CN109902543A - Target trajectory estimation method, device and Target Tracking System - Google Patents
Target trajectory estimation method, device and Target Tracking System Download PDFInfo
- Publication number
- CN109902543A CN109902543A CN201711306051.7A CN201711306051A CN109902543A CN 109902543 A CN109902543 A CN 109902543A CN 201711306051 A CN201711306051 A CN 201711306051A CN 109902543 A CN109902543 A CN 109902543A
- Authority
- CN
- China
- Prior art keywords
- target
- image
- convergence
- region
- algorithm
- 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
Abstract
The disclosure proposes a kind of target trajectory estimation method, device and Target Tracking System, is related to air vehicle technique field.A kind of target trajectory estimation method of the disclosure comprises determining that the region in image where target；Target area convergence algorithm is carried out based on the region where target, obtains the convergence location information of target；The spatial position of target is determined according to convergence location information；According to the spatial position of target, the motion profile of target is estimated by Kalman Filter Estimation algorithm.By such method, the region of the predetermined shape where target can be first determined in the picture, operation is further restrained by target area and improves the accuracy that region determines, so as to improve the accuracy of Target space position confirmation, improves the reliability of target following.
Description
Technical field
This disclosure relates to air vehicle technique field, especially a kind of target trajectory estimation method, device and target with
Track system.
Background technique
The target following of frame to frame has important application in unmanned plane landing guidance.The algorithm using in image by with
The contextual relation of track object, it is assumed that the movement of object is gentle, and the target position variation between adjacent two frame is little, Ke Yiyou
Effect improves the real-time of algorithm.
The target tracking algorism ROI (Region of Interest, area-of-interest) of frame to frame is selected and target positions
Situation is as shown in Figure 1A.In t frame, with the region bounding box Boundingbox (the solid line region in Figure 1A left figure) of t-1 frame
Based on, amplification k (k is positive number) times is as t frame unmanned plane target ROI region (dashed region in Figure 1A), in the ROI
The final Boundingbox for determining target in t frame is tracked in region, such as solid box position in Figure 1A right figure.
Summary of the invention
Inventors have found that in the related technology frame to frame track algorithm this assumes that be tracked object adjacent two
Change in location between frame image is little, and motion profile is smooth.In the practical application scene of unmanned plane landing guidance
In, due to the quick movement of aircraft, when turntable tracking effect is bad, and wide-angle is beated, the ROI region of traditional frame to frame
Method of determination may be because that ROI region does not include unmanned plane target and the tracking to unmanned plane is caused to fail, as shown in Figure 1B.
One purpose of the disclosure is to improve the reliability of target following by the accuracy for improving target positioning.
According to one embodiment of the disclosure, a kind of target trajectory estimation method is proposed, comprising: determine in image
Region where target；Target area convergence algorithm is carried out based on the region where target, obtains the convergence position letter of target
Breath；The spatial position of target is determined according to convergence location information；According to the spatial position of target, pass through Kalman Filter Estimation
The motion profile of algorithm estimation target.
Optionally, target is determined using TLD (Tracking-Learning-Detection, tracking study detection) algorithm
Region in image where target, the shape in region include rectangle or circle.
Optionally, rectangle frame is based on using active contour following algorithm and carries out target area convergence algorithm, obtain target
Restrain location information.
Optionally, further includes: morphologic filtering pretreatment is carried out to image, so as to true according to image after treatment
Rectangle frame where setting the goal.
Optionally, to image carry out morphologic filtering pretreatment include: respectively obtain image opening operation processing result and
Closed operation processing result；Closed operation processing result and opening operation processing result are subtracted each other, image after treatment is obtained.
Optionally, determine that the spatial position of target includes: to carry out cross-correlation according to convergence position according to convergence location information
Filtering, determines the center of target；The spatial position of target is determined according to center.
Optionally, carrying out cross correlation filter according to convergence position includes: to change to filter by translation to determine the newest of target
Position；The dimensional variation situation for determining target is filtered by dimensional variation.
By such method, the region that can first determine the predetermined shape where target in the picture, further passes through
Target area convergence operation improves the accuracy that region determines and mentions so as to improve the accuracy of Target space position confirmation
The reliability of high target following.
According to another embodiment of the present disclosure, a kind of target trajectory estimation device is proposed, comprising: region determines
Unit is configured to determine that in image the region where target；Convergence algorithm unit is configured as based on the area where target
Domain carries out target area convergence algorithm, obtains the convergence location information of target；Space orientation unit, is configured as according to convergence
Location information determines the spatial position of target；Track estimation unit is configured as the spatial position according to target, passes through karr
The motion profile of graceful filtering algorithm for estimating estimation target.
Optionally, area determination unit is configured as determining the region in target image where target using TLD algorithm,
The shape in region includes rectangle or circle.
Optionally, tracking study detection is configured as carrying out target area based on rectangle frame using active contour following algorithm
Domain convergence algorithm obtains the convergence location information of target.
Optionally, further includes: pretreatment unit, for carrying out morphologic filtering pretreatment to image, so as to according to process
Treated, and image determines the rectangle frame where target.
Optionally, pretreatment unit includes: opening operation subelement, for obtaining the opening operation processing result of image；Close fortune
Operator unit, for obtaining the closed operation processing result of image；It handles image and obtains subelement, tied for handling closed operation
Fruit is subtracted each other with opening operation processing result, obtains image after treatment.
Optionally, space orientation unit includes: cross correlation filter subelement, for carrying out cross-correlation according to convergence position
Filtering, determines the center of target；Locator unit, for determining the spatial position of target according to center.
Optionally, cross correlation filter subelement is used for: being changed by translation and is filtered the latest position for determining target；Pass through
Dimensional variation filters the dimensional variation situation for determining target.
According to another embodiment of the disclosure, a kind of target trajectory estimation device is proposed, comprising: memory；With
And it is coupled to the processor of memory, what processor was configured as being mentioned above based on the instruction execution for being stored in memory
Any one target trajectory estimation method.
Such device can first determine the region of the predetermined shape where target in the picture, further pass through target
Region convergence operation improves the accuracy that region determines, so as to improve the accuracy of Target space position confirmation, improves mesh
Mark the reliability of tracking.
According to the further embodiment of the disclosure, proposes a kind of computer readable storage medium, be stored thereon with computer
Program instruction realizes any one the target trajectory estimation method being mentioned above when the instruction is executed by processor
Step.
Such computer readable storage medium can be improved Target space position confirmation by executing instruction thereon
Accuracy, improve the reliability of target following.
In addition, proposing a kind of Target Tracking System according to one embodiment of the disclosure, comprising: what is be mentioned above appoints
It anticipates a kind of target trajectory estimation device；With, image acquiring device, it is configured as obtaining the image of target；Turntable is matched
It is set to carrying image acquiring device, target rotational is followed according to the estimated result of target trajectory estimation device.
Such Target Tracking System can be improved the accuracy of Target space position confirmation, improve target trajectory
The accuracy of estimation improves mesh to avoid ROI region from not including unmanned plane target and the tracking to unmanned plane is caused to fail
Mark the reliability of tracking.
Detailed description of the invention
Attached drawing described herein is used to provide further understanding of the disclosure, constitutes a part of this disclosure, this public affairs
The illustrative embodiments and their description opened do not constitute the improper restriction to the disclosure for explaining the disclosure.In the accompanying drawings:
Figure 1A is to track successful schematic diagram using target tracking algorism in the related technology.
Figure 1B is in the related technology using the schematic diagram of target tracking algorism tracking failure.
Fig. 2 is the flow chart of one embodiment of the target trajectory estimation method of the disclosure.
Fig. 3 is the flow chart of another embodiment of the target trajectory estimation method of the disclosure.
Fig. 4 is the schematic diagram of one embodiment of the target trajectory estimation device of the disclosure.
Fig. 5 is the schematic diagram of another embodiment of the target trajectory estimation device of the disclosure.
Fig. 6 is the schematic diagram of another embodiment of the target trajectory estimation device of the disclosure.
Fig. 7 is the schematic diagram of the further embodiment of the target trajectory estimation device of the disclosure.
Fig. 8 is the schematic diagram of one embodiment of the Target Tracking System of the disclosure.
Fig. 9 is the schematic diagram of one embodiment of the Target Tracking System operational process of the disclosure.
Specific embodiment
Below by drawings and examples, the technical solution of the disclosure is described in further detail.
The flow chart of one embodiment of the target trajectory estimation method of the disclosure is as shown in Figure 2.
In step 201, the region in image where target is determined.The shape in region can be predetermined shape.At one
In embodiment, predetermined shape can be rectangle, or be circle.In one embodiment, mesh can be determined using TLD algorithm
Region in logo image where target.
In step 202, target area convergence algorithm is carried out based on the region where target, obtains the convergence position of target
Confidence breath.In one embodiment, it can be based on rectangle frame using active contour following algorithm and carries out target area convergence fortune
It calculates.
In step 203, the spatial position of target is determined according to convergence location information.In one embodiment, if image
For binocular vision image, then can be being acquired in conjunction with image capture device by fixed to rower is clicked through in two images accordingly
Angle when image determines the space coordinate of target.
In step 204, according to the spatial position of target, the movement of target is estimated by Kalman Filter Estimation algorithm
Track.
By such method, the region that can first determine the predetermined shape where target in the picture, further passes through
Target area convergence operation improves the accuracy that region determines and mentions so as to improve the accuracy of Target space position confirmation
The reliability of high target following.
It in one embodiment, can be right using TLD algorithm during region in determining image where target
Unique objects in continuous image sequence are uninterruptedly tracked, which includes following three parts:
1. tracker (Tracker), the movement of tracked object is estimated by sequential frame image, it is assumed here that frame with
The relative motion of target is limited between frame, and tracks target and keep visible always in the picture.Tracker is it is possible that chase after
The situation of track failure, and can not restore to track after target jumps out camera fields of view.
2. each frame image is considered as independence, and all carries out full surface sweeping to each frame image by detector (Detector), with
Find out all candidate samples similar with target appearance in image.The detector can generate Type Ⅰ Ⅱ error: the positive sample of mistake
The negative sample (False Negatives) of (False Positives) and mistake.
3. learning (Learning), learning process observes the execution of tracker and detector in real time, and estimates detector
Mistake generates training examples to enable in future and avoids similar mistake.Learning object hypothesis tracker and detector have can
Can occur unsuccessfully executing.The introducing of learning object can make detector generate more tracking target appearances, to distinguish back
Scape.
By such method, no matter object picture whether away from keyboard, or be blocked and deformation occurs, not
It will affect tracking effect, to reduce the probability of tracking failure.
It in one embodiment, can be with during carrying out target area convergence algorithm based on the region where target
The further toe-in of area information for being exported TLD method using active contour tracing method, to obtain more accurate position
Confidence breath.
Traditional active contour method is mainly used for image segmentation, and the initial position of Active contour models is often identical bits
Set a circle or rectangle.And the initial position during target identification is in addition to the rectangle frame of TLD method offer before use
Except, the revised active contour model of previous frame can also be used.
In practical iterative process, to guarantee real-time, it is also necessary to be set to termination condition.Iteration is considered first
Stopping, if the length variable quantity of the Active contour models of two continuous frames be less than certain threshold value, need to stop iteration, at this time
Think contour line near expectation target.Secondly as unmanned plane target is in remote imaging, target is smaller and clear
Clear degree is poor, it is therefore desirable to stop in time when Active contour models length is less than a certain threshold value, Active contour models is avoided to restrain
Excessively, or even it can be retracted to a pixel, influences to calculate.
Due to the particularity of unmanned plane imaging, in determining target area, the corresponding pixel of the geometric center in region is past
Toward the position for being not target movement guide point (such as unmanned plane head), therefore certain error usually is brought to resolving, and this
A little errors are likely to result in that target next can not be being successfully tracked.Contour line can be made close to target wheel by convergence algorithm
Exterior feature, to improve the accuracy that target position determines.
It in one embodiment, can also be by the method for cross correlation filter come further true after obtaining convergence position
Set the goal center, while by the analysis to target transverse and longitudinal direction scale, further increasing mark precision.Cross-correlation filter
Wave can be using DSST (Discriminative Scale Space Tracking differentiates scale space tracking) method.The party
Method devises two different filters, and core concept is to describe the translation of target respectively using the feature of multiple dimensions
Transformation and change of scale:
(1) it translates change filter: estimating the translation transformation situation of target, which calculates the one of each pixel
A one-dimensional gray scale and with 20 dimension FHOG (Fusion Histogram of Oriented Gradient, the direction gradient of fusion are straight
Side's figure) feature, finally calculate the latest position of target.Wherein FHOG feature refers to a kind of quickly calculating HOG (Histogram
Of Oriented Gradient, histograms of oriented gradients) feature method, this method can be quickly according to current color figure
As calculating 9 directions, the characteristics of image of 32 dimensions.In order to improve the operation efficiency of subsequent processing, can only use it is therein before
20 are used as main feature.
(2) dimensional variation filter: estimating the dimensional variation situation of target, which carries out 30 kinds of differences for sample
The transformation of scale, by extracting 20 dimension FHOG features in above-mentioned every kind of variation, the current size for finally calculating target becomes
Change.
Cross correlation filter needs to carry out measuring similarity to input picture and template using computing cross-correlation.It is general fixed
The computing cross-correlation of adopted input picture and Filtering Template is
Wherein g indicates that the response diagram of output, h indicate that Filtering Template, f indicate input picture,Indicate computing cross-correlation.
The Fast Fourier Transform (FFT) for defining cross-correlation function, that is, calculate time-consuming convolution algorithm and be converted to common dot product
Operation
Wherein, F indicates Fourier transformation, and ⊙ indicates the dot product of each element, the conjugation of h* representative function h.
By the new sign flag of above formula are as follows:
Since the operation on the right side of above formula is the dot product for each element, it is hereby achieved that the analytic solutions of H*
Since filter needs to scan for (m is positive integer) tracking m, target area periphery block, design
Objective function need to carry out operation to each region, and compared with reality output.Objective function is mathematically represented as
I.e. to the image F of each inputiWith reality output GiIt is the operation result in frequency domain, target is to find suitably
Filter H, so that the filter and input picture carry out between output result and reality output result after computing cross-correlation
Error is minimum.
Change filter mentality of designing is translated according to DSST method, d (d is positive integer) Wei Te can be used in a sub-picture
Sign is to be described, and the size for being defined on one block of image of target area is f, for the feature f of different dimensionslMark, l
For 1,2 ... d:
It differentiates to above formula, the calculation formula of available optimal solution
It is independently updated to the molecule of above-mentioned formula and denominator respectively
Wherein μ is scale factor.
After obtaining updated molecule and denominator, calculated in response region, and solve target position.
For the selection that dimensional variation filter converts image, following method is generallyd use.Define current goal
Scale be P × R, then define a=1:05 as horizontal scaling factor, b=1:02 is as vertical scaling factor, S=30 conduct
The width of scaling filter, then the variation of sequence size can be calculated by following formula
The selection of two factors of above-mentioned a and b relies primarily on the original size of unmanned plane.
By such method, target center caused by the diversity that different location is imaged can be overcome really
Fixed unstable problem improves the accuracy that center determines.
In one embodiment, before the region in determining image where target, first image can be located in advance
Reason.In one embodiment, morphologic filtering algorithm can be used, closed operation and opening operation processing result are subtracted each other as pre-
The result of processing.
The characteristics of opening operation is first to carry out erosion operation, then carry out dilation operation, can eliminate small objects, protrusion
Body edge, that is, the region revealed increase.The characteristics of closed operation is first to carry out dilation operation, then carry out erosion operation, can be eliminated
Minuscule hole, so that the object closed on is interconnected, i.e., the region of dark place increases.
The advantages of result by the way that the result of closed operation to be subtracted to opening operation can be using opening operation and closed operation result,
Dark place and the region revealed are screened and retained, further characteristics of image of the prominent small drone at distal end, solution
Certainly unmanned plane is apart from video camera remote position, since the influence of wing color and illumination makes the imaging of unmanned plane become one
The problem of bright spot or a dim spot, lays good basis for subsequent target following.
In one embodiment, after determining target's center position in the picture, it can further obtain target and be sat in space
Position in mark system.In one embodiment, if image is binocular vision image, it can use binocular calculation method and obtain three
Coordinate system is tieed up, space coordinates is obtained in conjunction with binocular camera rotation angle in space and pitch angle, then obtains nothing
Man-machine spatial position.
It by such method, can be the seat in space coordinates by coordinate transformation of the target in image coordinate system
Mark, further progress location estimation and tracking make target trajectory indicate that result is not influenced by camera photography angle,
Under the coordinate system unification to space coordinates of motion profile, accuracy is further increased, also can be improved and repaired according to motion profile
Change the efficiency of picture catching operating angle.
In one embodiment, when obtain target spatial position after, can by the way of Kalman Filter Estimation into
One step estimated motion track.Kalman filter method is a kind of time domain approach, and the theory of state space is introduced by it to be estimated at random
In the theory of meter, signal process is considered as the output of a linear system under white noise effect, describes this with state equation
Input/output relation is planted, state equation, observational equation and the white-noise excitation of system, the i.e. mistake of system are utilized in estimation procedure
The statistical property of journey noise and random noise forms filtering algorithm.
Assuming that X is the state variable of system, y is the observation of system.The recurrence equation of Extended Kalman filter mainly by
Five equations are completed below.
State one-step prediction:
One-step prediction covariance matrix:
P(k|k-1)=FP(k-1|k-1)FR+Q
Filtering gain matrix:
Kk=P(k|k-1)HT(HP(k|k-1)HT+R)-1
State updates:
Covariance matrix updates:
P(k|k)=(1-KkH)P(k|k-1)
Utilize the state of previous momentAnd system inputs ukThe state of ' estimation system at this time
F is state-transition matrix, wherein Q is process noise covariance battle array.Above five equations are to utilize Kalman filter
Estimate the overall process of k moment whole system state.
Object vector:
X=[Poss, Vs, CAtt, Cw]T
Wherein, PossFor Target space position, VsFor target velocity, CAttFor camera posture, CwFor angular speed.Wherein own
Element be all the expression under world coordinate system, they are respectively indicated are as follows:
Poss=[x1 y1 z1]
Vs=[vx vy vz]
CAtt=[lpan ltilt rpan rtilt]
Cw=[wlpan wlitilt wrpan wrtilt]
In formula, lpan, ltilt, rpan, rtilt respectively indicate the yaw of left and right camera, pitch angle, wlpan,
Wltilt, wrpan and wrtil are corresponding angular speed.
The measured value of filter defines:
Wherein Pos 'LWith Pos 'RIndicate position of the target in left images, respectivelyWithPTUAttIndicate the posture of PTU, including pitch angle and yaw angle.
Assuming that this moment is the k moment, first according to k-1 moment statePredict the state at k moment
Wherein, FkIt is state-transition matrix.The spatial position of target and the posture of camera are predicted according to following equation:
Wherein Δ t indicates the time interval between continuous two frame.So state-transition matrix FkIt can indicate are as follows:
Wherein Δ tn×nBe diagonal line be t, other elements be 0 n × n matrix, it can be seen that process model is linear
's.Predict the covariance matrix at k moment:
The wherein Gaussian noise of process noise setting position zero-mean, Q are the covariance matrix of process noise.
After process model f () is established, measurement model h () is next established.First by the object space of prediction
Position is projected into image.This chapter uses projection model of the pin-hole model as video camera.By the target space of points in space
Coordinate Poss (k|k-1)It should set to as the Pos ' point in plane:
When following turntable to rotate, optical center position has almost no change camera, therefore translation vector t is also experiment
The calibration process of preceding outer ginseng obtains.
So measurement model can indicate are as follows:
Wherein the effect of J () is the vector for taking the first two element of input vector to be thought of as one, from this, measurement
Model is nonlinear model, according to kalman filtering theory, needs to calculate the Jacobian matrix H (k) of measurement model h ()
():
Calculate kalman gain KkIt is as follows:
Sk=HkP(k|k-1)Hk T+R
Kk=P(k|k-1)Hk T(Sk)-1
Wherein R is that zero-mean gaussian measures noise covariance matrix.More new stateAnd covariance matrix
P(k|k):
P(k|k)=(1-KkHk)P(k|k-1)
By such method, it is capable of the influence of Removing Random No, further increases the accurate of motion profile estimation
Property, improve the robustness of target following.
The flow chart of another embodiment of the target trajectory estimation method of the disclosure is as shown in Fig. 3.
In step 301, the image including target of acquisition is pre-processed, such as uses morphologic filtering algorithm, it will
Closed operation and opening operation processing result are subtracted each other as pretreated result.
In step 302, the region in pretreated image where target is determined using TLD algorithm.
In step 303, it using active contour tracing method after TLD method output area information, further inwardly receives
It holds back, obtains more accurate location information.
In step 304, further determine that target's center position by the method for cross correlation filter, at the same by pair
The analysis of target transverse and longitudinal direction scale, further increases the mark precision of rectangle frame.
In step 305, position of the target in space coordinates is obtained.
Within step 306, the motion profile of target is estimated by Kalman Filter Estimation algorithm.
By such method, it can be realized the pretreatment operation to the image of acquisition, improve image definition, it can
The accuracy that target position determines is improved by the cooperation of TLD algorithm and Active contour models track algorithm, passes through cross correlation algorithm
It determines the center of target, the influence of random noise is further excluded by Kalman filtering, thus from multiple angle levels
Property improve target's center position determine accuracy, improve target trajectory estimation accuracy, improve target following
Reliability.
The schematic diagram of one embodiment of the target trajectory estimation device of the disclosure is as shown in Figure 4.Region determines single
Member 401 can determine the region in image where target.Region can be predetermined shape, such as rectangle, or circle.At one
In embodiment, the region in target image where target can be determined using TLD algorithm.Convergence algorithm unit 402 can be based on
Region where target carries out target area convergence algorithm, obtains the convergence location information of target.In one embodiment, may be used
Target area convergence algorithm is carried out to be based on rectangle frame using active contour following algorithm.Space orientation unit 403 being capable of basis
Convergence location information determines the spatial position of target.In one embodiment, if image is binocular vision image, can lead to
It is fixed to rower is clicked through in two images accordingly to cross, and determines target in conjunction with angle of the image capture device when acquiring image
Space coordinate.Track estimation unit 404 can estimate mesh by Kalman Filter Estimation algorithm according to the spatial position of target
Target motion profile.
Such device can first determine the region of the predetermined shape where target in the picture, further pass through target
Region convergence operation improves the accuracy that region determines, so as to improve the accuracy of Target space position confirmation, improves mesh
Mark the reliability of tracking.
The schematic diagram of another embodiment of the target trajectory estimation device of the disclosure is as shown in Fig. 5.Region determines
Unit 51, convergence algorithm unit 52, space orientation unit 53 and track estimation unit 54 are similar to embodiment illustrated in fig. 4.
Space orientation unit 53 may include cross correlation filter subelement 531 and locator unit 532.Cross correlation filter
Subelement 531 can further determine that target's center position using cross correlation algorithm, while by target transverse and longitudinal direction scale
Analysis, further increase mark precision；Locator unit 532 can determine in image in cross correlation filter subelement 531
Behind target's center position, position of the target in space coordinates is obtained.
Such target trajectory estimation device can overcome target caused by the diversity that different location is imaged in
The unstable problem of the determination of heart position improves the accuracy that center determines；Target trajectory can be made to indicate result
It is not influenced by camera photography angle, under the coordinate system unification to space coordinates of motion profile, it is accurate to further increase
Property, it also can be improved the efficiency that picture catching operating angle is modified according to motion profile.
In one embodiment, as shown in figure 5, target trajectory estimation device can also include pretreatment unit 55,
Image can be pre-processed before region in determining image where target.Pretreatment unit 55 may include out fortune
Operator unit 551, closed operation subelement 552 and processing image obtain subelement 553.Opening operation subelement 551 can be carried out first
Erosion operation, then dilation operation is carried out, small objects are eliminated, prominent object edge increases the region revealed；Closed operation
Unit 552 can first carry out dilation operation, then carry out erosion operation, minuscule hole be eliminated, so that the object closed on mutually interconnects
Logical, the region of dark place increases.Processing image, which obtains subelement 553, can subtract the result of closed operation the result of opening operation.
Such target trajectory estimation device can utilize the advantages of opening operation and closed operation result, by dark place and
The region revealed is screened and is retained, and further characteristics of image of the prominent small drone at distal end, solves unmanned plane
Apart from video camera remote position, since the influence of wing color and illumination makes the imaging of unmanned plane become a bright spot or one
The problem of a dim spot, lays good basis for subsequent Target Tracking Problem.
The structural schematic diagram of one embodiment of disclosure target trajectory estimation device is as shown in Fig. 6.Target movement
Track estimation device includes memory 610 and processor 620.Wherein: memory 610 can be disk, flash memory or other any
Non-volatile memory medium.Memory is used to store the finger in the above corresponding embodiment of target trajectory estimation method
It enables.Processor 620 is coupled to memory 610, can be used as one or more integrated circuits to implement, for example, microprocessor or
Microcontroller.The processor 620 can be improved the standard of Target space position confirmation for executing the instruction stored in memory
Exactness improves the reliability of target following.
It in one embodiment, can be as shown in fig. 7, target trajectory estimation device 700 includes memory 710
With processor 720.Processor 720 is coupled to memory 710 by BUS bus 730.The target trajectory estimation device 700
External memory 750 can also be connected to by memory interface 740 to call external data, can also be connect by network
Mouth 760 is connected to network or an other computer system (not shown).It no longer describes in detail herein.
In this embodiment, it is instructed by memory stores data, then above-metioned instruction, Neng Goushi is handled by processor
The accuracy for now improving Target space position confirmation, improves the reliability of target following.
In another embodiment, a kind of computer readable storage medium, is stored thereon with computer program instructions, should
The step of method in target trajectory estimation method corresponding embodiment is realized when instruction is executed by processor.In the art
Technical staff it should be appreciated that embodiment of the disclosure can provide as method, apparatus or computer program product.Therefore, this public affairs
Open the form that complete hardware embodiment, complete software embodiment or embodiment combining software and hardware aspects can be used.And
And the disclosure can be used can use non-transient in the computer that one or more wherein includes computer usable program code
The computer program product implemented on storage medium (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.)
Form.
The schematic diagram of one embodiment of the Target Tracking System of the disclosure is as shown in Figure 8.Target trajectory estimation dress
Setting 81 can be any one the target trajectory estimation device being mentioned above.Image acquiring device 82 can be target
Motion profile estimation device 81 provides image.In one embodiment, image acquiring device 82 can be binocular camera, or
Two monocular-cameras.Turntable 83 can carrying image acquisition device 82, image is changed by the adjustment of the angle of turntable 83 and is obtained
The shooting angle of device 82 is taken, turntable follows target rotational according to the estimated result of target trajectory estimation device, thus real
Now to the tracking of target.
Such Target Tracking System can be improved the accuracy of Target space position confirmation, improve target trajectory
The accuracy of estimation improves mesh to avoid ROI region from not including unmanned plane target and the tracking to unmanned plane is caused to fail
Mark the reliability of tracking.
The schematic diagram of one embodiment of the Target Tracking System operational process of the disclosure is as shown in Figure 9.It is carried on turntable
Image acquiring device obtain original image 901, by obtain that treated image, the root of morphologic filtering image preprocessing 902
According to the image performance objective location estimation.Estimation procedure include target with the TLD target tracking algorism 914 in middle algorithm 904 at
Reason and Active contour models track algorithm 924 are handled, and are then handled, are obtained in target by cross correlation filter track algorithm 905
Heart point position 916, can also obtain unmanned plane target rectangle frame information 926.Cooperate turntable pitch angle, azimuth information 927 true
The space coordinates for determining the target in image realize the estimation of motion profile in conjunction with Kalman filtering algorithm.Turntable is according to estimation
As a result carrying image acquiring device rotation, to realize target following.
Such Target Tracking System can be realized the pretreatment operation to the image of acquisition, improve image definition, energy
The accuracy that target position determines enough is improved by the cooperation of TLD algorithm and Active contour models track algorithm, is calculated by cross-correlation
Method determines the center of target, the influence of random noise is further excluded by Kalman filtering, thus from multiple angle layers
Secondary property improve target's center position determine accuracy, improve target trajectory estimation accuracy, improve target with
The reliability of track.
The disclosure is referring to method, the process of equipment (system) and computer program product according to the embodiment of the present disclosure
Figure and/or block diagram describe.It should be understood that can be realized by computer program instructions each in flowchart and/or the block diagram
The combination of process and/or box in process and/or box and flowchart and/or the block diagram.It can provide these computer journeys
Sequence instructs the processor to general purpose computer, special purpose computer, Embedded Processor or other programmable data processing devices
To generate a machine, so that being generated by the instruction that computer or the processor of other programmable data processing devices execute
For realizing the function of being specified in one or more flows of the flowchart and/or one or more blocks of the block diagram
Device.
These computer program instructions, which may also be stored in, is able to guide computer or other programmable data processing devices with spy
Determine in the computer-readable memory that mode works, so that instruction stored in the computer readable memory generation includes
The manufacture of command device, the command device are realized in one box of one or more flows of the flowchart and/or block diagram
Or the function of being specified in multiple boxes.
These computer program instructions also can be loaded onto a computer or other programmable data processing device, so that
Series of operation steps are executed on computer or other programmable devices to generate computer implemented processing, thus calculating
The instruction executed on machine or other programmable devices is provided for realizing in one or more flows of the flowchart and/or side
The step of function of being specified in block diagram one box or multiple boxes.
So far, the disclosure is described in detail.In order to avoid covering the design of the disclosure, this field institute is not described
Well known some details.Those skilled in the art as described above, completely it can be appreciated how implementing skill disclosed herein
Art scheme.
Disclosed method and device may be achieved in many ways.For example, can by software, hardware, firmware or
Person's software, hardware, firmware any combination realize disclosed method and device.The step of for the method
Sequence is stated merely to be illustrated, the step of disclosed method is not limited to sequence described in detail above, unless with other
Mode illustrates.In addition, in some embodiments, the disclosure can be also embodied as recording program in the recording medium,
These programs include for realizing according to the machine readable instructions of disclosed method.Thus, the disclosure also covers storage and is used for
Execute the recording medium of the program according to disclosed method.
Finally it should be noted that: above embodiments are only to illustrate the technical solution of the disclosure rather than its limitations；To the greatest extent
Pipe is described in detail the disclosure referring to preferred embodiment, it should be understood by those ordinary skilled in the art that: still
It can modify to the specific embodiment of the disclosure or some technical features can be equivalently replaced；Without departing from this
The spirit of public technology scheme should all cover in the claimed technical proposal scope of the disclosure.
Claims (15)
1. a kind of target trajectory estimation method, comprising:
Determine the region in image where target；
Target area convergence algorithm is carried out based on the region where the target, obtains the convergence location information of the target；
The spatial position of target is determined according to the convergence location information；
According to the spatial position of the target, the motion profile of the target is estimated by Kalman Filter Estimation algorithm.
2. according to the method described in claim 1, wherein:
Detection TLD algorithm is learnt using tracking and determines the region in the target image where target, the shape packet in the region
Include rectangle or circle；And/or
The rectangle frame is based on using active contour following algorithm and carries out target area convergence algorithm, obtains the convergence of the target
Location information.
3. according to the method described in claim 1, further include:
Morphologic filtering pretreatment is carried out to described image, to determine the rectangle where target according to image after treatment
Frame.
4. according to the method described in claim 3, wherein, described pre-process to described image progress morphologic filtering includes:
The opening operation processing result and closed operation processing result of described image are obtained respectively；
Opening operation processing result and closed operation processing result are subtracted each other, image after treatment is obtained.
5. according to the method described in claim 1, wherein, the spatial position that target is determined according to the convergence location information
Include:
Cross correlation filter is carried out according to the convergence position, determines the center of the target；
The spatial position of target is determined according to the center.
It is described cross correlation filter is carried out according to the convergence position to include: 6. according to the method described in claim 5, wherein
Changed by translation and filters the latest position for determining target；
The dimensional variation situation of estimation target is filtered by dimensional variation.
7. a kind of target trajectory estimation device, comprising:
Area determination unit is configured to determine that in image the region where target；
Convergence algorithm unit is configured as carrying out target area convergence algorithm based on the region where the target, described in acquisition
The convergence location information of target；
Space orientation unit is configured as determining the spatial position of target according to the convergence location information；
Track estimation unit is configured as the spatial position according to the target, estimates institute by Kalman Filter Estimation algorithm
State the motion profile of target.
8. device according to claim 7, in which:
The area determination unit is configured as determining in the target image where target using tracking study detection TLD algorithm
Region, the shape in the region includes rectangle or circle；
And/or
The tracking study detection is configured as being based on rectangle frame progress target area receipts using active contour following algorithm
Operation is held back, the convergence location information of the target is obtained.
9. device according to claim 7, further includes:
Pretreatment unit, for carrying out morphologic filtering pretreatment to described image, so as to true according to image after treatment
Rectangle frame where setting the goal.
10. device according to claim 9, wherein the pretreatment unit is configured as:
Opening operation subelement, for obtaining the opening operation processing result of described image；
Closed operation subelement, for obtaining the closed operation processing result of described image；
It handles image and obtains subelement, for subtracting each other opening operation processing result and closed operation processing result, obtain by processing
Image afterwards.
11. device according to claim 7, wherein the space orientation unit includes:
Cross correlation filter subelement determines the centre bit of the target for carrying out cross correlation filter according to the convergence position
It sets；
Locator unit, for determining the spatial position of target according to the center.
12. device according to claim 11, wherein the cross correlation filter subelement is used for:
Changed by translation and filters the latest position for determining target；
The dimensional variation situation of estimation target is filtered by dimensional variation.
13. a kind of target trajectory estimation device, comprising:
Memory；And
It is coupled to the processor of the memory, the processor is configured to based on the instruction execution for being stored in the memory
Such as method as claimed in any one of claims 1 to 6.
14. a kind of computer readable storage medium, is stored thereon with computer program instructions, real when which is executed by processor
The step of method described in existing claim 1 to 6 any one.
15. a kind of Target Tracking System, comprising:
Target trajectory estimation device described in claim 7~13 any one；With,
Image acquiring device is configured as obtaining the image of target；
Turntable is configured as carrying described image acquisition device, is followed according to the estimated result of target trajectory estimation device
The target rotational.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711306051.7A CN109902543A (en) | 2017-12-11 | 2017-12-11 | Target trajectory estimation method, device and Target Tracking System |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711306051.7A CN109902543A (en) | 2017-12-11 | 2017-12-11 | Target trajectory estimation method, device and Target Tracking System |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109902543A true CN109902543A (en) | 2019-06-18 |
Family
ID=66942052
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711306051.7A Pending CN109902543A (en) | 2017-12-11 | 2017-12-11 | Target trajectory estimation method, device and Target Tracking System |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109902543A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110570463A (en) * | 2019-09-11 | 2019-12-13 | 深圳市道通智能航空技术有限公司 | target state estimation method and device and unmanned aerial vehicle |
| CN110827321A (en) * | 2019-10-16 | 2020-02-21 | 天津大学 | Multi-camera cooperative active target tracking method based on three-dimensional information |
| CN114399528A (en) * | 2021-11-29 | 2022-04-26 | 深圳先进技术研究院 | Three-dimensional space moving target tracking method based on two-dimensional image and related device |
| CN114939216A (en) * | 2022-05-30 | 2022-08-26 | 深圳英鸿骏智能科技有限公司 | Equipment and method for assisting rehabilitation exercise |
| CN113075086B (en) * | 2021-03-25 | 2024-02-02 | 苏州未艾视智能科技有限公司 | Wood chip detection and screening device based on X-rays |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102043964A (en) * | 2010-12-30 | 2011-05-04 | 复旦大学 | Tracking algorithm and tracking system for taking-off and landing of aircraft based on tripod head and camera head |
| CN103150740A (en) * | 2013-03-29 | 2013-06-12 | 上海理工大学 | Method and system for moving target tracking based on video |
| CN103413330A (en) * | 2013-08-30 | 2013-11-27 | 中国科学院自动化研究所 | Method for reliably generating video abstraction in complex scene |
| CN104281837A (en) * | 2014-09-26 | 2015-01-14 | 哈尔滨工业大学深圳研究生院 | Pedestrian tracking method combining Kalman filtering with ROI expansion between adjacent frames |
| CN105574895A (en) * | 2016-01-05 | 2016-05-11 | 浙江博天科技有限公司 | Congestion detection method during the dynamic driving process of vehicle |
| CN106709938A (en) * | 2016-11-18 | 2017-05-24 | 电子科技大学 | Multi-target tracking method based on improved TLD (tracking-learning-detected) |
-
2017
- 2017-12-11 CN CN201711306051.7A patent/CN109902543A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102043964A (en) * | 2010-12-30 | 2011-05-04 | 复旦大学 | Tracking algorithm and tracking system for taking-off and landing of aircraft based on tripod head and camera head |
| CN103150740A (en) * | 2013-03-29 | 2013-06-12 | 上海理工大学 | Method and system for moving target tracking based on video |
| CN103413330A (en) * | 2013-08-30 | 2013-11-27 | 中国科学院自动化研究所 | Method for reliably generating video abstraction in complex scene |
| CN104281837A (en) * | 2014-09-26 | 2015-01-14 | 哈尔滨工业大学深圳研究生院 | Pedestrian tracking method combining Kalman filtering with ROI expansion between adjacent frames |
| CN105574895A (en) * | 2016-01-05 | 2016-05-11 | 浙江博天科技有限公司 | Congestion detection method during the dynamic driving process of vehicle |
| CN106709938A (en) * | 2016-11-18 | 2017-05-24 | 电子科技大学 | Multi-target tracking method based on improved TLD (tracking-learning-detected) |
Non-Patent Citations (1)
| Title |
|---|
| 徐杨: "面向视频监控的动态目标检测_跟踪与识别关键技术研究", 《中国博士学位论文全文数据库信息科技辑》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110570463A (en) * | 2019-09-11 | 2019-12-13 | 深圳市道通智能航空技术有限公司 | target state estimation method and device and unmanned aerial vehicle |
| CN110570463B (en) * | 2019-09-11 | 2023-04-07 | 深圳市道通智能航空技术股份有限公司 | Target state estimation method and device and unmanned aerial vehicle |
| CN110827321A (en) * | 2019-10-16 | 2020-02-21 | 天津大学 | Multi-camera cooperative active target tracking method based on three-dimensional information |
| CN110827321B (en) * | 2019-10-16 | 2023-05-30 | 天津大学 | Multi-camera collaborative active target tracking method based on three-dimensional information |
| CN113075086B (en) * | 2021-03-25 | 2024-02-02 | 苏州未艾视智能科技有限公司 | Wood chip detection and screening device based on X-rays |
| CN114399528A (en) * | 2021-11-29 | 2022-04-26 | 深圳先进技术研究院 | Three-dimensional space moving target tracking method based on two-dimensional image and related device |
| CN114939216A (en) * | 2022-05-30 | 2022-08-26 | 深圳英鸿骏智能科技有限公司 | Equipment and method for assisting rehabilitation exercise |
| CN114939216B (en) * | 2022-05-30 | 2023-11-10 | 深圳英鸿骏智能科技有限公司 | Device and method for assisting rehabilitation exercise |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109902543A (en) | Target trajectory estimation method, device and Target Tracking System | |
| Simo-Serra et al. | A joint model for 2d and 3d pose estimation from a single image | |
| KR100860640B1 (en) | Method and system for multi-modal component-based tracking of an object using robust information fusion | |
| US7706571B2 (en) | Flexible layer tracking with weak online appearance model | |
| JP4766495B2 (en) | Object tracking device and object tracking method | |
| CN106780542A (en) | A kind of machine fish tracking of the Camshift based on embedded Kalman filter | |
| JP2015181042A (en) | detection and tracking of moving objects | |
| US20120296202A1 (en) | Method and System for Registration of Ultrasound and Physiological Models to X-ray Fluoroscopic Images | |
| JP2008304269A (en) | Information processor, information processing method, and computer program | |
| CN106846367B (en) | A kind of Mobile object detection method of the complicated dynamic scene based on kinematic constraint optical flow method | |
| JP2005339288A (en) | Image processor and its method | |
| Gall et al. | Drift-free tracking of rigid and articulated objects | |
| CN112927251A (en) | Morphology-based scene dense depth map acquisition method, system and device | |
| CN107194950B (en) | Multi-person tracking method based on slow feature analysis | |
| Tang et al. | Retinal image registration based on robust non-rigid point matching method | |
| CN113379789B (en) | Moving target tracking method in complex environment | |
| Lee et al. | Human body tracking with auxiliary measurements | |
| US9672412B2 (en) | Real-time head pose tracking with online face template reconstruction | |
| Li et al. | RD-VIO: Robust visual-inertial odometry for mobile augmented reality in dynamic environments | |
| Arnaud et al. | Partial linear gaussian models for tracking in image sequences using sequential monte carlo methods | |
| CN115170621A (en) | Target tracking method and system under dynamic background based on relevant filtering framework | |
| Kitsikidis et al. | Unsupervised dance motion patterns classification from fused skeletal data using exemplar-based HMMs | |
| Ginhoux et al. | Model-based object tracking using stereo vision | |
| JP2018116147A (en) | Map creation device, map creation method and map creation computer program | |
| JP2008541859A (en) | 3D-CT registration with guidance method based on 3D-2D pose estimation and application to raw bronchoscopy |
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 | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20210302 Address after: Room a1905, 19 / F, building 2, No. 18, Kechuang 11th Street, Daxing District, Beijing, 100176 Applicant after: Beijing Jingdong Qianshi Technology Co.,Ltd. Address before: 101, 1st floor, building 2, yard 20, Suzhou street, Haidian District, Beijing 100080 Applicant before: Beijing Jingbangda Trading Co.,Ltd. Effective date of registration: 20210302 Address after: 101, 1st floor, building 2, yard 20, Suzhou street, Haidian District, Beijing 100080 Applicant after: Beijing Jingbangda Trading Co.,Ltd. Address before: 100195 Beijing Haidian Xingshikou Road 65 West Cedar Creative Garden 4 District 11 Building East 1-4 Floor West 1-4 Floor Applicant before: BEIJING JINGDONG SHANGKE INFORMATION TECHNOLOGY Co.,Ltd. Applicant before: BEIJING JINGDONG CENTURY TRADING Co.,Ltd. |
|
| TA01 | Transfer of patent application right |