CN109446942A - Method for tracking target, device and system - Google Patents
Method for tracking target, device and system Download PDFInfo
- Publication number
- CN109446942A CN109446942A CN201811195130.XA CN201811195130A CN109446942A CN 109446942 A CN109446942 A CN 109446942A CN 201811195130 A CN201811195130 A CN 201811195130A CN 109446942 A CN109446942 A CN 109446942A
- Authority
- CN
- China
- Prior art keywords
- target
- tracking
- specified
- frame image
- blocked
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
- G06V20/53—Recognition of crowd images, e.g. recognition of crowd congestion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/48—Matching video sequences
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides a kind of method for tracking target, device and system；Wherein, this method comprises: detecting the target object in current frame image by detection model, and judge whether target object is specified target；If so, the tracking target that matches with specified target in designated frame image before obtaining current frame image；If the tracking target to match be it is multiple, according to it is multiple tracking targets detection datas relative position, from multiple tracking targets determine specify target；According in multiple tracking targets, position of the tracking target in designated frame image in addition to specified target, estimate that the tracking target in addition to specified target in the position of current frame image, determines the target that is blocked according to estimated result from the tracking target in addition to specified target.The tracking of the target object not being blocked and the target that is blocked may be implemented in the present invention, the probability of happening that tracking is lost is reduced, to improve the accuracy rate of target following.
Description
Technical field
The present invention relates to technical field of image processing, more particularly, to a kind of method for tracking target, device and system.
Background technique
In monitoring field, carrying out tracking to pedestrian has important value.Limited by shooting angle, when pedestrian is more or
When being closer between pedestrian, it is easy to happen on video and mutually blocks.It, can be based on tradition filter in relevant pedestrian tracting method
Wave device, as KFPPK (Kalman filter and the probability product kernel, Kalman filter with
Probability product core) algorithm, slight occlusion issue is handled by algorithm estimation；Pass through the Shandong based on Kalman filtering there are also a kind of
Stick multi-target detection track algorithm, the algorithm carry out shadowing using Pixel-level gap and length-width ratio；But if hiding mutually
Figure between the pedestrian of gear has differences or the positional relationship of pedestrian is not just or when having other extraneous non-pedestrian factors interference, on
That states these modes blocks that treatment effect is poor, and then causes the accuracy rate of pedestrian tracking lower.
Summary of the invention
In view of this, not hidden the purpose of the present invention is to provide a kind of method for tracking target, device and system with realizing
The tracking of the target object of gear and the target that is blocked reduces the probability of happening that tracking is lost, to improve the accurate of target following
Rate.
In a first aspect, the embodiment of the invention provides a kind of method for tracking target, this method comprises: obtaining present frame figure
Picture；The target object in current frame image is detected by preset detection model, and judges whether target object is specified target；
Specified target is the target not being blocked, and specifies the target occlusion target that is blocked；If so, before obtaining current frame image
Designated frame image in the tracking target that matches with specified target；If the tracking target to match be it is multiple, according to multiple
Track target detection data relative position, from multiple tracking targets determine specify target, with to specified target carry out with
Track；According in multiple tracking targets, position of the tracking target in designated frame image in addition to specified target, estimation is except specified
Tracking target other than target is in the position of current frame image, according to estimated result from the tracking target in addition to specified target
The target that is blocked is determined, to track to the target that is blocked.
In preferred embodiments of the present invention, above by the target pair in preset detection model detection current frame image
As, and the step of whether target object is specified target judged, comprising: by first network model inspection current frame image
Target object；Whether the target that is blocked is blocked by the second network model detected target object, if so, target object is true
It is set to specified target.
In preferred embodiments of the present invention, in the designated frame image before above-mentioned acquisition current frame image with specified target
The step of tracking target to match, comprising: the tracking target in designated frame image before acquisition current frame image；If with
Track target be it is multiple, one by one by it is each tracking target detection data matched with the detection data of specified target；It will matching
As a result it is determined as the tracking target that specified target matches more than or equal to the tracking target of preset matching threshold.
It is above-mentioned one by one by the inspection of the detection data and specified target of each tracking target in preferred embodiments of the present invention
The step of measured data is matched, comprising: calculate the detection data of each tracking target and the detection data of specified target one by one
Between friendship and ratio；It, will be current if currently tracking the corresponding friendship of target and than being greater than or equal to preset first fractional threshold
The detection data of tracking target is matched with the detection data of specified target.
In preferred embodiments of the present invention, above-mentioned detection data is identified by detection block；According to multiple tracking targets
The relative position of detection data determines the step of specifying target, comprising: know from multiple tracking targets from multiple tracking targets
The detection block bottom edge position of other detection data is located at the tracking target of bottommost；The tracking target that will identify that is determined as specified mesh
Mark.
In preferred embodiments of the present invention, above-mentioned designated frame image includes multiple image；According in multiple tracking targets,
Position of the tracking target in designated frame image in addition to specified target estimates that the tracking target in addition to specified target is being worked as
The step of position of prior image frame, comprising: according in multiple tracking targets, the tracking target in addition to specified target is in multiframe figure
Position as in carries out estimation to the tracking target in addition to specified target by Kalman Algorithm, obtains except specified mesh
Outer tracking target is marked in the estimated location of current frame image.
It is above-mentioned true from the tracking target in addition to specified target according to estimated result in preferred embodiments of the present invention
Surely the step of target that is blocked, comprising: judge the friendship of estimated location and the detection data of specified target and than whether be greater than or wait
In preset second fractional threshold；If so, determining that the corresponding tracking target of estimated location is the target that is blocked, by estimated location
It is determined as position of the target in current frame image that be blocked；If not, according to the position of the detection data of specified target, adjustment
Estimated location adjusted is determined as position of the target in current frame image that be blocked by estimated location.
Second aspect, the embodiment of the invention provides a kind of target tracker, which includes: image collection module,
For obtaining current frame image；Detection module, for detecting the target object in current frame image by preset detection model,
And judge whether target object is specified target；Specified target is the target not being blocked, and target occlusion is specified to be blocked
Target；First tracking module, sets the goal if referred to for target object, in the designated frame image before acquisition current frame image
The tracking target to match with specified target；If the tracking target to match be it is multiple, according to it is multiple tracking targets detections
The relative position of data determines specified target, from multiple tracking targets to track to specified target；Second tracking mould
Block, for according in multiple tracking targets, position of the tracking target in designated frame image in addition to specified target, estimation is removed
Tracking target other than specified target is in the position of current frame image, according to estimated result from the tracking mesh in addition to specified target
The target that is blocked is determined in mark, to track to the target that is blocked.
The third aspect, the embodiment of the invention provides a kind of Target Tracking System, the system include: image capture device,
Processing equipment and storage device；Image capture device, for obtaining preview video frame or image data；It is stored on storage device
Computer program, computer program execute such as above-mentioned method for tracking target when equipment processed is run.
Fourth aspect, the embodiment of the invention provides a kind of computer readable storage medium, computer readable storage mediums
On be stored with computer program, the step of when computer program equipment operation processed executes above-mentioned method for tracking target.
The embodiment of the present invention bring it is following the utility model has the advantages that
Above-mentioned method for tracking target provided in an embodiment of the present invention, device and system pass through preset detection model first
The target object in current frame image is detected, and judges whether target object is specified target；The specified target is not to be blocked
Target, and the specified target occlusion target that is blocked；It sets the goal if target object refers to；Then obtain current frame image it
The tracking target to match in preceding designated frame image with specified target；According to the opposite position of the detection data of multiple tracking targets
It sets, determines specified target, from multiple tracking targets to track to specified target；Further according in multiple tracking targets, remove
Position of the tracking target in designated frame image other than specified target, estimates the tracking target in addition to specified target current
The position of frame image determines the target that is blocked according to estimated result, to the quilt from the tracking target in addition to specified target
Shelter target is tracked.Aforesaid way can be realized simultaneously the tracking of the target object not being blocked and the target that is blocked, only
It detects that target object has blocked the target that is blocked, the mesh that is blocked can be realized by way of estimation between multiple image
Target tracking reduces the probability of happening that tracking is lost, to improve the accuracy rate of target following.
Especially in crowded video frame, blocking between target is more serious, through the above way can be very
The case where same tracking target is identified as multiple pedestrians is reduced well, improves the target following effect in crowded situation
Fruit.
Other features and advantages of the present invention will illustrate in the following description, alternatively, Partial Feature and advantage can be with
Deduce from specification or unambiguously determine, or by implementing above-mentioned technology of the invention it can be learnt that.
To enable the above objects, features and advantages of the present invention to be clearer and more comprehensible, better embodiment is cited below particularly, and match
Appended attached drawing is closed, is described in detail below.
Detailed description of the invention
It, below will be to specific in order to illustrate more clearly of the specific embodiment of the invention or technical solution in the prior art
Embodiment or attached drawing needed to be used in the description of the prior art be briefly described, it should be apparent that, it is described below
Attached drawing is some embodiments of the present invention, for those of ordinary skill in the art, before not making the creative labor
It puts, is also possible to obtain other drawings based on these drawings.
Fig. 1 is a kind of structural schematic diagram of electronic system provided in an embodiment of the present invention；
Fig. 2 is a kind of flow chart of method for tracking target provided in an embodiment of the present invention；
Fig. 3 is the flow chart of another method for tracking target provided in an embodiment of the present invention；
Fig. 4 is a kind of structural schematic diagram of target tracker provided in an embodiment of the present invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with attached drawing to the present invention
Technical solution be clearly and completely described, it is clear that described embodiments are some of the embodiments of the present invention, rather than
Whole embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art are not making creative work premise
Under every other embodiment obtained, shall fall within the protection scope of the present invention.
In view of it is existing block processing mode block that treatment effect is poor, and then cause the accuracy rate of pedestrian tracking compared with
Low problem, the embodiment of the invention provides a kind of method for tracking target, device and system；The technology can be applied to video prison
In the scenes such as control, pedestrian tracking；The technology can be used corresponding software and hardware and realize, carry out below to the embodiment of the present invention detailed
It is thin to introduce.
Embodiment one:
Firstly, describing the example of method for tracking target for realizing the embodiment of the present invention, device and system referring to Fig.1
Electronic system 100.
A kind of structural schematic diagram of electronic system as shown in Figure 1, electronic system 100 include one or more processing equipments
102, one or more storage devices 104, input unit 106, output device 108 and one or more image capture devices
110, these components pass through the interconnection of bindiny mechanism's (not shown) of bus system 112 and/or other forms.It should be noted that Fig. 1 institute
The component and structure for the electronic system 100 shown be it is illustrative, and not restrictive, as needed, the electronic system
It can have other assemblies and structure.
The processing equipment 102 can be gateway, or intelligent terminal, or include central processing unit
It (CPU) or the equipment of the processing unit of the other forms with data-handling capacity and/or instruction execution capability, can be to institute
The data for stating other components in electronic system 100 are handled, and other components in the electronic system 100 can also be controlled
To execute desired function.
The storage device 104 may include one or more computer program products, and the computer program product can
To include various forms of computer readable storage mediums, such as volatile memory and/or nonvolatile memory.It is described easy
The property lost memory for example may include random access memory (RAM) and/or cache memory (cache) etc..It is described non-
Volatile memory for example may include read-only memory (ROM), hard disk, flash memory etc..In the computer readable storage medium
On can store one or more computer program instructions, processing equipment 102 can run described program instruction, to realize hereafter
The client functionality (realized by processing equipment) in the embodiment of the present invention and/or other desired functions.Institute
Various application programs and various data can also be stored by stating in computer readable storage medium, such as the application program uses
And/or various data generated etc..
The input unit 106 can be the device that user is used to input instruction, and may include keyboard, mouse, wheat
One or more of gram wind and touch screen etc..
The output device 108 can export various information (for example, image or sound) to external (for example, user), and
It and may include one or more of display, loudspeaker etc..
Described image acquisition equipment 110 can acquire preview video frame or image data, and collected preview is regarded
Frequency frame or image data are stored in the storage device 104 for the use of other components.
Illustratively, for realizing method for tracking target according to an embodiment of the present invention, the exemplary electron of device and system
Each device in system can integrate setting, can also be with scattering device, such as by processing equipment 102, storage device 104, input
Device 106 and output device 108 are integrally disposed in one, and image capture device 110, which is set to, can collect frame image
Designated position.When each device in above-mentioned electronic system is integrally disposed, the electronic system may be implemented as such as camera,
The intelligent terminals such as smart phone, tablet computer, computer.
Embodiment two:
A kind of method for tracking target is present embodiments provided, this method is executed by the processing equipment in above-mentioned electronic system；
The processing equipment can be any equipment or chip with data-handling capacity.The processing equipment can be independently to receiving
Information is handled, and can also be connected with server, is analyzed and processed jointly to information, and processing result is uploaded to cloud
End.
As shown in Fig. 2, the method for tracking target includes the following steps:
Step S202 obtains current frame image；
It, can sequentially in time, from the video data when carrying out target following to it for one section of video data
First frame image starts, and target object therein is detected and identified；Same target pair between multiple image, will be identified as
The object of elephant is identified by the same serial number or symbol, to track to the target object.During actual tracking,
The detection and identification that can carry out target object to every frame image in above-mentioned video data one by one, at this time can be according to video counts
Current frame image is obtained according to middle putting in order for frame image；In another way, sample frequency can be preset, according to video
In data frame image put in order and preset sample frequency, obtain current frame image.
Step S204 detects the target object in current frame image by preset detection model, and judges the target pair
As if it is no for specified target；The specified target is the target not being blocked, and specifies the target occlusion target that is blocked；
Above-mentioned preset detection model can be obtained by neural metwork training；Current frame image is input in detection model
Afterwards, it can specifically be detected by way of Face datection or humanoid detection in current frame image with the presence or absence of target object；If
There are target objects, and the detection data of the target object can be identified by detection block；The detection block is generally rectangular, the detection
The face area of target object or the body region by head to foot are generally comprised in frame.
Above-mentioned detection model can be trained in advance only detect be not blocked or coverage extent be less than preset threshold mesh
Mark object；Specifically, it is only marked in training sample using detection block (such as hough transform frame) in advance and is not blocked or blocks
The lesser pedestrian of degree, then the training sample is input in neural network and is trained, above-mentioned detection model can be obtained.
It should be noted that in the prior art, the pedestrian detection model realized based on neural network or deep neural network
Usually ignore the pedestrian that is blocked, therefore, if detection model in the present embodiment using pedestrian detection model in the prior art,
It is not trained by above-mentioned training sample, the target object detected is also not to be blocked or coverage extent is smaller mostly
Pedestrian, and after being trained by above-mentioned training sample, can make the target object detected is the journey that is not blocked or blocks
The accounting for spending lesser pedestrian is higher, and testing result is more accurate.
Seen from the above description, the target object that detection model detects may be considered the target not being blocked, but not
It must be the specified object for having blocked the target that is blocked；In order to obtain specified object from target object；It also needs to these mesh
It marks object to judge whether to have blocked other objects one by one, which is the above-mentioned target that is blocked.Specific deterministic process
It can also be realized by above-mentioned detection model, or be realized by a discrete discrimination model that blocks；When passing through detection model
When realization, the above-mentioned target object detected can be through the realization of the first sub-network of the detection model, exports and detects
Target object is as intermediate result；After obtaining target object, passing through the second sub-network in the detection model to these targets
Object is detected, and specified target is filtered out from these target objects.When by it is discrete block discrimination model and realize when, will
The target object that detection model detects, which is input to, to be blocked discrimination model and is screened, and output obtains specified target.
Above-mentioned second sub-network or it is above-mentioned block discrimination model and can be trained in advance for identification target object whether
Block the model of other objects；Specifically, it is marked out in training sample using detection block in advance and carrys out pedestrian itself and do not hidden
Gear, but the pedestrian of other pedestrians has been blocked, then the training sample is input in neural network and is trained, it can be obtained above-mentioned
Second sub-network blocks discrimination model.
Step S206, if so, obtain current frame image before designated frame image in specified target match with
Track target；If the tracking target to match be it is multiple, according to the relative position of the detection datas of multiple tracking targets, from multiple
It tracks and determines specified target in target, to be tracked to specified target；
Seen from the above description, target tracking is sequentially in time since the first frame image of video data, to it
In target object detected and identified；When getting current frame image, the frame image before the current frame image is usual
Have been completed the process of target tracking；That is, passing through in the video frame images for a frame image before current frame image
Detection block designates each tracking target, the corresponding tracking mark of each tracking target；Multiple frames before current frame image
Between image, same tracking target corresponds to the same tracking mark.In order to by the target object identified in current frame image with
The tracking target in frame image before current frame image is corresponding, it usually needs before the target object and current frame image
Frame image in tracking target matched, the tracking target of successful match is associated with the target object, with complete
The tracking of tracking target in the current frame.
Designated frame image before above-mentioned current frame image can be the former frame figure of current frame image in video data
Picture, or after being sampled according to setpoint frequency, the previous frame image of current frame image, or before current frame image
Other specified frame images.Certainly, which may also include multiple image；But this refers to during most tracking
Framing image is single-frame images.For above-mentioned specified target, before being usually also required to obtain matched current frame image
Designated frame image in tracking target, due to the above-mentioned specified target occlusion target that is blocked, the testing number of the specified target
At least part detection data for the target that is blocked may be included according to middle；Thus specified target may in designated frame image with
The multiple tracking targets of track target match, for example, this is specified if the target that is blocked of the specified target occlusion is one
Target may match with two tracking targets in designated frame image.
In multiple tracking targets of the specified object matching, usually an only tracking target and the specified target are same
A target；Seen from the above description, which is the target not being blocked, and camera is clapped at a particular angle mostly
Take the photograph video；Therefore, which tracking target and the specified target can be determined according to the relative position of matched multiple tracking targets
It is the same target；For example, specifying target relative to the target that is blocked in frame figure when camera shoots video with depression angle
Position as in more on the lower, is based on this, and specified target can be determined from multiple tracking targets；And if camera is with other
In the case where angle shot or sideways inclined, specify target may relative to the position in frame image for the target that is blocked
It can change, the relative position of specified target and the target that is blocked is determined i.e. according to shooting angle or sideways inclined angle at this time
Can, and then specified target is determined from multiple tracking targets.So far, specified target and the present frame in current frame image are realized
The association that target is tracked in designated frame image before image, completes the tracking target in the tracking of current frame image.
It should be noted that the specified target may also only match with one of tracking target, should it may be the case that
Since this specifies target is more serious to the coverage extent for the target that is blocked to cause, but due to having been detected by specified target proximity
In the presence of the target that is blocked, can be blocked position of the target in designated frame image according to the location estimation of specified target, if
There is tracking target in estimated location, it can be by the tracking target and the target association that is blocked；If be not present in estimated location
Target is tracked, illustrates that the target that is blocked is emerging target, the tracking mesh can be marked using a new mark at this time
Mark, and tracked in subsequent video frame images.
Step S208, according in multiple tracking targets, the tracking target in addition to specified target is in designated frame image
The tracking target in addition to specified target is estimated in the position of current frame image, according to estimated result from except specified target in position
The target that is blocked is determined in tracking target in addition, to track to the target that is blocked.
Object tracking process in the present embodiment not only needs to track specified target, it is also necessary to specify mesh to this
The target that is blocked blocked is marked to be tracked；Since target following is continuous process between video frame images, in the present embodiment
Designated frame image before default current frame image, which has been completed, hides the specified target not being blocked and the specified target
The target that is blocked of gear is tracked.
Above-mentioned multiple tracking targets are with specified object matching, and there may be this in the tracking target in addition to specified target
The target that is blocked of specified target occlusion；At this point it is possible to estimation is carried out to the tracking target in addition to specified target, with
To estimated location of the tracking result in current frame image；In actual implementation, the process of above-mentioned estimation may need
The position of same tracking target in multiple frame images before current frame image, above-mentioned designated frame image contains multiframe figure at this time
Picture.
If the estimated location of the tracking target is located at specified target proximity, and the detection of the estimated location and specified target
Data are overlapped that degree is higher, illustrate that the tracking target may be the target that is blocked in current frame image, at this time can by this with
Track target and the target association that is blocked, to complete the tracking target in the tracking of current frame image.
If the estimated location of the above-mentioned tracking target in addition to specified target not in specified target proximity, or with finger
The detection data to set the goal is not overlapped；And the specified target proximity in current frame image detects the presence of the target that is blocked, this
When the adjustable corresponding estimated location of tracking target in addition to specified target, by estimated location adjusted and the mesh that is blocked
Mark association, to complete the tracking target in the tracking of current frame image.
Above-mentioned method for tracking target provided in an embodiment of the present invention detects present frame figure by preset detection model first
Target object as in, and judge whether target object is specified target；The specified target is the target not being blocked, and this refers to
It sets the goal and has blocked the target that is blocked；It sets the goal if target object refers to；Then obtain the designated frame figure before current frame image
The tracking target to match as in specified target；According to it is multiple tracking targets detection datas relative position, from it is multiple with
Specified target is determined in track target, to track to specified target；Further according in multiple tracking targets, in addition to specified target
Position of the tracking target in designated frame image, estimate the tracking target in addition to specified target in the position of current frame image
Set, determined and be blocked target from the tracking target in addition to specified target according to estimated result, with to this be blocked target into
Line trace.Aforesaid way can be realized simultaneously the tracking of the target object not being blocked and the target that is blocked, as long as detecting mesh
Mark object has blocked the target that is blocked, and the tracking for the target that is blocked can be realized by way of estimation between multiple image,
The probability of happening that tracking is lost is reduced, to improve the accuracy rate of target following.
Especially in crowded video frame, blocking between target is more serious, through the above way can be very
The case where same tracking target is identified as multiple pedestrians is reduced well, improves the target following effect in crowded situation
Fruit.
Embodiment three:
Another method for tracking target is present embodiments provided, this method is realized on the basis of the above embodiments；The party
In method, the method for tracking target is further described, the specified target especially blocked and target occlusion is specified by this
The tracking mode for the target that is blocked；As shown in figure 3, this method comprises the following steps:
Step S302 obtains current frame image；
Step S304 passes through the target object in first network model inspection current frame image；
The first network model can by way of Face datection or humanoid detection detected target object；In general, face
It detects obtained detection data and generally includes face or head, the humanoid obtained detection data that detects generally includes entire human body.
By above-described embodiment description it is found that subsequent need based on each relative positional relationship realization target following for tracking target, and people
Height gap is larger between people, thus is not easy to realize the target following in the present embodiment, therefore this implementation by Face datection
Detected target object, the target object detected are identified by detection block by the way of humanoid detection in example, such as rectangle frame.
The first network model can be obtained by training；Specifically, it is only marked in training sample using detection block in advance
Note is not blocked or the lesser pedestrian of coverage extent, then the training sample is input in neural network model and is trained, i.e.,
Above-mentioned first network model can be obtained；Wherein, which is specifically as follows Faster-RCNN model, RetinaNet
Model etc..
Whether step S306 has blocked the target that is blocked by the second network model detected target object, if so, by mesh
Mark object is determined as specified target.
Second network model can be trained to the model whether target object for identification blocks other objects in advance；Tool
Body is in advance marked out using detection block in training sample and to carry out pedestrian itself and be not blocked but blocked other pedestrians'
Pedestrian, then the training sample is input in neural network model and is trained, above-mentioned second network model can be obtained, this
Two network models are referred to as blocking discrimination model in above-described embodiment.The neural network model is specifically as follows small-sized point
Class network, such as resnet10, resnet18.
Above-mentioned first network model and the second network model can be used light-weighted neural metwork training and realize.Due to
Neural network usually only receives the image data of fixed size, therefore, the instruction of above-mentioned first network model and the second network model
Practice in sample, after marking out pedestrian using detection block, will test Image Adjusting to the corresponding neural network that frame includes can be connect
The fixed size received, then the image that the detection block includes is input in neural network and is trained.Wherein, detection block includes
The specific adjustment mode of image can be stretched, delete redundance for image, portion's grading mode of plugging a gap.
The training sample of above-mentioned second network model can specifically be obtained by following manner: from existing pedestrian's data set
The detection block of middle selection and the disjoint pedestrian of other detection blocks are as negative sample；Choose the detection block work for having blocked other pedestrians
For positive sample.Since positive sample is smaller relative to the quantity of negative sample, then from choosing, pedestrian is very more, blocks more serious view
Positive sample is chosen in frequency, specifically, if two detection blocks exist be more than some hand over and than the friendship of threshold value (such as 0.3) and than when,
The detection block of frame bottom edge on the lower be will test as positive sample.
After obtaining training sample through the above way, pass through the process of these training samples the second network model of training
In, influence caused by positive sample and negative sample quantity gap in training sample is adjusted again using difficult negative sample digging technology.
Specifically, in the training process, the negative sample with positive sample with the order of magnitude can be first randomly selected, in entire negative sample after training
In set excavate can not correct decision negative sample as " difficult negative sample " continue training the second network model.
It, can first data set (such as imagenet number in generic object in the training process of above-mentioned second network model
According to collection) on carry out training for the first time, then carry out second training (i.e. on normal pedestrian's data set (such as caltech data set)
The process of fine tuning), then on training data (training sample of i.e. above-mentioned second network model) third is carried out containing blocking in construction
Secondary training (process finely tuned again), and then obtain the second final network model.
Step S308, obtain current frame image before designated frame image in tracking target；
As can be seen from the above embodiments, before the designated frame image before current frame image can be set to current frame image
Arbitrary frame image, but usually before current frame image, the frame image being closer with the current frame image.Due to designated frame
Image is located at before current frame image, and when carrying out target following to current frame image, designated frame image is had been completed
The process of target following is identified each tracking target in designated frame image by detection block, and assign tracking mark.Cause
This, obtains the tracking target in designated frame image, in specific available designated frame image the position of each detection block and
Corresponding tracking mark.
Step S310, if tracking target be it is multiple, one by one by the detection data and specified target of each tracking target
Detection data is matched；The tracking target that matching result is greater than or equal to preset matching threshold is determined as specified target phase
Matched tracking target.
In actual implementation, the number for tracking mark in designated frame image can be counted, and then knows the designated frame image
The number of middle tracking target.If track target be it is multiple, each tracking target be possible to specified target be same a line
People, therefore one by one match the detection data of each tracking target with the detection data of specified target.Due to detection data
It is the partial region image identified by detection block, therefore, tracking target can be realized by relevant image matching algorithm
Matching between detection data and the detection data of specified target；The image matching algorithm is it can be appreciated that image similarity ratio
To algorithm, primal dual algorithm, the delta algorithm of dynamic tree maintenance shortest path tree, histogram method, image mould are specifically included
Plate matching process, perceptual hash algorithm etc..
It tracks after being matched between the detection data and the detection data of specified target of target, the matching result of generation can be with table
Levy similarity degree between the two；For example, the matching result can be a percentage, matching result is higher, between the two
Similarity degree is bigger；When matching result is 100%, illustrate that the two is identical.Since tracking target is in movement shape mostly
State, therefore a matching threshold can be rule of thumb set, such as 70%, if the detection data of tracking target and specified target
Matching result between detection data is greater than or equal to 70%, illustrates to track target and specified target is same a group traveling together, the tracking
Target is the tracking target that specified target matches.
If tracking target is one, the tracking target and the specified target in current frame image are matched, if
It is greater than or equal to preset matching threshold with result, which is the tracking target that specified target matches；And if
Matching result be less than preset matching threshold, the tracking target can in current frame image, in addition to above-mentioned specified target
Other target objects (not blocking the target object of other targets) are matched.It is appreciated that the description of the present embodiment emphasis
The object tracking process of specified target (target not being blocked, and this specifies the target occlusion target that is blocked), for not having
There is the target object for blocking other targets, target following can be carried out using existing way.
In addition, if the number of tracking target is more, one by one by the detection data and specified target of each tracking target
Detection data, which carries out matching, may result in operation higher cost；It in order to solve this problem, can will be more before being matched
A tracking target is screened, specifically, in above-mentioned steps S310, one by one by the detection data of each tracking target and specified
The detection data of target carries out matched process, can be realized by following step 1- step 2:
Step 1, the friendship between the detection data of each tracking target and the detection data of specified target and ratio are calculated one by one；
Track the friendship between the detection data of target and the detection data of specified target and than can be understood as tracking target
Detection data and specified target detection data between overlapping rate, in general, hand over and ratio it is higher, illustrate track target detection
A possibility that detection data overlapping degree of data and specified target is higher, which with specified target is same a group traveling together is got over
Greatly.
Friendship between the detection data of above-mentioned tracking target and the detection data of specified target simultaneously compares IOU
(Intersection Over Union) can be calculated by following formula: (the detection data ∩ of tracking target is specified by IOU=
The detection data of target)/(the detection data ∪ of tracking target specifies the detection data of target).
It step 2, will be current if currently tracking the corresponding friendship of target and than being greater than or equal to preset first fractional threshold
The detection data of tracking target is matched with the detection data of specified target.
Above-mentioned friendship and ratio can be a percentage；If tracking the detection data of target and the detection data of specified target
Between friendship and smaller, even zero, illustrate to track positioning remote from for target and specified target；It is same as current frame image
It is adjacent or be closer between designated frame image, with a group traveling together in the moving distance of two frame images usually not too large, base
In this, if the friendship and smaller between the detection data of tracking target and the detection data of specified target, it is more likely that the two
It is not same a group traveling together, just no longer needs to carry out the matching between detection data yet.First fractional threshold specifically can be set, this
One fractional threshold can be 50%, when handing over and than being greater than or equal to first fractional threshold, by the detection of current tracking target
Data are matched with the detection data of specified target；If hand over and compare be less than first fractional threshold, no longer will currently with
The detection data of track target is matched with the detection data of specified target.
Therefore, by the above-mentioned means, can track target it is more in the case where, first according to hand over and than filtered out away from
It is matched from the closer tracking target of specified target, then by these tracking targets with specified target, so as to save image
Matched operation cost.
Step S312, if the tracking target that specified target matches is multiple, the recognition detection number from multiple tracking targets
According to detection block bottom edge position be located at the tracking target of bottommost；The tracking target that will identify that is determined as specified target.
For example, if tracking target A and tracking target B illustrate the testing number of the specified target with specified object matching
According to all or at least part of detection data comprising tracking target A, also comprising all or at least part of inspection of tracking target B
Measured data；Therefore, in the multiple tracking targets to match with specified target, one of tracking target and specified target belong to together
One target, other tracking targets are the target that is blocked of the specified target occlusion.
It is closer apart from camera since the cameras setting of most of acquisition video frames is in the higher position of opposite crowd
Pedestrian is not blocked usually by other pedestrians, and easily blocks other pedestrians；Also, it is located at video frame apart from the closer pedestrian of camera
In position more on the lower.Based on the principle, the position of multiple tracking targets can be compared, extreme lower position will be in
Tracking target is determined as specified target.In view of the height between pedestrian is different, in order to accurately obtain the position of each tracking target
It sets, the detection block bottom edge of data can be will test as benchmark, the detection block bottom edge position that will test data is located at bottommost
Target is tracked, specified target is determined as.
So far, the present embodiment realizes the specified target in current frame image and the designated frame image before current frame image
The association of middle tracking target, completes the tracking target in the tracking of current frame image.But since specified target occlusion is hidden
Target is kept off, therefore, it is also desirable to carry out target following to the target that is blocked, is realized especially by following step.
Step S314, according to position of the tracking target in multiple image in multiple tracking targets, in addition to specified target
It sets, estimation is carried out to the tracking target in addition to specified target by Kalman Algorithm, is obtained in addition to specified target
Target is tracked in the estimated location of current frame image.
Since above-mentioned multiple tracking targets match with the specified target in current frame image, removed in multiple tracking targets
With specified target be with the tracking target of a group traveling together other than, remaining is the tracking target of designated target occlusion；Due to these
The detection data that the detection data for the tracking target being blocked largely is designated target is blocked, so being difficult to through matched side
Position of the tracking target that formula is blocked in current frame image；Therefore, in above-mentioned steps, by way of estimation
Estimate position of the tracking target being blocked in current frame image.
It is appreciated that the motion state of tracking target is relatively continuous, therefore can be according to tracking target in present frame
The motion profile in multiple frame images before image, predicts the tracking target in the position of current frame image.Above-mentioned Kalman
Algorithm is specifically as follows least-squares estimation, and linear minimum-variance estimation, minimum variance estimate, Recursive Least Squares Estimation etc. are calculated
Method.
Step S316 judges the friendship of the detection data of above-mentioned estimated location and specified target and than whether being greater than or equal in advance
If the second fractional threshold；If so, executing step S318；If not, executing step S320；
The step may further determine that the reasonability of estimated location.Since above-mentioned estimated location is to be designated target occlusion
Tracking target position, therefore, which needs that there are certain overlapping, overlapping journeys with the detection data of specified target
Degree can be by above-mentioned friendship and than measuring.If the friendship of estimated location and the detection data of specified target is simultaneously more pre- than being greater than or equal to
If the second fractional threshold (second fractional threshold can be rule of thumb arranged), illustrate the estimated location be designated target inspection
Measured data " is blocked ", i.e., the estimated location is relatively reasonable；If the friendship of estimated location and the detection data of specified target and than small
In preset second fractional threshold, illustrate that the estimated location is not designated the detection data " blocking " of target, the estimated location
Less rationally.
Step S318 determines that the corresponding tracking target of estimated location is the target that is blocked, estimated location is determined as being hidden
Keep off position of the target in current frame image；
When estimated location is relatively reasonable, the target that is blocked which can be determined as to specified target occlusion exists
Position in current frame image；The corresponding tracking target of the estimated location is the target that is blocked, which can be with
Same tracking mark is arranged in tracking target corresponding with estimated location, to represent the target that is blocked with tracking target as same a line
People.
Step S320 adjusts estimated location, by estimated location adjusted according to the position of the detection data of specified target
It is determined as position of the target in current frame image that be blocked.
When estimated location is unreasonable, i.e., the estimated location, which should be located at, is designated the area that the detection data of target is blocked
Domain, and the estimated location be practically in distance specify target detection data remote position, at this time can by estimated location into
Row adjustment.
Specific adjustment process, which can be such that, to be obtained in the central point A of estimated location and the detection data of specified target
The heart point B, central point A and central point B form a line；The central point A of estimated location is moved along the line, is being moved through
The friendship of the detection data of estimated location and specified target and ratio are calculated in journey, when the friendship and than being greater than or equal to preset second ratio
When being worth threshold value, estimated location stops movement, and estimated location is position of the target in current frame image that be blocked at this time.
As long as by the above-mentioned means, detecting that (i.e. the target is not blocked, and also hides in the presence of specified target in current frame image
Keep off other targets), the target that is blocked of the specified target would not be tracked loss.But if a certain target is not detected among out
Come, or blocked the targets of other targets originally and be not detected among out and refer to and set the goal, it is likely that will cause tracking target
Loss.
Aforesaid way can be realized simultaneously the tracking of the target object not being blocked and the target that is blocked, as long as detecting mesh
Mark object has blocked the target that is blocked, and the tracking for the target that is blocked can be realized by way of estimation between multiple image,
The probability of happening that tracking is lost is reduced, to improve the accuracy rate of target following.
Example IV:
Corresponding to above method embodiment, a kind of structural schematic diagram of target tracker shown in Figure 4, the device
Include:
Image collection module 40, for obtaining current frame image；
Detection module 41 for detecting the target object in current frame image by preset detection model, and judges mesh
Whether mark object is specified target；Specified target is the target not being blocked, and specifies the target occlusion target that is blocked；
First tracking module 42, sets the goal if referred to for target object, obtains the designated frame before current frame image
The tracking target to match in image with specified target；If the tracking target to match be it is multiple, according to multiple tracking targets
Detection data relative position, determine specified target, from multiple tracking targets to track to specified target；
Second tracking module 43, for according in multiple tracking targets, the tracking target in addition to specified target to be specified
Position in frame image estimates that in the position of current frame image, estimated result is determined for the tracking target in addition to specified target
For position of the target in current frame image that be blocked, to be tracked to the target that is blocked.
Above-mentioned target tracker provided in an embodiment of the present invention detects present frame figure by preset detection model first
Target object as in, and judge whether target object is specified target；The specified target is the target not being blocked, and this refers to
It sets the goal and has blocked the target that is blocked；It sets the goal if target object refers to；Then obtain the designated frame figure before current frame image
The tracking target to match as in specified target；According to it is multiple tracking targets detection datas relative position, from it is multiple with
Specified target is determined in track target, to track to specified target；Further according in multiple tracking targets, in addition to specified target
Position of the tracking target in designated frame image, estimate the tracking target in addition to specified target in the position of current frame image
Set, determined and be blocked target from the tracking target in addition to specified target according to estimated result, with to this be blocked target into
Line trace.Aforesaid way can be realized simultaneously the tracking of the target object not being blocked and the target that is blocked, as long as detecting mesh
Mark object has blocked the target that is blocked, and the tracking for the target that is blocked can be realized by way of estimation between multiple image,
The probability of happening that tracking is lost is reduced, to improve the accuracy rate of target following.
Further, above-mentioned detection module is also used to: passing through the target pair in first network model inspection current frame image
As；Whether the target that is blocked is blocked by the second network model detected target object, if so, target object is determined as referring to
It sets the goal.
Further, above-mentioned first tracking module is also used to: obtain current frame image before designated frame image in
Track target；If track target be it is multiple, one by one by it is each tracking target detection data and specified target detection data into
Row matching；The tracking target that matching result is greater than or equal to preset matching threshold is determined as the tracking that specified target matches
Target.
Further, above-mentioned first tracking module is also used to: being calculated the detection data of each tracking target one by one and is specified
Friendship and ratio between the detection data of target；If the currently corresponding friendship of tracking target is simultaneously compared than being greater than or equal to preset first
It is worth threshold value, the detection data of current tracking target is matched with the detection data of specified target.
Further, above-mentioned detection data is identified by detection block；Above-mentioned first tracking module is also used to: from multiple tracking
The detection block bottom edge position of recognition detection data is located at the tracking target of bottommost in target；The tracking target that will identify that determines
To specify target.
Further, above-mentioned designated frame image includes multiple image；Above-mentioned second tracking module is also used to: according to it is multiple with
In track target, position of the tracking target in multiple image in addition to specified target, by Kalman Algorithm to except specified mesh
It is marked with outer tracking target and carries out estimation, obtain the tracking target in addition to specified target in the estimation position of current frame image
It sets.
Further, above-mentioned second tracking module is also used to: judging the friendship of the detection data of estimated location and specified target
And than whether being greater than or equal to preset second fractional threshold；If so, determining that the corresponding tracking target of estimated location is to be hidden
Target is kept off, estimated location is determined as position of the target in current frame image that be blocked；If not, according to the inspection of specified target
The position of measured data adjusts estimated location, is determined as being blocked target in current frame image for estimated location adjusted
Position.
The technical effect of device provided by the present embodiment, realization principle and generation is identical with previous embodiment, for letter
It describes, Installation practice part does not refer to place, can refer to corresponding contents in preceding method embodiment.
Embodiment five:
The embodiment of the invention provides a kind of Target Tracking System, the system include: image capture device, processing equipment and
Storage device；Image capture device, for obtaining preview video frame or image data；Computer journey is stored on storage device
Sequence, computer program execute above-mentioned method for tracking target when equipment processed is run.
It is apparent to those skilled in the art that for convenience and simplicity of description, the target of foregoing description
The specific work process of tracking system, can refer to corresponding processes in the foregoing method embodiment, and details are not described herein.
Further, the present embodiment additionally provides a kind of computer readable storage medium, on computer readable storage medium
It is stored with computer program, the step of when computer program equipment operation processed executes above-mentioned method for tracking target.
The computer program product of method for tracking target, device and system provided by the embodiment of the present invention, including storage
The computer readable storage medium of program code, the instruction that said program code includes can be used for executing previous methods embodiment
Described in method, specific implementation can be found in embodiment of the method, details are not described herein.
In addition, in the description of the embodiment of the present invention unless specifically defined or limited otherwise, term " installation ", " phase
Even ", " connection " shall be understood in a broad sense, for example, it may be being fixedly connected, may be a detachable connection, or be integrally connected；It can
To be mechanical connection, it is also possible to be electrically connected；It can be directly connected, can also can be indirectly connected through an intermediary
Connection inside two elements.For the ordinary skill in the art, above-mentioned term can be understood at this with concrete condition
Concrete meaning in invention.
It, can be with if the function is realized in the form of SFU software functional unit and when sold or used as an independent product
It is stored in a computer readable storage medium.Based on this understanding, technical solution of the present invention is substantially in other words
The part of the part that contributes to existing technology or the technical solution can be embodied in the form of software products, the meter
Calculation machine software product is stored in a storage medium, including some instructions are used so that a computer equipment (can be a
People's computer, server or network equipment etc.) it performs all or part of the steps of the method described in the various embodiments of the present invention.
And storage medium above-mentioned includes: that USB flash disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), arbitrary access are deposited
The various media that can store program code such as reservoir (RAM, Random Access Memory), magnetic or disk.
In the description of the present invention, it should be noted that term " center ", "upper", "lower", "left", "right", "vertical",
The orientation or positional relationship of the instructions such as "horizontal", "inner", "outside" be based on the orientation or positional relationship shown in the drawings, merely to
Convenient for description the present invention and simplify description, rather than the device or element of indication or suggestion meaning must have a particular orientation,
It is constructed and operated in a specific orientation, therefore is not considered as limiting the invention.In addition, term " first ", " second ",
" third " is used for descriptive purposes only and cannot be understood as indicating or suggesting relative importance.
Finally, it should be noted that embodiment described above, only a specific embodiment of the invention, to illustrate the present invention
Technical solution, rather than its limitations, scope of protection of the present invention is not limited thereto, although with reference to the foregoing embodiments to this hair
It is bright to be described in detail, those skilled in the art should understand that: anyone skilled in the art
In the technical scope disclosed by the present invention, it can still modify to technical solution documented by previous embodiment or can be light
It is readily conceivable that variation or equivalent replacement of some of the technical features；And these modifications, variation or replacement, do not make
The essence of corresponding technical solution is detached from the spirit and scope of technical solution of the embodiment of the present invention, should all cover in protection of the invention
Within the scope of.Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. a kind of method for tracking target, which is characterized in that the described method includes:
Obtain current frame image；
Detect the target object in the current frame image by preset detection model, and judge the target object whether be
Specified target；The specified target is the target not being blocked, and the specified target occlusion target that is blocked；
If so, the tracking mesh to match in designated frame image before obtaining the current frame image with the specified target
Mark；If the tracking target to match be it is multiple, according to it is multiple it is described tracking targets detection datas relative position, from
The specified target is determined in multiple tracking targets, to track to the specified target；
According to position of the tracking target in the designated frame image in multiple tracking targets, in addition to the specified target
Set, the estimation tracking target in addition to the specified target in the position of the current frame image, according to estimated result from
Be blocked target described in determining in tracking target in addition to the specified target, to track to the target that is blocked.
2. the method according to claim 1, wherein detecting the current frame image by preset detection model
In target object, and the step of whether target object is specified target judged, comprising:
Pass through the target object in current frame image described in first network model inspection；
Detect whether the target object has blocked the target that is blocked by the second network model, if so, by the target pair
As being determined as specified target.
3. the method according to claim 1, wherein in designated frame image before obtaining the current frame image
The step of tracking target to match with the specified target, comprising:
The tracking target in designated frame image before acquisition current frame image；
If the tracking target be it is multiple, one by one by it is each it is described tracking target detection data and the specified target inspection
Measured data is matched；
By the tracking target that matching result is greater than or equal to preset matching threshold be determined as that the specified target matches with
Track target.
4. according to the method described in claim 3, it is characterized in that, one by one by the detection data of each tracking target and institute
State the step of detection data of specified target is matched, comprising:
The friendship between the detection data of each tracking target and the detection data of the specified target and ratio are calculated one by one；
If currently tracking the corresponding friendship of target and than being greater than or equal to preset first fractional threshold, by the current tracking mesh
Target detection data is matched with the detection data of the specified target.
5. the method according to claim 1, wherein the detection data is identified by detection block；
According to the relative position of the detection data of multiple tracking targets, determined from multiple tracking targets described specified
The step of target, comprising:
The detection block bottom edge position of recognition detection data is located at the tracking target of bottommost from multiple tracking targets；
The tracking target that will identify that is determined as the specified target.
6. the method according to claim 1, wherein the designated frame image includes multiple image；
According to position of the tracking target in the designated frame image in multiple tracking targets, in addition to the specified target
It sets, the estimation tracking target in addition to the specified target is the position of the current frame image the step of, comprising:
According to position of the tracking target in the multiple image in multiple tracking targets, in addition to the specified target
It sets, estimation is carried out to the tracking target in addition to the specified target by Kalman Algorithm, is obtained except the specified mesh
Outer tracking target is marked in the estimated location of the current frame image.
7. according to the method described in claim 6, it is characterized in that, according to estimated result from addition to the specified target with
The step of target that is blocked described in being determined in track target, comprising:
Judge the friendship of the detection data of the estimated location and the specified target and than whether being greater than or equal to preset second
Fractional threshold；
If so, determining that the corresponding tracking target of the estimated location is the target that is blocked, the estimated location is determined
For the position of the target in the current frame image that be blocked；
If not, adjust the estimated location according to the position of the detection data of the specified target, described estimate adjusted
Meter position is determined as the position of the target in the current frame image that be blocked.
8. a kind of target tracker, which is characterized in that described device includes:
Image collection module, for obtaining current frame image；
Detection module, for detecting the target object in the current frame image by preset detection model, and described in judgement
Whether target object is specified target；The specified target is the target not being blocked, and the specified target occlusion is hidden
Keep off target；
First tracking module, if being the specified target for the target object, before obtaining the current frame image
The tracking target to match in designated frame image with the specified target；If the tracking target to match is multiple, root
According to the relative position of the detection data of multiple tracking targets, the specified target is determined from multiple tracking targets,
To be tracked to the specified target；
Second tracking module, for according in multiple tracking targets, the tracking target in addition to the specified target to be in institute
The position in designated frame image is stated, the estimation tracking target in addition to the specified target is in the position of the current frame image
It sets, be blocked target according to estimated result determination from the tracking target in addition to the specified target, to the quilt
Shelter target is tracked.
9. a kind of Target Tracking System, which is characterized in that the system comprises: image capture device, processing equipment and storage dress
It sets；
Described image acquires equipment, for obtaining preview video frame or image data；
Computer program is stored on the storage device, the computer program executes such as when being run by the processing equipment
The described in any item methods of claim 1 to 7.
10. a kind of computer readable storage medium, computer program, feature are stored on the computer readable storage medium
It is, the computer program equipment processed executes the step of the described in any item methods of the claims 1 to 7 when running
Suddenly.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811195130.XA CN109446942B (en) | 2018-10-12 | 2018-10-12 | Target tracking method, device and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811195130.XA CN109446942B (en) | 2018-10-12 | 2018-10-12 | Target tracking method, device and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109446942A true CN109446942A (en) | 2019-03-08 |
| CN109446942B CN109446942B (en) | 2020-10-16 |
Family
ID=65544987
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811195130.XA Active CN109446942B (en) | 2018-10-12 | 2018-10-12 | Target tracking method, device and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109446942B (en) |
Cited By (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110223320A (en) * | 2019-05-08 | 2019-09-10 | 北京百度网讯科技有限公司 | Object detection tracking and detecting and tracking device |
| CN110335293A (en) * | 2019-07-12 | 2019-10-15 | 东北大学 | A kind of long-time method for tracking target based on TLD frame |
| CN110414443A (en) * | 2019-07-31 | 2019-11-05 | 苏州市科远软件技术开发有限公司 | A kind of method for tracking target, device and rifle ball link tracking |
| CN110688896A (en) * | 2019-08-23 | 2020-01-14 | 北京正安维视科技股份有限公司 | Pedestrian loitering detection method |
| CN110717414A (en) * | 2019-09-24 | 2020-01-21 | 青岛海信网络科技股份有限公司 | Target detection tracking method, device and equipment |
| CN110910428A (en) * | 2019-12-05 | 2020-03-24 | 江苏中云智慧数据科技有限公司 | Real-time multi-target tracking method based on neural network |
| CN110929639A (en) * | 2019-11-20 | 2020-03-27 | 北京百度网讯科技有限公司 | Method, apparatus, device and medium for determining position of obstacle in image |
| CN111127358A (en) * | 2019-12-19 | 2020-05-08 | 苏州科达科技股份有限公司 | Image processing method, device and storage medium |
| CN111145214A (en) * | 2019-12-17 | 2020-05-12 | 深圳云天励飞技术有限公司 | Target tracking method, device, terminal equipment and medium |
| CN111339855A (en) * | 2020-02-14 | 2020-06-26 | 睿魔智能科技（深圳）有限公司 | Vision-based target tracking method, system, equipment and storage medium |
| WO2020215552A1 (en) * | 2019-04-26 | 2020-10-29 | 平安科技（深圳）有限公司 | Multi-target tracking method, apparatus, computer device, and storage medium |
| CN112069879A (en) * | 2020-07-22 | 2020-12-11 | 深圳市优必选科技股份有限公司 | Target person following method, computer-readable storage medium and robot |
| CN112166436A (en) * | 2019-12-24 | 2021-01-01 | 商汤国际私人有限公司 | Image screening method and device and electronic equipment |
| CN112419368A (en) * | 2020-12-03 | 2021-02-26 | 腾讯科技（深圳）有限公司 | Method, device and equipment for tracking track of moving target and storage medium |
| CN112597917A (en) * | 2020-12-25 | 2021-04-02 | 太原理工大学 | Vehicle parking detection method based on deep learning |
| CN112700494A (en) * | 2019-10-23 | 2021-04-23 | 北京灵汐科技有限公司 | Positioning method, positioning device, electronic equipment and computer readable storage medium |
| CN113012190A (en) * | 2021-02-01 | 2021-06-22 | 河南省肿瘤医院 | Hand hygiene compliance monitoring method, device, equipment and storage medium |
| CN113129337A (en) * | 2021-04-14 | 2021-07-16 | 桂林电子科技大学 | Background perception tracking method, computer readable storage medium and computer device |
| CN113168532A (en) * | 2020-07-27 | 2021-07-23 | 深圳市大疆创新科技有限公司 | Target detection method and device, unmanned aerial vehicle and computer readable storage medium |
| CN113160149A (en) * | 2021-03-31 | 2021-07-23 | 杭州海康威视数字技术股份有限公司 | Target display method and device, electronic equipment and endoscope system |
| CN113326715A (en) * | 2020-02-28 | 2021-08-31 | 初速度（苏州）科技有限公司 | Target association method and device |
| WO2021217450A1 (en) * | 2020-04-28 | 2021-11-04 | 深圳市大疆创新科技有限公司 | Target tracking method and device, and storage medium |
| CN113674318A (en) * | 2021-08-16 | 2021-11-19 | 支付宝(杭州)信息技术有限公司 | Target tracking method, device and equipment |
| CN114092515A (en) * | 2021-11-08 | 2022-02-25 | 国汽智控(北京)科技有限公司 | Target tracking detection method, device, equipment and medium for obstacle blocking |
| CN114299115A (en) * | 2021-12-28 | 2022-04-08 | 天翼云科技有限公司 | Method and device for multi-target tracking, storage medium and electronic equipment |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102214301A (en) * | 2011-07-26 | 2011-10-12 | 西南交通大学 | Multi-target tracking method for associated cooperation of adaptive motion |
| CN102663743A (en) * | 2012-03-23 | 2012-09-12 | 西安电子科技大学 | Multi-camera cooperative character tracking method in complex scene |
| CN103646407A (en) * | 2013-12-26 | 2014-03-19 | 中国科学院自动化研究所 | Video target tracking method based on ingredient and distance relational graph |
| CN104424638A (en) * | 2013-08-27 | 2015-03-18 | 深圳市安芯数字发展有限公司 | Target tracking method based on shielding situation |
| CN104732187A (en) * | 2013-12-18 | 2015-06-24 | 杭州华为企业通信技术有限公司 | Method and equipment for image tracking processing |
| CN106920253A (en) * | 2017-02-10 | 2017-07-04 | 华中科技大学 | It is a kind of based on the multi-object tracking method for blocking layering |
| WO2017199840A1 (en) * | 2016-05-18 | 2017-11-23 | 日本電気株式会社 | Object tracking device, object tracking method, and recording medium |
| CN107564034A (en) * | 2017-07-27 | 2018-01-09 | 华南理工大学 | The pedestrian detection and tracking of multiple target in a kind of monitor video |
| CN107580199A (en) * | 2017-09-08 | 2018-01-12 | 深圳市伊码泰珂电子有限公司 | The target positioning of overlapping ken multiple-camera collaboration and tracking system |
| CN108521554A (en) * | 2018-03-01 | 2018-09-11 | 西安电子科技大学 | Large scene multi-target cooperative tracking method, intelligent monitor system, traffic system |
-
2018
- 2018-10-12 CN CN201811195130.XA patent/CN109446942B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102214301A (en) * | 2011-07-26 | 2011-10-12 | 西南交通大学 | Multi-target tracking method for associated cooperation of adaptive motion |
| CN102663743A (en) * | 2012-03-23 | 2012-09-12 | 西安电子科技大学 | Multi-camera cooperative character tracking method in complex scene |
| CN104424638A (en) * | 2013-08-27 | 2015-03-18 | 深圳市安芯数字发展有限公司 | Target tracking method based on shielding situation |
| CN104732187A (en) * | 2013-12-18 | 2015-06-24 | 杭州华为企业通信技术有限公司 | Method and equipment for image tracking processing |
| CN103646407A (en) * | 2013-12-26 | 2014-03-19 | 中国科学院自动化研究所 | Video target tracking method based on ingredient and distance relational graph |
| WO2017199840A1 (en) * | 2016-05-18 | 2017-11-23 | 日本電気株式会社 | Object tracking device, object tracking method, and recording medium |
| CN106920253A (en) * | 2017-02-10 | 2017-07-04 | 华中科技大学 | It is a kind of based on the multi-object tracking method for blocking layering |
| CN107564034A (en) * | 2017-07-27 | 2018-01-09 | 华南理工大学 | The pedestrian detection and tracking of multiple target in a kind of monitor video |
| CN107580199A (en) * | 2017-09-08 | 2018-01-12 | 深圳市伊码泰珂电子有限公司 | The target positioning of overlapping ken multiple-camera collaboration and tracking system |
| CN108521554A (en) * | 2018-03-01 | 2018-09-11 | 西安电子科技大学 | Large scene multi-target cooperative tracking method, intelligent monitor system, traffic system |
Non-Patent Citations (3)
| Title |
|---|
| 卢钢: "面向监视视频序列中运动目标的跟踪算法研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 权伟: "可视对象跟踪算法研究及应用", 《中国博士学位论文全文数据库信息科技辑》 * |
| 肖义涵: "以NAO机器人为平台的人机互动技术研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (37)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020215552A1 (en) * | 2019-04-26 | 2020-10-29 | 平安科技（深圳）有限公司 | Multi-target tracking method, apparatus, computer device, and storage medium |
| CN110223320A (en) * | 2019-05-08 | 2019-09-10 | 北京百度网讯科技有限公司 | Object detection tracking and detecting and tracking device |
| CN110335293A (en) * | 2019-07-12 | 2019-10-15 | 东北大学 | A kind of long-time method for tracking target based on TLD frame |
| CN110414443A (en) * | 2019-07-31 | 2019-11-05 | 苏州市科远软件技术开发有限公司 | A kind of method for tracking target, device and rifle ball link tracking |
| CN110688896A (en) * | 2019-08-23 | 2020-01-14 | 北京正安维视科技股份有限公司 | Pedestrian loitering detection method |
| CN110717414A (en) * | 2019-09-24 | 2020-01-21 | 青岛海信网络科技股份有限公司 | Target detection tracking method, device and equipment |
| CN110717414B (en) * | 2019-09-24 | 2023-01-03 | 青岛海信网络科技股份有限公司 | Target detection tracking method, device and equipment |
| CN112700494A (en) * | 2019-10-23 | 2021-04-23 | 北京灵汐科技有限公司 | Positioning method, positioning device, electronic equipment and computer readable storage medium |
| CN110929639B (en) * | 2019-11-20 | 2023-09-19 | 北京百度网讯科技有限公司 | Method, apparatus, device and medium for determining the position of an obstacle in an image |
| CN110929639A (en) * | 2019-11-20 | 2020-03-27 | 北京百度网讯科技有限公司 | Method, apparatus, device and medium for determining position of obstacle in image |
| CN110910428A (en) * | 2019-12-05 | 2020-03-24 | 江苏中云智慧数据科技有限公司 | Real-time multi-target tracking method based on neural network |
| CN110910428B (en) * | 2019-12-05 | 2022-04-01 | 江苏中云智慧数据科技有限公司 | Real-time multi-target tracking method based on neural network |
| CN111145214A (en) * | 2019-12-17 | 2020-05-12 | 深圳云天励飞技术有限公司 | Target tracking method, device, terminal equipment and medium |
| CN111127358B (en) * | 2019-12-19 | 2022-07-19 | 苏州科达科技股份有限公司 | Image processing method, device and storage medium |
| CN111127358A (en) * | 2019-12-19 | 2020-05-08 | 苏州科达科技股份有限公司 | Image processing method, device and storage medium |
| CN112166436A (en) * | 2019-12-24 | 2021-01-01 | 商汤国际私人有限公司 | Image screening method and device and electronic equipment |
| CN112166436B (en) * | 2019-12-24 | 2024-09-24 | 商汤国际私人有限公司 | Image screening method and device and electronic equipment |
| CN111339855B (en) * | 2020-02-14 | 2023-05-23 | 睿魔智能科技（深圳）有限公司 | Vision-based target tracking method, system, equipment and storage medium |
| CN111339855A (en) * | 2020-02-14 | 2020-06-26 | 睿魔智能科技（深圳）有限公司 | Vision-based target tracking method, system, equipment and storage medium |
| CN113326715B (en) * | 2020-02-28 | 2022-06-10 | 魔门塔(苏州)科技有限公司 | Target association method and device |
| CN113326715A (en) * | 2020-02-28 | 2021-08-31 | 初速度（苏州）科技有限公司 | Target association method and device |
| WO2021217450A1 (en) * | 2020-04-28 | 2021-11-04 | 深圳市大疆创新科技有限公司 | Target tracking method and device, and storage medium |
| CN112069879A (en) * | 2020-07-22 | 2020-12-11 | 深圳市优必选科技股份有限公司 | Target person following method, computer-readable storage medium and robot |
| CN112069879B (en) * | 2020-07-22 | 2024-06-07 | 深圳市优必选科技股份有限公司 | Target person following method, computer-readable storage medium and robot |
| WO2022021028A1 (en) * | 2020-07-27 | 2022-02-03 | 深圳市大疆创新科技有限公司 | Target detection method, device, unmanned aerial vehicle, and computer-readable storage medium |
| CN113168532A (en) * | 2020-07-27 | 2021-07-23 | 深圳市大疆创新科技有限公司 | Target detection method and device, unmanned aerial vehicle and computer readable storage medium |
| CN112419368A (en) * | 2020-12-03 | 2021-02-26 | 腾讯科技（深圳）有限公司 | Method, device and equipment for tracking track of moving target and storage medium |
| CN112597917A (en) * | 2020-12-25 | 2021-04-02 | 太原理工大学 | Vehicle parking detection method based on deep learning |
| CN113012190B (en) * | 2021-02-01 | 2024-02-06 | 河南省肿瘤医院 | Hand hygiene compliance monitoring method, device, equipment and storage medium |
| CN113012190A (en) * | 2021-02-01 | 2021-06-22 | 河南省肿瘤医院 | Hand hygiene compliance monitoring method, device, equipment and storage medium |
| CN113160149A (en) * | 2021-03-31 | 2021-07-23 | 杭州海康威视数字技术股份有限公司 | Target display method and device, electronic equipment and endoscope system |
| CN113160149B (en) * | 2021-03-31 | 2024-03-01 | 杭州海康威视数字技术股份有限公司 | Target display method and device, electronic equipment and endoscope system |
| CN113129337A (en) * | 2021-04-14 | 2021-07-16 | 桂林电子科技大学 | Background perception tracking method, computer readable storage medium and computer device |
| CN113674318A (en) * | 2021-08-16 | 2021-11-19 | 支付宝(杭州)信息技术有限公司 | Target tracking method, device and equipment |
| CN114092515B (en) * | 2021-11-08 | 2024-03-05 | 国汽智控(北京)科技有限公司 | Target tracking detection method, device, equipment and medium for obstacle shielding |
| CN114092515A (en) * | 2021-11-08 | 2022-02-25 | 国汽智控(北京)科技有限公司 | Target tracking detection method, device, equipment and medium for obstacle blocking |
| CN114299115A (en) * | 2021-12-28 | 2022-04-08 | 天翼云科技有限公司 | Method and device for multi-target tracking, storage medium and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109446942B (en) | 2020-10-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109446942A (en) | Method for tracking target, device and system | |
| CN112041848B (en) | System and method for counting and tracking number of people | |
| CN109117827B (en) | Video-based method for automatically identifying wearing state of work clothes and work cap and alarm system | |
| CN109076198B (en) | Video-based object tracking occlusion detection system, method and equipment | |
| CN108986064B (en) | People flow statistical method, equipment and system | |
| EP2801078B1 (en) | Context aware moving object detection | |
| US9008365B2 (en) | Systems and methods for pedestrian detection in images | |
| CN102663452B (en) | Suspicious act detecting method based on video analysis | |
| Luber et al. | People tracking in rgb-d data with on-line boosted target models | |
| Stalder et al. | Cascaded confidence filtering for improved tracking-by-detection | |
| CN102542289B (en) | Pedestrian volume statistical method based on plurality of Gaussian counting models | |
| CN109145708B (en) | Pedestrian flow statistical method based on RGB and D information fusion | |
| JP5604256B2 (en) | Human motion detection device and program thereof | |
| CN201255897Y (en) | Human flow monitoring device for bus | |
| CN106295638A (en) | Certificate image sloped correcting method and device | |
| US20130243343A1 (en) | Method and device for people group detection | |
| US20150334267A1 (en) | Color Correction Device, Method, and Program | |
| CN103870824B (en) | A kind of face method for catching and device during Face datection tracking | |
| CN103093212A (en) | Method and device for clipping facial images based on face detection and face tracking | |
| CN110991397B (en) | Travel direction determining method and related equipment | |
| CN106447701A (en) | Methods and devices for image similarity determining, object detecting and object tracking | |
| CN104281839A (en) | Body posture identification method and device | |
| CN104079798B (en) | Image detecting method, device and a kind of video monitoring system | |
| CN108197585A (en) | Recognition algorithms and device | |
| CN110412566A (en) | A kind of fine granularity human arm motion's recognition methods based on Doppler radar time and frequency domain characteristics |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |