CN106683113A - Characteristic point tracking method and device - Google Patents
Characteristic point tracking method and device Download PDFInfo
- Publication number
- CN106683113A CN106683113A CN201611001873.XA CN201611001873A CN106683113A CN 106683113 A CN106683113 A CN 106683113A CN 201611001873 A CN201611001873 A CN 201611001873A CN 106683113 A CN106683113 A CN 106683113A
- Authority
- CN
- China
- Prior art keywords
- field picture
- characteristic point
- image
- point
- new
- 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
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a characteristic point tracking method and device. The method is applied to electronic equipment with an image tracking unit, and the electronic equipment is used to track a target object on the basis of the image tracking unit. The method comprises that the loss quantity n of characteristic points of a first frame of image relative to characteristic points of a second frame of image is determined, n represents an integer greater than 1, and the second frame of image is a last frame of image of the first frame of image in the time dimension; first processing is carried out on the first frame of image till n new characteristic points are found, pixel points are selected randomly from the image and whether the selected pixel points are characteristic points is determined according to the first processing, and the new characteristic points are different from characteristic points, not lost, in the first frame of image; and the characteristic points, not lost, in the first frame of image are combined with the n new characteristic points to obtain all characteristic points of the first frame of image. According to the invention, the problem that a characteristic point tracking method in the related art tends to cause too large time cost of the system is solved.
Description
Technical field
The present invention relates to image processing field, in particular to a kind of feature point tracking method and apparatus.
Background technology
It is a kind of conventional image characteristic extracting method that characteristic point (angle point) is extracted, by carrying out to the characteristic point in image
Extraction can detect, extract, recognize and track the target in image sequence.FAST(Features from Accelerated
Segment Test) it is a kind of angular-point detection method, it can be used for the extraction of characteristic point, and complete to track and map object.
FAST Corner Detection Algorithms are proposed by Edward Rosten and Tom Drummond, and the most prominent advantage of the algorithm is
Its computational efficiency.
In prior art, the operation principle of feature point tracking method is as follows:
The all characteristic points in present viewing field are extracted in initialization, the higher feature of wherein a number of quality is chosen
Point is tracked.As the change of camera position or the mobile scene that may cause in image of object in camera view are sent out
Raw to change, when image changes, the Partial Feature point in image may be moved in the picture or be lost.It is every afterwards
In one frame, visual system can utilize sparse optical flow method characteristic point is tracked, according to previous frame image and characteristic point
The position that characteristic point before finding is matched in the current frame.During tracking characteristics point, Partial Feature point may be tracked
Failure, in causing image, total feature points are reduced.Ensure that characteristic point maintains certain quantity as algorithm is usually required that, be
The FAST characteristic points traced in ensureing present frame maintain certain amount, when characteristic point disappears because of the motion of robot,
Require supplementation with characteristic point.In prior art, visual system can call method traversal present image during initialization again, find out and work as
All of characteristic point in forward view, and wherein preferably characteristic point is chosen with the quantity of complementary features point.
In said extracted present viewing field, the algorithm of all characteristic points can travel through whole image, result in larger time consumption
Take, in real-time system, for example, robotic vision system, the time overhead of traversing graph picture are excessive, can affect robot entirety
Performance.
The excessive problem of system time expense is easily caused for the feature point tracking method in correlation technique, at present not yet
Propose effective solution.
The content of the invention
Present invention is primarily targeted at a kind of feature point tracking method and apparatus is provided, to solve the spy in correlation technique
Levy point-tracking method and be easily caused the excessive problem of system time expense.
To achieve these goals, according to an aspect of the invention, there is provided a kind of feature point tracking method.The method
It is applied to the electronic equipment with image trace unit, electronic equipment is for performing to destination object based on image trace unit
Tracking, the method include:Determine characteristic point loss quantity n of the characteristic point compared to the second two field picture of the first two field picture, wherein,
N is the integer more than 1, and the second two field picture is previous frame image of first two field picture on time dimension;First two field picture is held
The process of row first until find the new characteristic points of n, wherein, first is processed as randomly selecting pixel in the picture and detecting choosing
Whether the pixel for taking is characterized a little, the characteristic point that new characteristic point is not lost in being different from the first two field picture;By the first frame figure
The new characteristic point combination of the characteristic point do not lost as in and n obtains whole characteristic points of the first two field picture.
Further, it is determined that the first two field picture characteristic point compared to the second two field picture characteristic point lose quantity n it
Afterwards, the method also includes:Judge that whether quantity n exceed predetermined threshold value, wherein, judge quantity n whether more than predetermined threshold value it
Afterwards, the method also includes:If it is judged that quantity n exceedes predetermined threshold value, perform second processing to obtain the to the first two field picture
The all of characteristic point of one two field picture, wherein, whether second processing is characterized a little for all of pixel in detection image;Obtaining
Pre-conditioned characteristic point is determined for compliance with all of characteristic point of the first two field picture got;If it is judged that quantity n does not surpass
Predetermined threshold value is crossed, first is performed to the first two field picture and is processed.
Further, it is determined that the first two field picture characteristic point compared to the second two field picture characteristic point lose quantity n it
Before, the method also includes:Initial two field picture is performed second processing to obtain all of characteristic point of initial two field picture;Obtaining
To initial two field picture all of characteristic point in be determined for compliance with pre-conditioned characteristic point.
Further, it is determined that the characteristic point of the first two field picture includes compared to characteristic point loss quantity n of the second two field picture:
The characteristic point of the second two field picture is tracked by presetting track algorithm in the first two field picture, wherein, the characteristic point of the second two field picture
The quantity that successful characteristic point is not tracked in the first two field picture is n, and it is in the second two field picture not track successful characteristic point
The characteristic point for existing but losing in the first two field picture.
Further, the first two field picture and the second two field picture are the image obtained by image trace unit, to first
Two field picture performs first and processes up to before finding n new characteristic point, and the method also includes:Obtain the fortune of image trace unit
Dynamic information;Estimate the first two field picture compared to the n feature that the second two field picture is lost according to the movable information of image trace unit
The region X o'clock being located in the first two field picture;Accordingly, perform first to process until finding n new spy to the first two field picture
Levy a little, including:Perform first to process until finding n new characteristic point to the region X in the first two field picture.
To achieve these goals, according to an aspect of the invention, there is provided a kind of feature point tracking device.The device
It is applied to the electronic equipment with image trace unit, electronic equipment is for performing to destination object based on image trace unit
Tracking, the device include:First determining unit, for determining the feature of the characteristic point compared to the second two field picture of the first two field picture
Point loses quantity n, wherein, n is the integer more than 1, and the second two field picture is former frame figure of first two field picture on time dimension
Picture;First performance element, for the first process is performed to the first two field picture until finding n new characteristic point, wherein, at first
Manage to randomly select pixel in the picture and detecting whether the pixel of selection is characterized a little, new characteristic point is different from first
The characteristic point do not lost in two field picture;Assembled unit, for characteristic point that will not lose in the first two field picture and n new feature
Point combination obtains whole characteristic points of the first two field picture.
Further, the device also includes:Judging unit, for it is determined that the characteristic point of the first two field picture is compared to second
After the characteristic point of two field picture loses quantity n, judge whether quantity n exceedes predetermined threshold value, the second performance element, for judging
After whether quantity n exceedes predetermined threshold value, if it is judged that quantity n exceedes predetermined threshold value, the first two field picture is performed at second
Manage to obtain all of characteristic point of the first two field picture, wherein, second processing is that whether all of pixel is in detection image
Characteristic point;Pre-conditioned characteristic point is determined for compliance with all of characteristic point of the first two field picture for getting, wherein, first
Performance element is additionally operable to if it is judged that quantity n performs first to the first two field picture and processes not less than predetermined threshold value.
Further, the second performance element is additionally operable to it is determined that the characteristic point of the first two field picture is compared to the second two field picture
Before characteristic point loses quantity n, initial two field picture is performed second processing to obtain all of characteristic point of initial two field picture, should
Device also includes:Second determining unit, in all of characteristic point of the initial two field picture for getting being determined for compliance with presetting
The characteristic point of condition.
Further, the first determining unit includes:Tracking module, for by default track algorithm in the first two field picture
The characteristic point of the second two field picture is tracked, wherein, the characteristic point of the second two field picture does not track successful feature in the first two field picture
The quantity of point is n, and it is the feature for existing in the second two field picture but losing in the first two field picture not track successful characteristic point
Point.
Further, the first two field picture and the second two field picture are the image obtained by image trace unit, and the device is also
Including:Acquiring unit, for, before the first process is performed to the first two field picture until finding n new characteristic point, obtaining figure
As the movable information of tracking cell;Evaluation unit, for estimating the first two field picture phase according to the movable information of image trace unit
The region X that the n characteristic point lost compared with the second two field picture is located in the first two field picture;Wherein, the first performance element is also used
Region X in the first two field picture performs first and processes until finding n new characteristic point.
Characteristic point of the present invention by the characteristic point of the first two field picture of determination compared to the second two field picture loses quantity n, its
In, n is the integer more than 1, and the second two field picture is previous frame image of first two field picture on time dimension;To the first two field picture
The first process is performed until finding n new characteristic point, wherein, first is processed as randomly selecting pixel in the picture and detecting
Whether the pixel of selection is characterized a little, the characteristic point that new characteristic point is not lost in being different from the first two field picture;By the first frame
The new characteristic point combination of the characteristic point do not lost in image and n obtains whole characteristic points of the first two field picture, solves correlation
Feature point tracking method in technology is easily caused the excessive problem of system time expense, and then has reached reduction feature point tracking
The effect of the system time expense that method takes.
Description of the drawings
The accompanying drawing for constituting the part of the application is used for providing a further understanding of the present invention, the schematic reality of the present invention
Apply example and its illustrate, for explaining the present invention, not constituting inappropriate limitation of the present invention.In the accompanying drawings:
Fig. 1 is the flow chart of feature point tracking method according to a first embodiment of the present invention;
Fig. 2 is the flow chart of feature point tracking method according to a second embodiment of the present invention;
Fig. 3 is the flow chart of feature point tracking method according to a third embodiment of the present invention;And
Fig. 4 is the schematic diagram of feature point tracking device according to embodiments of the present invention.
Specific embodiment
It should be noted that in the case where not conflicting, the feature in embodiment and embodiment in the application can phase
Mutually combine.Below with reference to the accompanying drawings and in conjunction with the embodiments describing the present invention in detail.
In order that those skilled in the art more fully understand application scheme, below in conjunction with the embodiment of the present application
Accompanying drawing, is clearly and completely described to the technical scheme in the embodiment of the present application, it is clear that described embodiment is only
The embodiment of the application part, rather than the embodiment of whole.Based on the embodiment in the application, ordinary skill people
The every other embodiment obtained under the premise of creative work is not made by member, should all belong to the model of the application protection
Enclose.
It should be noted that the description and claims of this application and the term " first " in above-mentioned accompanying drawing, "
Two " it is etc. for distinguishing similar object, without for describing specific order or precedence.It should be appreciated that so using
Data can exchange in the appropriate case, so as to embodiments herein described herein.Additionally, term " including " and " tool
Have " and their any deformation, it is intended that cover it is non-exclusive includes, for example, contain series of steps or unit
Process, method, system, product or equipment are not necessarily limited to those steps clearly listed or unit, but may include without clear
Other steps list to Chu or intrinsic for these processes, method, product or equipment or unit.
The embodiment provides a kind of feature point tracking method.
The method can apply to the electronic equipment with image trace unit, and electronic equipment can be based on image trace list
Unit performs the tracking to destination object.For example, electronic equipment can be robot (such as ground mobile robot, unmanned plane etc.),
Robot can be by photographic head photographic subjects object, according to the distance between image determination and destination object for shooting and direction
Etc. information and according to these information trace destination objects.
Fig. 1 is the flow chart of feature point tracking method according to embodiments of the present invention.As shown in figure 1, the method include with
Lower step:
Step S101, determines characteristic point loss quantity n of the characteristic point compared to the second two field picture of the first two field picture.
Wherein, n is the integer more than 1.
First two field picture and the second two field picture can be the figures arrived by the image trace unit photographs of electronic equipment
As, or the image that collected by other image sampling devices send to electronic equipment.
Second two field picture is previous frame image of first two field picture on time dimension.First two field picture and the second two field picture
Can be adjacent in time, the time interval between the first two field picture and the second two field picture can be carried according to the embodiment
For tracking determine time interval, the time interval between the first two field picture and the second two field picture can be with image sampling
Time interval it is different.
For example, the time interval of image sampling is 1ms, and the time interval of the tracking that the embodiment is provided is 5ms,
Then the second two field picture is the image collected after the first two field picture 5ms.
When destination object is tracked, the image that can pass through to collect performs tracking to electronic equipment, specifically, can extract
Characteristic point in per two field picture, characteristic point of the characteristic point of every two field picture with previous frame image is compared, the first frame is determined
The characteristic point of image loses quantity n compared to the characteristic point of the second two field picture.
Step S102, performs first and processes until finding n new characteristic point to the first two field picture.
It is determined that the first two field picture characteristic point compared to the second two field picture characteristic point lose quantity n after, to first
Two field picture performs first and processes until finding n new characteristic point.
First is processed as randomly selecting pixel in the picture, and detects whether the pixel of selection is characterized a little.Wherein,
The characteristic point that new characteristic point is not lost in being different from the first two field picture.
It is pointed out that the residing location of pixels in the first two field picture of the characteristic point do not lost in the first two field picture
May be different with the location of pixels residing for corresponding characteristic point in the second two field picture, image trace unit can be according to not losing
The change of the location of pixels residing for characteristic point determines the direction of destination object movement and distance, and calculates and control electronics
The direction of motion and speed are tracking destination object.
The characteristic point do not lost in first two field picture and n new characteristic point combination is obtained the first frame figure by step S103
Whole characteristic points of picture.
After first being performed to the first two field picture and is processed until finding the new characteristic points of n, by the first two field picture not
A characteristic point and n new characteristic point combination of loss obtains whole characteristic points of the first two field picture.
The feature point tracking method that the embodiment is provided, by determining the characteristic point of the first two field picture compared to the second frame figure
The characteristic point of picture loses quantity n, wherein, n is the integer more than 1, and the second two field picture is the first two field picture on time dimension
Previous frame image;The first process is performed to the first two field picture until finding n new characteristic point, wherein, first is processed as in figure
Pixel is randomly selected as in and detects whether the pixel of selection is characterized a little, new characteristic point is different from the first two field picture
The characteristic point do not lost;The characteristic point do not lost in first two field picture and n new characteristic point combination is obtained into the first two field picture
Whole characteristic points, it is excessive that the feature point tracking method in correlation technique that solves the problems, such as is easily caused system time expense,
And then reached the effect for reducing the system time expense that feature point tracking method takes.
As a preferred embodiment of above-described embodiment, it is determined that the characteristic point of the first two field picture is compared to the second frame figure
After the characteristic point of picture loses quantity n, the method can also include:Judge whether quantity n exceedes predetermined threshold value, wherein, sentencing
After whether disconnected quantity n exceedes predetermined threshold value, the method also includes:If it is judged that quantity n exceedes predetermined threshold value, to the first frame
Image performs second processing to obtain all of characteristic point of the first two field picture, wherein, second processing is all in detection image
Pixel whether be characterized a little;Pre-conditioned spy is determined for compliance with all of characteristic point of the first two field picture for getting
Levy a little;If it is judged that quantity n performs first to the first two field picture and processes not less than predetermined threshold value.
That is, it is determined that the first two field picture characteristic point compared to the second two field picture characteristic point lose quantity n, if lost
Quantity n of mistake is less, for example, when losing less than 5 characteristic points, first can be adopted to process and find new characteristic point.If lost
Quantity n of mistake is larger, then can update all of feature by the way of whether being characterized a little with all of pixel in the traversing graph picture
Point.
As a preferred embodiment of above-described embodiment, it is determined that the characteristic point of the first two field picture is compared to the second frame figure
Before the characteristic point of picture loses quantity n, the method can also include:Initial two field picture is performed second processing to obtain initial frame
The all of characteristic point of image;Pre-conditioned feature is determined for compliance with all of characteristic point of the initial two field picture for getting
Point.
That is, when the characteristic point of initial two field picture is obtained, can be obtained by the method to initial two field picture second processing
All of characteristic point is taken, and pre-conditioned characteristic point is determined for compliance with all characteristic points for getting.
Pre-conditioned can be provided with predetermined number, and the characteristic point of predetermined number is randomly choosed in all characteristic points.
It is pre-conditioned can also be according to preset algorithm choose in all characteristic points for obtaining the quality of predetermined number compared with
Characteristic point of the high characteristic point as initial two field picture.
Specifically, preset algorithm could be for the algorithm of the quality for calculating characteristic point, and the quality of characteristic point can be with spy
Levy a little that region is related in the picture, for example, if characteristic point occurs in central area, then it is assumed that its quality is higher, feature
The quality of point can also be related to the position of the characteristic point not disappeared, for example, if characteristic point and the characteristic point distance not disappeared
Farther out, then it is assumed that its quality is higher, " of quality for the quality of quantization characteristic point, can be calculated by specific algorithm
Point ", a quality evaluation object function is such as can determine, the value of function is obtained as matter as dependent variable according to related factor
The result that amount is evaluated is using as the standard for judging quality height.
Alternatively, it is pre-conditioned be additionally may included in all characteristic points in choose predetermined number, for example, by pre- imputation
After method calculates the quality evaluation result of characteristic point, it is ranked up according to quality from high to low, chooses front 10 characteristic point conducts
New a collection of characteristic point.
Determine that to lose quantity n can be by default compared to the characteristic point of the second two field picture for the characteristic point of the first two field picture
Track algorithm tracks the characteristic point of the second two field picture in the first two field picture, and the quantity for tracking the characteristic point lost is characteristic point
Lose quantity n.
Wherein, the characteristic point of the second two field picture do not track in the first two field picture successful characteristic point quantity be n, not with
The successful characteristic point of track is the characteristic point for existing in the second two field picture but losing in the first two field picture.Default track algorithm can
Being sparse optical flow tracking etc..The quantity that successful characteristic point is not tracked by presetting track algorithm is lost for current frame image
The characteristic point quantity of mistake.
Used as a preferred embodiment of above-described embodiment, the first two field picture and the second two field picture are by image trace list
The image that unit obtains, is processed up to before finding n new characteristic point first is performed to the first two field picture, and the method is also wrapped
Include:Obtain the movable information of image trace unit.
The movable information for obtaining image trace unit can be according to existing in the first two field picture and the second two field picture
The change of characteristic point residing location of pixels in the picture determines the movable information of image trace unit, if image trace unit is solid
Determine on an electronic device, the movable information of the movable information namely electronic equipment of image trace unit, movable information include direction
And distance.
After the movable information for obtaining image trace unit, the first frame is estimated according to the movable information of image trace unit
The region X that image is located in the first two field picture compared to the n characteristic point that the second two field picture is lost, performs to the first two field picture
First is processed until it is to perform first in region X in the first two field picture to process until finding n to find n new characteristic points
New characteristic point.
The scope for randomly selecting region can be reduced by said method, it is new to determine whether pixel is randomly selected
Characteristic point when faster can more efficiently search out the new characteristic points of n.
Fig. 2 is the flow chart of feature point tracking method according to a second embodiment of the present invention.The embodiment can be used as upper
The preferred implementation of first embodiment is stated, as shown in Fig. 2 the method is comprised the following steps:
Step S201, initialization detect certain amount FAST characteristic point.Extracted using FAST algorithms in initial two field picture
All characteristic points, preserve the preferable characteristic point of certain amount quality for tracking afterwards.
After a new two field picture is obtained, execution step S202 performs sparse optical flow tracking, obtains to a new two field picture
To the characteristic point of a new two field picture, the characteristic point of a new two field picture is n compared with the characteristic point quantity of previous frame missing image.
Sparse optical flow tracking is a kind of track algorithm, for the characteristic point in previous frame image is found in a new two field picture new
Corresponding characteristic point in one two field picture.After the characteristic point that tracking obtains a new two field picture, statistics obtains tracking failure
The number of characteristic point be defined as n.
Whether step S203, judge n less than threshold value.If it is, execution step S204, if not, execution step S205.
To new two field picture execution step S204 for obtaining, step S204 is to take a detection at random.If the spy for losing
Levy a little less than threshold value, then sufficient amount and up-to-standard is extracted in current frame image using taking a method for inspection at random
Characteristic point, wherein, it can randomly select a point and the point is detected to take at random a little and check, and then randomly select one again
Point the pointwise detection method that detected to the point, or multiple points are first randomly selected, then it is many to what is selected one by one
Individual point detects whether to be characterized a little, and whether the quality of this feature point is qualified, and the characteristic point for determining qualified is added to characteristic point
Sequence, keeps stable with the quantity for keeping all characteristic points in every two field picture.After execution step S204, execution step
S206。
Step S205, if the characteristic point lost is more than threshold value, directly using institute in FAST algorithms traversal current frame image
Some pixels, extract all of characteristic point in the two field picture.In the two field picture is extracted after all of characteristic point, can be by
Current signature point sequence is not added to the current point for having characteristic point repetition in the characteristic point that extraction is obtained, make characteristic point sequence
In point be maintained at certain quantity;Or, in the two field picture is extracted after all of characteristic point, extraction can be obtained
Characteristic point carries out quality evaluation, if N number of characteristic point is needed per two field picture, then chooses quality in the characteristic point for obtaining is extracted and comes
The characteristic point of front N.
After execution step S204, execution step S206, step S206 are a number of characteristic point of supplement.Will be current
The characteristic point quantity polishing of two field picture.
After execution step S206, two field picture execution step S202 of the renewal to obtaining so is circulated.
When the feature point tracking method that the embodiment is provided can be significantly shorter required during supplementary FAST characteristic points
Between, it is ensured that the real-time of algorithm, when there is minority FAST characteristic point to disappear from the visual field, by being taken in current frame image at random
Pixel and check its be whether characteristic point method, new-found characteristic point is added into the characteristic point sequence of current frame image
In, it is to avoid the hundreds thousand of pixels traversal detections to entire image.Due to generally only needing to when the characteristic point for disappearing is less
The supplement that characteristic point is completed by 1000 points is extracted, therefore, the method significantly reduces time spending.
Fig. 3 is the flow chart of feature point tracking method according to a third embodiment of the present invention.The embodiment can be used as upper
The preferred implementation of first embodiment is stated, as shown in figure 3, the method is comprised the following steps:
Step S301, initialization detect certain amount FAST characteristic point.Extracted using FAST algorithms in initial two field picture
All characteristic points, preserve the preferable characteristic point of certain amount quality for tracking afterwards.
After a new two field picture is obtained, execution step S302 performs sparse optical flow tracking, obtains to a new two field picture
To the characteristic point of a new two field picture, the characteristic point of a new two field picture is n compared with the characteristic point quantity of previous frame missing image.
Sparse optical flow tracking is a kind of track algorithm, for the characteristic point in previous frame image is found in a new two field picture new
Corresponding characteristic point in one two field picture.After the characteristic point that tracking obtains a new two field picture, statistics obtains tracking failure
The number of characteristic point be defined as n.
Whether step S303, judge n less than threshold value.If it is, execution step S304, if not, execution step S305.
To new two field picture execution step S304 for obtaining, step S304 is to take a detection at random in X regions.
Wherein, before new two field picture execution step S304 to obtaining, first carry out step S307 to determine X areas
Domain.
Step S307 is determination camera motion information.Camera is the camera of electronic equipment, for shooting image.Specifically
Ground, when camera motion information is calculated, can be calculated using VIO (Visual Inertial Odometry visions inertial navigation range finding)
Method, VIO algorithms are that one kind determines robot location and direction by analysis robot photographic head and internal inertial sensor
Algorithm.
It is determined that in front and back between two field pictures after the movable information of camera, judging phase according to the relative movement information of camera
Region of the emerging region in current frame image in machine coverage.For example, if between two field pictures, camera is left
Turn, then original characteristic point in visual field right side edge can disappear from the visual field, new region on the left of the visual field, occurs, order newly goes out
Existing region is X regions, takes a detection at random in X regions.
If the characteristic point lost is less than threshold value, in X regions using taking the method for inspection at random in current frame image
In extract sufficient amount and up-to-standard characteristic point.After execution step S304, execution step S306.
Step S305, if the characteristic point lost is more than threshold value, directly using institute in FAST algorithms traversal current frame image
Some pixels, extract all of characteristic point in the two field picture.In the two field picture is extracted after all of characteristic point, can be by
Current signature point sequence is not added to the current point for having characteristic point repetition in the characteristic point that extraction is obtained, make characteristic point sequence
In point be maintained at certain quantity;Or, in the two field picture is extracted after all of characteristic point, extraction can be obtained
Characteristic point carries out quality evaluation, if N number of characteristic point is needed per two field picture, then chooses quality in the characteristic point for obtaining is extracted and comes
The characteristic point of front N.
After execution step S304, execution step S306, step S306 are a number of characteristic point of supplement.Will be current
The characteristic point quantity polishing of two field picture.
After execution step S306, two field picture execution step S302 of the renewal to obtaining so is circulated.
Whether the method that the embodiment is provided exceedes according to the quantity of the characteristic point lost during sparse optical flow tracking
The difference of the judged result of threshold value performs different processing modes, when the characteristic point lost is less, by taking at random a little and examining
The method complementary features point tested, as the point lost is little, the process can be quickly completed;When the characteristic point lost is a lot,
The method that all characteristic points of direct detection in former algorithm can then be kept.
In the movable information of known machine people, can predict emerging part in the visual field be present frame which
Region, then just more in emerging region can take a little in the point process that takes at random afterwards, according to the rotation of photographic head
Direction and angle, the emerging region in anticipation image simultaneously take a detection in emerging region, so have bigger general
Rate gets suitable characteristic point, while can also make that characteristic point is more uniform to be distributed in whole visual field.
The change in the visual field in robot kinematics is that than shallower, when most of, the characteristic point quantity of loss is all
Fewer, then taking a compensation process for detection in the method at random can be quickly completed.Due to by traversing graph picture in original method
The process for finding all characteristic points is almost avoided completely, and the method significantly reduces time consumption, it is ensured that the real-time of system.
It should be noted that the Feature Points Extraction adopted in above-mentioned feature point tracking method is not limited to above-mentioned reality
The FAST Feature Points Extractions described in example are applied, the track algorithm adopted in above-mentioned feature point tracking method is not limited to above-mentioned
Sparse optical flow track algorithm described in embodiment.
It should be noted that can be in such as one group of computer executable instructions the step of the flow process of accompanying drawing is illustrated
Perform in computer system, and, although show logical order in flow charts, but in some cases, can be with not
The order being same as herein performs shown or described step.
Embodiments of the invention additionally provide a kind of feature point tracking device.It should be noted that the embodiment of the present invention
Feature point tracking device can be used for the feature point tracking method for performing the present invention.
Fig. 4 is the schematic diagram of feature point tracking device according to embodiments of the present invention.The device is can apply to figure
As the electronic equipment of tracking cell, electronic equipment is for the tracking based on the execution of image trace unit to destination object.
As shown in figure 4, the device includes the first determining unit 10, for determining the characteristic point of the first two field picture compared to the
The characteristic point of two two field pictures loses quantity n, wherein, n is the integer more than 1, and the second two field picture is the first two field picture in time dimension
Previous frame image on degree;First performance element 20, is processed until finding n new spy for first is performed to the first two field picture
Levy a little, wherein, first is processed as randomly selecting pixel in the picture and detects whether the pixel of selection is characterized a little, new
The characteristic point that characteristic point is not lost in being different from the first two field picture;Assembled unit 30, for will not lose in the first two field picture
The new characteristic point combination of characteristic point and n obtains whole characteristic points of the first two field picture.
Preferably, the device can also include:Judging unit, for it is determined that the characteristic point of the first two field picture is compared to
After the characteristic point of two two field pictures loses quantity n, judge whether quantity n exceedes predetermined threshold value, the second performance element, for sentencing
After whether disconnected quantity n exceedes predetermined threshold value, if it is judged that quantity n exceedes predetermined threshold value, second is performed to the first two field picture
Process to obtain all of characteristic point of the first two field picture, wherein, second processing is that whether all of pixel in detection image
It is characterized a little;Pre-conditioned characteristic point is determined for compliance with all of characteristic point of the first two field picture for getting, wherein, the
One performance element is additionally operable to if it is judged that quantity n performs first to the first two field picture and processes not less than predetermined threshold value.
Preferably, the second performance element is can be also used for it is determined that the characteristic point of the first two field picture is compared to the second two field picture
Characteristic point lose quantity n before, to initial two field picture perform second processing to obtain all of characteristic point of initial two field picture,
The device can also include:Second determining unit, for symbol is determined in all of characteristic point of the initial two field picture for getting
Close pre-conditioned characteristic point.
Preferably, the first determining unit can include:Tracking module, for by presetting track algorithm in the first two field picture
The characteristic point of the second two field picture of middle tracking, wherein, the characteristic point of the second two field picture does not track successful spy in the first two field picture
The quantity levied a little is n, and it is the spy for existing in the second two field picture but losing in the first two field picture not track successful characteristic point
Levy a little.
Preferably, the first two field picture and the second two field picture are the image obtained by image trace unit, and the device may be used also
To include:Acquiring unit, for, before the first process is performed to the first two field picture until finding n new characteristic point, obtaining
The movable information of image trace unit;Evaluation unit, for estimating the first two field picture according to the movable information of image trace unit
The region X that the n characteristic point lost compared to the second two field picture is located in the first two field picture;Wherein, the first performance element is also
Process until finding n new characteristic point for first is performed to the region X in the first two field picture.
Obviously, those skilled in the art should be understood that each module or each step of the above-mentioned present invention can be with general
Computing device realizing, they can be concentrated on single computing device, or are distributed in multiple computing devices and are constituted
Network on, alternatively, they can be realized with the executable program code of computing device, it is thus possible to they are stored
In the storage device by computing device performing, or they are fabricated to each integrated circuit modules respectively, or by they
In multiple modules or step be fabricated to single integrated circuit module to realize.So, the present invention is not restricted to any specific
Hardware and software is combined.
The preferred embodiments of the present invention are the foregoing is only, the present invention is not limited to, for the skill of this area
For art personnel, the present invention can have various modifications and variations.It is all within the spirit and principles in the present invention, made any repair
Change, equivalent, improvement etc., should be included within the scope of the present invention.
Claims (10)
1. a kind of feature point tracking method, it is characterised in that methods described is applied to the electronic equipment with image trace unit,
For the tracking to destination object is performed based on described image tracking cell, methods described includes the electronic equipment:
Determine characteristic point loss quantity n of the characteristic point compared to the second two field picture of the first two field picture, wherein, n is whole more than 1
Number, second two field picture is previous frame image of first two field picture on time dimension;
The first process is performed to first two field picture until finding n new characteristic point, wherein, described first is processed as in figure
Pixel is randomly selected as in and detects whether the pixel of selection is characterized a little, the new characteristic point is different from described first
The characteristic point do not lost in two field picture;
The characteristic point do not lost in first two field picture and the n new characteristic point combination is obtained into first two field picture
Whole characteristic points.
2. method according to claim 1, it is characterised in that it is determined that the characteristic point of the first two field picture is compared to the second frame
After the characteristic point of image loses quantity n, methods described also includes:
Judge whether quantity n exceedes predetermined threshold value,
Wherein, after judging whether quantity n exceedes predetermined threshold value, methods described also includes:
If it is judged that quantity n exceedes the predetermined threshold value, second processing is performed to first two field picture to obtain
The all of characteristic point of the first two field picture is stated, wherein, whether the second processing is special for all of pixel in detection image
Levy a little;Pre-conditioned characteristic point is determined for compliance with all of characteristic point of first two field picture for getting;
If it is judged that quantity n performs described first to first two field picture and processes not less than the predetermined threshold value.
3. method according to claim 2, it is characterised in that it is determined that the characteristic point of the first two field picture is compared to the second frame
Before the characteristic point of image loses quantity n, methods described also includes:
Initial two field picture is performed the second processing to obtain all of characteristic point of the initial two field picture;
The pre-conditioned characteristic point is determined for compliance with all of characteristic point of the described initial two field picture for getting.
4. method according to claim 1, it is characterised in that determine the characteristic point of the first two field picture compared to the second frame figure
The characteristic point of picture loses quantity n to be included:
The characteristic point of second two field picture is tracked in first two field picture by presetting track algorithm, wherein, described the
It is n that the characteristic point of two two field pictures does not track the quantity of successful characteristic point in first two field picture, described not track successfully
Characteristic point be exist but the characteristic point lost in first two field picture in second two field picture.
5. method according to claim 4, it is characterised in that first two field picture and second two field picture are to pass through
The image that described image tracking cell is obtained, is processed until finding n new feature first is performed to first two field picture
Before point, methods described also includes:
Obtain the movable information of described image tracking cell;According to the movable information of described image tracking cell estimation described first
The region X that two field picture is located in first two field picture compared to the n characteristic point that second two field picture is lost;
Accordingly, the first process is performed to first two field picture until finding n new characteristic point, including:To described first
Region X in two field picture performs described first and processes until finding n new characteristic point.
6. a kind of feature point tracking device, it is characterised in that described device is applied to the electronic equipment with image trace unit,
For the tracking to destination object is performed based on described image tracking cell, described device includes the electronic equipment:
First determining unit, for determining characteristic point loss quantity n of the characteristic point of the first two field picture compared to the second two field picture,
Wherein, n is the integer more than 1, and second two field picture is previous frame image of first two field picture on time dimension;
First performance element, for the first process is performed to first two field picture until finding n new characteristic point, wherein,
Described first is processed as randomly selecting pixel in the picture and detects whether the pixel of selection is characterized a little, the new spy
Levy the characteristic point for being a little different from do not lose in first two field picture;
Assembled unit, for the characteristic point do not lost in first two field picture and the n new characteristic point combination is obtained
Whole characteristic points of first two field picture.
7. device according to claim 6, it is characterised in that described device also includes:
Judging unit, for it is determined that the first two field picture characteristic point compared to the second two field picture characteristic point lose quantity n it
Afterwards, judge whether quantity n exceedes predetermined threshold value,
Second performance element, for after judging whether quantity n exceedes predetermined threshold value, if it is judged that quantity n
More than the predetermined threshold value, first two field picture is performed second processing to obtain all of feature of first two field picture
Point, wherein, whether the second processing is characterized a little for all of pixel in detection image;In first frame for getting
Pre-conditioned characteristic point is determined for compliance with all of characteristic point of image,
Wherein, first performance element is additionally operable to if it is judged that quantity n is not less than the predetermined threshold value, to described
One two field picture performs described first and processes.
8. device according to claim 7, it is characterised in that
Second performance element is additionally operable to it is determined that the characteristic point of the first two field picture is lost compared to the characteristic point of the second two field picture
Before losing quantity n, the second processing is performed to initial two field picture to obtain all of characteristic point of the initial two field picture,
Described device also includes:Second determining unit, in all of characteristic point of the described initial two field picture for getting
It is determined for compliance with the pre-conditioned characteristic point.
9. device according to claim 6, it is characterised in that first determining unit includes:
Tracking module, for tracking the feature of second two field picture by presetting track algorithm in first two field picture
Point, wherein, it is n that the characteristic point of second two field picture does not track the quantity of successful characteristic point in first two field picture,
It is described that not track successful characteristic point be exist but the feature lost in first two field picture in second two field picture
Point.
10. device according to claim 9, it is characterised in that first two field picture and second two field picture are logical
The image of described image tracking cell acquisition is crossed, described device also includes:
Acquiring unit, for, before the first process is performed to first two field picture until finding n new characteristic point, obtaining
The movable information of described image tracking cell;
Evaluation unit, for estimating first two field picture compared to described the according to the movable information of described image tracking cell
The region X that the n characteristic point that two two field pictures are lost is located in first two field picture;
Wherein, first performance element be additionally operable to in first two field picture region X perform it is described first process until
Find n new characteristic point.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2016109588249 | 2016-10-27 | ||
CN201610958824 | 2016-10-27 |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106683113A true CN106683113A (en) | 2017-05-17 |
CN106683113B CN106683113B (en) | 2020-02-04 |
Family
ID=58840201
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201611001873.XA Active CN106683113B (en) | 2016-10-27 | 2016-11-14 | Feature point tracking method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106683113B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107633526A (en) * | 2017-09-04 | 2018-01-26 | 腾讯科技(深圳)有限公司 | A kind of image trace point acquisition methods and equipment, storage medium |
CN109461174A (en) * | 2018-10-25 | 2019-03-12 | 北京陌上花科技有限公司 | Video object area tracking method and video plane advertisement method for implantation and system |
CN111397609A (en) * | 2020-05-13 | 2020-07-10 | 广东星舆科技有限公司 | Route planning method, mobile machine and computer readable medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101923718A (en) * | 2009-06-12 | 2010-12-22 | 新奥特(北京)视频技术有限公司 | Optimization method of visual target tracking method based on particle filtering and optical flow vector |
US20130257720A1 (en) * | 2012-03-27 | 2013-10-03 | Sony Corporation | Information input apparatus, information input method, and computer program |
CN103593679A (en) * | 2012-08-16 | 2014-02-19 | 北京大学深圳研究生院 | Visual human-hand tracking method based on online machine learning |
CN103996220A (en) * | 2014-05-26 | 2014-08-20 | 江苏大学 | Three-dimensional reconstruction method and system in intelligent transportation |
-
2016
- 2016-11-14 CN CN201611001873.XA patent/CN106683113B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101923718A (en) * | 2009-06-12 | 2010-12-22 | 新奥特(北京)视频技术有限公司 | Optimization method of visual target tracking method based on particle filtering and optical flow vector |
US20130257720A1 (en) * | 2012-03-27 | 2013-10-03 | Sony Corporation | Information input apparatus, information input method, and computer program |
CN103593679A (en) * | 2012-08-16 | 2014-02-19 | 北京大学深圳研究生院 | Visual human-hand tracking method based on online machine learning |
CN103996220A (en) * | 2014-05-26 | 2014-08-20 | 江苏大学 | Three-dimensional reconstruction method and system in intelligent transportation |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107633526A (en) * | 2017-09-04 | 2018-01-26 | 腾讯科技(深圳)有限公司 | A kind of image trace point acquisition methods and equipment, storage medium |
US11164323B2 (en) | 2017-09-04 | 2021-11-02 | Tencent Technology (Shenzhen) Company Limited | Method for obtaining image tracking points and device and storage medium thereof |
CN109461174A (en) * | 2018-10-25 | 2019-03-12 | 北京陌上花科技有限公司 | Video object area tracking method and video plane advertisement method for implantation and system |
CN111397609A (en) * | 2020-05-13 | 2020-07-10 | 广东星舆科技有限公司 | Route planning method, mobile machine and computer readable medium |
Also Published As
Publication number | Publication date |
---|---|
CN106683113B (en) | 2020-02-04 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US20240064391A1 (en) | Methods for refining rgbd camera poses | |
US10260862B2 (en) | Pose estimation using sensors | |
EP2858008A2 (en) | Target detecting method and system | |
Chen et al. | Asynchronous tracking-by-detection on adaptive time surfaces for event-based object tracking | |
CN111340749B (en) | Image quality detection method, device, equipment and storage medium | |
GB2594657A (en) | Damage detection from multi-view visual data | |
CN106447701A (en) | Methods and devices for image similarity determining, object detecting and object tracking | |
CN111383252B (en) | Multi-camera target tracking method, system, device and storage medium | |
CN109902675B (en) | Object pose acquisition method and scene reconstruction method and device | |
CN106683113A (en) | Characteristic point tracking method and device | |
Hadviger et al. | Feature-based event stereo visual odometry | |
EP3035242A1 (en) | Method and electronic device for object tracking in a light-field capture | |
CN113298707B (en) | Image frame splicing method, video inspection method, device, equipment and storage medium | |
CN110930436B (en) | Target tracking method and device | |
Ma et al. | Depth assisted occlusion handling in video object tracking | |
CN112802112B (en) | Visual positioning method, device, server and storage medium | |
CN115294493A (en) | Visual angle path acquisition method and device, electronic equipment and medium | |
Yu et al. | Accurate motion detection in dynamic scenes based on ego-motion estimation and optical flow segmentation combined method | |
JP2009116686A (en) | Imaging target detection apparatus and method | |
CN113033551A (en) | Object detection method, device, equipment and storage medium | |
Zhang et al. | A traffic flow detection system combining optical flow and shadow removal | |
JP2000011134A (en) | Device and method for detecting moving amount | |
Chan et al. | Image deblurring using a hybrid optimization algorithm | |
Pushpa et al. | Multiple object detection and tracking in cluttered region with rational mobile area | |
He et al. | Dynamic Objects Detection Based on Stereo Visual Inertial System in Highly Dynamic Environment |
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 |