CN106342332B - Target following keeping method when switch visual field under airborne moving condition - Google Patents

Abstract

The present invention is target following keeping method when switch visual field under airborne moving condition, the method is to convert infrared the optical signals of extraneous scenery or TV sensor to the signal of telecommunication, after image acquisition, become digital image array discrete on time-space domain, calculated again the interframe side-play amount of institute's tracking target by computer, be sent to following control system, by following control system according to this side-play amount driver framework, and then drive is arranged on the infrared or TV sensor on framework, target is withdrawn on the visual axis of infrared or TV, it is characterized in that: image is processed with the step of servo-actuated control procedure and is: (1) determines target best match position, (2) image processing is switched in visual field. the utilization of the method, has solved the problem of visual field track rejection while switching effectively, provides powerful guarantee for the accurate tracking of Onboard searching system strikes target.

Description

Target following keeping method when switch visual field under airborne moving condition

Technical field

The present invention is target following keeping method when switch visual field under airborne moving condition, relates to targetThe target detection that message handler changes visual field.

Background technology

Traditional Search tracking system, the single visual field of some employings, visual field need not be switched, but like thisObservable visual field can reduce; And first get To Template during for double-view field system keeps track, switch visual fieldAfter from new crawl To Template, the problem of its existence is that (Δ is t) blink while switching because of tracking field of viewInside there is no image information, and easily make track rejection.

In airborne circumstance, as shown in Figure 1, the infrared or TV sensor of the optical signals of extraneous sceneryConvert the signal of telecommunication to, after image acquisition, become digital image array discrete on time-space domain, then byComputer calculates the interframe side-play amount of institute's tracking target, is sent to following control system, by follow-up control systemUnite according to this side-play amount driver framework, and then drive the infrared or TV sensor being arranged on framework,Target is withdrawn on the visual axis of infrared or TV, thereby completes tracing task.

Summary of the invention

The present invention designs and provides a kind of airborne for the problem existing in art methods justWhen under moving condition, switch visual field, target following keeping method, its objective is by image correlation and follows the tracks ofAlgorithm solves under airborne moving condition visual field track rejection problem while switching.

The object of the invention is to realize by following measures:

Target following keeping method when switch visual field under this kind of airborne moving condition, the method be by outsideInfrared or the TV sensor of the optical signals of boundary's scenery converts the signal of telecommunication to, after image acquisition, becomesDiscrete digital image array on time-space domain, then be offset by the interframe that computer calculates institute's tracking targetAmount, be sent to following control system, by following control system according to this side-play amount driver framework, and then bandMove the infrared or TV sensor being arranged on framework, target is withdrawn into the optical axis of infrared or TVOn line, it is characterized in that:

Image is processed with the step of servo-actuated control procedure:

(1) determine target best match position

Target setting template is T, and size is N × N, in sequence image fn (x, y), gets the field of searchS, n is frame number, S size in the field of search is M × M, M > N, template T within the scope of the S of the field of search byRow is sliding window search by column, to each sliding window position, correspondence position in seek template T and field of search SThe calculation of correlation of image, calculation of correlation be made as D (Δ x, y), computing formula is as follows for Δ:

$D(\mathrm{\Δ}x,\mathrm{\Δ}y)=\underset{i,j\∈T}{\Σ}abs\left(T\right(i,j)-S(i+\mathrm{\Δ}x,j+\mathrm{\Δ}y\left)\right)$

(Δ x, Δ is y) sliding window position, 0≤Δ x, Δ y≤M-N+1;

D (Δ x, Δ is y) position (Δ x, Δ sliding window y) and the calculation of correlation of template T;

T is To Template, and T (i, j) is the gray value of the pixel that in To Template T, position (i, j) locates, 0≤i，j≤N+1；

(i+ Δ x, j+ Δ is y) (i+ Δ x, the pixel that j+ Δ y) is located of position in the field of search to SGray value, 0≤i+ Δ x, j+ Δ y≤M+1;

D (Δ x, Δ y) the corresponding sliding window position of minimum of a value (Δ x, Δ y), is target the bestMatched position;

(2) image processing is switched in visual field

In the time that tracking field of view switches, (Δ does not have image information, requirement in t) to there will be blinkImage tracking algorithm suspends in section at this moment, waits visual field to switch complete while again having image informationContinue to maintain the tracking to target. But the scenery in visual field will expand (wide visual field becomes narrow visual field)Or dwindle (narrow visual field becomes wide visual field), will lose if carry out relevant matches with former To Template againLose target, this just need to zoom in or out conversion in proportion to former To Template, new to adapt toThe variation of scene size, follows the tracks of the continuation of original target thereby maintain.

If fresh target template is T ', by near (N/k) * (N/k) old To Template T centerRegional percentage be amplified to the fresh target template T ' of N*N, taking the center of old To Template as coordinate is formerPoint, transformation relation is as follows:

T′(x，y)＝T(x/k，y/k)

Wherein 0≤x < N, 0≤y < N, k is visual field multiplication factor, by the field size of two visual fieldsRatio obtain.

(3), in the time that tracking field of view switches, it is last that servomechanism keeps the angular speed of current each frameworkMoment is calculated the angular speed of gained, does not lose for target after visual field is switched, and requires at N (N >=ΔT) in time, framework should keep the angular speed before switching, and the value of N depends on the running software cycle of controlling.Interrupt for t if control software timing, 2 Δ t >=N >=Δ t, occurrence is definite by testing.

Image tracking algorithm is exactly will in sequence image, find and the optimum Match position of To TemplatePut, thereby realize the tracking to target. First get a To Template, then in subsequent image framesNear target location, set up the field of search, by To Template, in the field of search, sliding window is relevant line by lineFind target best match position.

In the design of carrying out in correlation tracking algorithm adaptation visual field switching, calculate with traditional correlation trackingThe processing method difference of method. Tradition correlation tracking algorithm is taked first to get during for double-view field system keeps trackTo Template, captures the scheme of To Template after waiting visual field to switch again, easily causes target to followThere is larger skew in track point. In the present invention, we have adopted the To Template to correlation tracking algorithmThe method of carrying out To Template transformation of scale according to visual field variation ratio, has retained former target mould completelyThe information of plate, therefore the skew of tracking position can not caused in visual field after switching.

Brief description of the drawings

Fig. 1 is image tracking system theory diagram

Fig. 2 is the schematic diagram of track algorithm of the present invention

Fig. 3 is the schematic diagram of To Template conversion of the present invention

Fig. 4 is the software flow pattern of track algorithm of the present invention

Fig. 5 is the flow chart of the control software of visual field of the present invention switching

Detailed description of the invention

Below with reference to drawings and Examples, technical solution of the present invention is further described:

Shown in accompanying drawing 1, target following maintenance side when switch visual field under this kind of airborne moving conditionMethod, the method is to convert infrared the optical signals of extraneous scenery or TV sensor to the signal of telecommunication, warpAfter image acquisition, become digital image array discrete on time-space domain, then calculate institute by computer and follow the tracks ofThe interframe side-play amount of target, is sent to following control system, is driven according to this side-play amount by following control systemMoving framework, and then drive the infrared or TV sensor being arranged on framework, target is withdrawn into redOn the visual axis of outer or TV, image is processed with the step of servo-actuated control procedure and is:

(1) determine target best match position

Shown in accompanying drawing 2, target setting template is T, and size is N × N, sequence image fn (x,Y) in, get field of search S, n is frame number, and S size in the field of search is M × M, M > N, and template T is in searchSliding window search line by line within the scope of district S, to each sliding window position, T and field of search S seek templateThe calculation of correlation of middle correspondence position image, calculation of correlation be made as D (Δ x, y), computing formula is as follows for Δ:

$D(\mathrm{\&Delta;}x,\mathrm{\&Delta;}y)=\underset{i,j\&Element;T}{\&Sigma;}abs\left(T\right(i,j)-S(i+\mathrm{\&Delta;}x,j+\mathrm{\&Delta;}y\left)\right)$

(Δ x, Δ is y) sliding window position, 0≤Δ x, Δ y≤M-N+1;

D (Δ x, Δ is y) position (Δ x, Δ sliding window y) and the calculation of correlation of template T;

T is To Template, and T (i, j) is the gray value of the pixel that in To Template T, position (i, j) locates, 0≤i，j≤N+1；

(i+ Δ x, j+ Δ is y) (i+ Δ x, the pixel that j+ Δ y) is located of position in the field of search to SGray value, 0≤i+ Δ x, j+ Δ y≤M+1;

D (Δ x, Δ y) the corresponding sliding window position of minimum of a value (Δ x, Δ y), is target the bestMatched position;

Visual field proportionate relationship k is obtained by the ratio of the field size of He Zhai visual field, wide visual field; Cut visual fieldChanging time Δ t obtains by the actual measurement to visual field switching time.

Convert that later fresh target template just can continue to maintain to target after has switched visual fieldFollow the tracks of. In like manner, do similar processing for the situation that is become wide visual field from narrow visual field.

Shown in accompanying drawing 4, the tracing process under the switch instances of visual field is to have controlled by softwareBecome, software adopts C Programming with Pascal Language, the CCS2.1 version that compiling development environment is TI, operationEnvironment is that TMS320C6701 is the signal-processing board of core.

(2) image processing is switched in visual field

Shown in accompanying drawing 3, in the time that tracking field of view switches, (Δ does not have in t) to there will be blinkHave image information, require image tracking algorithm at this moment between section in suspend, wait visual field to switch completeAgain while having image information, continue to maintain the tracking to target. But it is (wide that the scenery in visual field will expandVisual field becomes narrow visual field) or dwindle (narrow visual field becomes wide visual field), if entered with former To Template againLine correlation coupling will lose objects, and this just need to amplify in proportion or contract former To TemplateLittle conversion, to adapt to the variation of new scene size, follows the tracks of the continuation of original target thereby maintain.

If fresh target template is T ', by near (N/k) * (N/k) old To Template T centerRegional percentage be amplified to the fresh target template T ' of N*N, taking the center of old To Template as coordinate is formerPoint, transformation relation is as follows:

T′(x，y)＝T(x/k，y/k)

Wherein 0≤x < N, 0≤y < N, k is visual field multiplication factor, by the field size of two visual fieldsRatio obtain.

(3), in the time that tracking field of view switches, it is last that servomechanism keeps the angular speed of current each frameworkMoment is calculated the angular speed of gained, does not lose for target after visual field is switched, and requires at N (N >=ΔT) in time, framework should keep the angular speed before switching, and the value of N depends on the controlling run cycle, establishesControl Interruption is t, 2 Δ t >=N >=Δ t.

The size of supposing the To Template T of correlation tracking algorithm is 32 × 32, sequence image fn (x,Y) in (n is frame number), getting field of search S size is 64 × 64, and template T is within the scope of the S of the field of searchSliding window search is line by line relevant, and the calculation of correlation D at each sliding window position place (Δ x, y) count by ΔCalculate as follows:

Wherein (Δ x, Δ is y) sliding window position, 0≤Δ x, Δ y≤33.

(Δ x, Δ y) is worth minimum sliding window position, and (Δ x, Δ y) is target best match position to D.

In the time that tracking field of view switches, in time Δ t, there is no image information, correlation tracking algorithm is at this momentBetween section in suspend, at this moment need To Template T to carry out transformation of scale, suppose You Kuan visual field, visual fieldBecome narrow visual field, visual field scenery all expands 3 times (k=3) in the x and y direction, now with targetTemplate center is initial point, does and can obtain new To Template T ' as down conversion:

T′(x，y)＝T(x/3，y/3)

Wherein 0≤x <, 32,0≤y < 32.

In the correlation tracking algorithm of fresh target template after conversion after has switched visual field, replace former orderMark template T, continues to maintain the tracking to target, and the process that switch correlation tracking algorithm adaptation visual field justComplete.

Shown in accompanying drawing 5, visual field handoff procedure has been controlled by software, and software is adoptedUse C Programming with Pascal Language, compiling development environment is IDE386 version, and running environment is 80486 for coreServo-actuated control panel.

Advantage of the present invention and implementation result are;

The utilization of the method, has solved the problem of visual field track rejection while switching, effectively for airborneSearch system accurate tracking strikes target provides powerful guarantee.

Claims (1)

1. target following keeping method when switch visual field under airborne moving condition, the method is by the external worldInfrared or the TV sensor of the optical signals of scenery converts the signal of telecommunication to, after image acquisition, while becomingDiscrete digital image array on spatial domain, then calculated the interframe side-play amount of institute's tracking target by computer,Be sent to following control system, by following control system according to this side-play amount driver framework, and then drive peaceBe contained in the infrared or TV sensor on framework, target be withdrawn on the visual axis of infrared or TV,It is characterized in that: image is processed with the step of servo-actuated control procedure and is:

(1) determine target best match position

Target setting template is T, and size is N × N, in sequence image fn (x, y), gets the field of searchS, n is frame number, S size in the field of search is M × M, M > N, template T within the scope of the S of the field of search byRow is sliding window search by column, to each sliding window position, correspondence position in seek template T and field of search SThe calculation of correlation of image, calculation of correlation be made as D (Δ x, y), computing formula is as follows for Δ:

$D(\mathrm{\&Delta;}x,\mathrm{\&Delta;}y)=\underset{i,j\&Element;T}{\&Sigma;}abs\left(T\right(i,j)-S(i+\mathrm{\&Delta;}x,j+\mathrm{\&Delta;}y\left)\right)$

(Δ x, Δ is y) sliding window position, 0≤Δ x, Δ y≤M-N+1;

D (Δ x, Δ is y) position (Δ x, Δ sliding window y) and the calculation of correlation of template T;

T is To Template, and T (i, j) is the gray value of the pixel that in To Template T, position (i, j) locates, 0≤i，j≤N+1；

(i+ Δ x, j+ Δ is y) (i+ Δ x, the pixel that j+ Δ y) is located of position in the field of search to SGray value, 0≤i+ Δ x, j+ Δ y≤M+1;

D (Δ x, Δ y) the corresponding sliding window position of minimum of a value (Δ x, Δ y), is target the bestMatched position;

(2) image processing is switched in visual field

In the time that tracking field of view switches, former To Template is zoomed in or out to conversion in proportion, establishFresh target template is T ', by the region of (N/k) * (N/k) old To Template T center nearRatio is amplified to the fresh target template T ' of N*N, taking the center of old To Template as the origin of coordinates, becomesThe relation of changing is as follows:

T′(x，y)＝T(x/k，y/k)

Wherein 0≤x < N, 0≤y < N, k is visual field multiplication factor, by the field size of two visual fieldsRatio obtain;

(3), in the time that tracking field of view switches, it is last that servomechanism keeps the angular speed of current each frameworkMoment is calculated the angular speed of gained, does not lose for target after visual field is switched, and requires at N (N >=ΔT) in time, framework should keep the angular speed before switching, and the value of N depends on the controlling run cycle, establishesControl Interruption is t, 2 Δ t >=N >=Δ t.