CN104616322A - Onboard infrared target image identifying and tracking method and device - Google Patents

Abstract

The invention discloses an onboard infrared target image identifying and tracking method and an onboard infrared target image identifying and tracking device. The identifying and tracking device comprises an up-working controller, a servo control slave computer, a rotating holder, a video gathering card, a searching lamp, an infrared camera, and a tracking control software. Through the up-working controller and the tracking control software, identifying and tracking commands of a moving object are generated and the rotating angle of the holder is controlled, the control commands of the up-working controller are received by the servo control slave computer so that the holder is driven to rotate with a certain angle to track the object, the searching lamp and the infrared camera are fixed to the holder, and objective infrared images are gathered by the video gathering card via the infrared camera and uploaded to the up-working controller so that the tracked object can be in real time identified by the searching lamp. According to the method and the device, the infrared moving objects can be extracted, tracked and locked, and the moving object can be quickly extracted, the tracked object cannot be lost and the rotation of the holder can be in advance estimated, thus the method and the device can be widely used for identifying and warning the moving objects in the port and offshore, for early warning.

Description

Boat-carrying infrared target image identification tracking and device thereof

Technical field

The present invention relates to a kind of boat-carrying infrared target image identification tracking and device thereof.

Background technology

Sea moving object detection, recognition and tracking technology have important purposes in occasions such as coast defence safety, customs management, marine anti-smuggling and harbour allocation of ships.Based on the target travel detection and tracking of infrared imaging, all have broad application prospects in dual-use field.But having that motion detects, the high-intelligentization supervisory system of Target Segmentation and target following must relate to image operation consuming time in a large number simultaneously, being difficult to well take into account reliability and real-time under active computer processing speed simultaneously.The technical barrier of current existence is mainly: the target recognition and tracking 1) in complicated dynamic background situation; 2) hysteresis quality of cloud platform rotation and Video processing time delay cannot meet real-time; 4) the blocking and again search for of moving target.These problems all constrain the application in practice of sea Algorithms for Moving Target Recogntion above.

Summary of the invention

The technical problem to be solved in the present invention is to provide a kind of boat-carrying infrared target image identification tracking means.

The another one technical matters that the present invention will solve is to provide a kind of boat-carrying infrared target image identification tracking.

For boat-carrying infrared target image identification tracking means, the technical solution used in the present invention is: comprise upper industrial computer, servocontrol slave computer, rotary head, video frequency collection card, search lamp, infrared camera;

Upper industrial computer is connected with servocontrol slave computer, and servocontrol slave computer is connected with rotary head and controls rotary head makes all-direction rotation; Rotary head is fixedly equipped with search lamp, infrared camera, infrared camera is connected with upper industrial computer by video frequency collection card, and search lamp is synchronous with the sensing of infrared camera.

As preferably, video frequency collection card is connected with upper industrial computer by USB serial ports.

As preferably, search lamp is high-power sodium vapor lamp.

For boat-carrying infrared target image identification tracking, the technical solution used in the present invention comprises the following steps:

(1) motion estimate

First infrared camera is utilized to carry out image sequence acquisition, obtain the original image of target, target image signal is sent in industrial computer, operator selectes target to be locked, by industrial computer as CPU (central processing unit), Tracking Control Software adopts corresponding image processing method to process to image respectively, extracts comparatively complete target to be tracked from original image; Frame differential method is used to judge whether there is moving target in video sequence, and determine the position of target, frame differential method adopts the difference based on pixel grey scale to two two field pictures before and after a Minimum-time interval, then extracts the moving region in image with thresholding method, and algorithm is:

$M(x,y,t)=\left\{\begin{array}{c}1,f(x,y,t)-f(x,y,t-\mathrm{\Δt})>\mathrm{\τ}\\ 0,f(x,y,t)-f(x,y,t-\mathrm{\Δt})\≤\mathrm{\τ}\end{array}\right.$

In above formula, Δ t is inter frame temporal interval, f (x, y, t) and f (x, y, t-Δ t)) be respectively the picture frame of t and t-Δ t; If the absolute difference of two two field picture respective pixel gray scales is greater than threshold tau, then shows that this pixel is sport foreground, then its value is put 1, otherwise for static background, its value is set to 0, bianry image M (x, y, t) can be obtained;

(2) tracking of moving target

Adopt CamShift algorithm to carry out motion target tracking, algorithm is as follows:

The first step, calculates the 1D histogram of H passage (color channel) component in image HSI space;

Second step, utilizes this 1D histogram that former figure is reconstructed into 2D probability distribution graph;

3rd step, calculates the center of target area;

4th step, utilizes CamShift algorithm, and continuous translation adjustment window center overlaps to target barycentric;

5th step, by the window size of previous frame and center, as the initial value of next frame CamShift algorithm search window;

6th step, continues CamShift computing in the next frame;

(3) rotary head rotates and controls

Kalman filtering algorithm is adopted to carry out the prediction of moving target state, to realize the rotation in advance of rotary head; Target subsequent time position can be turned in advance according to prediction rotary head, adopt Kalman filtering algorithm to carry out the motion state of predicted motion target, utilize the recurrence estimation ability of filtering to predict subsequent time target location; System noise and observation noise are considered as white Gaussian noise, measure as systematic perspective using the position of moving target, with institute's syllabus target location estimation value for exporting, moving target current motion state being made and estimates and Future movement state is made prediction;

State equation and the observation equation of Target Tracking System system are respectively:

$\left\{\begin{array}{c}x(k+1)=\mathrm{\φx}\left(k\right)+w\left(k\right)\\ z\left(k\right)=\mathrm{Hx}\left(k\right)+v\left(k\right)\end{array}\right.$

In above formula: x (k) ∈ R ⁶for state vector, represent the position of moving target, speed and acceleration, its expression formula is x (k)=[x y x _vy _vx _ay _a] ^t; Z (k) ∈ R ²for observation vector, represent the position of moving target; W (k) ∈ R ⁶it is systematic procedure noise vector; V (k) ∈ R ⁶it is observation noise vector; φ ∈ R ^{6 × 6}be state-transition matrix, its expression formula is:

$\mathrm{\φ}=\left[\begin{array}{cccccc}1& 0& T& 0& T^2/2& 0\\ 0& 1& 0& T& 0& T^2/2\\ 0& 0& 1& 0& T& 0\\ 0& 0& 0& 1& 0& T\\ 0& 0& 0& 0& 1& 0\\ 0& 0& 0& 0& 0& 1\end{array}\right]$

H is observing matrix, and need the position observing moving target here, its expression formula is

$H=\left[\begin{array}{cccccc}1& 0& 0& 0& 0& 0\\ 0& 1& 0& 0& 0& 0\end{array}\right]$

System noise and observation noise are mutual incoherent white Gaussian noise;

Employing time equal interval sampling is for the target following of infrared camera; When observing a certain moment target location, pre-estimation in time the most probable position that next moment target occurs can be gone out, so that infrared camera adjusts angle in time, quick tracking target; When known system original state, obtain x (1/1) according to the obserred coordinate value Kalman filtering of current kinetic target, and then prediction x (2/1), concrete Kalman filter equation is as follows:

x(k/k)＝x(k/k-1)+K(k)(z(k)-Hx(k/k-1))

x(k/k-1)＝φx(k-1/k-1))

K(k)＝P(k/k-1)H ^T(HP(k/k-1)H ^T+R(k)) ^-1

P(k/k-1)＝φP(k-1/k-1)φ ^T+Q(k-1)

P(k/k)＝(I-K(k)H)P(k/k-1)

In various above, k=1,2,3 ... x (k/k-1) and P (k/k-1) is status predication value and minimum prediction Square Error matrix thereof, K (k) is Kalman filtering gain matrix, and z (k) ceases for observation is new, and x (k/k) and P (k/k) is the state estimation and error co-variance matrix thereof revised;

Obtain the position of target during k moment, then can obtain the position angle of target according to positional information, according to kalman filter forecasting moving target after the position in k+1 moment, the position angle in k+1 moment can be calculated to obtain equally, Δ θ=θ _k+1-θ _k, the rotational angular velocity of known rotary head, therefore the control to rotary head can be reached by the rotation time arranging rotary head; When moving target enters the observation scope of infrared camera, infrared camera, can according to the coordinate (x, y) of each moment moving target of these information acquisitions after photographing image; Before not obtaining new measured value z (k) in k moment, the estimation can only made from the k-1 moment signal, estimates status signal x (k) in k moment; After the present co-ordinate position recording k moment target, just can obtain the time of day optimal estimation value x (k|k) of target; Corrected Calculation is carried out to predicted value and goes out the observed reading x (k+1|k) of target at k-1 moment most probable position, the deflection angle of target can be obtained according to coordinate position, control the rotation time of rotary head according to deflection angle.

The invention has the beneficial effects as follows:

The present invention is showing on the basis of image in real time, and real-time detection and tracking moving target, detection and tracking have higher robustness to the complex background under marine environment and all kinds of disturbance factor.In tracing process, even if target keeps static, program also can not lose the tracking to target.By hysteresis quality and the Video processing time delay of kalman filtering algorithm process cloud platform rotation, host computer sends steering order by serial ports to The Cloud Terrace, The Cloud Terrace accepts the instruction that host computer sends, camera and search lamp is driven to turn to target location, target is made to be in the immediate vicinity of shooting screen all the time, thus the Real time identification realized sea moving target and tracking.

Accompanying drawing explanation

Below in conjunction with the drawings and specific embodiments, the present invention is further detailed explanation.

Fig. 1 is the structural representation of boat-carrying infrared target image identification tracking means embodiment of the present invention.

Fig. 2 is the Motion estimation software block diagram of the employing Kalman filter of boat-carrying infrared target image identification tracking embodiment of the present invention.

Mark in Fig. 1: the upper industrial computer of 1-, 2-servocontrol slave computer, 3-rotary head, 4-infrared camera, 5-searches for lamp, 6-video frequency collection card.

Embodiment

Fig. 1 is a kind of boat-carrying infrared target image identification tracking means, is made up of upper industrial computer 1, servocontrol slave computer 2, rotary head 3, video frequency collection card 6, search lamp 5, infrared camera 4 and corresponding Tracking Control Software.

Upper industrial computer 1 is connected with servocontrol slave computer 2, and servocontrol slave computer 2 is connected with rotary head 3 and controls rotary head makes all-direction rotation.Search lamp 5, infrared camera 4 rotary head 3 be fixedly equipped with, wherein search for lamp 5 and adopt high-power sodium vapor lamp, and the top of thermal camera 4 is supported on support, make infrared camera 4 and search lamp 5 can make omnibearing synchronous axial system along with rotary head, and it is synchronous with the sensing of infrared camera to search for lamp.Infrared camera 4 is connected with upper industrial computer 1 by video frequency collection card 6, and wherein video frequency collection card 6 is connected with upper industrial computer by USB serial ports.

Whole tracing process comprise image collection, image procossing target localization, target prediction, adjust The Cloud Terrace according to side-play amount and complete tracking.Upper industrial computer and Tracking Control Software complete the identification of moving target and trace command generates, control the anglec of rotation of The Cloud Terrace, servocontrol slave computer receive host computer steering order drive The Cloud Terrace rotate to an angle realize target follow the tracks of, search lamp and infrared camera are fixed on The Cloud Terrace, video frequency collection card gathers Infrared Targets image by infrared camera, and uploading to upper industrial computer, search lamp realizes the Real time identification to tracking target.

Concrete steps are as follows:

(1) motion estimate

First utilize infrared camera to carry out image sequence acquisition, obtain the original image of target, target image signal is sent in industrial computer, and operator selectes target to be locked.By industrial computer as CPU (central processing unit), Tracking Control Software adopts corresponding image processing method to process to image respectively, extracts comparatively complete target to be tracked from original image.Tracking Control Software uses frame differential method to judge whether there is moving target in video sequence, and determines the position of target.

Frame differential method, by adjacent two frames in sequence of video images or multiframe are done calculus of differences to cut down the region that there is not change, extracts the moving target in image.Frame differential method adopts the difference based on pixel grey scale to two two field pictures before and after a Minimum-time interval, then extracts the moving region in image with thresholding method, and algorithm is:

$M(x,y,t)=\left\{\begin{array}{c}1,f(x,y,t)-f(x,y,t-\mathrm{\Δt})>\mathrm{\τ}\\ 0,f(x,y,t)-f(x,y,t-\mathrm{\Δt})\≤\mathrm{\τ}\end{array}\right.$

In above formula, Δ t is inter frame temporal interval, f (x, y, t) and f (x, y, t-Δ t)) be respectively the picture frame of t and t-Δ t.If the absolute difference of two two field picture respective pixel gray scales is greater than threshold tau, then shows that this pixel is sport foreground, then its value is put 1, otherwise for static background, its value is set to 0, bianry image M (x, y, t) can be obtained.

After completing the segmentation of target image, the position of target current time can be obtained from image information, and compare with image center location, when the motion state of known target current time, by Forecasting Methodology, reasonable estimation is made to the target location of target subsequent time and speed, according to the target location side-play amount produced, by serial ports, to rotary head, steering order is sent to servo controller, rotary head accepts the instruction that servo controller sends, timely drive infrared camera turns to target predicted position, and target is in the immediate vicinity of shooting screen all the time.

(2) tracking of moving target

The present embodiment adopts CamShift algorithm to carry out motion target tracking, can overcome the large shortcoming with being subject to block interference of the calculated amount existed in conventional correlation algorithm preferably.Make full use of the result of moving object detection, improve matching efficiency during target following, and adopt threshold cutoff technology, reduce the hunting zone of target, realize the quick and precisely tracking to target.

Algorithm is as follows:

The first step, calculates the 1D histogram of H passage (color channel) component in image HSI space;

Second step, utilizes this 1D histogram that former figure is reconstructed into 2D probability distribution graph;

3rd step, calculates the center of target area;

4th step, utilizes CamShift algorithm, and continuous translation adjustment window center overlaps to target barycentric;

5th step, by the window size of previous frame and center, as the initial value of next frame CamShift algorithm search window;

6th step, continues CamShift computing in the next frame.

The algorithm operation quantity of the moving object segmentation that the present embodiment proposes and tracking is little, and tracking effect is good.As long as capture target initial errorless, and object and background has certain deviation on color space, video tracking just can reach suitable accuracy.

(3) rotary head rotates and controls

There is sampling time delay in the image taking mechanism due to infrared camera, the existence of the comparatively slow and process time delay of the velocity of rotation of rotary head, all may make to drive infrared photography cephalomotor rotary head cannot tracking target in time, particularly for movement velocity target faster.To real-time follow-up target, will look-ahead target travel.The present invention proposes to adopt kalman filtering algorithm to carry out the prediction of moving target state, and to realize the rotation in advance of rotary head, control flow chart is as Fig. 2.

The present embodiment can turn to target subsequent time position in advance according to prediction rotary head, adopts Kalman filtering algorithm to carry out the motion state of predicted motion target, utilizes the recurrence estimation ability of filtering to predict subsequent time target location.System noise and observation noise are considered as white Gaussian noise, measure as systematic perspective using the position of moving target, with institute's syllabus target location estimation value for exporting, moving target current motion state being made and estimates and Future movement state is made prediction.

State equation and the observation equation of Target Tracking System system are respectively:

$\left\{\begin{array}{c}x(k+1)=\mathrm{\φx}\left(k\right)+w\left(k\right)\\ z\left(k\right)=\mathrm{Hx}\left(k\right)+v\left(k\right)\end{array}\right.$

In above formula: x (k) ∈ R ⁶for state vector, represent the position of moving target, speed and acceleration, its expression formula is x (k)=[x y x _vy _vx _ay _a] ^t; Z (k) ∈ R ²for observation vector, represent the position of moving target; W (k) ∈ R ⁶it is systematic procedure noise vector; V (k) ∈ R ⁶it is observation noise vector; φ ∈ R ^{6 × 6}be state-transition matrix, its expression formula is:

$\mathrm{\φ}=\left[\begin{array}{cccccc}1& 0& T& 0& T^2/2& 0\\ 0& 1& 0& T& 0& T^2/2\\ 0& 0& 1& 0& T& 0\\ 0& 0& 0& 1& 0& T\\ 0& 0& 0& 0& 1& 0\\ 0& 0& 0& 0& 0& 1\end{array}\right]$

H is observing matrix, and need the position observing moving target here, its expression formula is

$H=\left[\begin{array}{cccccc}1& 0& 0& 0& 0& 0\\ 0& 1& 0& 0& 0& 0\end{array}\right]$

System noise and observation noise are mutual incoherent white Gaussian noise.

The present embodiment adopts time equal interval sampling for the target following of infrared camera.When observing a certain moment target location, pre-estimation in time the most probable position that next moment target occurs can be gone out, so that infrared camera adjusts angle in time, quick tracking target.When known system original state, obtain x (1/1) according to the obserred coordinate value Kalman filtering of current kinetic target, and then prediction x (2/1), concrete Kalman filter equation is as follows:

x(k/k)＝x(k/k-1)+K(k)(z(k)-Hx(k/k-1))

x(k/k-1)＝φx(k-1/k-1))

K(k)＝P(k/k-1)H ^T(HP(k/k-1)H ^T+R(k)) ^-1

P(k/k-1)＝φP(k-1/k-1)φ ^T+Q(k-1)

P(k/k)＝(I-K(k)H)P(k/k-1)

In various above, k=1,2,3 ... x (k/k-1) and P (k/k-1) is status predication value and minimum prediction Square Error matrix thereof, K (k) is Kalman filtering gain matrix, and z (k) ceases for observation is new, and x (k/k) and P (k/k) is the state estimation and error co-variance matrix thereof revised.

Obtain the position of target during k moment, then can obtain the position angle of target according to positional information, according to kalman filter forecasting moving target after the position in k+1 moment, the position angle in k+1 moment can be calculated to obtain equally, Δ θ=θ _k+1-θ _k, the rotational angular velocity of known rotary head, therefore the control to rotary head can be reached by the rotation time arranging rotary head.When moving target enters the observation scope of infrared camera, infrared camera, can according to the coordinate (x, y) of each moment moving target of these information acquisitions after photographing image.Before not obtaining new measured value z (k) in k moment, the estimation can only made from the k-1 moment signal, estimates status signal x (k) in k moment.After the present co-ordinate position recording k moment target, just can obtain the time of day optimal estimation value x (k|k) of target.Corrected Calculation is carried out to predicted value and goes out the observed reading x (k+1|k) of target at k-1 moment most probable position, the deflection angle of target can be obtained according to coordinate position, control the rotation time of rotary head according to deflection angle.Because the motion of now rotary head has certain lead, thus eliminate tracking lag phenomenon preferably.

Global search problem, by target of prediction object position in the next frame, is converted into Local Search by the present embodiment, improves the real-time of algorithm; Meanwhile, utilize predicted value effectively can prevent the interference of similar features object in background, improve the accuracy of algorithm.

Above-described embodiment of the present invention, does not form limiting the scope of the present invention.Any amendment done within the spirit and principles in the present invention, equivalent replacement and improvement etc., all should be included within claims of the present invention.

Claims (5)

1. boat-carrying infrared target image identification tracking means, is characterized in that: comprise upper industrial computer, servocontrol slave computer, rotary head, video frequency collection card, search lamp, infrared camera;

Described upper industrial computer is connected with servocontrol slave computer, and described servocontrol slave computer is connected with rotary head and controls rotary head and rotates; Described rotary head is fixedly equipped with search lamp, infrared camera, described infrared camera is connected with upper industrial computer by video frequency collection card, and described search lamp is synchronous with the sensing of infrared camera.

2. boat-carrying infrared target image identification tracking means according to claim 1, is characterized in that: described video frequency collection card is connected with upper industrial computer by USB serial ports.

3. boat-carrying infrared target image identification tracking means according to claim 1, is characterized in that: described search lamp is high-power sodium vapor lamp.

4. boat-carrying infrared target image identification tracking means according to claim 1, is characterized in that: described servocontrol slave computer controls rotary head and does upper and lower, left and right full angle rotation.

5. boat-carrying infrared target image identification tracking, is characterized in that comprising the following steps:

(1) motion estimate

First infrared camera is utilized to carry out image sequence acquisition, obtain the original image of target, target image signal is sent in industrial computer, operator selectes target to be locked, by industrial computer as CPU (central processing unit), Tracking Control Software adopts corresponding image processing method to process to image respectively, extracts comparatively complete target to be tracked from original image; Frame differential method is used to judge whether there is moving target in video sequence, and determine the position of target, frame differential method adopts the difference based on pixel grey scale to two two field pictures before and after a Minimum-time interval, then extracts the moving region in image with thresholding method, and algorithm is:

$M(x,y,t)=\left\{\begin{array}{cc}1,& f(x,y,t)-f(x,y,t-\mathrm{\Δt})>\mathrm{\τ}\\ 0,& f(x,y,t)-f(x,y,t-\mathrm{\Δt})\≤\mathrm{\τ}\end{array}\right.$

In above formula, Δ t is inter frame temporal interval, f (x, y, t) and f (x, y, t-Δ t)) be respectively the picture frame of t and t-Δ t; If the absolute difference of two two field picture respective pixel gray scales is greater than threshold tau, then shows that this pixel is sport foreground, then its value is put 1, otherwise for static background, its value is set to 0, bianry image M (x, y, t) can be obtained;

(2) tracking of moving target

Adopt CamShift algorithm to carry out motion target tracking, algorithm is as follows:

The first step, calculates the 1D histogram of H passage (color channel) component in image HSI space;

Second step, utilizes this 1D histogram that former figure is reconstructed into 2D probability distribution graph;

3rd step, calculates the center of target area;

4th step, utilizes CamShift algorithm, and continuous translation adjustment window center overlaps to target barycentric;

5th step, by the window size of previous frame and center, as the initial value of next frame CamShift algorithm search window;

6th step, continues CamShift computing in the next frame;

(3) rotary head rotates and controls

Kalman filtering algorithm is adopted to carry out the prediction of moving target state, to realize the rotation in advance of rotary head; Target subsequent time position can be turned in advance according to prediction rotary head, adopt Kalman filtering algorithm to carry out the motion state of predicted motion target, utilize the recurrence estimation ability of filtering to predict subsequent time target location; System noise and observation noise are considered as white Gaussian noise, measure as systematic perspective using the position of moving target, with institute's syllabus target location estimation value for exporting, moving target current motion state being made and estimates and Future movement state is made prediction;

State equation and the observation equation of Target Tracking System system are respectively:

$\left\{\begin{array}{c}x(k+1)=\mathrm{\φx}\left(k\right)+w\left(k\right)\\ z\left(k\right)=\mathrm{Hx}\left(k\right)+v\left(k\right)\end{array}\right.$

In above formula: x (k) ∈ R ⁶for state vector, represent the position of moving target, speed and acceleration, its expression formula is x (k)=[x y x _vy _vx _ay _a] ^t; Z (k) ∈ R ²for observation vector, represent the position of moving target; W (k) ∈ R ⁶it is systematic procedure noise vector; V (k) ∈ R ⁶it is observation noise vector; φ ∈ R ^{6 × 6}be state-transition matrix, its expression formula is:

$\mathrm{\φ}=\left[\begin{array}{cccccc}1& 0& T& 0& T^2/2& 0\\ 0& 1& 0& T& 0& T^2/2\\ 0& 0& 1& 0& T& 0\\ 0& 0& 0& 1& 0& T\\ 0& 0& 0& 0& 1& 0\\ 0& 0& 0& 0& 0& 1\end{array}\right]$

H is observing matrix, and need the position observing moving target here, its expression formula is

$H=\left[\begin{array}{cccccc}1& 0& 0& 0& 0& 0\\ 0& 1& 0& 0& 0& 0\end{array}\right]$

System noise and observation noise are mutual incoherent white Gaussian noise;

Employing time equal interval sampling is for the target following of infrared camera; When observing a certain moment target location, pre-estimation in time the most probable position that next moment target occurs can be gone out, so that infrared camera adjusts angle in time, quick tracking target; When known system original state, obtain x (1/1) according to the obserred coordinate value Kalman filtering of current kinetic target, and then prediction x (2/1), concrete Kalman filter equation is as follows:

x(k/k)＝x(k/k-1)+K(k)(z(k)-Hx(k/k-1))

x(k/k-1)＝φx(k-1/k-1))

K(k)＝P(k/k-1)H ^T(HP(k/k-1)H ^T+R(k)) ^-1

P(k/k-1)＝φP(k-1/k-1)φ ^T+Q(k-1)

P(k/k)＝(I-K(k)H)P(k/k-1)

In various above, k=1,2,3 ... x (k/k-1) and P (k/k-1) is status predication value and minimum prediction Square Error matrix thereof, K (k) is Kalman filtering gain matrix, and z (k) ceases for observation is new, and x (k/k) and P (k/k) is the state estimation and error co-variance matrix thereof revised;

Obtain the position of target during k moment, then can obtain the position angle of target according to positional information, according to kalman filter forecasting moving target after the position in k+1 moment, the position angle in k+1 moment can be calculated to obtain equally, Δ θ=θ _k+1-θ _k, the rotational angular velocity of known rotary head, therefore the control to rotary head can be reached by the rotation time arranging rotary head; When moving target enters the observation scope of infrared camera, infrared camera, can according to the coordinate (x, y) of each moment moving target of these information acquisitions after photographing image; Before not obtaining new measured value z (k) in k moment, the estimation can only made from the k-1 moment signal, estimates status signal x (k) in k moment; After the present co-ordinate position recording k moment target, just can obtain the time of day optimal estimation value x (k|k) of target; Corrected Calculation is carried out to predicted value and goes out the observed reading x (k+1|k) of target at k-1 moment most probable position, the deflection angle of target can be obtained according to coordinate position, control the rotation time of rotary head according to deflection angle.