CN101923717B - Method for accurately tracking characteristic points of quick movement target - Google Patents

Abstract

The invention belongs to video and image processing technology and particularly relates to a method for accurately tracking characteristic points of a quick movement target. The method comprises the following steps of: selecting M characteristic points in a tracked target selection rectangular region at the moment of t-1 according to characteristic matrixes; iterating the M characteristic points at the moment of t by using a KLT luminous flux vector formula to solve the optimal solution so as to acquire the new positions of the current frames of all the characteristic points; and estimating the proximity positions of all the characteristic points by using a checker in a tracked target searching rectangular region if the optimal solution cannot be solved so as to acquire new positions. Due to the adoption of the method, the proximity positions can be found, then the optimal position can be solved by the luminous flux vector calculation formula, and a search region is partitioned by using the checker, so that search time is saved greatly, and the proximity positions with approximate gray scale variation to that of the tracked target can be searched. The method avoids the situation of losing the tracking of the characteristic points and is very suitable for tracking the characteristic points of the quick movement target.

Description

A kind of method of the accurately tracking characteristic points to fast-moving target

Technical field

The invention belongs to video and image processing techniques, be specifically related to a kind of method of the accurately tracking characteristic points to fast-moving target.

Background technology

In image/video post-processed software, the pixel characteristic zone of moving image to be followed the tracks of, tracking data can be used for controlling the motion and the stable motion object of other object, and this has demand widely.Use KLT light stream feature point tracking method to follow the tracks of unique point, KLT light stream feature point tracking algorithm is selected a group of feature point usually in reference frame, suppose that the unique point texture remains unchanged in interframe, accomplishes tracing task through local match search then.The quadratic sum (SSD) of Kanade-Lucas-Tomasi (KLT) algorithm use gradation of image difference is as the matching criterior of unique point.Concrete KLT feature point tracking algorithm can be consulted: B_D_Lucas and T_Kanade_An iterative image registration technique with an application tostereo vision_IJCAI_1981.

But, when using existing KLT light stream feature point tracking algorithm that fast target is followed the tracks of, be easy to cause feature point tracking to be lost because that brightness changes is discontinuous.Therefore, be necessary the feature point tracking method of existing fast-moving target is improved.

Summary of the invention

The objective of the invention is to the defective to prior art, a kind of method of the accurately tracking characteristic points to fast-moving target is provided, thereby realizes correct tracking the high-speed moving object unique point.

Technical scheme of the present invention is following: a kind of method of the accurately tracking characteristic points to fast-moving target comprises the steps:

(1) being engraved in tracking target during t-1 selects in the rectangular area according to eigenmatrix $\left[\begin{array}{cc}{\∫\∫}_w{g}_x^2& {\∫\∫}_w{g}_x{g}_y\\ {\∫\∫}_w{g}_x{g}_y& {\∫\∫}_w{g}_y^2\end{array}\right]$ Select M unique point, g in the matrix _xPresentation video is at the shade of gray of directions X, g _yPresentation video is at the shade of gray of Y direction,

Be illustrated in tracking target select the rectangular area discrete data with;

(2) use KLT light stream vector formula iteration to try to achieve optimum solution to M unique point constantly at t, obtain the reposition of all unique points at present frame;

(3) if in step (2), tried to achieve optimum solution, then change step (5) over to; Otherwise expression has certain unique point to lose, and gets into step (4);

(4) in tracking target search rectangular area, use gridiron pattern that all unique points are carried out apparent position and estimate that obtain reposition, concrete grammar is following:

Tracking target is searched for the rectangular area carry out the gridiron pattern division; Each lattice is identical with the unique point area size; Have K image-region like this; Ask in this K zone the minimum zone of quadratic sum of unique point area image gray scale difference constantly, get into step (2) with this regional center again as the iteration initial position of light stream vector formula and calculate with t-1;

(5) make t=t+1, return step (2), carry out cycling.

Further, the method for aforesaid accurately tracking characteristic points to fast-moving target, wherein, the number M of unique point gets 40.

Further, the method for aforesaid accurately tracking characteristic points to fast-moving target, wherein, the KLT light stream vector formula described in the step (2) is: Zd=e,

Wherein, $Z={\∫\∫}_{w}g\left(\stackrel{\→}{X}\right)g^T\left(\stackrel{\→}{X}\right)W\left(\stackrel{\→}{X}\right)d\stackrel{\→}{X},$

$e={\∫\∫}_w[I\left(\stackrel{\→}{X}\right)-J\left(\stackrel{\→}{X}\right)]g\left(\stackrel{\→}{X}\right)W\left(\stackrel{\→}{X}\right)d\stackrel{\→}{X},$

$g={[\frac{\∂}{\∂x}\left(\frac{I+J}{2}\right),\frac{\∂}{\∂y}\left(\frac{(I+J)}{2}\right)]}^T,$

d＝[dx，dy] ^T，

J, I represent the luminance function of t-1 and t two width of cloth images constantly respectively;

representation feature kinematic parameter, the displacement of d representation feature point.

Beneficial effect of the present invention is following: feature point tracking method provided by the present invention is the improvement of existing KLT light stream feature point tracking method; If unique point is present in the present image; Method provided by the present invention can find apparent position earlier, finds the solution the optimum position through the light stream vector computing formula again; Use gridiron pattern to divide the region of search, can improve greatly and save search time, can search the apparent position approaching again with the tracking target grey scale change.This method can the situation that feature point tracking is lost not occur because of the discontinuous of brightness of image variation, is highly suitable for the feature point tracking of fast-moving target.

Description of drawings

Fig. 1 is a method flow diagram of the present invention.

Embodiment

Below in conjunction with accompanying drawing and specific embodiment the present invention is carried out detailed description.

The method of the accurately tracking characteristic points to fast-moving target provided by the present invention relates to confirming of two zones; Selecting the rectangular area for tracking target for one is used for confirming in which scope of image, to produce unique point; It is that the user selects that tracking target is selected the rectangular area, such as a human face region; Another is used for confirming in which scope, to search for the unique point of losing in advance for tracking target search rectangular area, and tracking target search rectangular area generally is the entire image zone.The idiographic flow of this method is as shown in Figure 1, may further comprise the steps:

(1) being engraved in tracking target during t-1 selects in the rectangular area according to eigenmatrix $\left[\begin{array}{cc}{\∫\∫}_w{g}_x^2& {\∫\∫}_w{g}_x{g}_y\\ {\∫\∫}_w{g}_x{g}_y& {\∫\∫}_w{g}_y^2\end{array}\right]$ Select M unique point.g _xPresentation video is at the shade of gray of directions X, g _yPresentation video is at the shade of gray of Y direction.

be illustrated in tracking target select the rectangular area discrete data with.Eigenwert according to eigenmatrix draws is selected optimum M unique point (eigenwert is excellent more greatly more), and the unique point number M gets 40 among the present invention.

(2) use KLT light stream vector formula iteration to try to achieve optimum solution to M unique point constantly at t, obtain the reposition of all unique points at present frame.

KLT light stream vector formula is: Zd=e,

Wherein, $Z={\∫\∫}_{w}g\left(\stackrel{\→}{X}\right)g^T\left(\stackrel{\→}{X}\right)W\left(\stackrel{\→}{X}\right)d\stackrel{\→}{X},$

$e={\∫\∫}_w[I\left(\stackrel{\→}{X}\right)-J\left(\stackrel{\→}{X}\right)]g\left(\stackrel{\→}{X}\right)W\left(\stackrel{\→}{X}\right)d\stackrel{\→}{X},$

$g={[\frac{\∂}{\∂x}\left(\frac{(I+J)}{2}\right),\frac{\∂}{\∂y}\left(\frac{(I+J)}{2}\right)]}^T,$

d＝[dx，dy] ^T，

J, I represent the luminance function of t-1 and t two width of cloth images constantly respectively;

representation feature kinematic parameter; Such as two dimensional motion is (x; Y), the displacement of d representation feature point.

Approach through the Newton process of iteration and to try to achieve optimum solution.

(3) if can't try to achieve optimum solution in step (2); Expression has certain unique point to lose; If this unique point zone might also exist in the present image; Just since object of which movement make soon shade of gray discontinuous cause the light stream vector formula can't converge to optimum solution, at this time can attempt in tracking target search rectangular area, searching for the apparent position of tracking target.

Tracking target is searched for the rectangular area carry out the gridiron pattern division; Each lattice is identical with the unique point area size; Have K image-region like this; Ask in this K zone the minimum zone of quadratic sum (SSD) of unique point area image gray scale difference constantly, get into step (2) with this regional center again as the iteration initial position of light stream vector formula and calculate with t-1.So divide gridiron pattern; Theory is if tracking characteristics point exists in this time chart picture; We can find a zone and t-1 similarity to be arranged the unique point zone constantly on grey scale change, carry out vector tracking with this zone like this and must iterate to best match position.Formulate with concrete is following:

$S(x,y)=\left({\∫\∫}_w\right|(J\left(X\right)-I\left(X\right))\left|\right)$

$(x_t,y_i)=\min \left(S(x,y)\right),(x,y)\⋐\mathrm{searchregion}$

S representes the brightness of this position and the luminance difference of template in the above-mentioned formula, and J, I represent the luminance function of t-1 and t two width of cloth images constantly respectively.

(4) if still do not find unique point, characterization point does not exist, and just abandons this unique point, otherwise unique point is noted at t reposition constantly, returns to the user.

(5) at t+1 constantly, make t=t+1, return step (2), carry out cycling.

Method of the present invention is not limited to the embodiment described in the embodiment, and those skilled in the art's technical scheme according to the present invention draws other embodiment, belongs to innovation scope of the present invention equally.

Claims (3)

1. the method to the accurately tracking characteristic points of fast-moving target comprises the steps:

(1) being engraved in tracking target during t-1 selects in the rectangular area according to eigenmatrix $\left[\begin{array}{cc}{\∫\∫}_w{g}_x^2& {\∫\∫}_w{g}_x{g}_y\\ {\∫\∫}_w{g}_x{g}_y& {\∫\∫}_w{g}_y^2\end{array}\right]$ Select M unique point, g in the matrix _xPresentation video is at the shade of gray of directions X, g _yPresentation video is at the shade of gray of Y direction,

Be illustrated in tracking target select the rectangular area discrete data with;

(2) use KLT light stream vector formula iteration to try to achieve optimum solution to M unique point constantly at t, obtain the reposition of all unique points at present frame;

(3) if in step (2), tried to achieve optimum solution, then change step (5) over to; Otherwise expression has certain unique point to lose, and gets into step (4);

(4) in tracking target search rectangular area, use gridiron pattern that all unique points are carried out apparent position and estimate that obtain reposition, concrete grammar is following:

Tracking target is searched for the rectangular area carry out the gridiron pattern division; Each lattice is identical with the unique point area size; Have K image-region like this; Ask in this K zone the minimum zone of quadratic sum of unique point area image gray scale difference constantly, get into step (2) with this regional center again as the iteration initial position of light stream vector formula and calculate with t-1;

(5) make t=t+1, return step (2), carry out cycling.

2. the method for the accurately tracking characteristic points to fast-moving target as claimed in claim 1, it is characterized in that: the number M of unique point gets 40.

3. the method for the accurately tracking characteristic points to fast-moving target as claimed in claim 1, it is characterized in that: the KLT light stream vector formula described in the step (2) is: Zd=e,

Wherein, $Z={\∫\∫}_{w}g\left(\stackrel{\→}{X}\right)g^T\left(\stackrel{\→}{X}\right)W\left(\stackrel{\→}{X}\right)d\stackrel{\→}{X},$

$e={\∫\∫}_w[I\left(\stackrel{\→}{X}\right)-J\left(\stackrel{\→}{X}\right)]g\left(\stackrel{\→}{X}\right)W\left(\stackrel{\→}{X}\right)d\stackrel{\→}{X},$

$g={[\frac{\∂}{\∂x}\left(\frac{(I+J)}{2}\right),\frac{\∂}{\∂y}\left(\frac{(I+J)}{2}\right)]}^T,$

d＝[dx，dy] ^T，

J, I represent the luminance function of t-1 and t two width of cloth images constantly respectively;

representation feature kinematic parameter, the displacement of d representation feature point.

Images