CN109685832A - A kind of motion target tracking method, device and computer equipment - Google Patents
A kind of motion target tracking method, device and computer equipment Download PDFInfo
- Publication number
- CN109685832A CN109685832A CN201811596674.7A CN201811596674A CN109685832A CN 109685832 A CN109685832 A CN 109685832A CN 201811596674 A CN201811596674 A CN 201811596674A CN 109685832 A CN109685832 A CN 109685832A
- Authority
- CN
- China
- Prior art keywords
- target
- size
- change
- filter
- frame
- 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 56
- 230000008859 change Effects 0.000 claims abstract description 88
- 230000004044 response Effects 0.000 claims abstract description 42
- 238000012549 training Methods 0.000 claims abstract description 32
- 238000005315 distribution function Methods 0.000 claims abstract description 11
- 238000004590 computer program Methods 0.000 claims description 6
- 230000006870 function Effects 0.000 claims 2
- 238000004422 calculation algorithm Methods 0.000 abstract description 16
- 238000004364 calculation method Methods 0.000 abstract description 5
- 238000005516 engineering process Methods 0.000 abstract description 2
- 230000008569 process Effects 0.000 description 7
- 238000001914 filtration Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 230000007423 decrease Effects 0.000 description 3
- 238000001514 detection method Methods 0.000 description 3
- 230000000149 penetrating effect Effects 0.000 description 3
- 230000006978 adaptation Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 239000004744 fabric Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000005352 clarification Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000001360 synchronised 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
- G06T7/251—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments involving models
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/262—Analysis of motion using transform domain methods, e.g. Fourier domain methods
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
-
- 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
-
- 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
- G06T2207/20048—Transform domain processing
- G06T2207/20056—Discrete and fast Fourier transform, [DFT, FFT]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Multimedia (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Mathematical Physics (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to field of computer technology more particularly to a kind of motion target tracking methods, device and equipment.This method comprises: obtaining the first FHOG feature of target, and construct the position response output and size response output of gauss of distribution function distribution；Corresponding filter is obtained according to FHOG feature and position response output, size response output training；The second frame of video selection area is obtained, and obtains its characteristic information, carries out characteristic information and change in location filter to calculate acquisition target movement position according to corresponding model；The image-region of multiple and different sizes is chosen centered on the center of second frame target position, obtains multiple sub- FHOG features；Multiple sub- FHOG features calculate according to corresponding model with change in size filter respectively and obtain target size.The method increase the adaptability changed to target scale, can be with accurately tracking to target；And algorithm reduces calculation amount, improves the real-time of algorithm.
Description
Technical field
The present invention relates to field of computer technology more particularly to a kind of motion target tracking method, device and computer to set
It is standby.
Background technique
Video frequency object tracking can be used in vehicle monitoring, intelligent security guard etc., have become computer vision
Important research content.
Target following can be mainly divided into 5 parts, be respectively motion model, feature extraction, display model, target positioning and
Model root, wherein motion model can predict that the region that present frame target is likely to occur is existing according to the position of previous frame target
Target movement is modeled using the method for particle filter or correlation filtering in major part, wherein tracking based on correlation filtering
Method, using circular matrix time domain convert special nature, budget is transformed into frequency domain and is calculated, classifier can be accelerated
Training, while obtaining all response images that must be grouped as of circulation samples, target positioning carried out according to most legal system position.
But current algorithm cannot well adapt to the transformation of target scale itself and the transformation of movement position, thus not
It can guarantee the accuracy of target following.
Summary of the invention
Based on this, the present invention is directed to above-mentioned problem, provides a kind of motion target tracking method, the technical solution is such as
Under:
A kind of motion target tracking method, specifically includes,
The first FHOG feature of video first frame tracking target is obtained, and constructs the position response of gauss of distribution function distribution
Output and size response output；
Position is obtained according to the training of preset training rules according to the first FHOG feature and position response output
Change filter obtains size according to training rules training according to the first FHOG feature and size response output
Change filter；
The second frame of video is obtained, by the first area that presumptive area selection rule is selected, and obtains the first area
Characteristic information；
The characteristic information and the change in location filter calculate according to predetermined position variation model and obtain mesh
Mark movement position；
According to change in location filter prediction the second frame target position center, and in second frame target position
Selection principle chooses the image-region of multiple and different sizes according to predetermined dimensions centered on the heart, and obtains described image region
Multiple sub- FHOG features；
By the multiple sub- FHOG feature respectively with the change in size filter according to predetermined dimensions variation model into
Row, which calculates, obtains target size.
The present invention also provides a kind of motion target tracking devices, comprising:
Position feature obtains module, for obtaining the first FHOG feature of video first frame tracking target, and constructs Gauss
The position response output of distribution function distribution；
Change in location filter training module, for according to the first FHOG feature and position response output training
Obtain change in location filter；
First area selecting module, for obtaining the second frame of video, by the first area that pre-defined rule is chosen, to obtain
State the characteristic information of first area；
Target movement position obtains module, is used for the characteristic information and the change in location filter according to pre-determined bit
Variation model is set to carry out calculating acquisition target movement position.
The present invention also provides another motion target tracking devices, further includes:
Size characteristic obtains module, for obtaining the first FHOG feature of video first frame tracking target, and constructs Gauss
The size of distribution function distribution responds output；
Change in size filter training module, for according to the first FHOG feature and size response output training
Obtain change in size filter；
Subcharacter obtains module, is used for according to change in location filter prediction the second frame target position center, and with
Selection principle chooses the image-region of multiple and different sizes according to predetermined dimensions centered on the center of second frame target position,
To obtain multiple sub- FHOG features in described image region；
Target size determining module, by the multiple sub- FHOG feature respectively with the change in size filter according to predetermined
Change in size model carry out calculate obtain target size.
The present invention also provides a kind of computer equipment, including memory and processor, the memory is stored with calculating
The step of machine program, the processor realizes motion target tracking method when executing the computer program.
Motion target tracking scheme provided in an embodiment of the present invention is rung by obtaining the FHOG feature of target and constructing position
It should export to respond with size and export, corresponding filter is obtained according to FHOG feature and response output, is obtaining the second frame image
When according to scheduled regular selection area and obtain characteristic information, finally calculated according to scheduled change in location model and obtain target
Movement position, while the image-region of multiple and different sizes is chosen according to scheduled selection of dimension principle, and obtain multiple sons
FHOG feature, finally variation model calculates and obtains target size according to predetermined dimensions, and this method propose dimensional variation filtering
Device, the adaptation to tracking target scale variation is realized by individually establishing dimensional variation filter, and above-mentioned algorithm comments position
Estimate filter and dimensional variation filter combines, first passes through position assessment filter assessment tracking target in a new frame
Position, then scale size of the target in a new frame is determined with dimensional variation filter, which is guaranteeing tracking effect
The complexity for reducing calculating simultaneously, improves the real-time of algorithm.
It should be understood that above general description and following detailed description be only it is exemplary and explanatory, not
The disclosure can be limited.
Detailed description of the invention
The drawings herein are incorporated into the specification and forms part of this specification, and shows the implementation for meeting the disclosure
Example, and together with specification for explaining the principles of this disclosure.
Fig. 1 is a kind of motion target tracking method flow chart that the embodiment of the present invention one provides；
Fig. 2 be a kind of second frame of acquisitions video provided by Embodiment 2 of the present invention, by pre-defined rule selection first area,
To obtain the detail flowchart of the characteristic information method of the first area；
Fig. 3 be the embodiment of the present invention three provide it is a kind of by the characteristic information with the change in location filter according to pre-
Positioning sets variation model and carries out calculating the detail flowchart for obtaining target movement position method；
Fig. 4 be the embodiment of the present invention four provide it is a kind of by the multiple sub- FHOG feature respectively with the change in size filter
Variation model carries out calculating the detail flowchart for obtaining target size method wave device according to predetermined dimensions；
Fig. 5 be the embodiment of the present invention five provide it is a kind of by the multiple sub- FHOG feature respectively with the change in size filter
Variation model carries out calculating the detail flowchart for obtaining target size method wave device according to predetermined dimensions；
Fig. 6 is a kind of motion target tracking method flow chart that the embodiment of the present invention six provides；
Fig. 7 is a kind of structural schematic diagram for motion target tracking device that the embodiment of the present invention seven provides；
Fig. 8 is a kind of structural schematic diagram for motion target tracking device that the embodiment of the present invention eight provides.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.
Fig. 1 shows a kind of motion target tracking method flow chart of the offer of the embodiment of the present invention one, and details are as follows:
In step s101, the first FHOG feature of video first frame tracking target is obtained, and constructs gauss of distribution function
The position response output and size response output of distribution.
In embodiments of the present invention, the target tracked is chosen in a frame of video image, in the mesh chosen
The FHOG feature of extracted region target is marked, wherein there are multiple dimensions for FHOG feature；Building is with gauss of distribution function point simultaneously
The position response output and size response output of cloth, the intermediate response value of the two is maximum and successively successively decreases around.
In step s 102, according to the first FHOG feature and position response output according to preset training rules
Training obtains change in location filter, according to the first FHOG feature and size response output according to the training rules
Training obtains change in size filter.
In embodiments of the present invention, according to the first FHOG feature, the position input of the first FHOG feature is determined
Sample F1, the dimension of the position input sample are d, and export G according to the position response1Obtain change in location filter
H1l。
In an embodiment of the present invention, according to the first FHOG feature, determine that the size of the first FHOG feature is defeated
Enter sample F2, the dimension of the size input sample is m, and responds output G according to the size2Obtain change in size filter
H2l。
In an embodiment of the present invention, λ is regularization coefficient, in order to control keeping away during filter frequency domain parametric solution
Exempt 0, on the other hand also can control filter parameter variation range, λ is smaller, and filter parameter variation range is bigger.
In step s 103, the second frame of video is obtained, by the first area that presumptive area selection rule is selected, and obtains institute
State the characteristic information of first area.
In an embodiment of the present invention, when obtaining the second frame of video, in the second frame according to scheduled region selection rule
Region where selected tracking target, and obtain the characteristic information in the region.
In step S104, by the characteristic information and the change in location filter according to predetermined position variation model into
Row, which calculates, obtains target movement position.
In step s105, according to change in location filter prediction the second frame target position center, and with described
Selection principle chooses the image-region of multiple and different sizes according to predetermined dimensions centered on two frame target positions center, and obtains
Multiple sub- FHOG features in described image region.
In an embodiment of the present invention, it is calculated first according to change in location filter by change in location model and obtains second
The center of frame target, and selection principle chooses the image-regions of multiple and different sizes according to predetermined dimensions with this center,
And from its sub- FHOG feature of the multiple images extracted region of selection, for predicting target in the size of the second frame.
In step s 106, by the multiple sub- FHOG feature respectively with the change in size filter according to scheduled ruler
Very little variation model, which calculate, obtains target size.
It should be understood that the first frame and the second frame in the present embodiment not merely represent the first frame and second of video
Frame, two frame of front and back of the target following of major embodiment, how according to mesh in the positions and dimensions of former frame target acquisition a later frame
Target positions and dimensions.This method combines change in location filter and change in size filter, is selected by different rules
Target area and different sized images are selected, and is carried out calculating the positions and dimensions for obtaining tracking target, energy according to corresponding model
It is enough that accurately target is tracked.
In an embodiment of the present invention, it Fig. 2 shows a kind of second frame of acquisition video provided by Embodiment 2 of the present invention, presses
The first area that pre-defined rule is chosen, to obtain the detail flowchart of the characteristic information method of the first area.
In step s 201, the second frame of video is obtained, tracks mesh to obtain first frame according to the change in location filter
Centered on target center, the first area having a size of 1.5-3.5 times of target size is chosen.
In an embodiment of the present invention, after obtaining the second frame of video, first frame tracking is obtained according to position filtering device
The center of target, response output maximum value is target's center, and is mesh in first frame as center chosen area size
Mark 1.5-3.5 times of first area of size.
In step S202, characteristic information is extracted in the first area.
In an embodiment of the present invention, after obtaining first area image, the image in the region is obtained by statistical method
Obtain the characteristic information in the region.
This method is quickly obtained the approximate region in the second frame where target by the selection rule of setting target area,
To realize the tracking of target.
In an embodiment of the present invention, Fig. 3 show the embodiment of the present invention three offer it is a kind of by the characteristic information with
The change in location filter carries out calculating the detailed process for obtaining target movement position method according to predetermined position variation model
Figure.
In step S301, characteristic information is obtained, the characteristic information includes gray feature and FHOG feature, and will be described
Characteristic information inputs Z1 as sample.
In an embodiment of the present invention, the characteristic information of the first area of acquisition includes gray feature and FHOG feature,
It includes 1 dimension gray feature and 27 dimension FHOG features that each pixel in the region, which calculates feature, and by all pictures of one's respective area
The characteristic information of vegetarian refreshments is inputted as sample, is denoted as Z1.
In step s 302, position assessment models are obtained according to the change in location filter, WhereinForB1tFor
In an embodiment of the present invention according to above-mentioned change in location filter, by moleculeIt is denoted asIn denominatorIt is denoted as B1t, and carried out Fourier transform and obtain position assessment models y1.
In step S303, the sample information Z1 is calculated by the position assessment models, is obtained max (y1),
To obtain target movement position.
In an embodiment of the present invention, by former frame Pt-1In A1t-1And B1t-1And input sample Z1 is substituting to position
In assessment models y1, y1 is obtained, and calculates max (y 1), obtains the new position of target.
This method obtains the new position of target, the calculation of the model by establishing position assessment models, by calculating its maximum value
Method is simple, and stability is good.
In an embodiment of the present invention, Fig. 4 shows a kind of by the multiple sub- FHOG of the offer of the embodiment of the present invention four
Variation model carries out calculating acquisition target size method feature according to predetermined dimensions with the change in size filter respectively
Detail flowchart.
In step S401, the wide P and high R of target in first frame are obtained, and the P and R is substituted into formula anP×anR into
Row calculates, wherein
A is constant scale factor and s scale quantity, obtains multiple and different sizes；
In an embodiment of the present invention, when choosing different scale sample centered on target position, it is contemplated that tracking mesh
The size being marked in adjacent 2 frame will not have big difference, therefore the sample-size selection principle of scale assessment are as follows:
Wherein, P and R is respectively that target is high in the width of former frame, and a=1.02 is scale factor, and S=33 is the number of scale
Amount.Above-mentioned scale not instead of linear relationship, by the detection process in precise and penetrating thick (direction from inside to outside).
In step S402, centered on the center of second frame target position, according to the multiple different selection of dimension
The image-region of multiple and different sizes.
In embodiments herein, the second frame target position center is obtained, and as center according to above-mentioned steps meter
The different sizes calculated, choose the region of multiple and different sizes, with the characteristic information for obtaining variant size area.
This method sets target size selection principle, and resulting scale is by precise and penetrating thick detection process, energy
It is enough fast and accurately to select multiple sizes, the size of enough and accurate target possibility is provided for the calculating of change in size,
Improve the accuracy rate of target size tracking.
In embodiments herein, Fig. 5 shows a kind of by the multiple sub- FHOG of the offer of the embodiment of the present invention five
Variation model carries out calculating acquisition target size method feature according to predetermined dimensions with the change in size filter respectively
Detail flowchart.
In step S501, the multiple sub- FHOG feature that will acquire inputs Z2 as sample.
In an embodiment of the present invention, multiple and different size areas of acquisition, each region are set as fixed size, point
31 dimension FHOG features are indescribably taken, is inputted as sample, is denoted as Z2.
In step S502, size assessment model is obtained according to the change in location filter WhereinForB2tFor
In an embodiment of the present invention according to above-mentioned change in size filter, by moleculeIt is denoted asIn denominatorIt is denoted as B1t, and carried out Fourier transform and obtain position assessment models y2.
In step S503, the sample information Z2 is calculated by the size assessment model, is obtained max (y2),
To obtain target size.
In an embodiment of the present invention, by former frame Pt-1In A2t-1And B2t-1And input sample Z2 is substituting to position
In assessment models y2, y2 is obtained, and calculates max (y2max), obtain the new size of target.
This method obtains target new size, the calculation of the model by establishing size assessment model, by calculating its maximum value
Method is simple, and stability is good.
In an embodiment of the present invention, Fig. 6 shows a kind of motion target tracking method of the offer of the embodiment of the present invention six
Flow chart.
It should be understood that the clarification of objective information of acquisition is constantly sent out when tracking process is during gradually carrying out
It is raw to change, therefore need to update filter model accordingly during tracking.
In step s 601, the first frame target FHOG feature and response output and the second frame target FHOG is special
Response output of seeking peace compares, and obtains filter learning efficiency.
In an embodiment of the present invention, by the first frame target FHOG feature and response output and the second frame target
FHOG feature and response output compare according to the following formula.
Wherein η indicates learning rate.
In step S602, position change filter and change in size filter are carried out more according to the learning efficiency
Newly.
In an embodiment of the present invention, position change filter and change in size filter are carried out more according to learning efficiency
The following formula of new algorithm.
It should be understood that change in location filter is different with the FHOG feature of change in size filter, learning efficiency
Difference, only the algorithm of the two is identical, so being indicated according to the same formula.
Change in location filter is identical with the algorithm of change in size filter in this method, reduces calculation amount, and with
Constantly each filter is updated accordingly during track, improves the real-time of algorithm.
Below to be tracked to same pedestrian target, this tracking is compared with other algorithms under the conditions of same video
Analysis, see Table 1 for details.
Table 1
| Algorithm | FTC | TLD | The present invention |
| Average every frame runing time/s | 0.027 | 0.030 | 0.025 |
| Average central deviation/pixel | 15.20 | 2.32 | 2.40 |
| Average Duplication | 0.5 | 0.6 | 0.72 |
As known from Table 1, inventive algorithm can achieve 40 frames/s in terms of real-time, reach requirement of real-time, in target
Also there is relatively good effect compared to other algorithms in terms of tracking stability.
Fig. 7 shows a kind of structural schematic diagram of motion target tracking device of the offer of the embodiment of the present invention seven, in order to just
In explanation, only the parts related to the present invention are shown.
In an embodiment of the present invention, motion target tracking device includes that position feature obtains module 710, change in location filter
Wave device training module 720, first area selecting module 730, target movement position obtain module 740, and details are as follows:
Position feature obtains module 710, for obtaining the first FHOG feature of video first frame tracking target, and constructs height
The position response output of this distribution function distribution.
In an embodiment of the present invention, the target tracked is chosen in a frame of video image, what is chosen
The FHOG feature of target is extracted in target area, and wherein there are multiple dimensions for FHOG feature；Building has gauss of distribution function simultaneously
The position response of distribution exports, and in-between response is maximum and successively successively decreases around.
Change in location filter training module 720, for being exported according to the first FHOG feature and the position response
Training obtains change in location filter.
In embodiments of the present invention, according to the first FHOG feature, the position input of the first FHOG feature is determined
Sample F1, the dimension of the position input sample are d, and export G according to the position response1Obtain change in location filter
H1l。
In an embodiment of the present invention, λ is regularization coefficient, in order to control keeping away during filter frequency domain parametric solution
Exempt 0, on the other hand also can control filter parameter variation range, λ is smaller, and filter parameter variation range is bigger.
First area selecting module 730, for obtaining the second frame of video, by the first area that pre-defined rule is chosen, to obtain
Obtain the characteristic information of the first area.
In an embodiment of the present invention, after obtaining the second frame of video, first frame tracking is obtained according to position filtering device
The center of target, response output maximum value is target's center, and is mesh in first frame as center chosen area size
Mark 1.5-3.5 times of first area of size；The image in the region is obtained to the characteristic information in the region by statistical method.
Target movement position obtains module 740, is used for the characteristic information with the change in location filter according to pre-
Positioning sets variation model and carries out calculating acquisition target movement position.
In an embodiment of the present invention, the characteristic information of the first area of acquisition includes gray feature and FHOG feature,
It includes 1 dimension gray feature and 27 dimension FHOG features that each pixel in the region, which calculates feature, and by all pictures of one's respective area
The characteristic information of vegetarian refreshments is inputted as sample, is denoted as Z1.
According to above-mentioned change in location filter, by moleculeIt is denoted asIn denominatorIt is denoted as B1t,
And it is carried out Fourier transform and obtains position assessment models y1.
By former frame Pt-1In A1t-1And B1t-1And input sample Z1 is substituting in the assessment models y1 of position, obtains y1,
And max (y1) is calculated, obtain the new position of target.
The device passes through position assessment models meter by establishing change in location filter and corresponding regional choice principle
Calculate obtain target new position, can the position to target accurately positioned.
Fig. 8 shows a kind of structural schematic diagram of motion target tracking device of the offer of the embodiment of the present invention eight, in order to just
In explanation, only the parts related to the present invention are shown.
In an embodiment of the present invention, motion target tracking device includes that size characteristic obtains module 810, change in size filter
Wave device training module 820, subcharacter obtain module 830, and target size determining module 840, details are as follows:
Size characteristic obtains module 810, for obtaining the first FHOG feature of video first frame tracking target, and constructs height
The size of this distribution function distribution responds output.
In embodiments of the present invention, the target tracked is chosen in a frame of video image, in the mesh chosen
The FHOG feature of extracted region target is marked, wherein there are multiple dimensions for FHOG feature；Building is with gauss of distribution function point simultaneously
The size of cloth responds output, and in-between response is maximum and successively successively decreases around.
Change in size filter training module 820, for according to the first FHOG feature and size response output
Training obtains change in size filter.
In an embodiment of the present invention, according to the first FHOG feature, determine that the size of the first FHOG feature is defeated
Enter sample F2, the dimension of the size input sample is m, and responds output G according to the size2Obtain change in size filter
H2l。
In an embodiment of the present invention, λ is regularization coefficient, in order to control keeping away during filter frequency domain parametric solution
Exempt 0, on the other hand also can control filter parameter variation range, λ is smaller, and filter parameter variation range is bigger.
Subcharacter obtains module 830, is used for according to change in location filter prediction the second frame target position center, and
Selection principle chooses the image district of multiple and different sizes according to predetermined dimensions centered on the center of second frame target position
Domain, to obtain multiple sub- FHOG features in described image region.
In an embodiment of the present invention, it is calculated first according to change in location filter by change in location model and obtains second
The center of frame target, when choosing different scale sample centered on target position, it is contemplated that tracking target is in adjacent 2 frame
In size will not have big difference, therefore the sample-size selection principle of scale assessment are as follows:
Wherein, P and R is respectively that target is high in the width of former frame, and a=1.02 is scale factor, and S=33 is the number of scale
Amount.Above-mentioned scale not instead of linear relationship, by the detection process in precise and penetrating thick (direction from inside to outside).
In embodiments herein, the second frame target position center is obtained, and as center according to above-mentioned steps meter
The different sizes calculated, choose the region of multiple and different sizes, with the characteristic information for obtaining variant size area, i.e.,
For multiple sub- FHOG features of image-region.
Target size determining module 840, by the multiple sub- FHOG feature respectively with the change in size filter according to
Scheduled change in size model, which calculate, obtains target size.
In an embodiment of the present invention, multiple and different size areas of acquisition, each region are set as fixed size, point
31 dimension FHOG features are indescribably taken, is inputted as sample, is denoted as Z2.According to above-mentioned change in size filter, by moleculeIt is denoted asIn denominatorIt is denoted as B1t, and carried out Fourier transform and obtain position assessment models y2.
In an embodiment of the present invention, by former frame Pt-1In A2t-1And B2t-1And input sample Z2 is substituting to position
In assessment models y2, y2 is obtained, and calculates max (y2max), obtain the new size of target.
The device passes through size assessment mould by establishing the algorithm of dimensional variation filter and the selection of different size areas
Type calculates the new size for obtaining target, can be improved the adaptability to target scale variation, and algorithm reduces calculation amount,
Improve the real-time of algorithm.
In an embodiment of the present invention, the present invention also provides a kind of computer equipment, including memory and processor, institutes
It states memory and is stored with computer program, which is characterized in that the processor realizes movement mesh when executing the computer program
The step of marking tracking.
The embodiments described above only express several embodiments of the present invention, and the description thereof is more specific and detailed, but simultaneously
Limitations on the scope of the patent of the present invention therefore cannot be interpreted as.It should be pointed out that for those of ordinary skill in the art
For, without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to guarantor of the invention
Protect range.Therefore, the scope of protection of the patent of the invention shall be subject to the appended claims.
Those skilled in the art after considering the specification and implementing the invention disclosed here, will readily occur to its of the disclosure
Its embodiment.This application is intended to cover any variations, uses, or adaptations of the disclosure, these modifications, purposes or
Person's adaptive change follows the general principles of this disclosure and including the undocumented common knowledge in the art of the disclosure
Or conventional techniques.The description and examples are only to be considered as illustrative, and the true scope and spirit of the disclosure are wanted by right
It asks and points out.
Although should be understood that various embodiments of the present invention flow chart in each step according to arrow instruction successively
It has been shown that, but these steps are not that the inevitable sequence according to arrow instruction successively executes.Unless expressly state otherwise herein,
There is no stringent sequences to limit for the execution of these steps, these steps can execute in other order.Moreover, each embodiment
In at least part step may include that perhaps these sub-steps of multiple stages or stage are not necessarily multiple sub-steps
Completion is executed in synchronization, but can be executed at different times, the execution in these sub-steps or stage sequence is not yet
Necessarily successively carry out, but can be at least part of the sub-step or stage of other steps or other steps in turn
Or it alternately executes.
Those of ordinary skill in the art will appreciate that realizing all or part of the process in above-described embodiment method, being can be with
Relevant hardware is instructed to complete by computer program, the program can be stored in a non-volatile computer and can be read
In storage medium, the program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein, provided herein
Each embodiment used in any reference to memory, storage, database or other media, may each comprise non-volatile
And/or volatile memory.Nonvolatile memory may include that read-only memory (ROM), programming ROM (PROM), electricity can be compiled
Journey ROM (EPROM), electrically erasable ROM (EEPROM) or flash memory.Volatile memory may include random access memory
(RAM) or external cache.By way of illustration and not limitation, RAM is available in many forms, such as static state RAM
(SRAM), dynamic ram (DRAM), synchronous dram (SDRAM), double data rate sdram (DDRSDRAM), enhanced SDRAM
(ESDRAM), synchronization link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) directly RAM (RDRAM), straight
Connect memory bus dynamic ram (DRDRAM) and memory bus dynamic ram (RDRAM) etc..
Each technical characteristic of embodiment described above can be combined arbitrarily, for simplicity of description, not to above-mentioned reality
It applies all possible combination of each technical characteristic in example to be all described, as long as however, the combination of these technical characteristics is not deposited
In contradiction, all should be considered as described in this specification.
Claims (10)
1. a kind of motion target tracking method, which is characterized in that it specifically includes,
The first FHOG feature of video first frame tracking target is obtained, and constructs the position response output of gauss of distribution function distribution
It responds and exports with size；
Change in location is obtained according to the training of preset training rules according to the first FHOG feature and position response output
Filter obtains change in size according to training rules training according to the first FHOG feature and size response output
Filter；
The second frame of video is obtained, by the first area that presumptive area selection rule is selected, and obtains the feature of the first area
Information；
The characteristic information and the change in location filter calculate according to predetermined position variation model and obtain target fortune
Dynamic position；
It is according to change in location filter prediction the second frame target position center, and with second frame target position center
Selection principle chooses the image-region of multiple and different sizes according to predetermined dimensions at center, and obtains the multiple of described image region
Sub- FHOG feature；
By the multiple sub- FHOG feature, variation model is counted according to predetermined dimensions with the change in size filter respectively
It calculates and obtains target size.
2. motion target tracking method according to claim 1, which is characterized in that described according to the first FHOG feature
Change in location filter is obtained according to the training of preset training rules with position response output, it is special according to the first FHOG
The method that the size response output of seeking peace obtains change in size filter according to training rules training；It specifically includes:
According to the first FHOG feature, the position input sample F1 of the first FHOG feature is determined, the position inputs sample
This dimension is d, and exports G according to the position response1Obtain change in location filter
According to the first FHOG feature, the size input sample F2 of the first FHOG feature is determined, the size inputs sample
This dimension is m, and responds output G according to the size2Obtain change in size filter
The λ is regularization coefficient.
3. motion target tracking method according to claim 1, which is characterized in that the second frame of the acquisition video, by pre-
The first area that set pattern is then chosen, the method to obtain the characteristic information of the first area specifically include,
The second frame of video is obtained, centered on the center for obtaining first frame tracking target according to the change in location filter, choosing
Take the first area having a size of 1.5-3.5 times of target size；
Characteristic information is extracted in the first area.
4. motion target tracking method according to claim 2, which is characterized in that it is described by the characteristic information with it is described
Change in location filter carries out calculating the method for obtaining target movement position according to predetermined position variation model, specifically includes:
Characteristic information is obtained, the characteristic information includes gray feature and FHOG feature, and using the characteristic information as sample
Input Z1；
Position assessment models are obtained according to the change in location filter
WhereinForB1tFor
The sample information Z1 is calculated by the position assessment models, is obtained max (y1), to obtain target motion bit
It sets.
5. motion target tracking method according to claim 1, which is characterized in that described with second frame target position
Selection principle chooses the image-region method of multiple and different sizes according to predetermined dimensions centered on center, specifically includes:
The wide P and high R of target in first frame are obtained, and the P and R is substituted into formula anP×anR is calculated, wherein
A is constant scale factor and s scale quantity, obtains multiple and different sizes；
Centered on the center of second frame target position, according to the image of the multiple and different sizes of the multiple different selection of dimension
Region.
6. motion target tracking method according to claim 2, which is characterized in that described by the multiple sub- FHOG feature
Variation model carries out calculating the method for obtaining target size according to predetermined dimensions with the change in size filter respectively, specifically
Include:
The multiple sub- FHOG feature that will acquire inputs Z2 as sample；
Size assessment model is obtained according to the change in location filter
WhereinForB2tFor
The sample information Z2 is calculated by the size assessment model, is obtained max (y2), to obtain target size.
7. the motion target tracking method according to claim 4 or 6, which is characterized in that the motion target tracking method
Further include,
The first frame target FHOG feature and response output are carried out with the second frame target FHOG feature and response output
Comparison obtains filter learning efficiency；
Position filter and size filter are updated according to the learning efficiency.
8. a kind of motion target tracking device characterized by comprising
Position feature obtains module, for obtaining the first FHOG feature of video first frame tracking target, and constructs Gaussian Profile
The position response output of function distribution；
Change in location filter training module, for being obtained according to the first FHOG feature and position response output training
Change in location filter；
First area selecting module, for obtaining the second frame of video, by the first area that pre-defined rule is chosen, to obtain described the
The characteristic information in one region；
Target movement position obtains module, for becoming the characteristic information according to predetermined position with the change in location filter
Change model to carry out calculating acquisition target movement position.
9. a kind of motion target tracking device, which is characterized in that further include:
Size characteristic obtains module, for obtaining the first FHOG feature of video first frame tracking target, and constructs Gaussian Profile
The size of function distribution responds output；
Change in size filter training module, for being obtained according to the first FHOG feature and size response output training
Change in size filter；
Subcharacter obtains module, is used for according to change in location filter prediction the second frame target position center, and with described
Selection principle chooses the image-region of multiple and different sizes according to predetermined dimensions centered on second frame target position center, to obtain
Take multiple sub- FHOG features in described image region；
Target size determining module, by the multiple sub- FHOG feature respectively with the change in size filter according to scheduled ruler
Very little variation model, which calculate, obtains target size.
10. a kind of computer equipment, including memory and processor, the memory are stored with computer program, feature exists
In the step of processor realizes any one of claims 1 to 7 the method when executing the computer program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811596674.7A CN109685832A (en) | 2018-12-26 | 2018-12-26 | A kind of motion target tracking method, device and computer equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811596674.7A CN109685832A (en) | 2018-12-26 | 2018-12-26 | A kind of motion target tracking method, device and computer equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109685832A true CN109685832A (en) | 2019-04-26 |
Family
ID=66188572
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811596674.7A Pending CN109685832A (en) | 2018-12-26 | 2018-12-26 | A kind of motion target tracking method, device and computer equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109685832A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111080675A (en) * | 2019-12-20 | 2020-04-28 | 电子科技大学 | Target tracking method based on space-time constraint correlation filtering |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015163830A1 (en) * | 2014-04-22 | 2015-10-29 | Aselsan Elektronik Sanayi Ve Ticaret Anonim Sirketi | Target localization and size estimation via multiple model learning in visual tracking |
| CN106683110A (en) * | 2015-11-09 | 2017-05-17 | 展讯通信(天津)有限公司 | User terminal and object tracking method and device thereof |
| CN107452022A (en) * | 2017-07-20 | 2017-12-08 | 西安电子科技大学 | A kind of video target tracking method |
| CN107644430A (en) * | 2017-07-27 | 2018-01-30 | 孙战里 | Target following based on self-adaptive features fusion |
| CN108846851A (en) * | 2018-04-25 | 2018-11-20 | 河北工业职业技术学院 | Motion target tracking method and terminal device |
-
2018
- 2018-12-26 CN CN201811596674.7A patent/CN109685832A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015163830A1 (en) * | 2014-04-22 | 2015-10-29 | Aselsan Elektronik Sanayi Ve Ticaret Anonim Sirketi | Target localization and size estimation via multiple model learning in visual tracking |
| CN106683110A (en) * | 2015-11-09 | 2017-05-17 | 展讯通信(天津)有限公司 | User terminal and object tracking method and device thereof |
| CN107452022A (en) * | 2017-07-20 | 2017-12-08 | 西安电子科技大学 | A kind of video target tracking method |
| CN107644430A (en) * | 2017-07-27 | 2018-01-30 | 孙战里 | Target following based on self-adaptive features fusion |
| CN108846851A (en) * | 2018-04-25 | 2018-11-20 | 河北工业职业技术学院 | Motion target tracking method and terminal device |
Non-Patent Citations (2)
| Title |
|---|
| MARTIN DANELLJAN ET AL.: ""Accurate Scale Estimation for Robust Visual Tracking"", 《ACCURATE SCALE ESTIMATION FOR RUBUST VISUAL TRACKING》 * |
| MARTIN DANELLJAN ET AL.: ""Discriminative Scale Space Tracking"", 《IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111080675A (en) * | 2019-12-20 | 2020-04-28 | 电子科技大学 | Target tracking method based on space-time constraint correlation filtering |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Haeusser et al. | Associative domain adaptation | |
| Lin et al. | Spatial-temporal residential short-term load forecasting via graph neural networks | |
| Chen et al. | 2-D regional short-term wind speed forecast based on CNN-LSTM deep learning model | |
| CN112348849B (en) | Twin network video target tracking method and device | |
| Xu et al. | Accurate and interpretable bayesian mars for traffic flow prediction | |
| CN112949828B (en) | Graph convolution neural network traffic prediction method and system based on graph learning | |
| CN108764325A (en) | Image-recognizing method, device, computer equipment and storage medium | |
| CN112437451B (en) | Wireless network flow prediction method and device based on generation countermeasure network | |
| CN112419710A (en) | Traffic congestion data prediction method, traffic congestion data prediction device, computer equipment and storage medium | |
| CN111340842B (en) | Correlation filtering target tracking method based on joint model | |
| CN111583294B (en) | Target tracking method combining scale self-adaption and model updating | |
| CN111047078B (en) | Traffic characteristic prediction method, system and storage medium | |
| Li et al. | A novel multichannel long short-term memory method with time series for soil temperature modeling | |
| Wilms et al. | Exploiting spatio-temporal dependencies for RNN-based wind power forecasts | |
| CN108537825B (en) | Target tracking method based on transfer learning regression network | |
| Akter et al. | Analysis and prediction of hourly energy consumption based on long short-term memory neural network | |
| Lu et al. | A multi-source transfer learning model based on LSTM and domain adaptation for building energy prediction | |
| Jain et al. | Data predictive control for peak power reduction | |
| CN109685832A (en) | A kind of motion target tracking method, device and computer equipment | |
| Fan | Research on deep learning energy consumption prediction based on generating confrontation network | |
| CN110751670B (en) | Target tracking method based on fusion | |
| Zhao et al. | Urban traffic flow forecasting based on memory time-series network | |
| Fan et al. | Generating high quality crowd density map based on perceptual loss | |
| CN116386312A (en) | Traffic prediction model construction method and system | |
| CN112308045B (en) | Method and device for detecting dense crowd and electronic equipment |
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: 20190426 |
|
| RJ01 | Rejection of invention patent application after publication |