CN105321162B - A kind of more mesh real-time optimization localization methods of the list of moving target - Google Patents
A kind of more mesh real-time optimization localization methods of the list of moving target Download PDFInfo
- Publication number
- CN105321162B CN105321162B CN201410350844.9A CN201410350844A CN105321162B CN 105321162 B CN105321162 B CN 105321162B CN 201410350844 A CN201410350844 A CN 201410350844A CN 105321162 B CN105321162 B CN 105321162B
- Authority
- CN
- China
- Prior art keywords
- target
- mesh
- video camera
- positioning
- real
- 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.)
- Expired - Fee Related
Links
Abstract
The present invention proposes a kind of more mesh real-time optimization localization methods of the list of moving target, its technical solution is 5 system modes of design, it is waited for including calculating, object initialization, more mesh positioning measurements, monocular positioning measurement and target prodiction state, in real time according to target search, more mesh tracking measurements, the measuring conditions such as monocular tracking measurement and target transient loss, measure state optimization transfer, to be positioned to moving target using distinct methods, accomplish that metrical information is made full use of to improve positioning accuracy and ensures the continuity measured as far as possible, realize the real-time optimization positioning measurement of target.
Description
Technical field
The present invention relates to the technical fields such as videographic measurment, moving target positioning, refer in particular to using single more mesh to moving mesh
Target real-time location method realizes that the real-time optimization of target positioning is handled according to measuring condition using state transition method.
Background technology
Moving target is located in Research on Target and scouts has important application with fields such as attacking and defending, aircraft experiment observing and controlling.According to
By radar emission electromagnetic wave and the major way that target echo realization is the positioning of current kinetic target to the positioning of target is received, but
Radar detection, which belongs to, actively to be measured, and is easy to be found and anti-strike by other side in confrontation on the battlefield.Videographic measurment is imaged using camera
Determine target location, be a kind of passive measurement, have the characteristics that precision is high, non-contact, at low cost, target measurement, target with
There are many more ripe applications in the fields such as track, target range observing and controlling.
When carrying out target measurement and tracking using photographing measurement method, fixation measuring base station or traverse measurement need to be generally set up
Platform shoots target following using optical measuring apparatus such as cinetheodollite, high-speed cameras, obtains image and data.Profit
It can be intersected to obtain target location by sight with double image machine or more cameras, can weakened to synchronization after merging target trajectory
The requirement of the conditions such as imaging, intersection angle.Traditional list camera imaging normally only can determine that the direction of visual lines of target relative measurement point, nothing
Method determines target location, by increasing the ancillary equipments such as ranging, utilizing object construction information or motion track information etc., Ke Yishi
Existing positioning of the monocular to multiple target.
The difficult point that moving target positions in real time is the variation that target movement leads to videographic measurment condition, it is difficult to using a kind of solid
Fixed positioning method, as video camera is lost, target causes to measure the reduction of camera number or even all video cameras are lost in the short time
Target；Video camera recapture target causes measurement camera number to increase, and metrical information increases.Especially in confrontation on the battlefield environment
In, some evading property of target are motor-driven to make video camera be more difficult to tenacious tracking, and the variation of measuring condition is more frequent.Therefore into
When row moving target positioning measurement, need according to measuring condition optimization processing target search, more range estimation amounts, monocular and target
A variety of situations such as transient loss make full use of metrical information to improve positioning accuracy and ensure the continuity measured as far as possible, realize
The real-time optimization of moving target position measures.This problem there are no open source literature research at present.
Invention content
The technical problem to be solved by the present invention is to the more mesh real-time optimization orientation problems of the list of moving target, for moving target
There may be a variety of situations such as target search, more mesh tracking measurements, monocular tracking measurement and target transient loss to change in positioning,
State conversion is measured according to real-time conditions, realizes the optimization processing of target positioning.
The technical scheme is to design multiple measuring states, including calculating waiting, object initialization, the positioning survey of more mesh
The measuring states such as amount, monocular positioning measurement, target prodiction measure state optimization transfer in real time according to measuring condition,
To be positioned to moving target using distinct methods, metrical information is made full use of, realizes the real-time optimization positioning measurement of target.
The real-time positioning measurement state of the more mesh of moving target list of the present invention is fixed comprising calculating waiting, object initialization, more mesh
5 position measurement, monocular positioning measurement, target prodiction states.Once there is camera shooting machine examination in the system for calculating wait state
It measures target and proceeds to object initialization state, object initialization completes target label, target measurement information data library initialization
Etc. after tasks according to detecting that the video camera number of target is transferred to monocular positioning measurement or more mesh positioning measurement states；If all take the photograph
Camera transient loss target then enters target prodiction state, according to target motion history trajectory predictions target location；If having
Video camera detects target again, then goes back to monocular positioning measurement or more mesh positioning measurement states；When target location is continuously predicted
After certain time threshold value, then the target following measurement is no longer carried out, into calculating wait state.System mode is arranged and turns
Shifting condition is as shown in Fig. 1.
The task description of each measuring state is as follows：
1, wait state is calculated
Show not measure target when in the state, system constantly according to camera review searching and detecting target, once
It was found that target, is transferred to object initialization state at once；
2, object initialization state
Object initialization is realized to target label, initialized target metrical information database task.According to target whether
It is fresh target, object initialization is divided into two classes：
Newly there is object initialization：It makes a mark to target, establishes new target information database and describe the fresh target, set
Target information is active state；
Lose reproducing target initialization：Since the target has made marks, and information database is established, only needs to restart
The target information is active state；
3, more mesh positioning measurement states
In the case where more mesh are imaged synchronous condition, directly target is positioned using the intersection of more line of sight, and is gone through using target location
History information reduces crossing location error to data smoothing；It is fully sharp under the conditions of the asynchronous or intersection angle of more mesh imaging is improper
With target trajectory information, it is assumed that target kinetic characteristic meets certain rule, and target is determined according to target motion history information
More line of sight are intersected with parametrization target trajectory, metrical information optimization are made full use of to measure target location by kinematic parameter；
4, monocular positioning measurement state
Since monocular sight only can determine that the direction of target, target range can not be determined, therefore needed in monocular positioning measurement
Target is positioned according to some prior informations of target.For cooperative target, if target image is more clear, target knot is utilized
Structure information (such as target feature point, target area, target length) and line of sight direction position target as common constraint；
If target (such as ground) in known altitude plane moves, is intersected with height face using line of sight direction and target is positioned；If
Assuming that target movement meets certain rule, then moving equation can be established according to target histories track and the characteristics of motion, according to
Sight observes Optimization Solution kinematic parameter, realizes and is positioned to target；
5, target prodiction state
Short time future position after target prodiction is lost for target can keep target to position continuous
Property, also allow for the recapture for realizing target；
Target prodiction is according to target histories track and target movement model future position, for maneuvering target,
Since target movement model is difficult to very accurately, generally to predict the long time, when by the way that time threshold control forecasting is arranged
Between, stop the positioning to target after continuous prediction is more than certain time threshold value, is transferred to calculating wait state.
Following technique effect can be reached using the present invention：
1) present invention is directed to the more mesh real-time optimization orientation problems of list of moving target, devises multiple measuring states, according to
Measuring condition carries out state transfer, selects suitable localization method, can be with optimization processing target search, more mesh tracking measurements, list
A variety of situations such as mesh tracking measurement and target transient loss realize that the real-time optimization of moving target position measures.
2) of the invention more mesh positioning measurements and monocular positioning measurement consider a variety of situations, can be previously according to when use
Measuring condition selects a kind of method, can also be in real time according to condition preferred method or a variety of method fusion treatments.
Description of the drawings
The real-time positioning measurment system state of the more mesh of Fig. 1 moving target lists and jump condition.
Specific implementation mode
The attached real-time positioning measurment system state of the more mesh of moving target list shown in FIG. 1 include calculate waiting, object initialization,
5 more mesh positioning measurements, monocular positioning measurement, target prodiction states.System is according to current state and detects target
The conditions such as video camera number, predicted time carry out state transfer.
1, waiting and object initialization are calculated
Positioning measurement of the wait state without target is calculated, each camera review is constantly detected using endless form in real time
In whether there is or not target, once finding target, be transferred to object initialization state at once；
Object initialization state is to target label, initialized target metrical information database task.For newly there is mesh
Mark makes a mark, and establishes new target information database and describes the fresh target, and it is active state to set target information；It is reproduced to losing
It is active state that target, which restarts the target information,.When being carried out at the same time positioning measurement to multiple moving targets, then mark more
A target is active state, and the information database of these targets all carries out real-time update；
2, more mesh locating measurement methods
When more mesh positioning measurements, the direct sight of the video camera of synchronous imaging can be intersected and determine target location, but considered
It is likely to be out of synchronization to the imaging of more mesh or intersection angle is improper, target movement model information need to be made full use of.If target movement model
For
Wherein p (t), v (t), a (t) are target respectively in the position of t moment, speed and acceleration, and ξ (t) is to accelerate
The perturbing vector of degree；
Under the coordinate condition of known position, posture, intrinsic parameter and the target for measuring video camera in the picture, it may be determined that
The direction of visual lines for going out target relative camera i, is set as li(t), then target location meets
WhereinIt is video camera i in the position of t moment, λi(t) it is distance of the target at a distance of video camera.More visual observations
Each video camera gives the target location observational equation of similar formula (2)；
Filter can be established according to formula (1) and (2), merge multiple video cameras observation filtering estimation target position,
Speed and acceleration realize the more mesh positioning measurements of target；
3, monocular locating measurement method
Monocular positioning measurement can be selected one of with the following method according to condition：
(1) structure feature point location method
If target image is clear, object construction information is it is known that can extract several characteristic points on target image, if characteristic point
Image coordinate be(subscript i is the number of video camera, and subscript j is characterized period j=1,2 ...), j-th
Physical location coordinate of the characteristic point in objective body structure is [xj yj zj]T, then according to camera shooting model, have
WhereinIt is characterized line projections on optical axis of the point j to video camera i optical centers,For the x of video camera i
The equivalent focal length in direction, the directions y,For the image principal point coordinate in the directions x of video camera i, the directions y, Ri(t) it is t moment
Object construction coordinate system is to the spin matrix of video camera i coordinate systems, Ti(t) it is that t moment object construction coordinate origin is imaging
The motion vector of machine coordinate system；
After extracting multiple characteristic points, the pose computation of computer vision field can be utilized according to formula (3)
Solve Ri(t) and Ti(t), the position of target relative camera is obtained, realizes target positioning；
In addition to point feature, target position can also be carried out using structure features such as line feature, target area, length in target
Set calculating；
(2) line of sight and height face intersection
If target (such as ground) in known altitude plane moves, intersected to mesh with height face using line of sight direction
Demarcate position；
If t moment object height is h (t), position for video camera is set toThe direction of visual lines of target relative camera is li
(t), target location p (t) meets formula (2), wherein λiIt (t) can be according to being highly calculated
Wherein []HIndicate the altitude channel of vector；
(3) target trajectory fusion positioning
Assuming that target movement meets formula (1), single camera observation at each moment can obtain 1 similar to formula (2)
Target location observational equation, establishes filter, and position, speed and the acceleration of filtering estimation target realize the more mesh positioning of target
It measures；
4, target prodiction method
Filtering recursion or the method for fitting of a polynomial may be used in target prodiction；
(1) recurrence method is filtered
Estimate that target loses preceding moment t according to target histories track optimizing0Position p (the t of target0), speed v (t0) and add
Speed a (t0), recursion future position p (t) is carried out according to following formula
Primary condition is：T=t0:P (t)=p (t0), v (t)=v (t0)；
(2) polynomial fitting method
According to the multinomial of target histories position settling time,
Fitting estimation polynomial parameters ak、bk、ck, (k=0,1 ..., n), according to formula (6) to current target position
It is predicted.
Claims (2)
1. a kind of more mesh real-time optimization localization methods of the list of moving target, which is characterized in that initial comprising calculating waiting, target
5 change, more mesh positioning measurements, monocular positioning measurement, target prodiction measuring states are surveyed in real time according to measuring condition
State optimization transfer is measured, realizes the real-time optimization positioning measurement to target, specially：
Once there is video camera to detect that target proceeds to object initialization state in the system for calculating wait state, target is initial
Change basis after completing target label, target measurement information database initialization task and detects that the video camera number of target is transferred to list
Mesh positioning measurement or more mesh positioning measurement states；If all video camera transient loss targets, enter target prodiction state,
According to target motion history trajectory predictions target location；If there is video camera to detect target again, monocular positioning measurement is gone back to
Or more mesh positioning measurement states；After target location continuously predicts to be more than certain time threshold value, then the target following is no longer carried out
It measures, into calculating wait state；
When more mesh positioning measurements, the direct sight of the video camera of synchronous imaging is intersected and determines target location, if target moves
Model is
Wherein p (t), v (t), a (t) are target respectively in the position of t moment, speed and acceleration, and ξ (t) is acceleration
Perturbing vector；
Under the coordinate condition of known position, posture, intrinsic parameter and the target for measuring video camera in the picture, target phase is determined
To the direction of visual lines of video camera i, it is set as li(t), then target location meets
WhereinIt is video camera i in the position of t moment, λi(t) be target at a distance of the distance of video camera, each of more visual observations
Video camera all provides the target location observational equation of formula (2)；
Filter can be established according to formula (1) and (2), merges position, the speed of the observation filtering estimation target of multiple video cameras
And acceleration, realize the more mesh positioning measurements of target；
The monocular positioning measurement is selected one of with the following method according to condition：
(a) structure feature point location method
If target image is clear, object construction information is it is known that can extract several characteristic points on target image, if the figure of characteristic point
As coordinate isSubscript i is the number of video camera, and subscript j is characterized period j=1,2 ..., j-th of feature
Physical location coordinate of the point in objective body structure is [xj yj zj]T, then according to camera shooting model, have
WhereinIt is characterized line projections on optical axis of the point j to video camera i optical centers,For the side x of video camera i
To the equivalent focal length in, directions y,For the image principal point coordinate in the directions x of video camera i, the directions y, Ri(t) it is t moment mesh
Structure coordinate system is marked to the spin matrix of video camera i coordinate systems, Ti(t) be t moment object construction coordinate origin in video camera
The motion vector of coordinate system；
After extracting multiple characteristic points, R is solved using the pose computation of computer vision field according to formula (3)i(t)
And Ti(t), the position of target relative camera is obtained, realizes target positioning；
In addition to point feature, target location calculating is carried out using line feature, target area, the length structure feature in target；
(b) line of sight and height face intersection
If target is intersected with height face in known altitude move in plane, using line of sight direction and is positioned to target,
If t moment object height is h (t), position for video camera is set toThe direction of visual lines of target relative camera is li(t), mesh
Cursor position p (t) meets formula (2), wherein λiIt (t) can be according to being highly calculated
Wherein []HIndicate the altitude channel of vector,
(c) target trajectory fusion positioning
Assuming that target movement meets formula (1), single camera at each moment is observed obtaining the target location of 1 similar formula (2)
Observational equation, establishes filter, and position, speed and the acceleration of filtering estimation target realize the more mesh positioning measurements of target.
2. a kind of more mesh real-time optimization localization methods of the list of moving target according to claim 1, which is characterized in that described
Positioning measurement of the wait state without target is calculated, constantly detects in each camera review that whether there is or not mesh in real time using endless form
Mark is transferred to object initialization state at once once finding target；
The object initialization state is to target label, initialized target metrical information database tasks, for newly there is target
It makes a mark, establishes new target information database and describe the fresh target, it is active state to set target information；Mesh is reproduced to losing
It is that active state then marks this more when being carried out at the same time positioning measurement to multiple moving targets that mark, which restarts the target information,
A target is active state, and the information database of these targets all carries out real-time update；
It needs to position target according to some prior informations of target in the monocular positioning measurement, for cooperative target, if mesh
Logo image is more clear, is positioned to target as common constraint using object construction information and line of sight direction；If target exists
Known altitude move in plane is intersected with height face using line of sight direction and is positioned to target；Assuming that target movement meets rule
Rule, then can establish moving equation according to target histories track and the characteristics of motion, and Optimization Solution movement ginseng is observed according to sight
Number is realized and is positioned to target；
Short time future position after the target prodiction is lost for target keeps the continuity of target positioning,
It is easy to implement the recapture of target.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410350844.9A CN105321162B (en) | 2014-07-23 | 2014-07-23 | A kind of more mesh real-time optimization localization methods of the list of moving target |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410350844.9A CN105321162B (en) | 2014-07-23 | 2014-07-23 | A kind of more mesh real-time optimization localization methods of the list of moving target |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105321162A CN105321162A (en) | 2016-02-10 |
| CN105321162B true CN105321162B (en) | 2018-07-27 |
Family
ID=55248482
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410350844.9A Expired - Fee Related CN105321162B (en) | 2014-07-23 | 2014-07-23 | A kind of more mesh real-time optimization localization methods of the list of moving target |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105321162B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110297234B (en) * | 2018-03-22 | 2023-03-14 | 西安航通测控技术有限责任公司 | Networked large-area passive air target intersection determination method and system |
| CN109272482B (en) * | 2018-07-20 | 2021-08-24 | 浙江浩腾电子科技股份有限公司 | Urban intersection vehicle queuing detection system based on sequence images |
| CN112836418B (en) * | 2021-01-15 | 2024-02-20 | 中国人民解放军91550部队 | Aircraft real-time positioning method and system based on incomplete measurement |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101706576A (en) * | 2009-11-13 | 2010-05-12 | 山东大学 | Radar image based moving target morphology detecting and tracking method |
| CN101894378A (en) * | 2010-06-13 | 2010-11-24 | 南京航空航天大学 | Moving target visual tracking method and system based on double ROI (Region of Interest) |
| CN103020983A (en) * | 2012-09-12 | 2013-04-03 | 深圳先进技术研究院 | Human-computer interaction device and method used for target tracking |
| CN103064086A (en) * | 2012-11-04 | 2013-04-24 | 北京工业大学 | Vehicle tracking method based on depth information |
| CN103631698A (en) * | 2013-12-20 | 2014-03-12 | 中安消技术有限公司 | Camera PTZ (pan/tilt/zoom) control method and device for target tracking |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| FR2944629B1 (en) * | 2009-04-17 | 2017-01-20 | Univ De Tech De Troyes | SYSTEM AND METHOD FOR TARGET LOCATION BY A CAMERA NETWORK |
-
2014
- 2014-07-23 CN CN201410350844.9A patent/CN105321162B/en not_active Expired - Fee Related
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101706576A (en) * | 2009-11-13 | 2010-05-12 | 山东大学 | Radar image based moving target morphology detecting and tracking method |
| CN101894378A (en) * | 2010-06-13 | 2010-11-24 | 南京航空航天大学 | Moving target visual tracking method and system based on double ROI (Region of Interest) |
| CN103020983A (en) * | 2012-09-12 | 2013-04-03 | 深圳先进技术研究院 | Human-computer interaction device and method used for target tracking |
| CN103064086A (en) * | 2012-11-04 | 2013-04-24 | 北京工业大学 | Vehicle tracking method based on depth information |
| CN103631698A (en) * | 2013-12-20 | 2014-03-12 | 中安消技术有限公司 | Camera PTZ (pan/tilt/zoom) control method and device for target tracking |
Non-Patent Citations (1)
| Title |
|---|
| 用滑动窗多项式拟合法实时预测运动目标轨迹;张尚剑 等;《光电工程》;20030828;第30卷(第4期);24-27 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105321162A (en) | 2016-02-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105894542B (en) | A kind of online method for tracking target and device | |
| Ai et al. | Visual SLAM in dynamic environments based on object detection | |
| CN108734103A (en) | The detection of moving target and tracking in satellite video | |
| CN110007313A (en) | Obstacle detection method and device based on unmanned plane | |
| Cai et al. | Mobile robot localization using gps, imu and visual odometry | |
| CN105321162B (en) | A kind of more mesh real-time optimization localization methods of the list of moving target | |
| Jia et al. | A Survey of simultaneous localization and mapping for robot | |
| CN112525197B (en) | Ultra-wideband inertial navigation fusion pose estimation method based on graph optimization algorithm | |
| CN115731268A (en) | Unmanned aerial vehicle multi-target tracking method based on visual/millimeter wave radar information fusion | |
| CN111753694B (en) | Unmanned vehicle target searching system and method | |
| Perez-Grau et al. | Long-term aerial robot localization based on visual odometry and radio-based ranging | |
| Li et al. | The application of Yolov4 and a new pedestrian clustering algorithm to implement social distance monitoring during the COVID-19 pandemic | |
| CN108646760B (en) | Monocular vision based mobile robot target tracking and platform control system and method | |
| Zhou et al. | Robust global localization by using global visual features and range finders data | |
| Dai et al. | Analysis of target detection based on UWB NLOS ranging modeling | |
| Zhong et al. | Vision-based 3D aerial target detection and tracking for maneuver decision in close-range air combat | |
| CN114882077A (en) | Target real-time tracking control method and system | |
| CN111780744B (en) | Mobile robot hybrid navigation method, equipment and storage device | |
| Park et al. | Long-term stealth navigation in a security zone where the movement of the invader is monitored | |
| Lu et al. | A new performance index for measuring the effect of single target tracking with Kalman particle filter | |
| Wang et al. | UAV navigation in large-scale GPS-denied bridge environments using fiducial marker-corrected stereo visual-inertial localisation | |
| Lee et al. | Fusion of monocular vision and radio-based ranging for global scale estimation and drift mitigation | |
| Zhang et al. | Deep learning based multi-modal sensing for tracking and state extraction of small quadcopters | |
| Barbaric et al. | Passive ranging using image intensity and contrast measurements | |
| Yuan et al. | Topological mapping and navigation based on visual sensor network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20180727 Termination date: 20210723 |