0, pi/2 ];
5.2 selecting a penalty function f (| delta phi |) to punish the orientation angle error; according to the penalty function, the rule that the larger the orientation angle error is, the more serious the penalty is, the larger the orientation angle error value is, and the steeper the penalty function curve is must be satisfied; selecting a penalty function as a cosine function in the trigonometric function according to the penalty rule, namely f (| delta phi |) = cos (| delta phi |);
5.3 combining the result of Iou value calculated in step (4), obtaining an evaluation index value after the angle error penalty, and calculating by the following formula:
Aep＝Iou·f(|Δφ|)
from the characteristics of this penalty function, it can be seen that: when | Δ Φ | =0,f (| Δ Φ |) =1, there is Aep = Iou, i.e., no penalty is incurred when the ellipse orientation error is estimated to be zero, i.e., to account for the absence of orientation error; when | Δ Φ | = pi/2,f (| Δ Φ |) =0, aep =0 exists, the estimated ellipse orientation error at this time is the maximum, which indicates that the motion direction of the estimated group/extended target is completely wrong with the actual direction, and the estimated group/extended target is crossed, and the index value is set to zero to reflect the wrong estimation filtering result.
4. The multi-group/extended-target elliptical shape estimation and evaluation method as claimed in claim 1, wherein the step (6) of determining the performance evaluation index value after the potential estimation error penalty in each cell according to the judgment specifically comprises the following steps:
(6a) Judging whether overestimation or underestimation exists or not according to the matching result in each unit; if yes, executing the step (6 b) for further judgment, otherwise, executing the step (6 e): the method for judging whether the over-estimation or the underestimation condition exists is as follows: when the shape of the real group/expanded ellipse target in the unit is only matched with 1 group of ellipse estimation parameters, the situation of overestimation or underestimation is considered to be absent; otherwise, the actual group/extended target is considered to have an overestimation or underestimation condition;
(6b) Determining whether to overestimate or underestimate by judging: when a real group/expansion target in the unit is matched with 2 or more groups of elliptical estimation parameters, the situation that overestimation occurs on the group/expansion target estimation filtering at the moment is considered; if a real group/extended target is not matched with any elliptical estimation parameter, the situation that the filtering of the real group/extended target estimation at the moment is missed is considered; and (3) when the over-estimation condition occurs, executing the step (6 c), and when the under-estimation condition occurs, executing the step (6 d):
(6c) And (3) carrying out over-estimation punishment: performing a mean operation on all the evaluation index values Aep in the unit, and using the result as the Cep value of the multi-cluster/extended target shape estimation and tracking algorithm for the single cluster/extended target in the scene;
(6d) Directly setting the performance evaluation index value of the single group/extension target to zero, namely Cep =0;
(6e) No penalty is made: cep = Aep output at this time.
5. The method according to claim 1, wherein the step (7) of evaluating the advantages and disadvantages of the multi-group/extended target elliptical shape estimation filter tracking by using the overall performance evaluation index value at that time is described as follows: performing mean calculation on the Cep index values in all the units obtained in the step (6), wherein the obtained result is the overall performance evaluation index value of the multi-group/extended target elliptical shape estimation tracking algorithm at the moment; and analyzing the tracking performance of the filter by comparing the magnitude of the NIS value and the smoothness of a curve formed by connecting a plurality of moments to obtain the indication of the superiority and inferiority of the filter under different conditions.
CN201811640647.5A 2018-12-29 2018-12-29 Multi-group/expansion target elliptical shape estimation evaluation method Active CN109683150B (en)

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 CN109683150A (en) 2019-04-26
CN109683150B true 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)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
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

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP3415037B2 (en) * 1998-08-27 2003-06-09 三菱電機株式会社 Sensor group management device
CN105373667B (en) * 2015-11-27 2016-08-24 西安交通大学 The multigroup cross section perturbation motion method of uncertainty analysis is calculated for reactor physics
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

Also Published As

Publication number Publication date
CN109683150A (en) 2019-04-26

Similar Documents

Publication Publication Date Title
CN106842165B (en) Radar centralized asynchronous fusion method based on different distance angular resolutions
CN109509210A (en) Barrier tracking and device
CN107507417B (en) A kind of smartway partitioning method and device based on microwave radar echo-signal
CN103729859A (en) Probability nearest neighbor domain multi-target tracking method based on fuzzy clustering
CN107066806B (en) Data Association and device
CN105761276B (en) Based on the iteration RANSAC GM-PHD multi-object tracking methods that adaptively newborn target strength is estimated
CN110515051A (en) A kind of measurement method and system of trailer-mounted radar established angle angle value
CN103017771B (en) Multi-target joint distribution and tracking method of static sensor platform
CN113075648B (en) Clustering and filtering method for unmanned cluster target positioning information
CN107390631B (en) A kind of track initial method and system for maneuvering target of turning
CN113781773B (en) Traffic operation evaluation method, device and system and electronic equipment
CN103678949A (en) Tracking measurement set partitioning method for multiple extended targets based on density analysis and spectrum clustering
CN104159297A (en) Multilateration algorithm of wireless sensor networks based on cluster analysis
CN112162246B (en) Complex electromagnetic environment effect analysis method based on Bayesian network radar system
CN109683150B (en) Multi-group/expansion target elliptical shape estimation evaluation method
CN109633628A (en) The method of anti-RGPO interference based on distributed networking Radar Data Fusion
CN110488273A (en) A kind of vehicle tracking detection method and device based on radar
CN106872722B (en) A kind of measurement method and device of speed
KR101354522B1 (en) Retrival method of high resolution wind fields of multiple-doppler radar using by variational and fgat method
Zhang et al. Improved interacting multiple model-new nearest neighbor data association algorithm
CN109405833B (en) Logic base track starting method, system, electronic device and storage medium
CN113673105A (en) Design method of true value comparison strategy
Pusadan et al. Optimum partition in flight route anomaly detection
Sun et al. A real-time similarity measure model for multi-source trajectories
Meng et al. An ellipse feature tracking method based on the kalman filter

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