CN108062763A - Method for tracking target and device, storage medium - Google Patents
Method for tracking target and device, storage medium Download PDFInfo
- Publication number
- CN108062763A CN108062763A CN201711485059.4A CN201711485059A CN108062763A CN 108062763 A CN108062763 A CN 108062763A CN 201711485059 A CN201711485059 A CN 201711485059A CN 108062763 A CN108062763 A CN 108062763A
- Authority
- CN
- China
- Prior art keywords
- tracking
- information
- image
- target
- current time
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- 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
- G06T2207/20084—Artificial neural networks [ANN]
Abstract
The embodiment of the invention discloses a kind of method for tracking target and devices and storage medium.The method for tracking target includes：Obtain the image currently gathered；Using the first track algorithm, information processing is carried out to the first tracking information of the first image information and previous moment, to obtain first tracking information at current time, wherein, described first image information is the image information of the described image currently gathered；Utilize the second track algorithm, first characteristics of image of the tracking target obtained at predetermined intervals to the second image information and historical juncture carries out information processing, and determine second tracking information at current time, wherein, second image information is the image information of the described image currently gathered；At the time of the historical juncture is before the current time；According to second tracking information at the current time, first tracking information at the current time is corrected.
Description
Technical field
The present invention relates to information technology field more particularly to a kind of method for tracking target and device and storage medium.
Background technology
Image Acquisition is carried out to tracking target to camera, the extraction of image information then is carried out to the image of acquisition, it is real
Now the vision for tracking target is tracked.
There are many vision tracking modes in the prior art, for example, correlation filtering etc. carries out target following, these
Track algorithm has the characteristics that calculation amount is small.But during tracking, the letter based on current frame image and previous frame image
It ceases into line trace, but this tracking can cause the accumulation of error with the accumulation of time, finally due to the error mistake of accumulation
Greatly, and then cause to track the problem of target with losing, is failed so as to cause tracking.Therefore in the prior art, a kind of raising is proposed
The method of tracking success rate is a problem to be solved.
The content of the invention
In view of this, an embodiment of the present invention is intended to provide a kind of method for tracking target and devices and storage medium, at least portion
Divide and solve the above problems.
In order to achieve the above objectives, the technical proposal of the invention is realized in this way：
In a first aspect, the embodiment of the present invention provides a kind of method for tracking target, including：
Obtain the image currently gathered；
Using the first track algorithm, to the first tracking information of the first image information and previous moment at row information
Reason, to obtain first tracking information at current time, wherein, described first image information is the figure of the described image currently gathered
As information；
Using the second track algorithm, the of the tracking target that is obtained at predetermined intervals to the second image information and historical juncture
One characteristics of image carries out information processing, and determines second tracking information at current time, wherein, second image information is to work as
The image information of the described image of preceding acquisition；At the time of the historical juncture is before the current time；
According to second tracking information at the current time, the first tracking information for correcting the current time utilizes first
Track algorithm, the first image information and the first tracking information to described image carry out information processing, to obtain current time
First tracking information.
Optionally, it is described to utilize the first track algorithm, the first image information and the first tracking information to described image into
Row information processing, to obtain first tracking information at current time, including：
According to first tracking information of previous moment, target area is determined；；
Extract the second characteristics of image of the target area in described image；
Second characteristics of image is determined to the location information and dimension information at the tracking target current time.
Optionally, the historical juncture utilizes the second track algorithm, at predetermined intervals to the second image information and the first figure
As feature progress information processing, and determine second tracking information at current time, including：
Target detection is carried out to described image, to obtain the 3rd characteristics of image of candidate target；
3rd characteristics of image and described first image feature are clustered；
According to cluster as a result, selecting the tracking target from the candidate target；
Determine the status information and dimension information of the tracking target.
Optionally, the method further includes：
If the tracking target based on first track algorithm is lost, using second track algorithm to image
Second image information and described first image feature carry out information processing, obtain second tracking at the tracking target current time
Information；
Using second tracking information as the first tracking that the current time is obtained using first track algorithm
Information.
Optionally, the method further includes：
Tracking start time selectes the tracking target.
Optionally, the tracking start time selectes the tracking target, including at least one of：
It is indicated to select the tracking target according to user；
The tracking target is automatically selected according to alternative condition.
Optionally, it is described that the tracking target is automatically selected according to alternative condition, including：
It is the tracking target to select the candidate target in the focusing area of camera；
According to the relative position relation between candidate target and camera, selection is located at the candidate immediately ahead of the camera
Target is the tracking target；
In the image gathered before tracking start time, the candidate target of imaging area maximum is selected as the tracking mesh
Mark.
Optionally, it is described to utilize the second track algorithm, the second image information and historical juncture are obtained at predetermined intervals
First characteristics of image of the tracking target carries out information processing, and determines second tracking information at current time, including：
Using second track algorithm, at predetermined intervals to described in second image information and historical juncture acquisition
The first characteristics of image for tracking target carries out asynchronous process, and determines second tracking information at current time.
Second aspect, the embodiment of the present invention provide a kind of target tracker, including：
Acquiring unit, for obtaining the image currently gathered；
First tracking cell, for utilizing the first track algorithm, to the first of the first image information and previous moment with
Track information carries out information processing, to obtain first tracking information at current time, wherein, described first image information is currently to adopt
The image information of the described image of collection；
Second tracking cell, for utilizing the second track algorithm, at predetermined intervals to the second image information and historical juncture
First characteristics of image of the tracking target of acquisition carries out information processing, and determines second tracking information at current time, wherein, institute
State the image information that the second image information is the described image currently gathered；The historical juncture is pervious for the current time
Moment；
Correct unit, for the second tracking information according to the current time, correct the first of the current time with
Track information.
Optionally, second tracking cell is lost if being additionally operable to the tracking target based on first track algorithm
It loses, information processing is carried out to the second image information and described first image feature of image using second track algorithm, is obtained
Obtain second tracking information at the tracking target current time；
First tracking cell is additionally operable to obtain using second tracking information as using first track algorithm
First tracking information at the current time.
The third aspect, the embodiment of the present invention provide a kind of computer storage media, and the computer storage media is stored with
Computer program；After the computer program is performed, can realize target that foregoing one or more technical solution provides with
Track method.
Method for tracking target and device and storage medium provided in an embodiment of the present invention, in the tracking into line trace target
When, it is tracked in real time using the first track algorithm, obtains the first tracking information；In order to avoid the error of the first tracking information
The problem of Loss Rate of tracking target is high caused by accumulation.The second track algorithm is also introduced in embodiments of the present invention according to pre-
Fixed be spaced into line trace obtains the second tracking information, and school is carried out at predetermined intervals using the second tracking information as the first tracking information
Positive control information tracks the problem of target Loss Rate is high caused by the first tracking information long-time error accumulation so as to reduce,
So as to promote the tracking success rate of tracking target.And usual first track algorithm tracking calculation amount it is small and tracking Loss Rate spy
Point.
Description of the drawings
Fig. 1 is the flow diagram of the first method for tracking target provided in an embodiment of the present invention；
Fig. 2 is the flow diagram of second of method for tracking target provided in an embodiment of the present invention；
Fig. 3 is a kind of structure diagram of target tracker provided in an embodiment of the present invention；
Fig. 4 is the flow diagram of the third method for tracking target provided in an embodiment of the present invention；
Fig. 5 is the flow diagram of the 4th kind of method for tracking target provided in an embodiment of the present invention；
Fig. 6 is the flow diagram of the 5th kind of method for tracking target provided in an embodiment of the present invention.
Specific embodiment
Technical scheme is further elaborated below in conjunction with Figure of description and specific embodiment.
As shown in Figure 1, the present embodiment provides a kind of method for tracking target, including：
Step S110：Obtain the image currently gathered；
Step S120：Using the first track algorithm, to the first tracking information of the first image information and previous moment into
Row information processing, to obtain first tracking information at current time, wherein, described first image information is described currently to gather
The image information of image；
Step S130：Using the second track algorithm, at predetermined intervals to the second image information and historical juncture obtain with
First characteristics of image of track target carries out information processing, and determines second tracking information at current time, wherein, second figure
As the image information that information is the described image currently gathered；At the time of the historical juncture is before the current time；
Step S140：According to second tracking information at the current time, the first tracking for correcting the current time is believed
Breath.
Method for tracking target provided in this embodiment can be the method applied in tracking system or apply to
Method in track equipment.
The tracking system may include：The camera of multiple fixed settings, each camera carry out in oneself region
Tracking then when tracking target moves out the scope of itself, is continued to track by next camera.
The tracking equipment can be the mobile equipment that is equipped with camera and can move.
The mobile equipment may include：On the vehicles or movable machine people.The vehicles can be fuel oil
Vapour, electric car, hybrid electric vehicle, man carrier or load-carrying vehicle etc..The movable machine people can be：Include mobile chassis
It can be in the robot on ground, alternatively, the air-robot that can be flown or slide in the air.
In the present embodiment using camera collection image, one or more cameras can be utilized in the present embodiment
Image is gathered, for example, gathering two dimensional image using common camera, depth camera can also be utilized to gather 3-D view,
So as to obtain tracking more accurate location information of target etc..
The tracking target may include in the present embodiment：Tracked people, tracked vehicle or tracked target
Etc. can be with the life entity or article of automatic moving.
It may include in step s 110：By video acquisition obtain image, in the step s 120 be pair in step S130
The picture frame of video frame carries out image procossing, obtains the first tracking information and the second tracking information respectively.Institute in the present embodiment
It states the first tracking information and second tracking information includes at least location information, in some cases, the second tracking letter
Breath further includes dimension information.
Image procossing is carried out to image in the step s 120, obtains the first image information of image, is believed using the first image
Breath obtains first tracking information at current time with reference to the first tracking information of previous moment.At the present embodiment current time
First tracking information is to rely on first tracking information at previous moment, if first tracking information at previous moment exists
Error, and since each first tracking information of moment all relies on first tracking information at previous moment, it is clear that have higher
Probability generate error accumulation the phenomenon that, so as to ultimately result in error accumulation it is excessive so that tracking target by the risk with losing
It is larger.
For example, in the step s 120 in order to reduce the calculation amount of image procossing, can according to the first of the previous moment with
Location information in track information in the image of current time acquisition, draws a circle to approve out the target of the imaging of the searching tracking target
Region, the area of the target area are typically less than the area of whole image, so as to fulfill the reduction of calculation amount and subtracting for delay
It is small.At this point, if the imaging of tracking target may not can cause the loss for tracking target in the target area.
Also the second image of image is believed according to predetermined space using the second track algorithm in the step S130 of the present embodiment
Breath and the first characteristics of image of historical juncture carry out information processing, and determine the second tracking letter at tracking target current time
Breath.Here historical juncture is any one pervious moment at current time.In the present embodiment, described first image feature can
For the CNN features of convolutional neural networks (CNN) extraction.
For example, in the present embodiment, the entire figure that the step S130 can gather current time according to predetermined space
As carrying out the information processings such as feature extraction, if the imaging for so tracking target still includes in the picture, just can centainly looking for
To the tracking target, further according to the position and the size that are imaged in the picture of tracking target, it is possible to know it compared with target
The position of the reference points such as tracks of device, so as to obtain tracking information.Obviously compared with only to target area progress image procossing
First track algorithm, it is possible to reduce the phenomenon that tracking is lost, but need to handle whole image, it may generate more
Calculation amount and bigger delay.
It the characteristics of both track algorithms can be combined in this embodiment, is persistently tracked using the first track algorithm, so as to subtract
Few calculation amount, and utilize the reference of the first tracking information of the second track algorithm interval tracking acquisition the first track algorithm of correction
Amount, so as to solve calculation amount and track the collision problem between success rate.
It is above-mentioned be to step S120 and step S130 for example, specific implementation be not limited to it is above-mentioned any one.
In some embodiments, first track algorithm can be any suitable track algorithm in short-term in short time tracking, for example, typical
Correlation filtering track algorithm, deterministic metric space track algorithm (Discriminant Scale Space Tracing,
DSST) etc..Second track algorithm can be the track algorithm of tracking for a long time, for example, Hungary Algorithm etc..
In some embodiments, the predetermined space can be predetermined time interval, for example, just carrying out profit at interval of t durations
Second tracking information is obtained with second track algorithm, is like this achieved that based on the second track algorithm
Periodically tracking.In the present embodiment, using first tracking information at the second tracking information correction current time, so it is equivalent to
First tracking information will be corrected every duration t, so as to reduce the error of first tracking information at current time, so as to subtract
Lack and do not corrected the phenomenon that the first tracking information causes the first tracking information error accumulation to cause tracking failure for a long time.
In further embodiments, the predetermined space can also be the acquisition image or picture frame of interval predetermined quantity.Example
Such as, N number of picture frame is spaced, an image procossing is carried out using second track algorithm, to obtain once second tracking
Information.The picture frame for being spaced predetermined number obtains the second tracking information of first tracking information of correction afterwards, so that
When next moment obtains follow-up first tracking information using first tracking information at current time, it is possible to avoid the length of error
Accumulated time, so as to fulfill accurate tracking, so as to reduce the phenomenon that tracking target is lost.
In having some embodiments, the predetermined space is static state setting, for example, preset interval duration or,
Preset partition image quantity.
In further embodiments, the method further includes：Obtain second tracking information at current time with
The diversity factor of first tracking information determines next predetermined space according to the diversity factor.If for example, second tracking
The diversity factor of information and first tracking information is located in the range of first, then the predetermined space is the first interval, if described
The diversity factor of second tracking information and second tracking information is located at outside the first scope, then between the predetermined space is second
Every second interval is more than the described second interval.Accordingly, target tracker can be after second interval be spaced, profit
Next second tracking information is obtained with second track algorithm.Therefore in the present embodiment, the predetermined space be can
It is dynamically determined with the current tracking error degree according to the first track algorithm, if the current tracking error based on the first track algorithm
It is small, it is possible to appropriate reduction obtains the number of operations of the second tracking information using the second track algorithm, so as to reduce target with
The calculation amount of track algorithm, if the current tracking based on the first track algorithm is big, appropriate increase is obtained using the second track algorithm
The number of operations of second tracking information is obtained, so as to reduce the tracking target by the risk with losing.By being set dynamically
Predetermined space is stated, so as to realize the balance between calculation amount and tracking success rate well.
Optionally, the step S120 may include：
According to first tracking information of previous moment, target area is determined；；
Extract the second characteristics of image of the target area in described image；
Second characteristics of image is determined to the location information and dimension information at the tracking target current time.
First tracking information may include：The location information and/or ruler that the tracking target is imaged in described image
Very little information.
The location information can be：The tracking target is equivalent to the position letter in the coordinate system of tracking equipment itself setting
Breath.For example, the coordinate system is two-dimensional Cartesian coordinate system, then the location information may include：Coordinate in x-axis and y-axis.If the seat
Mark system is three-dimensional cartesian coordinate system, then the location information may include：The coordinate of x-axis, y-axis and z-axis.The x, y and z-axis are hung down two-by-two
Directly.In further embodiments, if the coordinate system is spherical coordinate, the coordinate information may include：Relative angle
The length information of information and radius.In short, location information described in the present embodiment can reflect it is described tracking target compared with
Location information between reference point.The reference point can be for target following equipment in itself.
The dimension information may include：In the width of the first dimension and the width of the second dimension；First dimension perpendicular
In second dimension.For example, utilization orientation represents an imaging in the picture, then the length and width of the direction, can correspond to respectively
In the width of first dimension and the second dimension.In further embodiments, the dimension information may also include：Indicate area
Area information etc., be not limited to the indication informations such as length and width.
According to first tracking information of previous moment, target area is determined；；The is proposed from the target area in image
Two characteristics of image.Under normal conditions, the target area is only the subregion of image.The second characteristics of image is extracted, this
Two characteristics of image may include：The features such as color histogram and/or histograms of oriented gradients (HOG), then by the second characteristics of image
With the matching of the second characteristics of image in the image of previous moment extraction, it is possible to determine which imaging is tracking in image
The imaging of target in the location information based on previous moment, with reference to the variation of dimension information, determines that the position at current time is believed
Breath.
The image of the previous moment can be previous image frame in video in the present embodiment, and the image at current time can be
Present frame in video.It can carry out image procossing frame by frame in the step s 120, determine the first tracking information of tracking target.
In the present embodiment, by the matching of characteristics of image, it is known that the image at current time and previous moment
The size change over of the imaging of target is tracked in image, the uniformity based on size change over and evolution can be based on previous
Location information in picture frame is determined to track the location information of target described in current time.
Optionally, the step S130 may include：
Target detection is carried out to described image, to obtain the 3rd characteristics of image of candidate target；
3rd characteristics of image and described first image feature are clustered；
According to cluster as a result, selecting the tracking target from the candidate target；
Determine that the status information of the tracking target and dimension information determine the status information and size of the tracking target
Information.
In the present embodiment, the 3rd characteristics of image equally may include：The color histogram of each candidate target and/or
HOG etc..3rd characteristics of image is clustered with described first image feature, so that it is determined that going out to track target.Described
Three characteristics of image and the second characteristics of image can may determine that the 3rd from the characteristics of image that different dimensions extract by cluster
The similarity of characteristics of image and the first characteristics of image, so as to which the sufficiently high candidate target of similarity be selected to be considered as the tracking pair
As.Which candidate target is so equivalent to define in image as tracking target, a kind of mode be first determine target with
Then track object can combine the imaging position in the picture of tracking target and/or area of occupancy etc., it may be determined that go out with
The tracking informations such as the current location information of track target.Another way is, while some candidate target is determined, base
In the characteristics of image of candidate target, the status information and dimensional information of each candidate target are defined, then determines target object,
If target object determines, the status information and dimensional information of target object also determine that.
As shown in Fig. 2, the method further includes：
Step S101：If the tracking target based on first track algorithm is lost, calculated using the described second tracking
Method carries out information processing to the second image information and described first image feature of image, obtains the tracking target current time
The second tracking information；
Step S102：The current time is obtained using second tracking information as using first track algorithm
First tracking information.
In some embodiments, tracking target is possible to blocked by other objects alternatively, due to accelerating movement causing
In the image of current time acquisition, the imaging for tracking target is no longer at first tracking information based on the previous moment
In definite target area.At this point, cause the loss for tracking target.In the present embodiment, without etc. reach predetermined space just
It enables the second track algorithm and carries out image procossing, to obtain the second tracking information of the tracking target.If currently track target
It is to be blocked by other objects, then using processing of second track algorithm to multiple images, until finding tracking target in the picture
Imaging, and obtain its second tracking information.The second tracking information will be regained again and is assigned to the first tracking information, and rewriting is opened
The dynamic tracking based on the first track algorithm.Obviously like this, can track target centainly lose can quickly again with
Track to tracking target.
Optionally, the method further includes：Tracking start time selectes the tracking target.
The tracking start time selectes the tracking target, including at least one of：
It is indicated to select the tracking target according to user；
The tracking target is automatically selected according to alternative condition.
For example, camera acquires the image of an imaging for including multiple candidate targets, target tracker shows institute
Image is stated, acquisition user acts on the operation of the selection tracking target of described image, described in being selected according to user's instruction
Track target.
In another example the target tracker can also automatically select tracking target, for example, according to preconfigured selection
Condition selects the tracking target.Several optional modes for automatically selecting tracking target presented below：
Optional mode one：
It is the tracking target to select the candidate target in the focusing area of camera.Camera is clear in order to gather
Image be required for performing focus operation, most clearly region is normally referred to as focusing area in image after focus operation, if
The imaging of some candidate target is fully located at or part is located at focusing area, then it is assumed that the candidate target is the tracking mesh
Mark.If the imaging for having multiple candidate targets is located at the focusing area, can select positioned at the focusing area ratio most
Big or area maximum candidate target is as the tracking target.
Optional mode two：
According to the relative position relation between candidate target and camera, selection is located at the candidate immediately ahead of the camera
Target is the tracking target.Camera has certain acquisition direction or acquisition angles, if camera is towards some direction
Illustrate its image for wanting to gather people in this direction, object or scape, therefore can position be determined by image analysis in the present embodiment
Candidate target in front of camera is the tracking target.
Optional mode three：
In the image gathered before tracking start time, the candidate target of imaging area maximum is selected as the tracking mesh
Mark.If user wants to track that object, the acquisition parameter of camera can be adjusted to be suitable for gathering the object naturally.Therefore at this
Selected in embodiment from the image that camera gathers before tracking is formally started the candidate target of imaging area maximum as
The tracking target.
There are many modes for determining the tracking target in a word, when specific implementation be not limited to it is above-mentioned any one.
Optionally, the step S130 includes：
Using second track algorithm, at predetermined intervals to described in second image information and historical juncture acquisition
The first characteristics of image for tracking target carries out asynchronous process, and determines second tracking information at current time.
In the present embodiment, the second track algorithm carries out image procossing and obtains the second tracking information, compared with incessantly S120
In to obtain the first tracking information based on the first track algorithm be asynchronous progress, therefore the asynchronous process gathered in the present embodiment.
In the present embodiment, the asynchronous process may include：First track algorithm and the second track algorithm are performed using different threads
Respective processing when recycling the second track algorithm execution asynchronous process, will not occupy the calculating money of the first track algorithm execution
Source and storage resource etc., so as to directly be mutually independent.
As shown in figure 3, the present embodiment provides a kind of target tracker, including：
Acquiring unit 110, for obtaining the image currently gathered；
First tracking cell 120, for utilizing the first track algorithm, to the first of the first image information and previous moment
Tracking information carries out information processing, to obtain first tracking information at current time, wherein, described first image information is current
The image information of the described image of acquisition；
Second tracking cell 130, for utilize the second track algorithm, at predetermined intervals to the second image information and history when
The first characteristics of image for carving the tracking target obtained carries out information processing, and determines second tracking information at current time, wherein,
Second image information is the image information of the described image currently gathered；The historical juncture is before the current time
At the time of；
Unit 140 is corrected, for the second tracking information according to the current time, corrects the first of the current time
Tracking information.
The acquiring unit 110, first tracking cell 120, the second tracking cell 130 and correction unit 140 can be
Program module.
For example, the acquiring unit 110 may correspond to communication interface, it can be performed by program from other equipment and receive institute
Image is stated, may correspond to camera, by the execution of program code camera is controlled to carry out Image Acquisition.
First tracking cell, 120 and second tracking cell 130 may correspond to processor, and the processor can be center
Processor, microprocessor, digital signal processor, application processor, programmable array or application-specific integrated circuit etc. pass through calculating
The execution of the executable instructions such as machine program, the first tracking information of realization and the second tracking information determine.
Optionally, first tracking cell 120, specifically for first tracking information according to previous moment, really
Set the goal region；；Extract the second characteristics of image of the target area in described image；Second characteristics of image is determined into institute
State the location information and dimension information at tracking target current time.
Optionally, second tracking cell 130, specifically for carrying out target detection to described image, to obtain candidate
3rd characteristics of image of target；3rd characteristics of image and described first image feature are clustered；According to the knot of cluster
Fruit selects the tracking target from the candidate target；Determine that status information and the dimension information of the tracking target are true
The status information and dimension information of the fixed tracking target.
Optionally, second tracking cell 130, if being additionally operable to the tracking target based on first track algorithm
It loses, information processing is carried out to the second image information and described first image feature of image using second track algorithm,
Obtain second tracking information at the tracking target current time；
First tracking cell 120 is additionally operable to using second tracking information as utilization first track algorithm
Obtain first tracking information at the current time.
Optionally, described device further includes：Selecting unit selectes the tracking target for tracking start time.
Optionally, the selecting unit, specifically for performing at least one of：It is indicated to select the tracking according to user
Target；The tracking target is automatically selected according to alternative condition.
Further, the selecting unit, specifically for performing at least one of：Selection is positioned at the focusing area of camera
Candidate target in domain is the tracking target；According to the relative position relation between candidate target and camera, selection is located at
Candidate target immediately ahead of the camera is the tracking target；In the image gathered before tracking start time, selection
The candidate target of imaging area maximum is the tracking target.
The embodiment of the present invention, which also provides a kind of target tracker, to be included：Memory and processor；It can on the memory
It is stored with computer program；The processor is connected with the memory, for example, connected by IC bus etc., it is described
Processor can realize the target following that any one foregoing technical solution provides by the execution of the computer program
Method, for example, performing the method described in Fig. 1 and/or Fig. 2.
The embodiment of the present invention also provides a kind of computer storage media, and the computer storage media is stored with computer journey
Sequence；After the computer program is performed, the method for tracking target that foregoing one or more technical solutions provide, example can be realized
Such as, Fig. 1 and/or method for tracking target shown in Fig. 2.
Computer program medium provided in this embodiment may include：Movable storage device, read-only memory (ROM, Read-
Only Memory), random access memory (RAM, Random Access Memory), magnetic disc or CD etc. are various can be with
Store the medium of program code；It is chosen as non-moment storage medium.
Several specific examples are provided below in conjunction with above-mentioned any embodiment：
Example 1：
This example is a kind of multi-object tracking method of view-based access control model, and this method is into while line trace, every certain
Time can detect multiple candidate targets.And candidate target is clustered, so that it is determined that the state of tracking target and position.Most
The feature of multiple candidate targets and the result of cluster can be stored afterwards, using the foundation clustered as subsequent candidate target.
As shown in figure 4, this exemplary target following flow can be as follows：
Selected tracking target, including：Tracked target is selected by the interaction of user, for example, it is selected against camera
User as tracking target.Or user's directly selected tracking target on mobile phone.
Condition is preset to select tracking target, for example directly selects the target of region maximum as tracking target.
It tracks in short-term：After selected target, using tracking target the methods of DSST correlation filterings, lose when tracking target or
Person can perform following step when reaching certain time interval.Here tracking in short-term corresponds to foregoing using the first tracking calculation
What method carried out can with this；
Whether judgement tracks in short-term fails or then, if into target re-detection, returns track in short-term if not；
Target re-detection：In an asynchronous manner to target detection of carry out of present frame, multiple candidate targets are obtained.Inspection
The method of survey does not limit to specific object detection algorithms.
Multi-object clustering：The features such as color histogram or the HOG of candidate target are extracted in an asynchronous manner, and pass through breast
Tooth profit algorithm is clustered.The algorithm can match current candidate target and the tracking being recorded target signature, with true
The tracking the target whether fixed candidate target is traced.Hungary Algorithm is to find augmenting path, it is a kind of to use augmentation
Seek the algorithm of bipartite graph maximum matching in path.The algorithm can overcome the candidate target increase reduction during tracking to bring well
Cluster difficulty.Goal re-detection and multi-object clustering are equivalent to the foregoing tracking carried out using the second track algorithm.
The characteristics of image of tracking target can be recorded in target following equipment, when convenient tracking information next time obtains,
Tracking target is selected from candidate target.
Tracking correction：The tracking information of current tracking target can be updated in the submodule tracked in short-term, and is held again
Row tracks in short-term.
As described in Figure 5, this example includes the tracking equipment of target tracker, and the flow for carrying out target following is as follows：
1. robot selectes tracking target automatically, make for example, selecting unique candidate target automatically on incoming image
To follow target.
2. start the multiple target tracking of view-based access control model, for example, the multi-track algorithm of operation view-based access control model carries out target following.
3. robot follows corresponding position, for example, obtaining the location information of tracking target, controllable robot is run to
Corresponding position
Example 2：
This example for a long time can effectively track current goal with view-based access control model.
This example combines target re-detection and multi-object clustering algorithm on the basis of track algorithm in short-term, appropriate
When interval time in, target re-detection can be carried out to obtain multiple candidate targets, then candidate target cluster it is final really
Surely clarification of objective and position are tracked.In long-play, this example can constantly confirm tracked target, correction tracking mesh
Target position ensures the reliability of tracking result.
This example employs Hungary matching algorithm, and the clustering problem in multiple target tracking algorithm is converted for a figure
Matching problem.Hungary matching algorithm can be effectively overcome during tracking, the addition of candidate target, lost, can be very
The effectively locking tracking target from candidate target.
Originally the execution that exemplary tracking system can be smooth in real time, in conventional methods where when a tracking system has merged crowd
After more algorithms, these algorithms order perform when, can mutually block, so as to when make system execution efficiency and fluency all under
Drop, the mode that this example employs asynchronous multithreading are integrated with and track in short-term, and target re-detection and multi-object clustering algorithm make
Three algorithms stated can be individually performed in different threads without blocking other side, so that system smooth can perform.
Example 3:
As shown in fig. 6, this example provides a kind of method for tracking target, including：
The information of given tracking target A, for example, the position of target A and size, size here can be regarded as foregoing ruler
Spend information；
Update the information of current goal A；
Features of the extraction record target A on present image, for example, CNN features；
Input new images；
It tracks in short-term, then returnes to the information of update current goal A.
At the same time, every n frame asynchronous executions, object detection (obtains the information of multiple objects, for example, the letter of A, B or C
Breath)；Here object A, B, C is foregoing alternative objects；
The feature of A, B, C on the image is extracted,
Feature based, matching target A (Hungary matching algorithm)；
Correct or recall the information (location information, size information) of target A, these information that this algorithm obtains, compared with
DSST's is more accurate.
In several embodiments provided herein, it should be understood that disclosed device and method can pass through it
Its mode is realized.Apparatus embodiments described above are only schematical, for example, the division of the unit, is only
A kind of division of logic function can have other dividing mode, such as in actual implementation：Multiple units or component can combine or
It is desirably integrated into another system or some features can be ignored or does not perform.In addition, shown or discussed each composition portion
Point mutual coupling or direct-coupling or communication connection can be the INDIRECT COUPLINGs by some interfaces, equipment or unit
Or communication connection, can be electrical, mechanical or other forms.
The above-mentioned unit illustrated as separating component can be or may not be physically separate, be shown as unit
The component shown can be or may not be physical location, you can be located at a place, can also be distributed to multiple network lists
In member；Part or all of unit therein can be selected to realize the purpose of this embodiment scheme according to the actual needs.
In addition, each functional unit in various embodiments of the present invention can be fully integrated into a processing module, also may be used
To be each unit individually as a unit, can also two or more units integrate in a unit；It is above-mentioned
The form that hardware had both may be employed in integrated unit is realized, can also be realized in the form of hardware adds SFU software functional unit.
One of ordinary skill in the art will appreciate that：Realizing all or part of step of above method embodiment can pass through
The relevant hardware of program instruction is completed, and foregoing program can be stored in a computer read/write memory medium, the program
Upon execution, the step of execution includes above method embodiment.
The above description is merely a specific embodiment, but protection scope of the present invention is not limited thereto, any
Those familiar with the art in the technical scope disclosed by the present invention, can readily occur in change or replacement, should all contain
Lid is within protection scope of the present invention.Therefore, protection scope of the present invention should be based on the protection scope of the described claims.
Claims (11)
1. a kind of method for tracking target, which is characterized in that including：
Obtain the image currently gathered；
Using the first track algorithm, information processing is carried out to the first tracking information of the first image information and previous moment, with
First tracking information at current time is obtained, wherein, described first image information is the image letter of the described image currently gathered
Breath；
Using the second track algorithm, the of the tracking target that is obtained at predetermined intervals to the second image information and historical juncture
One characteristics of image carries out information processing, and determines second tracking information at current time, wherein, second image information is to work as
The image information of the described image of preceding acquisition；At the time of the historical juncture is before the current time；
According to second tracking information at the current time, first tracking information at the current time is corrected.
2. according to the method described in claim 1, it is characterized in that,
It is described to utilize the first track algorithm, to the first tracking information of the first image information and previous moment at row information
Reason, to obtain first tracking information at current time, including：
According to first tracking information of previous moment, target area is determined；；
Extract the second characteristics of image of the target area in described image；
The location information and dimension information at the tracking target current time are determined according to second characteristics of image.
3. according to the method described in claim 1, it is characterized in that,
It is described to utilize the second track algorithm, the of the tracking target obtained at predetermined intervals to the second image information and historical juncture
One characteristics of image carries out information processing, and determines second tracking information at current time, including：
Target detection is carried out to described image, to obtain the 3rd characteristics of image of candidate target；
3rd characteristics of image and described first image feature are clustered；
According to cluster as a result, selecting the tracking target from the candidate target；
Determine the status information and dimension information of the tracking target.
4. according to the method described in claim 1, it is characterized in that,
The method further includes：
If the tracking target based on first track algorithm is lost, using second track algorithm to the second of image
Image information and described first image feature carry out information processing, obtain the second tracking letter at the tracking target current time
Breath；
Using second tracking information as the first tracking information that the current time is obtained using first track algorithm.
5. method according to any one of claims 1 to 4, which is characterized in that
The method further includes：
Tracking start time selectes the tracking target.
6. according to the method described in claim 5, it is characterized in that,
The tracking start time selectes the tracking target, including at least one of：
It is indicated to select the tracking target according to user；
The tracking target is automatically selected according to alternative condition.
7. according to the method described in claim 6, it is characterized in that,
It is described that the tracking target is automatically selected according to alternative condition, including：
It is the tracking target to select the candidate target in the focusing area of camera；
According to the relative position relation between candidate target and camera, selection is located at the candidate target immediately ahead of the camera
For the tracking target；
In the image gathered before tracking start time, the candidate target of imaging area maximum is selected as the tracking target.
8. according to the method described in claim 1, it is characterized in that,
The tracking target that is described to utilize the second track algorithm, being obtained at predetermined intervals to the second image information and historical juncture
The first characteristics of image carry out information processing, and determine current time the second tracking information, including：
Using second track algorithm, the tracking obtained at predetermined intervals to second image information and historical juncture
First characteristics of image of target carries out asynchronous process, and determines second tracking information at current time.
9. a kind of target tracker, which is characterized in that including：
Acquiring unit, for obtaining the image currently gathered；
First tracking cell for utilizing the first track algorithm, is believed first tracking at the first image information and previous moment
Breath carries out information processing, to obtain first tracking information at current time, wherein, described first image information currently gathers
The image information of described image；
Second tracking cell for utilizing the second track algorithm, at predetermined intervals obtains the second image information and historical juncture
The first characteristics of image of tracking target carry out information processing, and determine second tracking information at current time, wherein, described the
Two image informations are the image information of the described image currently gathered；The historical juncture for the current time it is pervious when
It carves；
Unit is corrected, for the second tracking information according to the current time, the first tracking for correcting the current time is believed
Breath.
10. device according to claim 9, which is characterized in that second tracking cell, if being additionally operable to based on described the
The tracking target of one track algorithm is lost, using second track algorithm to the second image information of image and described the
One characteristics of image carries out information processing, obtains second tracking information at the tracking target current time；
First tracking cell is additionally operable to using second tracking information as using described in first track algorithm acquisition
First tracking information at current time.
11. a kind of computer storage media, the computer storage media is stored with computer program；The computer program quilt
After execution, the method for tracking target that any one of claim 1 to 8 provides can be realized.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711485059.4A CN108062763B (en) | 2017-12-29 | 2017-12-29 | Target tracking method and device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711485059.4A CN108062763B (en) | 2017-12-29 | 2017-12-29 | Target tracking method and device and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108062763A true CN108062763A (en) | 2018-05-22 |
| CN108062763B CN108062763B (en) | 2020-10-16 |
Family
ID=62140961
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711485059.4A Active CN108062763B (en) | 2017-12-29 | 2017-12-29 | Target tracking method and device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108062763B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110221924A (en) * | 2019-04-29 | 2019-09-10 | 北京云迹科技有限公司 | The method and device of data processing |
| CN111091593A (en) * | 2018-10-24 | 2020-05-01 | 深圳云天励飞技术有限公司 | Image processing method, image processing device, electronic equipment and storage medium |
| WO2020134935A1 (en) * | 2018-12-26 | 2020-07-02 | 中兴通讯股份有限公司 | Video image correction method, apparatus and device, and readable storage medium |
| CN111563913A (en) * | 2020-04-15 | 2020-08-21 | 上海摩象网络科技有限公司 | Searching method and device based on tracking target and handheld camera thereof |
| CN111935450A (en) * | 2020-07-15 | 2020-11-13 | 长江大学 | Intelligent suspect tracking method and system and computer readable storage medium |
| WO2020258187A1 (en) * | 2019-06-27 | 2020-12-30 | 深圳市大疆创新科技有限公司 | State detection method and apparatus and mobile platform |
| CN112330721A (en) * | 2020-11-11 | 2021-02-05 | 北京市商汤科技开发有限公司 | Three-dimensional coordinate recovery method and device, electronic equipment and storage medium |
| CN113409373A (en) * | 2021-06-25 | 2021-09-17 | 浙江商汤科技开发有限公司 | Image processing method, related terminal, device and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101141633A (en) * | 2007-08-28 | 2008-03-12 | 湖南大学 | Moving object detecting and tracing method in complex scene |
| CN101393609A (en) * | 2008-09-18 | 2009-03-25 | 北京中星微电子有限公司 | Target detection tracking method and device |
| CN101860729A (en) * | 2010-04-16 | 2010-10-13 | 天津理工大学 | Target tracking method for omnidirectional vision |
| CN106683123A (en) * | 2016-10-31 | 2017-05-17 | 纳恩博（北京）科技有限公司 | Method and device for tracking targets |
| CN107481270A (en) * | 2017-08-10 | 2017-12-15 | 上海体育学院 | Table tennis target following and trajectory predictions method, apparatus, storage medium and computer equipment |
-
2017
- 2017-12-29 CN CN201711485059.4A patent/CN108062763B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101141633A (en) * | 2007-08-28 | 2008-03-12 | 湖南大学 | Moving object detecting and tracing method in complex scene |
| CN101393609A (en) * | 2008-09-18 | 2009-03-25 | 北京中星微电子有限公司 | Target detection tracking method and device |
| CN101860729A (en) * | 2010-04-16 | 2010-10-13 | 天津理工大学 | Target tracking method for omnidirectional vision |
| CN106683123A (en) * | 2016-10-31 | 2017-05-17 | 纳恩博（北京）科技有限公司 | Method and device for tracking targets |
| CN107481270A (en) * | 2017-08-10 | 2017-12-15 | 上海体育学院 | Table tennis target following and trajectory predictions method, apparatus, storage medium and computer equipment |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111091593A (en) * | 2018-10-24 | 2020-05-01 | 深圳云天励飞技术有限公司 | Image processing method, image processing device, electronic equipment and storage medium |
| CN111091593B (en) * | 2018-10-24 | 2024-03-22 | 深圳云天励飞技术有限公司 | Image processing method, device, electronic equipment and storage medium |
| WO2020134935A1 (en) * | 2018-12-26 | 2020-07-02 | 中兴通讯股份有限公司 | Video image correction method, apparatus and device, and readable storage medium |
| CN111369586A (en) * | 2018-12-26 | 2020-07-03 | 中兴通讯股份有限公司 | Video image correction method, device, equipment and readable storage medium |
| CN110221924A (en) * | 2019-04-29 | 2019-09-10 | 北京云迹科技有限公司 | The method and device of data processing |
| WO2020258187A1 (en) * | 2019-06-27 | 2020-12-30 | 深圳市大疆创新科技有限公司 | State detection method and apparatus and mobile platform |
| CN111563913A (en) * | 2020-04-15 | 2020-08-21 | 上海摩象网络科技有限公司 | Searching method and device based on tracking target and handheld camera thereof |
| CN111563913B (en) * | 2020-04-15 | 2021-12-10 | 上海摩象网络科技有限公司 | Searching method and device based on tracking target and handheld camera thereof |
| CN111935450A (en) * | 2020-07-15 | 2020-11-13 | 长江大学 | Intelligent suspect tracking method and system and computer readable storage medium |
| CN112330721A (en) * | 2020-11-11 | 2021-02-05 | 北京市商汤科技开发有限公司 | Three-dimensional coordinate recovery method and device, electronic equipment and storage medium |
| CN112330721B (en) * | 2020-11-11 | 2023-02-17 | 北京市商汤科技开发有限公司 | Three-dimensional coordinate recovery method and device, electronic equipment and storage medium |
| CN113409373A (en) * | 2021-06-25 | 2021-09-17 | 浙江商汤科技开发有限公司 | Image processing method, related terminal, device and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108062763B (en) | 2020-10-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108062763A (en) | Method for tracking target and device, storage medium | |
| US11575824B2 (en) | Photographing control method, apparatus, and control device | |
| EP1505543A2 (en) | Video object tracking | |
| CN110490900A (en) | Binocular visual positioning method and system under dynamic environment | |
| CN108592919B (en) | Drawing and localization method, device, storage medium and terminal device | |
| CN110544268B (en) | Multi-target tracking method based on structured light and SiamMask network | |
| CN113689503B (en) | Target object posture detection method, device, equipment and storage medium | |
| CN111354022B (en) | Target Tracking Method and System Based on Kernel Correlation Filtering | |
| CN112950696A (en) | Navigation map generation method and generation device and electronic equipment | |
| JP2018195166A (en) | Posture determination program, posture determination device and posture determination method | |
| CN112308879A (en) | Image processing apparatus, method of tracking target object, and storage medium | |
| CN110084885A (en) | A kind of cloud and image optimization method, device, equipment and storage medium | |
| CN109859314A (en) | Three-dimensional rebuilding method, device, electronic equipment and storage medium | |
| CN116309704A (en) | Small target tracking method based on anchor-free frame detection network and feature re-fusion module | |
| CN103578102B (en) | Object detection device and subject detection method | |
| CN110335308B (en) | Binocular vision odometer calculation method based on parallax constraint and bidirectional annular inspection | |
| CN106683113A (en) | Characteristic point tracking method and device | |
| CN116045991A (en) | Method and device for planning collision-free path of outer contour of vehicle | |
| CN115014366A (en) | Target fusion method and device, vehicle and storage medium | |
| CN115115675A (en) | Method for detecting carry-over, electronic device, computer readable medium | |
| CN113971687A (en) | Object tracking method, device electronic equipment and storage medium | |
| CN111738085B (en) | System construction method and device for realizing automatic driving simultaneous positioning and mapping | |
| CN114387324A (en) | Depth imaging method, depth imaging device, electronic equipment and computer readable storage medium | |
| CN113469130A (en) | Shielded target detection method and device, storage medium and electronic device | |
| CN113808077A (en) | Target detection method, device, equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |