CN113311448A - Dynamic target tracking method and device based on multi-feature information - Google Patents
Dynamic target tracking method and device based on multi-feature information Download PDFInfo
- Publication number
- CN113311448A CN113311448A CN202010125268.3A CN202010125268A CN113311448A CN 113311448 A CN113311448 A CN 113311448A CN 202010125268 A CN202010125268 A CN 202010125268A CN 113311448 A CN113311448 A CN 113311448A
- Authority
- CN
- China
- Prior art keywords
- target object
- measured
- predicted
- minimum envelope
- measured object
- 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.)
- Withdrawn
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
- G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
- G01S17/66—Tracking systems using electromagnetic waves other than radio waves
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Electromagnetism (AREA)
- General Physics & Mathematics (AREA)
- Computer Networks & Wireless Communication (AREA)
- Automation & Control Theory (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to a dynamic target tracking method and device based on multi-feature information, and belongs to the technical field of intelligent vehicle environment perception. The method comprises the steps of firstly determining the minimum envelope rectangle of each measured object in a data frame at the current moment, and then obtaining the characteristic information of the minimum envelope rectangle. And meanwhile, according to the known target object in the data frame at the previous moment, predicting to obtain a predicted target object of the target object in the data frame at the current moment, determining the minimum envelope rectangle of the predicted target object, and obtaining the characteristic information of the minimum envelope rectangle. According to the characteristic information of the minimum enveloping rectangle of the predicted target object, a filter gate is established to carry out preliminary filtering on the measured objects to obtain candidate measured objects, then the correlation value of each candidate measured object and the predicted target object is obtained according to the characteristic information of the candidate measured object and the predicted target object in three aspects of position, length, width and motion direction, the correlation value is compared with a set correlation value threshold value, and the target object in the data frame at the current moment is determined.
Description
Technical Field
The invention relates to a dynamic target tracking method and device based on multi-feature information, and belongs to the technical field of environment perception of intelligent vehicles.
Background
In an environment sensing system of an intelligent vehicle, a sensor such as a laser radar not only needs to provide information such as the position and the size of a target in a road, but also needs to track a dynamic target so as to obtain information such as the speed of the target and the like for a decision-making system to call. The tracking principle of the dynamic target relates to the data association technology part, and at present, a commonly used data association algorithm is mainly a minimum distance-based Nearest Neighbor Data Association (NNDA) algorithm, which realizes data association by calculating the distance from each measurement position to the center of a tracking gate: when only one measurement exists in the tracking gate, the measurement is considered as the real measurement of the target; when a plurality of measurements fall into the tracking door, the distances between the measurements and the center in the tracking door are respectively calculated, and the measurement with the minimum distance is screened out through comparison to serve as the real measurement of the target. The algorithm only uses the condition of distance when judging whether the candidate measurement is the real measurement of the target, and does not set a related judgment threshold, so that the following problems of missing tracking or error tracking and the like easily occur in practical application: (1) when only one measurement falls into the tracking gate, the measurement may be a true measurement of the target, or may be a clutter or new target; (2) when there are multiple measurements falling into the tracking gate, the measurement closest to the center of the tracking gate is not necessarily the true measurement of the target.
In summary, the dynamic target tracking method in the prior art has a problem of insufficient tracking robustness.
Disclosure of Invention
The invention aims to provide a dynamic target tracking method and a dynamic target tracking device based on multi-feature information, and the dynamic target tracking method and the dynamic target tracking device are used for solving the problem that the dynamic target tracking robustness in the prior art is insufficient.
In order to achieve the above object, the present invention provides a dynamic target tracking method based on multi-feature information, which comprises the following steps:
1) determining the position, the size and the movement direction of each measured object according to the current time data frame;
2) predicting to obtain a predicted target object of the target object in the current time data frame according to the known data of the target object in the previous time data frame, and determining the position, the size and the motion direction of the predicted target object;
3) obtaining the association degree between the measured object and the prediction target object according to the difference of the measured object and the prediction target object in the positions, the sizes and the movement directions, and accordingly determining the measured object with the maximum association degree as the target object; wherein the smaller the difference between the position, the size or the movement direction of the measurement object and the prediction target object is, the greater the degree of association between the measurement object and the prediction target object is.
The invention has the beneficial effects that:
the method comprises the steps of firstly determining three information of the position, the size and the movement direction of each measured object in a data frame at the current moment, and predicting a predicted target object of the target object in the data frame at the current moment according to the known target object information in the data frame at the previous moment; the association degree of each measured object and the predicted target object in the current time data frame is calculated by integrating the information of the position, the size and the movement direction, so that the tracking missing or error tracking condition in the tracking process can be avoided, and the identification robustness of the dynamic target tracking is improved.
Further, in order to provide a more preferable method for determining the degree of association, the degree of association is determined by the following association value determination function:
where Cor is a correlation value characterizing the degree of correlation, Pcenter、PLWAnd PratioRespectively representing the differences in position, size and direction of movement, k, of the measured object and the predicted target object1、k2And k3Weight representing correspondence of three-dimensional differences in position, size, and direction of motion in the degree of association between the measurement object and the prediction target object, and k1+k2+k3＝1；
And calculating the correlation value between each measured object and the predicted target object according to the correlation value determining function, wherein the candidate measured object with the maximum correlation value is the target object in the current time data frame.
Further, in order to provide a more optimal size representation method, the size of the measured object is represented by a minimum envelope rectangle corresponding to the measured object; the size of the prediction target object is represented by a minimum envelope rectangle corresponding to the prediction target object.
Further, to give a more optimal position difference determination method, the difference in position between the measured object and the prediction target object is determined by the following position difference calculation function:
wherein (x)ti，yti) For the measured object, the coordinate of the center point of the minimum envelope rectangle is (x)t|t-1，yt|t-1) And the coordinates of the central point of the minimum envelope rectangle correspond to the predicted target object.
Further, to give a more optimal size difference determination method, the difference in size between the measurement object and the prediction target object is determined by the following length-width size difference calculation function:
wherein L istiAnd WtiRespectively the length and width, L, of the smallest envelope rectangle corresponding to the measured objectt|t-1And Wt|t-1The length and width of the minimum envelope rectangle are respectively corresponding to the predicted target object.
Further, to give a more optimal motion direction difference determination method, the difference between the measured object and the prediction target object in the motion direction is determined by the following motion direction difference calculation function:
Pratio＝|αti-αt|t-1|
wherein alpha istiFor measuring the angle, alpha, between the minimum envelope rectangle and the X-axist|t-1And predicting the included angle between the minimum envelope rectangle corresponding to the target object and the X axis.
Further, in order to reduce the calculation amount as much as possible and improve the dynamic target tracking efficiency, before the target object is determined by means of the association degree, the method further comprises the step of screening candidate measured objects from the measured objects:
establishing a filter door;
judging whether the central point of the minimum envelope rectangle corresponding to each measured object is positioned in the filter gate or not;
and the measuring objects in the filter gate are candidate measuring objects, and then the correlation degree of the candidate measuring objects is judged.
Further, in order to provide a more optimal filter door establishing method, the filter door establishing steps are as follows:
taking the central point of the minimum envelope rectangle of the predicted target object as the center of the ellipse, taking the length of the minimum envelope rectangle of the predicted target object as the major axis of the ellipse, taking the width of the minimum envelope rectangle of the predicted target object as the minor axis of the ellipse, and forming an inscribed ellipse in the minimum envelope rectangle of the predicted target object to obtain a filter gate:
wherein (x)t|t-1，yt|t-1) For predicting the center point of the smallest enveloping rectangle of the target object, Lt|t-1For predicting the length of the smallest enveloping rectangle of the target object, Wt|t-1The width of the smallest envelope rectangle for the predicted target object.
Further, in order to improve the robustness of dynamic target tracking, the method further comprises the step of optimizing the filter gate:
the major axis and the minor axis of the ellipse representing the filter gate are respectively enlarged by A1Multiple sum of A2Obtaining an optimized filter gate having a larger area than the filter gate:
wherein the magnification A is larger than that of the major axis and the minor axis1And A2Determined according to the real vehicle test.
Further, the step of determining the measurement object with the maximum degree of association as the target object includes:
if the association degree of only one measured object in the current time data frame is greater than the set association threshold, the measured object is the target object; if the association degree of a plurality of measured objects in the current time data frame is greater than the set association threshold, the measured object with the maximum association degree is the target object.
Further, the movement direction of the measurement object is expressed as an included angle value between the movement direction of the measurement object and the X axis; the motion direction of the predicted target object is expressed as an included angle value between the motion direction of the predicted target object and an X axis; the X-axis is determined by establishing a uniform X-Y coordinate system across all data frames.
In order to achieve the above object, the present invention further provides a dynamic target tracking apparatus based on multi-feature information, including a memory and a processor, where the processor is configured to execute a computer program stored in the memory, so as to implement the following dynamic target tracking method based on multi-feature information:
1) determining the position, the size and the movement direction of each measured object according to the current time data frame;
2) predicting to obtain a predicted target object of the target object in the current time data frame according to the known data of the target object in the previous time data frame, and determining the position, the size and the motion direction of the predicted target object;
3) obtaining the association degree between the measured object and the prediction target object according to the difference of the measured object and the prediction target object in the positions, the sizes and the movement directions, and accordingly determining the measured object with the maximum association degree as the target object; wherein the smaller the difference between the position, the size or the movement direction of the measurement object and the prediction target object is, the greater the degree of association between the measurement object and the prediction target object is.
The invention has the beneficial effects that:
the method comprises the steps of firstly determining three information of the position, the size and the movement direction of each measured object in a data frame at the current moment, and predicting a predicted target object of the target object in the data frame at the current moment according to the known target object information in the data frame at the previous moment; the association degree of each measured object and the predicted target object in the current time data frame is calculated by integrating the information of the position, the size and the movement direction, so that the tracking missing or error tracking condition in the tracking process can be avoided, and the identification robustness of the dynamic target tracking is improved.
Further, in order to provide a more preferable method for determining the degree of association, the degree of association is determined by the following association value determination function:
where Cor is a correlation value characterizing the degree of correlation, Pcenter、PLWAnd PratioRespectively representing the differences in position, size and direction of movement, k, of the measured object and the predicted target object1、k2And k3Weight representing correspondence of three-dimensional differences in position, size, and direction of motion in the degree of association between the measurement object and the prediction target object, and k1+k2+k3＝1；
And calculating the correlation value between each measured object and the predicted target object according to the correlation value determining function, wherein the candidate measured object with the maximum correlation value is the target object in the current time data frame.
Further, in order to provide a more optimal size representation method, the size of the measured object is represented by a minimum envelope rectangle corresponding to the measured object; the size of the prediction target object is represented by a minimum envelope rectangle corresponding to the prediction target object.
Further, to give a more optimal position difference determination method, the difference in position between the measured object and the prediction target object is determined by the following position difference calculation function:
wherein (x)ti，yti) For the measured object, the coordinate of the center point of the minimum envelope rectangle is (x)t|t-1，yt|t-1) And the coordinates of the central point of the minimum envelope rectangle correspond to the predicted target object.
Further, to give a more optimal size difference determination method, the difference in size between the measurement object and the prediction target object is determined by the following length-width size difference calculation function:
wherein L istiAnd WtiRespectively the length and width, L, of the smallest envelope rectangle corresponding to the measured objectt|t-1And Wt|t-1The length and width of the minimum envelope rectangle are respectively corresponding to the predicted target object.
Further, to give a more optimal motion direction difference determination method, the difference between the measured object and the prediction target object in the motion direction is determined by the following motion direction difference calculation function:
Pratio＝|αti-αt|t-1|
wherein alpha istiFor measuring the angle, alpha, between the minimum envelope rectangle and the X-axist|t-1And predicting the included angle between the minimum envelope rectangle corresponding to the target object and the X axis.
Further, in order to reduce the calculation amount as much as possible and improve the dynamic target tracking efficiency, before the target object is determined by means of the association degree, the method further comprises the step of screening candidate measured objects from the measured objects:
establishing a filter door;
judging whether the central point of the minimum envelope rectangle corresponding to each measured object is positioned in the filter gate or not;
and the measuring objects in the filter gate are candidate measuring objects, and then the correlation degree of the candidate measuring objects is judged.
Further, in order to provide a more optimal filter door establishing method, the filter door establishing steps are as follows:
taking the central point of the minimum envelope rectangle of the predicted target object as the center of the ellipse, taking the length of the minimum envelope rectangle of the predicted target object as the major axis of the ellipse, taking the width of the minimum envelope rectangle of the predicted target object as the minor axis of the ellipse, and forming an inscribed ellipse in the minimum envelope rectangle of the predicted target object to obtain a filter gate:
wherein (x)t|t-1，yt|t-1) For predicting the center point of the smallest enveloping rectangle of the target object, Lt|t-1For predicting the length of the smallest enveloping rectangle of the target object, Wt|t-1The width of the smallest envelope rectangle for the predicted target object.
Further, in order to improve the robustness of dynamic target tracking, the method further comprises the step of optimizing the filter gate:
the major axis and the minor axis of the ellipse representing the filter gate are respectively enlarged by A1Multiple sum of A2Obtaining an optimized filter gate having a larger area than the filter gate:
wherein the magnification A is larger than that of the major axis and the minor axis1And A2Determined according to the real vehicle test.
Further, the step of determining the measurement object with the maximum degree of association as the target object includes:
if the association degree of only one measured object in the current time data frame is greater than the set association threshold, the measured object is the target object; if the association degree of a plurality of measured objects in the current time data frame is greater than the set association threshold, the measured object with the maximum association degree is the target object.
Further, the movement direction of the measurement object is expressed as an included angle value between the movement direction of the measurement object and the X axis; the motion direction of the predicted target object is expressed as an included angle value between the motion direction of the predicted target object and an X axis; the X-axis is determined by establishing a uniform X-Y coordinate system across all data frames.
Drawings
Fig. 1 is a flowchart of the dynamic target identification method based on multi-feature information according to the present application.
Detailed Description
In the running process of the intelligent vehicle, the vehicle needs to track surrounding dynamic targets through sensing equipment on the vehicle, so that obstacles are avoided, and guidance is provided for planning a running route. However, the current tracking method for the dynamic target is often low in identification efficiency and poor in robustness, which causes tracking errors or tracking omission, so the following dynamic target tracking method embodiment and system embodiment based on multi-feature information are provided in the application.
The following further describes embodiments of the present invention with reference to the drawings.
The embodiment of the method provided by the invention comprises the following steps:
the method comprises the following specific processes:
1. according to data collected by a vehicle sensor, the minimum enveloping rectangle of each measured object in the current frame is obtained, characteristic information of the minimum enveloping rectangle is extracted, meanwhile, a predicted target object corresponding to the target object in the current frame is obtained through data prediction of a known target object in the previous frame, the minimum enveloping rectangle of the predicted target object is determined, and the characteristic information of the minimum enveloping rectangle is extracted.
In this embodiment, a data frame detected at the current time t and a data frame detected at the previous time t-1 are taken as an example for explanation.
Defining a known target object in the data frame at the time t-1 by the target object A to measure an object BiDefining a plurality of moving objects in the data frame at the time t, and establishing a uniform X-Y coordinate system in all the data frames, thereby uniformly processing the data frames at the time t and the time t-1, wherein i represents the number of the measured objects in the data frame at the time t.
Clustering the data in the data frame at the time t to obtain all the measured objects B included in the data frameiAnd corresponding clustering clusters, and extracting the minimum envelope rectangle of each clustering cluster.
Obtaining each measured object B in the current time data frameiAfter defining the minimum envelope rectangles, defining the coordinates of the center point of each minimum envelope rectangle as (x)ti，yti) Length L oftiWidth WtiAnd each of the measuring objects BiHas an included angle alpha between the moving direction and the X axisti。
Predicting the position of a target object A in a data frame at the time t-1 at the time t through a Kalman filtering algorithm to obtain a predicted target object A 'corresponding to the target object A in the data frame at the time t-1, then determining the minimum envelope rectangle of the predicted target object A', and defining the coordinate of the central point of the minimum envelope rectangle as (x)t|t-1，yt|t-1) Length of Lt|t-1Width of Wt|t-1Predicting the included angle between the moving direction of the target object A' and the X axis as alphat|t-1。
Since a method of performing cluster analysis on data to obtain a cluster, a method of determining a minimum envelope rectangle of an object according to the cluster, and a kalman filter algorithm have been described in the prior art, they are not described herein again.
In the present embodiment, the center point, length and width dimensions and motion direction of the minimum enveloping rectangle are used to sequentially represent each measured object BiOr predicting the characteristic information of the position, the size and the movement direction of the target object A'; the reason for the rectangle is thatThe purpose of using the minimum envelope rectangle in the calculation is to improve accuracy, and in other embodiments, the shape of each measured object or the predicted target object may be represented by other envelope shapes.
In this embodiment, specifically, a kalman filter algorithm is used to predict a position, corresponding to the current time data frame, of a known target object in the previous time data frame, and in other embodiments, any other feasible method may also be used.
2. Calculating each of the objects BiAnd the degree of association with the predicted target object a' and finally determining the target object.
Measuring object BiThe degree of association with the prediction target object a' is mainly reflected in three aspects of the position, size, and movement direction of the object.
The difference in position is reflected in the measured object BiIs calculated from the distance between the center point of (a) and the center point of the predicted target object a':
the difference in size can be measured by measuring the object BiThe difference in length and width compared to the predicted target object a' is integrated to yield:
the difference in the moving direction can be measured by measuring the object BiThe included angle value of the self-moving direction and the X axis, the included angle value of the predicted target object A' self-moving direction and the X axis, and the difference value between the two included angle values:
Pratio＝|αti-αt|t-1|
this example characterizes the measurement object B by the above methodiAnd the predicted target object a' in terms of three differences in the position, size, and direction of movement of the object.
In other embodiments, the difference between the positions of the measured object and the predicted target object can be further characterized by the distance difference between the respective vertices of the minimum envelope rectangle corresponding to the measured object and the minimum envelope rectangle corresponding to the predicted target object.
The difference between the sizes of the measured object and the predicted object can be represented by the difference between the areas of the minimum envelope rectangle corresponding to the measured object and the minimum envelope rectangle corresponding to the predicted object, or the difference between the areas of the two ellipses of the minimum envelope ellipse corresponding to the measured object and the minimum envelope ellipse corresponding to the predicted object, or the sum of the difference between the minimum envelope ellipses of the measured object and the predicted object in the major axis and the difference between the minimum envelope ellipses of the measured object and the predicted object in the minor axis.
And calculating the difference between the measured object and the predicted target object in the motion direction by a method of establishing a space vector coordinate system.
After determining the differences in position, size and direction of movement, the difference is then determined by a factor k1、k2And k3Indicating that the object B is measured according to the differences in position, size and moving directioniWeight in degree of association with the prediction target object A', where k1+k2+k31. The weight of the difference factors of the position, the size and the movement direction can be determined by real vehicle experiments.
After determining the calculation functions of the differences in the three aspects of the position, the size and the movement direction and the respective weight coefficients thereof, each measured object B in the data frame at the current time t can be establishediThe correlation value with the prediction target object a' determines a function:
from this correlation value determination function, it is apparent that the measurement object B is reduced regardless of which of the calculated values of the differences in the three aspects of the position, the size, and the moving direction is reducediThe correlation value with the predicted target object A' is obtainedIncreasing; on the contrary, the measurement object B is measured regardless of which of the calculated values of the differences in the three aspects of the position, the size and the moving direction is increasediThe correlation value with the predicted target object a' is decreased.
In this embodiment, the correlation value is calculated by the correlation value determining function, and in other embodiments, in order to more obviously highlight the change of the correlation degree between the measured object and the predicted target object caused by the change of the difference values in the three aspects of the position, the size and the motion direction, the weighting parameter k may be further used1、k2And k3Respectively as the above-mentioned difference determining function Pcenter、PLWAnd PratioThe definite index of (A) is obtained.
Subsequently, if the target object is not missed in the data frame at the current time t, a certain measured object B in the data frame at the current time tiA target object, the measurement object BiMust have the smallest difference with the position, size and direction of motion of the predicted target object a' in the current time t data frame, i.e. the measured object B in the correlation functioniThe correlation value with the predicted target object a' is the largest.
Calculating each measured object B through the correlation value determining functioniAfter the correlation value with the predicted target object A', each measured object B is judged by means of an artificially set correlation value threshold epsiloniWhether it is a target object. The specific content of the judgment is as follows:
1) if only one measured object B exists in the data frame at the current time tiIs greater than the threshold epsilon, the measured object B is considered to beiThe target object in the data frame at the current time t is obtained;
2) if a plurality of measured objects B exist in the data frame at the current time tiIf the correlation value is greater than the threshold value epsilon, the measured object B with the maximum correlation value is selectediThe measured object B with the largest correlation value is considered asiThe target object in the data frame at the current time t is obtained;
3) if the measured object does not exist in the data frame at the current time tBody BiOr all measuring objects BiIf all the correlation values are smaller than the threshold epsilon, the tracking is determined to fail, and the target object is lost in the data frame at the current time t.
The dynamic target tracking method provided by the embodiment of the invention comprehensively considers the position, the size and the moving direction of the target object, tracks the target object by a multi-feature information method, can avoid the conditions of tracking missing or error tracking in the tracking process, and improves the identification robustness of dynamic target tracking.
The embodiment of the method provided by the invention comprises the following steps:
as shown in FIG. 1, the embodiment of the method is particularly added to the measurement object B based on the content of the first embodiment of the methodiThe screening step is specifically that the following steps are added after step 1 of the first method embodiment:
establishing a filter gate based on the characteristic information of the minimum envelope rectangle corresponding to the predicted target object A', and performing filtering on each measured object B in the current time data frameiAnd (5) screening.
It can be known that each measured object B in the current t-moment data frameiOnly one is the target object, so the characteristic information of the predicted target object A' in the predicted data frame at the current time t is used as the standard to establish the filter gate for each measured object BiAnd carrying out primary screening to reduce the calculation amount of subsequent comparison.
The specific process of establishing the filter gate is as follows:
in the data frame at the time t, the central point (x) of the minimum envelope rectangle of the target object A' is predictedt|t-1，yt|t-1) Is the center of the ellipse and has the length L of the smallest envelope rectangle of the predicted target object At|t-1And width Wt|t-1Forming an inscribed ellipse within the minimum envelope rectangle of the prediction target object a' as the major axis and the minor axis of the ellipse, respectively:
however, in order not to lose every measurement object that may be a target object, the major and minor axes of the ellipse are enlarged by a, respectively1Multiple sum of A2And (3) obtaining an elliptical filter gate with an area larger than that of the inscribed ellipse:
wherein the magnification A is for the major and minor axes1And A2Determined according to the real vehicle test.
For all measured objects BiThe central coordinate of the elliptic filter door is judged, and a measured object B with the central coordinate positioned in the elliptic filter door is judgediAs candidate measuring object Bi', these candidate measuring objects Bi'constitute a set of candidate measurement objects that are closest to the prediction target object A', and one of the candidate measurement objects Bi' is the target object.
In the subsequent steps, the measured object B is directly calculated, compared to the first embodiment of the methodiThe method for predicting the degree of association between the target object A' and finally determining the target object is implemented by calculating each candidate measured object B after screeningi'degree of association with the predicted target object a' to finally determine the target object.
Since each candidate measured object B is calculated in the present embodimenti'method and process for finalizing object by correlating with predicted object A', and method embodiment one in which measuring object B is calculatediThe method and process for predicting the degree of correlation between the target objects a' and finally determining the target objects are essentially the same, and therefore, will not be described herein again.
In this embodiment, the object B is measured by the method of establishing the filter gateiThe screening is performed, in other embodiments, for example, the minimum enveloping ellipse of the predicted target object a' may be directly used as a filter gate to complete the screening of the measured object; as another example, it is also possibleThe screening of the measured object is finished according to one of the central point, the length and width dimensions or the motion direction of the predicted target object and the measured object corresponding to the minimum envelope rectangle; and, the screening of the measured object can be completed according to other existing methods, which are not described herein again.
On the basis of the first method embodiment, the embodiment of the invention adds the step of screening each measured object measured in the data frame at the current moment, can effectively reduce the calculated amount in the process of tracking the dynamic target, and further improves the identification efficiency of tracking the dynamic target.
Device embodiment
A dynamic target tracking device based on multi-feature information comprises a memory, a processor and a computer program which is stored in the memory and can run on the processor, wherein the processor executes the computer program to realize the dynamic target tracking method based on the multi-feature information in any method embodiment.
The device can be intelligent equipment such as a whole vehicle processor, a tablet personal computer, a mobile phone or a personal computer.
When designing the processor in this embodiment, since a person skilled in the art has complete capability to use the existing programming language (e.g., C language, JAVA, assembly language, C #, C + +, etc.) to perform corresponding programming, so that the processor in this embodiment completes the dynamic object tracking method based on the multi-feature information, this process is not described herein again.
The specific embodiments are given above, but the present invention is not limited to the described embodiments. The basic idea of the present invention is to provide the basic solution described above, and variations, modifications, replacements, and variations of the embodiments can be made without departing from the principle and spirit of the present invention, and still fall within the protection scope of the present invention.
Claims (10)
1. A dynamic target tracking method based on multi-feature information is characterized by comprising the following steps:
1) determining the position, the size and the movement direction of each measured object according to the current time data frame;
2) predicting to obtain a predicted target object of the target object in the current time data frame according to the known data of the target object in the previous time data frame, and determining the position, the size and the motion direction of the predicted target object;
3) obtaining the association degree between the measured object and the prediction target object according to the difference of the measured object and the prediction target object in the positions, the sizes and the movement directions, thereby determining the measured object with the maximum association degree as the target object: wherein the smaller the difference between the position, the size or the movement direction of the measurement object and the prediction target object is, the greater the degree of association between the measurement object and the prediction target object is.
2. The multi-feature information based dynamic target tracking method according to claim 1, wherein the degree of association is determined by the following association value determination function:
where Cor is a correlation value characterizing the degree of correlation, Pcenter、PLWAnd PratioRespectively representing the differences in position, size and direction of movement, k, of the measured object and the predicted target object1、k2And k3Weight representing correspondence of three-dimensional differences in position, size, and direction of motion in the degree of association between the measurement object and the prediction target object, and k1+k2+k3＝1；
And calculating the correlation value between each measured object and the predicted target object according to the correlation value determining function, wherein the candidate measured object with the maximum correlation value is the target object in the current time data frame.
3. The multi-feature information based dynamic target tracking method according to claim 1 or 2, wherein the size of the measured object is represented by a minimum envelope rectangle corresponding to the measured object; the size of the prediction target object is represented by a minimum envelope rectangle corresponding to the prediction target object.
4. The multi-feature information based dynamic target tracking method of claim 2, wherein the difference in position between the measured object and the predicted target object is determined by the following position difference calculation function:
wherein (x)ti，yti) For the measured object, the coordinate of the center point of the minimum envelope rectangle is (x)t|t-1，yt|t-1) And the coordinates of the central point of the minimum envelope rectangle correspond to the predicted target object.
5. The multi-feature information based dynamic target tracking method of claim 2, wherein the difference in size between the measured object and the predicted target object is determined by the following length-width difference calculation function:
wherein L istiAnd WtiRespectively the length and width, L, of the smallest envelope rectangle corresponding to the measured objectt|t-1And Wt|t-1The length and width of the minimum envelope rectangle are respectively corresponding to the predicted target object.
6. The multi-feature information based dynamic target tracking method of claim 2, wherein the difference between the measured object and the predicted target object in the moving direction is determined by the following moving direction difference calculation function:
Pratio＝|αti-αt|t-1|
wherein alpha istiFor measuring the angle, alpha, between the minimum envelope rectangle and the X-axist|t-1And predicting the included angle between the minimum envelope rectangle corresponding to the target object and the X axis.
7. The multi-feature information based dynamic target tracking method according to any one of claims 1 or 2, further comprising the step of screening candidate metrology objects from the metrology objects before determining the target object by means of the degree of correlation:
establishing a filter door;
judging whether the central point of the minimum envelope rectangle corresponding to each measured object is positioned in the filter gate or not;
and the measuring objects in the filter gate are candidate measuring objects, and then the correlation degree of the candidate measuring objects is judged.
8. The multi-feature information based dynamic target tracking method according to claim 7, wherein the filter gate is established by:
taking the central point of the minimum envelope rectangle of the predicted target object as the center of the ellipse, taking the length of the minimum envelope rectangle of the predicted target object as the major axis of the ellipse, taking the width of the minimum envelope rectangle of the predicted target object as the minor axis of the ellipse, and forming an inscribed ellipse in the minimum envelope rectangle of the predicted target object to obtain a filter gate:
wherein (x)t|t-1，yt|t-1) For predicting the center point of the smallest enveloping rectangle of the target object, Lt|t-1For predicting the length of the smallest enveloping rectangle of the target object, Wt|t-1Width of minimum envelope rectangle for predicting target objectAnd (4) degree.
9. The multi-feature information based dynamic target tracking method according to claim 1, wherein the step of determining the measurement object with the highest correlation degree as the target object comprises:
if the association degree of only one measured object in the current time data frame is greater than the set association threshold, the measured object is the target object; if the association degree of a plurality of measured objects in the current time data frame is greater than the set association threshold, the measured object with the maximum association degree is the target object.
10. A multi-feature information based dynamic target tracking apparatus, comprising a memory and a processor, wherein the processor is configured to execute a computer program stored in the memory to implement the multi-feature information based dynamic target tracking method according to any one of claims 1 to 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010125268.3A CN113311448A (en) | 2020-02-27 | 2020-02-27 | Dynamic target tracking method and device based on multi-feature information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010125268.3A CN113311448A (en) | 2020-02-27 | 2020-02-27 | Dynamic target tracking method and device based on multi-feature information |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113311448A true CN113311448A (en) | 2021-08-27 |
Family
ID=77370791
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010125268.3A Withdrawn CN113311448A (en) | 2020-02-27 | 2020-02-27 | Dynamic target tracking method and device based on multi-feature information |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113311448A (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2002341024A (en) * | 2001-05-17 | 2002-11-27 | Mitsubishi Electric Corp | Multiple target tracking device |
| JP2002341025A (en) * | 2001-05-11 | 2002-11-27 | Nec Corp | Target tracking device |
| CN103064086A (en) * | 2012-11-04 | 2013-04-24 | 北京工业大学 | Vehicle tracking method based on depth information |
| CN103875021A (en) * | 2011-10-19 | 2014-06-18 | 克朗设备公司 | Identifying and selecting objects that may correspond to pallets in an image scene |
| CN109521420A (en) * | 2018-12-20 | 2019-03-26 | 西安电子科技大学 | Based on the matched multi-object tracking method of multiple features |
| CN110244294A (en) * | 2019-06-24 | 2019-09-17 | 中国人民解放军空军工程大学航空机务士官学校 | A kind of correlating method of the metric data of multisensor |
| CN110780289A (en) * | 2019-10-23 | 2020-02-11 | 北京信息科技大学 | Multi-target vehicle tracking method and device based on scene radar |
-
2020
- 2020-02-27 CN CN202010125268.3A patent/CN113311448A/en not_active Withdrawn
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2002341025A (en) * | 2001-05-11 | 2002-11-27 | Nec Corp | Target tracking device |
| JP2002341024A (en) * | 2001-05-17 | 2002-11-27 | Mitsubishi Electric Corp | Multiple target tracking device |
| CN103875021A (en) * | 2011-10-19 | 2014-06-18 | 克朗设备公司 | Identifying and selecting objects that may correspond to pallets in an image scene |
| CN103064086A (en) * | 2012-11-04 | 2013-04-24 | 北京工业大学 | Vehicle tracking method based on depth information |
| CN109521420A (en) * | 2018-12-20 | 2019-03-26 | 西安电子科技大学 | Based on the matched multi-object tracking method of multiple features |
| CN110244294A (en) * | 2019-06-24 | 2019-09-17 | 中国人民解放军空军工程大学航空机务士官学校 | A kind of correlating method of the metric data of multisensor |
| CN110780289A (en) * | 2019-10-23 | 2020-02-11 | 北京信息科技大学 | Multi-target vehicle tracking method and device based on scene radar |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109782763B (en) | Mobile robot path planning method in dynamic environment | |
| CN112526513B (en) | Millimeter wave radar environment map construction method and device based on clustering algorithm | |
| CN111444767B (en) | Pedestrian detection and tracking method based on laser radar | |
| CN110400332A (en) | A kind of target detection tracking method, device and computer equipment | |
| CN106714110A (en) | Auto building method and system of Wi-Fi position fingerprint map | |
| CN109145836B (en) | Ship target video detection method based on deep learning network and Kalman filtering | |
| CN111553950B (en) | Steel coil centering judgment method, system, medium and electronic terminal | |
| CN110986956A (en) | Autonomous learning global positioning method based on improved Monte Carlo algorithm | |
| CN113436223B (en) | Point cloud data segmentation method and device, computer equipment and storage medium | |
| CN108985375B (en) | Multi-feature fusion tracking method considering particle weight spatial distribution | |
| KR20120112293A (en) | Apparatus and method for detecting position of moving unit | |
| CN114973195A (en) | Vehicle tracking method, device and system based on multi-information fusion | |
| CN114119659A (en) | Multi-sensor fusion target tracking method | |
| CN110717934A (en) | Anti-occlusion target tracking method based on STRCF | |
| CN112164093A (en) | Automatic person tracking method based on edge features and related filtering | |
| CN110851978B (en) | Camera position optimization method based on visibility | |
| CN112580438A (en) | Point cloud identification method in three-dimensional scene | |
| CN113205108A (en) | YOLOv 4-based multi-target vehicle detection and tracking method | |
| CN111669702A (en) | Coarse and fine matching positioning method based on WIFI address and field intensity | |
| CN113311448A (en) | Dynamic target tracking method and device based on multi-feature information | |
| CN112833815B (en) | Laser-based house measurement and wall flatness detection method | |
| CN112989946A (en) | Method, device and equipment for determining lane line and vehicle | |
| CN104965190A (en) | Axial decoupling indoor positioning method based on location fingerprint | |
| CN112017446B (en) | Radar measurement data and vehicle track association method, device, medium and electronic device | |
| CN117523428B (en) | Ground target detection method and device based on aircraft platform |
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 | ||
| CB02 | Change of applicant information |
Address after: No. 6, Yutong Road, Guancheng Hui District, Zhengzhou, Henan 450061 Applicant after: Yutong Bus Co.,Ltd. Address before: No.1, Shibali Heyu Road, Guancheng Hui District, Zhengzhou City, Henan Province Applicant before: ZHENGZHOU YUTONG BUS Co.,Ltd. |
|
| CB02 | Change of applicant information | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20210827 |
|
| WW01 | Invention patent application withdrawn after publication |