CN102004898A - Target tracking method based on template matching - Google Patents

Abstract

The invention belongs to the field of image processing and machine vision, in particular relates to a target tracking method based on the template matching. The template image and pixels with a matching area are arrayed into a plurality of sub windows according to the circle; and the target is guaranteed to have the translation and the rotational invariance by adopting the circular template matching principle. The edge strength value of the template and each pixel in the tracking window can be calculated by using a Kirsch operator. The summation of the gray matching valve and the edge strength matching value is used as the matching result. The position of the optimal matching value is determined as the position of the tracking target. The invention can be used for a target identifying and tracking system.

Description

A kind of method for tracking target based on template matches

Technical field

The invention belongs to Flame Image Process and field of machine vision, relate to a kind of target recognition and tracking method, particularly a kind of image object recognition and tracking method based on template matches.

Background technology

Identification of targets is the important research content of machine vision and image processing field with following the tracks of always.There is important use to be worth at aspects such as supervisory system, safety-protection system, military fields.Early stage track algorithm mainly is based on the contrast track algorithm of target and background.The contrast track algorithm utilizes the contrast between target and its background to discern and extract echo signal, is the tracker class methods that grow up the earliest, can be divided into edge tracking, centre of form tracking, peak time tracking etc. according to the difference to the target reference point.These class methods are simple, are easy to realize that processing speed is fast, but resists in disturbing ability.Along with electronic technology and development of computer, the image relevant matches is applied in tracker.

Correlation tracking is that the benchmark image with system is chosen to be template, with template in track window with different off-set value displacements, calculation template overlaps the correlation of area image then, and the optimum matching zone is defined as tracking position of object.For example, the absolute average algorithm of gray scale difference is exactly a kind of simple correlation matching algorithm.But these class methods only have translation invariance usually, and do not have rotational invariance.This makes algorithm can't accurately orient the target location sometimes, thereby has influenced the range of application of algorithm.

Therefore design a kind of novel method for tracking target and have important use value.

Summary of the invention

Technical matters to be solved by this invention is, designs a kind of target recognition and tracking method, realizes gray level image identification of targets and tracking.

The technical solution adopted in the present invention is: a kind of target recognition and tracking method, template size is elected n * n (n is an odd number) as, pixel in the template is the center of circle with the template center, form (n+1)/2 subwindow by circular arrangement, first subwindow is unique point itself, second subwindow then is 3 * 3 the window at center for unique point ....Adopt the Kirsch operator to ask for the edge intensity value computing of template pixel and track window pixel respectively.The matching criterior function is made up of two parts, first be template and each subwindow pixel grey scale of zone to be matched and difference, second portion be template and each subwindow pixel edge intensity of zone to be matched and difference.Two absolute values be defined as the target location with minimum position.

The objective of the invention is to propose a kind of target recognition and tracking method, adopt circular shuttering to guarantee that algorithm has rotational invariance, in the matching criterior function, increase the matching value of edge strength, utilize the detailed information of image to improve the accuracy of Target Recognition, finish gray level image identification of targets and tracking.

Description of drawings

Fig. 1 is a Kirsch operator template.

Fig. 2 is a pixel grey scale neighborhood window.

Fig. 3 is the target following result.

Embodiment

Below in conjunction with embodiment and accompanying drawing the present invention is described in further detail.

The present invention adopts the Kirsch operator to ask for the edge intensity value computing of each pixel in template image and the track window.The Kirsch operator is made up of 83 * 3 template windows, each template represent respectively a specific detection side to, its template operator is as shown in Figure 1.

If (x y) is the pixel of edge intensity value computing to be calculated to P, and the intensity profile of its 8 neighborhood as shown in Figure 2.If q _k(k=0,1 ..., 7) be the edge strength on the k direction that obtains after image is handled through k Kirsch operator template, following recurrence relation formula is then arranged:

q ₀＝5(p ₄+p ₅+p ₆)-3(p ₀+p ₁+p ₂+p ₃+p ₇) (1)

$q_i=\left\{\begin{array}{c}{q}_{i-1}+8(p_{8-i}-p_{3-i}),i=\mathrm{1,2,3}\\ q_{i-1}+8(p_{8-i}-p_{11-i}),i=\mathrm{4,5,6,7}\end{array}\right.---\left(2\right)$

P (x, y) pixel through the edge strength S that obtains this point after the Kirsch algorithm process (x y) is:

S(x，y)＝max{q _k}，k＝0，1，2，...，7 (3)

The object matching criterion function is defined as:

D(A，B)＝D ₁+D ₂ (4)

D wherein ₁Be gray scale matching value, D ₂Be the edge strength matching value.

If (h, q) grey scale pixel value of position is a to template image _{H, q}, (h, q) grey scale pixel value of position is b to mate the district in the track window _{H, q}By circular arrangement, then the gray scale matching value of i round interior pixel is e with image to be matched in template image and the tracking area _i, this value is asked for by following formula:

$e_i=\underset{h=1}{\overset{(2i-1)}{\mathrm{\Σ}}}\underset{q=1}{\overset{(2i-1)}{\mathrm{\Σ}}}(a_{h,q}-b_{h,q})---\left(5\right)$

The gray scale matching value D in template and district to be matched ₁Ask for by following formula:

$D_1=\underset{i=1}{\overset{(n+1)/2}{\mathrm{\Σ}}}{w}_i\left|e_i\right|---\left(6\right)$

W wherein _iRepresent the contribution weights of different subwindow interior pixels to the gray scale matching value.

Because the gray scale matching value is the difference sum of the subwindow interior pixel gray-scale value of template image and district to be matched image, this detailed information to image has played certain filtration, thereby influences the accuracy of matching result.In order in matching process, to utilize detailed information such as edge of image, in the matching criterior function, increased edge strength occurrence D ₂Asking for and gray scale matching value D of this ₁Method similar.

If (h, q) the pixel edge intensity level of position is a to template image ^s _{H, q}, (h, q) the pixel edge intensity level of position is b to mate the district in the track window ^s _{H, q}Still by circular arrangement, then the edge strength matching value of i round interior pixel is e with image to be matched in template image and the tracking area ^s _i, this value is asked for by following formula:

$e_i^s=\underset{h=1}{\overset{(2i-1)}{\mathrm{\Σ}}}\underset{q=1}{\overset{(2i-1)}{\mathrm{\Σ}}}(a_{h,q}^s-b_{h,q}^s)---\left(7\right)$

The gray scale matching value D in template and district to be matched ₂Ask for by following formula:

$D_2=\underset{i=1}{\overset{(n+1)/2}{\mathrm{\Σ}}}{w}_i\left|e_i^s\right|---\left(8\right)$

W wherein _iRepresent the contribution weights of different subwindow interior pixels to edge strength matching value.

Template image is moved in tracking area, and the zone of matching criterior function minimum is defined as the position of target.

Embodiment

Fig. 3 has provided and has adopted the inventive method to seek the situation of target location in a few width of cloth consecutive images.The target's center position of cross position among the figure for adopting the inventive method to try to achieve, the target location that round dot is tried to achieve for the absolute average algorithm that adopts gray scale difference.By sample result as can be seen, the inventive method can be carried out comparatively accurate in locating tracking to the target in the consecutive image.Environmental background comparatively complexity or target taken place under the situation that translation, rotation etc. change, the inventive method is still effective.

Claims (6)

1. the method for tracking target based on template matches is characterized in that best match position is defined as tracking position of object according to the image calculation matching value in coupling district in trace template image and the track window.

2. the method for tracking target based on template matches according to claim 1 is characterized in that, the pixel of template image and district to be matched image is all pressed circular arrangement, has rotational invariance to guarantee algorithm.

3. the method for tracking target based on template matches according to claim 1 is characterized in that, adopts the edge intensity value computing of pixel in Kirsch operator calculation template and the track window.

4. the method for tracking target based on template matches according to claim 1 is characterized in that the matching criterior function is made up of two parts, and a part is the gray scale matching value, and another part is the edge strength matching value.

5. the method for tracking target based on template matches according to claim 1 is characterized in that, subwindow gray scale matching value is tried to achieve by the difference sum of template image and the corresponding subwindow interior pixel of district to be matched image gray-scale value.

6. the method for tracking target based on template matches according to claim 1 is characterized in that, subwindow edge strength matching value is tried to achieve by the difference sum of template image and the corresponding subwindow interior pixel of district to be matched image edge intensity value computing.

Images