CN108038868A - Across the visual field method for tracking target of substation's complex environment based on three-dimensional digital model - Google Patents
Across the visual field method for tracking target of substation's complex environment based on three-dimensional digital model Download PDFInfo
- Publication number
- CN108038868A CN108038868A CN201710966899.6A CN201710966899A CN108038868A CN 108038868 A CN108038868 A CN 108038868A CN 201710966899 A CN201710966899 A CN 201710966899A CN 108038868 A CN108038868 A CN 108038868A
- Authority
- CN
- China
- Prior art keywords
- target
- equation
- state
- substation
- visual field
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 36
- 230000000007 visual effect Effects 0.000 title claims abstract description 23
- 238000012544 monitoring process Methods 0.000 claims abstract description 17
- 239000011159 matrix material Substances 0.000 claims description 33
- 230000009897 systematic effect Effects 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 5
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000005183 dynamical system Methods 0.000 claims description 3
- 238000005530 etching Methods 0.000 claims description 3
- 238000002474 experimental method Methods 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 238000012546 transfer Methods 0.000 claims 1
- 230000009466 transformation Effects 0.000 claims 1
- 238000003384 imaging method Methods 0.000 description 3
- 230000007812 deficiency Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/277—Analysis of motion involving stochastic approaches, e.g. using Kalman filters
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
- G06T2207/10021—Stereoscopic video; Stereoscopic image sequence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20076—Probabilistic image processing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30196—Human being; Person
Abstract
The invention discloses a kind of across visual field method for tracking target of substation's complex environment based on three-dimensional digital model, its technical solution is：Monitoring camera is arranged at intervals with the region that substation need to monitor, the adjacent monitoring camera intersects position and the state of shooting operation personnel and maintained equipment, nonlinear system is linearized using Taylor series expansion using Extended Kalman filter, in the case where ignoring high-order error, original system model is changed into the state equation and observational equation represented with errors of form；The equation of motion of moving target is built under substation's three-dimensional system of coordinate using second-order dynamic model, after target monitors for the first time, by the use of multiple monitoring camera initial alignments and be used as EKF tracking initial value, the site error of three-dimensional coordinate is converted using the detected value of corresponding image space target afterwards, and substitute into iteration in state equation, when visual field or visual field switching are temporarily removed for target, target location is estimated by wave filter, higher tracking accuracy can be obtained.
Description
Technical field：
The present invention relates to a kind of field operation in transformer substation method for tracking target, and 3-dimensional digital mould is based on more particularly to one kind
Across the visual field method for tracking target of substation's complex environment of type.
Background technology：
Substation field security management and control has safe prison when substation is worked by the way of manually observing at present
Whether the person of superintending and directing, Real Time Observation working site situation, observation field personnel with charging equipment keep enough safe distance, profit
The method that safe fence is laid used in work of transformer substation region carries out management and control to operating area.But show in substation's actual job
, access way is long in power station, can not carry out complete closure substantially, and staff may reach work by abnormal path
Make region, the risk for adding personnel's mistakenly entering charged chamber and getting an electric shock；When working area is excessive in substation, setting is more
When, site safety care provider deficiency, it is difficult to real-time oversight is carried out to operating area, adds security risk.
In work of transformer substation environment, various equipment are stood in great numbers, and there may be mutually block between equipment and staff
Problem, this will be to accurately identify each target increase difficulty.Simultaneously as the movement of target, can bring target in different phases
The problem of machine visual field switches, this carrys out larger difficulty for the identification of target and continuous track band.Even if using GPS positioning or other areas
Domain location technology, its precision or cost cannot be met the requirements, and can not also solve headroom computational problem.
The content of the invention：
The technical problems to be solved by the invention are：Overcome the deficiencies of the prior art and provide a kind of easily implementation, identification essence
Across the visual field method for tracking target of substation's complex environment based on three-dimensional digital model that is true and reducing human input.
The technical scheme is that：
A kind of across visual field method for tracking target of substation's complex environment based on three-dimensional digital model, needs to monitor in substation
Region be arranged at intervals with monitoring camera, the adjacent monitoring camera intersects position and the shape of shooting operation personnel and maintained equipment
State, realizes that target three-dimensional position tracks using the method for Extended Kalman filter, and the image transmitting of shooting is passed through to central processing unit
Processing and extraction are crossed, with the target of monitoring results personnel, then carries out the tracking of dynamic object.
The Extended Kalman filter linearizes nonlinear system using Taylor series expansion, is missed ignoring high-order
In the case of difference, original system model is changed into the state equation and observational equation represented with errors of form；Moved using second order
Mechanical model builds the equation of motion of moving target under substation's three-dimensional system of coordinate, after target monitors for the first time, using more
A monitoring camera initial alignment and the initial value tracked as EKF, it is three-dimensional using the detected value conversion of corresponding image space target afterwards
The site error of coordinate, and iteration in state equation is substituted into, when visual field or visual field switching are temporarily removed for target, by wave filter
Estimate target location.
The wave filter is the algorithm that a status switch to dynamical system carries out Linear Minimum Variance estimation error, is led to
Dynamic state equation and observational equation are crossed to describe system, it starts to observe using any point as starting point, is filtered using recurrence
The method of ripple calculates；If the state equation and observational equation of linear system are respectively：
State equation：
xk=Axk-1+wk-1 (1-1)
Observational equation：
zk=Hxk+vk (1-2)
Wherein, xkIt is that system mode vector is tieed up in k moment n × 1；zkIt is that observation vector is tieed up in k moment m × 1；A is that n × n is maintained
System state-transition matrix；HkIt is dimension of m m systematic observation matrix；wkIt is the dimension random disturbances noise vector of n × 1 of k etching process；vk
It is the systematic observation noise vector that k moment m × 1 is tieed up；
Herein, wk, vkUsually assume that as independent zero mean Gaussian white noise vector mutually, make QkAnd RkRespectively they
Covariance matrix：
Qk=E { wkwk T} (1-3)
Rk=E { vkvk T} (1-4)
Due to system it has been determined that then A and HkIt is known that and wk-1And vkMeet certain it is assumed that and being known；If PkIt is xk's
Covariance matrix, Pk' it is xkWithError co-variance matrix.
The wave filter is minimized the error covariance of the posterior estimate of the system mode of each moment point k, it
Completed by predicting and correcting two parts, Kalman filter equation is as follows：
(1) predicted portions
Status predication equation：
Error covariance predictive equation：
(2) part is corrected
Kalman gain coefficient equations：
State revision equation：
Covariance update equation：
Pk=[I-KkHk]Pk,k-1 (1-9)
Kalman filter estimates motion state using feedback control system, it is estimated that state sometime, and
Obtain the predicted value of the state；Kalman filter formula divides two parts：Prediction and amendment.Wherein, predicted portions are responsible for utilizing and are worked as
The state of preceding state and error covariance estimation subsequent time, obtains prior estimate；Correct part and be responsible for feedback, by new reality
Border observation considers together with priori estimates, so as to obtain Posterior estimator；After each completion prediction and correcting, by posteriority
Estimate predicts the prior estimate of subsequent time, repeats above step.
Assuming that substation staff motion state parameters for a certain moment target position and speed, tracking
In object procedure, since the time interval of adjacent two field pictures is shorter, target state change in so short time interval
It is smaller, it can be assumed that target is uniform motion in unit interval；
Define Kalman filter system mode xkFor a 4 dimensional vector xk=(xsk,ysk,xvk,yvk)T., xsk,ysk,
xvk,yvkRespectively position and speed of the pixel in X-axis and Y direction, by images match, can only obtain the position of target
Information, so defining two-dimensional observation vector zk=(xpk,ypk)T, represent the coordinate that matching obtains；
Since target is uniform motion in unit interval, definition status shift-matrix A is:
Wherein Δ t represents the time interval between two continuous frames image；
From system mode and the relation of observation state, observing matrix HkFor：
W is assumed abovek, vkUsually mutual independent zero mean Gaussian white noise vector, therefore set their covariance square
Battle array be respectively：
It is divided into four-stage using the movement of Kalman filter estimation target during tracking, is respectively wave filter
Initialization, status predication, matching and state revision.
The specific implementation step of four-stage is as follows:
The first step：Initialization, will initialize wave filter when first time is using Kalman filter, by x0Assign just
It is worth the initial position and speed for target, in the case where speed is unknown, O can be set to, and record the present image moment, at the same time
If initial error covariance P0=0；
Second step：Prediction, before carrying out matching search in every the two field picture newly inputted, record and previous frame image when
Between interval of delta t, substitute into state-transition matrix, and by itself and x0Status predication equation is substituted into, predicts the motion state of current goalThe error of prediction is denoted as Δ pk=wk-sk, for the calculating of region of search in next frame, by state-transition matrix and association side
Poor matrix substitutes into error covariance predictive equation together, predicts new error covariance；
3rd step：Matching, set withIn (xsk,ysk) centered on region be region of search, sought in the region
Best match position is looked for, finds most suitable moving target, target area image is copied to Tk+1, and the target area upper left corner
First pixel coordinate is two-dimensional observation vector (xpk,ypk), substitute into state revision equation and obtain (xsk+1,ysk+1), count at the same time
Calculate the measuring speed v of targetk+1=(sk+1-sk)/Δt；
4th step：Correct, the coefficient of Kalman filter gain coefficient equation is obtained, by zk=(xpk,ypk)TSubstitute into shape
In state update equation, obtain by the currently practical revised state vector of observation, while error is corrected by covariance update equation
Covariance matrix.
Completed using iir filter to velocity deviation Δ vkAnd offset deviation Δ pkAmendment, its formula is：
In formula, α, β are constant, and 0≤α, β≤1.
The region of search in every frame is greatly reduced after Kalman filter is predicted, improves the effect of processing in next step
Rate, region of search are defined as：
Search starting point abscissa：The position of x filter predictions subtracts offset deviation Δ pk+1ρ times of x-axis component；
Search starting point ordinate：The position of Y filter predictions subtracts offset deviation Δ pk+1ρ times of y-axis component；
Region of search x-axis direction width：Sreachwidth=2 ρ x Δs pk+1The width of+template；
Region of search y-axis direction height：Sreachheight=2 ρ y Δs pk+1The height of+template；
Experiments verify that take 1/2 can tenacious tracking by wherein ρ.Search range is maximum no more than 1/4 image in principle
Frame sign, the minimum target sizes that cannot be less than 2 times.Second step is then return to, prediction, matching, makeover process is repeated, completes fortune
The tracing task of moving-target.
The beneficial effects of the invention are as follows：
1st, the difference of the invention for same point image space on two or more cameras, to determine its coordinate position,
Dynamic is determined with static state, so as to know the position coordinates of dynamic point, realizes the purpose of tracking.
2nd, the present invention realizes that target three-dimensional position tracks using the method for Extended Kalman filter (EKF), utilizes target
The equation of motion (dynamics constraint condition) and image motion monitoring method joint are realized, obtain higher tracking accuracy, while right
The moving target of coincidence has stronger separating capacity.
3rd, the data that the present invention is obtained by multiple phase machine simultaneously imagings, to position target, to monitor identification personnel positions,
It has a wide range of application, implementation easy to spread, with good economic efficiency.
Brief description of the drawings：
Fig. 1 is the binocular monitoring camera positioning schematic of the present invention.
Embodiment：
Embodiment：Referring to Fig. 1.
Across the visual field method for tracking target of substation's complex environment based on three-dimensional digital model, its technical solution are：Becoming
The region that power station need to monitor is arranged at intervals with monitoring camera, and the adjacent monitoring camera intersects shooting operation personnel and maintained equipment
Position and state, using the method for Extended Kalman filter realize target three-dimensional position track, the image transmitting of shooting is in
Central processor, by handling and extracting, with the target of monitoring results personnel, then carries out the tracking of dynamic object.
Based on single picture (two dimension) can not be realized, it is necessary to take since target positioning needs three-dimensional data
The method of more mesh cameras, i.e., the data obtained by two or more phase machine simultaneously imagings, to position target.Its basic principle
It is：For the difference of same point image space on two or more cameras, to determine its coordinate position (as shown in Figure 1).Its
Important premise is：First, the coordinate position and direction (posture) of known camera, two are to determine same target in different cameral
Imaging point (same place).In power plant application, due to higher to positioning accuracy request, using Extended Kalman filter (EKF)
Method realizes that target three-dimensional position tracks.Its basic principle is the equation of motion (dynamics constraint condition) and image using target
Motion monitoring method joint is realized.
Extended Kalman filter linearizes nonlinear system using Taylor series expansion, is ignoring high-order error
In the case of, original system model is changed into the state equation and observational equation represented with errors of form；Using second-order dynamic
Model builds the equation of motion of moving target under substation's three-dimensional system of coordinate, after target monitors for the first time, utilizes multiple prisons
Control camera initial alignment and as the initial value of EKF tracking, utilize the detected value conversion three-dimensional coordinate of corresponding image space target afterwards
Site error, and substitute into iteration in state equation, temporarily remove visual field for target or when visual field switches, estimated by wave filter
Go out target location.
Wave filter is the algorithm that a status switch to dynamical system carries out Linear Minimum Variance estimation error, by dynamic
The state equation and observational equation of state describes system, it starts to observe using any point as starting point, using recursive filtering
Method calculates；If the state equation and observational equation of linear system are respectively：
State equation：
xk=Axk-1+wk-1 (1-1)
Observational equation：
zk=Hxk+vk (1-2)
Wherein, xkIt is that system mode vector is tieed up in k moment n × 1；zkIt is that observation vector is tieed up in k moment m × 1；A is that n × n is maintained
System state-transition matrix；HkIt is dimension of m m systematic observation matrix；wkIt is the dimension random disturbances noise vector of n × 1 of k etching process；vk
It is the systematic observation noise vector that k moment m × 1 is tieed up；
Herein, wk, vkUsually assume that as independent zero mean Gaussian white noise vector mutually, make QkAnd RkRespectively they
Covariance matrix：
Qk=E { wkwk T} (1-3)
Rk=E { vkvk T} (1-4)
Due to system it has been determined that then A and HkIt is known that and wk-1And vkMeet certain it is assumed that and being known；If PkIt is xk's
Covariance matrix, Pk' it is xkWithError co-variance matrix.
Wave filter is minimized the error covariance of the posterior estimate of the system mode of each moment point k, it is by pre-
Survey and correct two parts to complete, Kalman filter equation is as follows：
(1) predicted portions
Status predication equation：
Error covariance predictive equation：
(2) part is corrected
Kalman gain coefficient equations：
State revision equation：
Covariance update equation：
Pk=[I-KkHk]Pk,k-1 (1-9)
Kalman filter estimates motion state using feedback control system, it is estimated that state sometime, and
Obtain the predicted value of the state；Kalman filter formula divides two parts：Prediction and amendment.Wherein, predicted portions are responsible for utilizing and are worked as
The state of preceding state and error covariance estimation subsequent time, obtains prior estimate；Correct part and be responsible for feedback, by new reality
Border observation considers together with priori estimates, so as to obtain Posterior estimator；After each completion prediction and correcting, by posteriority
Estimate predicts the prior estimate of subsequent time, repeats above step.
Assuming that substation staff motion state parameters for a certain moment target position and speed, tracking
In object procedure, since the time interval of adjacent two field pictures is shorter, target state change in so short time interval
It is smaller, it can be assumed that target is uniform motion in unit interval；
Define Kalman filter system mode xkFor a 4 dimensional vector xk=(xsk,ysk,xvk,yvk)T., xsk,ysk,
xvk,yvkRespectively position and speed of the pixel in X-axis and Y direction, by images match, can only obtain the position of target
Information, so defining two-dimensional observation vector zk=(xpk,ypk)T, represent the coordinate that matching obtains；
Since target is uniform motion in unit interval, definition status shift-matrix A is:
Wherein Δ t represents the time interval between two continuous frames image；
From system mode and the relation of observation state, observing matrix HkFor：
W is assumed abovek, vkUsually mutual independent zero mean Gaussian white noise vector, therefore set their covariance square
Battle array be respectively：
It is divided into four-stage using the movement of Kalman filter estimation target during tracking, is respectively wave filter
Initialization, status predication, matching and state revision.
The specific implementation step of four-stage is as follows:
The first step：Initialization, will initialize wave filter when first time is using Kalman filter, by x0Assign just
It is worth the initial position and speed for target, in the case where speed is unknown, O can be set to, and record the present image moment, at the same time
If initial error covariance P0=0；
Second step：Prediction, before carrying out matching search in every the two field picture newly inputted, record and previous frame image when
Between interval of delta t, substitute into state-transition matrix, and by itself and x0Status predication equation is substituted into, predicts the motion state of current goalThe error of prediction is denoted as Δ pk=wk-sk, for the calculating of region of search in next frame, by state-transition matrix and association side
Poor matrix substitutes into error covariance predictive equation together, predicts new error covariance；
3rd step：Matching, set withIn (xsk,ysk) centered on region be region of search, sought in the region
Best match position is looked for, finds most suitable moving target, target area image is copied to Tk+1, and the target area upper left corner
First pixel coordinate is two-dimensional observation vector (xpk,ypk), substitute into state revision equation and obtain (xsk+1,ysk+1), count at the same time
Calculate the measuring speed v of targetk+1=(sk+1-sk)/Δt；
4th step：Correct, the coefficient of Kalman filter gain coefficient equation is obtained, by zk=(xpk,ypk)TSubstitute into shape
In state update equation, obtain by the currently practical revised state vector of observation, while error is corrected by covariance update equation
Covariance matrix.
Completed using iir filter to velocity deviation Δ vkAnd offset deviation Δ pkAmendment, its formula is：
In formula, α, β are constant, and 0≤α, β≤1.
The region of search in every frame is greatly reduced after Kalman filter is predicted, improves the effect of processing in next step
Rate, region of search are defined as：
Search starting point abscissa：The position of x filter predictions subtracts offset deviation Δ pk+1ρ times of x-axis component；
Search starting point ordinate：The position of Y filter predictions subtracts offset deviation Δ pk+1ρ times of y-axis component；
Region of search x-axis direction width：Sreachwidth=2 ρ x Δs pk+1The width of+template；
Region of search y-axis direction height：Sreachheight=2 ρ y Δs pk+1The height of+template；
Experiments verify that take 1/2 can tenacious tracking by wherein ρ.Search range is maximum no more than 1/4 image in principle
Frame sign, the minimum target sizes that cannot be less than 2 times.Second step is then return to, prediction, matching, makeover process is repeated, completes fortune
The tracing task of moving-target.
The above described is only a preferred embodiment of the present invention, not make limitation in any form to the present invention, it is all
It is any simple modification, equivalent change and modification made according to the technical spirit of the present invention to above example, still falls within
In the range of technical solution of the present invention.
Claims (7)
1. a kind of across visual field method for tracking target of substation's complex environment based on three-dimensional digital model, it is characterized in that：In power transformation
The region that standing to monitor is arranged at intervals with monitoring camera, and the adjacent monitoring camera intersects shooting operation personnel and maintained equipment
Position and state, realize that target three-dimensional position tracks using the method for Extended Kalman filter, and the image transmitting of shooting is to central
Processor, by handling and extracting, with the target of monitoring results personnel, then carries out the tracking of dynamic object.
2. substation's complex environment across visual field method for tracking target according to claim 1 based on three-dimensional digital model,
It is characterized in that：The Extended Kalman filter linearizes nonlinear system using Taylor series expansion, is ignoring high-order
In the case of error, original system model is changed into the state equation and observational equation represented with errors of form；Using second order
Kinetic model builds the equation of motion of moving target under substation's three-dimensional system of coordinate, after target monitors for the first time, utilizes
Multiple monitoring camera initial alignments and the initial value tracked as EKF, utilize the detected value conversion three of corresponding image space target afterwards
The site error of dimension coordinate, and iteration in state equation is substituted into, when visual field or visual field switching are temporarily removed for target, by filtering
Device estimates target location.
3. substation's complex environment across visual field method for tracking target according to claim 2 based on three-dimensional digital model,
It is characterized in that：The wave filter is the algorithm that a status switch to dynamical system carries out Linear Minimum Variance estimation error,
System is described by dynamic state equation and observational equation, it starts to observe using any point as starting point, using recurrence
The method of filtering calculates；If the state equation and observational equation of linear system are respectively：
State equation：
xk=Axk-1+wk-1 (1-1)
Observational equation：
zk=Hxk+vk (1-2)
Wherein, xkIt is that system mode vector is tieed up in k moment n × 1；zkIt is that observation vector is tieed up in k moment m × 1；A is that n × n maintains system shape
State transfer matrix；HkIt is dimension of m m systematic observation matrix；wkIt is the dimension random disturbances noise vector of n × 1 of k etching process；vkIt is k
The systematic observation noise vector that moment m × 1 is tieed up；
Herein, wk, vkUsually assume that as independent zero mean Gaussian white noise vector mutually, make QkAnd RkRespectively their covariances
Matrix：
Qk=E { wkwk T} (1-3)
Rk=E { vkvk T} (1-4)
Due to system it has been determined that then A and HkIt is known that and wk-1And vkMeet certain it is assumed that and being known；If PkIt is xkCovariance
Matrix, Pk' it is xkWithError co-variance matrix.
4. substation's complex environment across visual field method for tracking target according to claim 3 based on three-dimensional digital model,
It is characterized in that：The wave filter is minimized the error covariance of the posterior estimate of the system mode of each moment point k, it
Completed by predicting and correcting two parts, Kalman filter equation is as follows：
(1) predicted portions
Status predication equation：
Error covariance predictive equation：
(2) part is corrected
Kalman gain coefficient equations：
State revision equation：
Covariance update equation：
Pk=[I-KkHk]Pk,k-1 (1-9)
Kalman filter estimates motion state using feedback control system, it is estimated that state sometime, and obtain
The predicted value of the state；Kalman filter formula divides two parts：Prediction and amendment.Wherein, predicted portions are responsible for utilizing currently
The state of state and error covariance estimation subsequent time, obtains prior estimate；Correct part and be responsible for feedback, by new actual sight
Measured value considers together with priori estimates, so as to obtain Posterior estimator；After each completion prediction and correcting, by Posterior estimator
The prior estimate of value prediction subsequent time, repeats above step.
5. substation's complex environment across visual field method for tracking target according to claim 4 based on three-dimensional digital model,
It is characterized in that：Assuming that substation staff motion state parameters for a certain moment target position and speed, with
In track object procedure, since the time interval of adjacent two field pictures is shorter, target state in so short time interval becomes
Change smaller, it can be assumed that target is uniform motion in unit interval；
Define Kalman filter system mode xkFor a 4 dimensional vector xk=(xsk,ysk,xvk,yvk)T., xsk,ysk,xvk,
yvkRespectively position and speed of the pixel in X-axis and Y direction, by images match, can only obtain the position letter of target
Breath, so defining two-dimensional observation vector zk=(xpk,ypk)T, represent the coordinate that matching obtains；
Since target is uniform motion in unit interval, definition status shift-matrix A is:
Wherein Δ t represents the time interval between two continuous frames image；
From system mode and the relation of observation state, observing matrix HkFor：
W is assumed abovek, vkUsually mutual independent zero mean Gaussian white noise vector, therefore set their covariance matrix point
It is not：
It is divided into four-stage using the movement of Kalman filter estimation target during tracking, is respectively the initial of wave filter
Change, status predication, matching and state revision.
6. substation's complex environment across visual field method for tracking target according to claim 5 based on three-dimensional digital model,
It is characterized in that：The specific implementation step of four-stage is as follows:
The first step：Initialization, will initialize wave filter when first time is using Kalman filter, by x0Assigning initial value is
The initial position and speed of target, in the case where speed is unknown, can be set to O, and record the present image moment, while set just
Beginning error covariance P0=0；
Second step：Prediction, before carrying out matching search in the every two field picture newly inputted, between the time of record and previous frame image
Every Δ t, state-transition matrix is substituted into, and by itself and x0Status predication equation is substituted into, predicts the motion state of current goalWill
The error of prediction is denoted as Δ pk=wk-sk, for the calculating of region of search in next frame, by state-transition matrix and covariance square
Battle array substitutes into error covariance predictive equation together, predicts new error covariance；
3rd step：Matching, set withIn (xsk,ysk) centered on region be region of search, found in the region optimal
Matched position, finds most suitable moving target, and target area image is copied to Tk+1, and first, the target area upper left corner
Pixel coordinate is two-dimensional observation vector (xpk,ypk), substitute into state revision equation and obtain (xsk+1,ysk+1), while calculate target
Measuring speed vk+1=(sk+1-sk)/Δt；
4th step：Correct, the coefficient of Kalman filter gain coefficient equation is obtained, by zk=(xpk,ypk)TSubstitute into state revision
In equation, obtain by the currently practical revised state vector of observation, while error covariance is corrected by covariance update equation
Matrix.
7. substation's complex environment across visual field method for tracking target according to claim 6 based on three-dimensional digital model,
It is characterized in that：Completed using iir filter to velocity deviation Δ vkAnd offset deviation Δ pkAmendment, its formula is：
In formula, α, β are constant, and 0≤α, β≤1.
The region of search in every frame is greatly reduced after Kalman filter is predicted, improves the efficiency of processing in next step,
Region of search is defined as：
Search starting point abscissa：The position of x filter predictions subtracts offset deviation Δ pk+1ρ times of x-axis component；
Search starting point ordinate：The position of Y filter predictions subtracts offset deviation Δ pk+1ρ times of y-axis component；
Region of search x-axis direction width：Sreachwidth=2 ρ x Δs pk+1The width of+template；
Region of search y-axis direction height：Sreachheight=2 ρ y Δs pk+1The height of+template；
Experiments verify that take 1/2 can tenacious tracking by wherein ρ.Search range is maximum big no more than 1/4 picture frame in principle
It is small, the minimum target sizes that cannot be less than 2 times.Second step is then return to, repeats prediction, matching, makeover process, completes movement mesh
Target tracing task.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710966899.6A CN108038868A (en) | 2017-10-17 | 2017-10-17 | Across the visual field method for tracking target of substation's complex environment based on three-dimensional digital model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710966899.6A CN108038868A (en) | 2017-10-17 | 2017-10-17 | Across the visual field method for tracking target of substation's complex environment based on three-dimensional digital model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108038868A true CN108038868A (en) | 2018-05-15 |
Family
ID=62093112
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710966899.6A Pending CN108038868A (en) | 2017-10-17 | 2017-10-17 | Across the visual field method for tracking target of substation's complex environment based on three-dimensional digital model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108038868A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109003292A (en) * | 2018-06-25 | 2018-12-14 | 华南理工大学 | A kind of motion target tracking method based on switch Kalman filter |
| CN109520519A (en) * | 2018-10-17 | 2019-03-26 | 安徽立卓智能电网科技有限公司 | A kind of substation's fire-fighting safety route planning method |
| CN110442831A (en) * | 2019-07-31 | 2019-11-12 | 中国人民解放军国防科技大学 | Space non-cooperative target space-based search method based on nonlinear deviation evolution |
| CN110466790A (en) * | 2019-08-30 | 2019-11-19 | 上海电机学院 | A kind of unmanned plane target tracking system based on machine vision |
| CN111781556A (en) * | 2020-05-18 | 2020-10-16 | 中国民用航空总局第二研究所 | Distributed time difference positioning method and device based on TDOA filtering of target signal |
| CN112630729A (en) * | 2020-12-11 | 2021-04-09 | 杭州博镨科技有限公司 | Method for positioning and tracking indoor human target based on thermopile sensor |
| CN114390431B (en) * | 2022-01-11 | 2024-04-26 | 上海则芯半导体科技有限公司 | Two-dimensional relative positioning method and device for two base stations based on ultra-wideband |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103281476A (en) * | 2013-04-22 | 2013-09-04 | 中山大学 | Television image moving target-based automatic tracking method |
| US20160265903A1 (en) * | 2013-11-20 | 2016-09-15 | Tianjiin University | Accuracy traceability method based on precision coordinate control network for workshop measurement positioning system |
| CN106780542A (en) * | 2016-12-29 | 2017-05-31 | 北京理工大学 | A kind of machine fish tracking of the Camshift based on embedded Kalman filter |
-
2017
- 2017-10-17 CN CN201710966899.6A patent/CN108038868A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103281476A (en) * | 2013-04-22 | 2013-09-04 | 中山大学 | Television image moving target-based automatic tracking method |
| US20160265903A1 (en) * | 2013-11-20 | 2016-09-15 | Tianjiin University | Accuracy traceability method based on precision coordinate control network for workshop measurement positioning system |
| CN106780542A (en) * | 2016-12-29 | 2017-05-31 | 北京理工大学 | A kind of machine fish tracking of the Camshift based on embedded Kalman filter |
Non-Patent Citations (2)
| Title |
|---|
| 姜文涛等: "一种基于特征点跟踪的电子稳像算法", 《应用光学》, vol. 31, no. 1, 31 January 2010 (2010-01-31), pages 73 - 77 * |
| 梁雪梅: "无人值守变电站智能视频监测系统设计", 《中国优秀硕士学位论文全文数据库》, 15 September 2010 (2010-09-15) * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109003292A (en) * | 2018-06-25 | 2018-12-14 | 华南理工大学 | A kind of motion target tracking method based on switch Kalman filter |
| CN109003292B (en) * | 2018-06-25 | 2022-01-18 | 华南理工大学 | Moving target tracking method based on switch Kalman filter |
| CN109520519A (en) * | 2018-10-17 | 2019-03-26 | 安徽立卓智能电网科技有限公司 | A kind of substation's fire-fighting safety route planning method |
| CN110442831A (en) * | 2019-07-31 | 2019-11-12 | 中国人民解放军国防科技大学 | Space non-cooperative target space-based search method based on nonlinear deviation evolution |
| CN110466790A (en) * | 2019-08-30 | 2019-11-19 | 上海电机学院 | A kind of unmanned plane target tracking system based on machine vision |
| CN111781556A (en) * | 2020-05-18 | 2020-10-16 | 中国民用航空总局第二研究所 | Distributed time difference positioning method and device based on TDOA filtering of target signal |
| CN111781556B (en) * | 2020-05-18 | 2023-08-18 | 中国民用航空总局第二研究所 | Distributed time difference positioning method and device based on target signal TDOA filtering |
| CN112630729A (en) * | 2020-12-11 | 2021-04-09 | 杭州博镨科技有限公司 | Method for positioning and tracking indoor human target based on thermopile sensor |
| CN114390431B (en) * | 2022-01-11 | 2024-04-26 | 上海则芯半导体科技有限公司 | Two-dimensional relative positioning method and device for two base stations based on ultra-wideband |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108038868A (en) | Across the visual field method for tracking target of substation's complex environment based on three-dimensional digital model | |
| Comport et al. | Robust model-based tracking for robot vision | |
| Winterhalter et al. | Accurate indoor localization for RGB-D smartphones and tablets given 2D floor plans | |
| Chen et al. | Localization for multirobot formations in indoor environment | |
| Wu et al. | Recovery of the 3-d location and motion of a rigid object through camera image (an Extended Kalman Filter approach) | |
| McLauchlan et al. | Active camera calibration for a head-eye platform using the variable state-dimension filter | |
| Voigt et al. | Robust embedded egomotion estimation | |
| CN102763132A (en) | Three-dimensional measurement apparatus, processing method, and non-transitory computer-readable storage medium | |
| KR20110092502A (en) | Apparatus and method for surveillance and tracking | |
| CN112288815B (en) | Target die position measurement method, system, storage medium and device | |
| Ibarguren et al. | Multiple target tracking based on particle filtering for safety in industrial robotic cells | |
| CN113733100B (en) | Target positioning method, device, equipment and storage medium of inspection operation robot | |
| Yoon et al. | 3D position estimation of drone and object based on QR code segmentation model for inventory management automation | |
| CN106127119B (en) | Joint probabilistic data association method based on color image and depth image multiple features | |
| Hashemi et al. | Particle filter based localization of the nao biped robots | |
| Zaccaria et al. | Multi-robot multiple camera people detection and tracking in automated warehouses | |
| CN107818587A (en) | A kind of machine vision high-precision locating method based on ROS | |
| Babic et al. | Indoor RFID localization improved by motion segmentation | |
| Bishop et al. | Vision-based target tracking and surveillance with robust set-valued state estimation | |
| Lasguignes et al. | Icp localization and walking experiments on a talos humanoid robot | |
| Saeedi et al. | 3D localization and tracking in unknown environments | |
| Cretual et al. | Image-based visual servoing by integration of dynamic measurements | |
| CN113989335A (en) | Method for automatically positioning workers in factory building | |
| Chen et al. | The evaluation of the gaussian mixture probability hypothesis density approach for multi-target tracking | |
| Chen et al. | High precision pose measurement for humanoid robot based on PnP and OI algorithms |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180515 |
|
| RJ01 | Rejection of invention patent application after publication |