CN103727931A - Improved logic-based track initiation method - Google Patents

Abstract

The present invention proposes a track initiation method based on improved logic. Take the measurement point of each tracking target in its first frame of image as the initial point to register a temporary track for each tracking target, and use the correlation coefficient of each tracking target as the correlation coefficient of the corresponding temporary track; The measurement point of the measurement point of the second frame image in the circular associated area of the measurement point of the first frame image is added as the second point to the corresponding temporary track; every time a measurement point is added to the temporary track of the tracking target When the point is reached, the correlation coefficient of the temporary track is accumulated once, and when the cumulative sum of the correlation coefficient is greater than or equal to the preset judgment threshold, the temporary track is added to the queue of the temporary successful track for track pruning Then output the valid track. The invention meets the requirement that different targets have different track initiations, improves the accuracy of track initiation, and reduces calculation time.

Description

A kind of track initiation method based on improving logic

Technical field

The invention belongs to image detection and process field, be specifically related to a kind of track initiation method based on improving logic.

Background technology

During due to track initiation, target range is far away, and sensor detection resolving power is low, measuring accuracy is poor, adds the appearance of true and false target, causes without real statistical law, especially, under complex background condition, all can affect the track initiation quality under conventional processing.Relevant in order to complete reliably target, filtering clutter point, in numerous targets, complete the reliably initial of real goal flight path separately, just need to have certain processing convergence time, so flight path initial requirement fast and high track initiation success ratio are conflicting.

Existing track initiation method mainly comprises heuristic rule method, logical approach and Hough converter technique etc.Document one (Huo Hangyu.Fast target track initiation research [D], Harbin Institute of Technology, 2006) the described method based on heuristic rule in the intensive many clutters situation that has a large amount of clutters and noise, improper if judgment rule or algorithm are selected, will cause higher false-alarm or false dismissal; (kingdom is grand, Su Feng for document two.Fast Track Initiation Algorithm [J] based on Hough conversion and logic under clutter environment, Journal of System Simulation, 2002,14(7)) the described method based on Hough conversion is in the process of Track forming, need accumulation and the complex calculations of long period, for the high application of requirement of real-time, need further improvement.

In actual applications, different targets has different track initiation requirements, for example, for expansion target, just wish the response time fast, otherwise just may cause target to depart from field range, causes track rejection; And for little target, in the situation that clutter is intensive, easily causing associated errors, track initiation may bring higher false alarm rate fast, at this moment just requires to increase the time of track initiation, in order to avoid the flight path of generation error.According to document three (Yan Kang, multiple-sensor and multiple-object Track In Track and Study on Fusion [D], Institutes Of Technology Of Nanjing, 2012) plot-track Association Algorithm of described general logical approach fixes track initiation length and the response time of target, can not dynamically change the response time, very possible lose objects or flase drop target.

Summary of the invention

The present invention proposes a kind of track initiation method based on improving logic, and reaching different target has the initial requirement of Different Flight, has improved the precision of track initiation, has reduced computing time.

In order to solve the problems of the technologies described above, the present invention proposes a kind of track initiation method based on improving logic, comprises the following steps:

Step 1: carry out the image sequence after context update processing acquisition is upgraded by taking the video obtaining, gauge point take each tracking target in its first two field picture is a temporary transient flight path of each tracking target registration as initial point, and calculate the correlation coefficient of each tracking target, the correlation coefficient of the temporary transient flight path using the correlation coefficient of each tracking target as correspondence; The computing method of described each tracking target correlation coefficient are as shown in Equation (1):

$w_i=\left\{\begin{array}{c}0.4|G_i-G_{\mathrm{ave}}|>100\\ 0.350<|G_i-G_{\mathrm{ave}}|<100\\ 0.210<|G_i-G_{\mathrm{ave}}|<\text{50}\\ 0.1|G_i-G_{\mathrm{ave}}|<10\end{array}\right.---\left(1\right)$

In formula (1), w _ibe i tracking target M ⁱcorrelation coefficient, G _irepresent tracking target M ⁱgray scale in current frame image, G _averepresent the average background gray scale of current frame image;

Step 2: judge that the gauge point of each tracking target in its second two field picture is whether in its first two field picture in the annular associated region of gauge point, if the gauge point of the second two field picture is in the annular associated region of the first two field picture gauge point, tracking target is effective dose measuring point and using this gauge point as second point, joins in corresponding temporary transient flight path at the gauge point of the second two field picture, then skips to step 5; If the gauge point of the second two field picture is not in the annular associated region of the first two field picture gauge point, corresponding temporary transient flight path is false track, and false track is nullified;

Step 3: according to the measuring value of each temporary transient flight path the first two point, use state estimation and prediction covariance matrix to predict state estimation and the covariance matrix of the next point of temporary transient flight path, then use elliptical wave door mode judge the gauge point of each tracking target in its next frame image whether with temporary transient track association, if with temporary transient track association; retain the gauge point in its next frame image; If not with temporary transient track association, utilize Kalman Prediction mode to extrapolate a point as the gauge point in its next frame image;

Step 4: utilize least square fitting calculate each tracking target in the horizontal direction with vertical direction on speed, within gauge point in the next frame image that determining step three the obtains deviation angle scope whether line direction allows before tracking target, if within the deviation angle scope that line direction allows before tracking target, think that the measurement point in this next frame image is effective, set it as next point and join in corresponding temporary transient flight path; Otherwise thinking that the measurement point in this next frame image is invalid, is not the subsequent point of temporary transient flight path; If there are multiple effective gauge points within the scope of a certain tracking target deviation angle that front line direction allows in its next frame image, further according to the scope of the position deviation of tracking target and feature deviation, determine the measurement degree of confidence of each measurement point, get the gauge point of degree of confidence maximum as the optimum subsequent point that adds temporary transient flight path;

Step 5: just the correlation coefficient of this temporary transient flight path is accumulated once when the temporary transient flight path of tracking target often adds a gauge point, when the cumulative sum of correlation coefficient is more than or equal to predefined judgment threshold, temporary transient flight path is joined in the queue of interim successful flight path and carries out exporting effective flight path after flight path beta pruning; If when the cumulative sum of current correlation coefficient does not reach predefined judgment threshold, skip to step 3.

The present invention compared with prior art, its remarkable advantage is, (1) the present invention adds the characteristic information of tracking target, be equivalent to have clarification of objective prior imformation, in track association, can weed out the noise spot that falls into temporary transient Trajectory Prediction Bo Mennei according to the feature of tracking target, thereby the correctness of tracking target association is improved; (2) the present invention is in track initiation process, utilize the characteristic information of tracking target to calculate the correlation coefficient of tracking target, then target is classified, according to the correlation coefficient difference of the target of different qualities, dynamically adjust the length of track initiation time, system can be identified according to the fundamental characteristics of tracking target adaptively, the real-time of assurance system, has improved the initial precision of multi-target traces; (3) the present invention is in track initiation process, utilize the characteristic information of tracking target to calculate the correlation coefficient of tracking target, then target is classified, according to the target dynamic of different qualities, adjust track initiation time span, system can be identified according to the essential characteristic of tracking target adaptively, the real-time of the system guaranteeing, has improved the initial precision of multi-target traces.

Accompanying drawing explanation

Fig. 1 is the inventive method simple flow chart.

Fig. 2 is the inventive method detail flowchart.

Fig. 3 is the track plot that the track initiation method described in use background technology document three obtains.

Fig. 4 be use the inventive method obtain track plot.

Embodiment

As depicted in figs. 1 and 2, based on the track initiation method of improving logic, comprise the following steps:

Step 1: carry out the image sequence after context update processing acquisition is upgraded by taking the video obtaining, gauge point take each tracking target in its first two field picture is a temporary transient flight path of each tracking target registration as initial point, and calculate the correlation coefficient of each tracking target, the correlation coefficient of the temporary transient flight path using the correlation coefficient of each tracking target as correspondence; The computing method of described each tracking target correlation coefficient are as shown in Equation (1):

$w_i=\left\{\begin{array}{c}0.4|G_i-G_{\mathrm{ave}}|>100\\ 0.350<|G_i-G_{\mathrm{ave}}|<100\\ 0.210<|G_i-G_{\mathrm{ave}}|<\text{50}\\ 0.1|G_i-G_{\mathrm{ave}}|<10\end{array}\right.---\left(1\right)$

In formula (1), w _ibe i tracking target M ⁱcorrelation coefficient, G _irepresent tracking target M ⁱgray scale in current frame image, G _averepresent the average background gray scale of current frame image;

In the present invention, each tracking target M ⁱgauge point in the first two field picture can be used

represent, wherein,

represent i tracking target M ⁱgauge point in the first two field picture coordinate figure in the horizontal direction,

represent i tracking target M ⁱgauge point in the first two field picture coordinate figure in vertical direction, i.e. the measuring value of gauge point, i=1,2 ...

Step 2: judge that the gauge point of each tracking target in its second two field picture is whether in its first two field picture in the annular associated region of gauge point, if the gauge point of the second two field picture is in the annular associated region of the first two field picture gauge point, tracking target is effective dose measuring point and using this gauge point as second point, joins in corresponding temporary transient flight path at the gauge point of the second two field picture, then skips to step 5; If the gauge point of the second two field picture is not in the annular associated region of the first two field picture gauge point, corresponding temporary transient flight path is false track, and false track is nullified.

Above-mentioned deterministic process specifically, judges each tracking target M exactly ⁱthe measuring value of gauge point in its second two field picture

horizontal coordinate value

and vertical coordinate value

whether all gauge point in its first two field picture

the horizontal associated region of annular associated region vertical with annular in, if gauge point

horizontal coordinate value

and vertical coordinate value all at gauge point the horizontal associated region of annular in associated region vertical with annular, gauge point

for effective dose measuring point and set it as second point and join in corresponding temporary transient flight path; If gauge point

horizontal coordinate value

not at gauge point

the horizontal associated region of annular in or vertical coordinate value not at gauge point

the vertical associated region of annular in, this gauge point

corresponding temporary transient flight path is false track, is nullified; Described gauge point

the horizontal associated region of annular refer to gauge point

centered by, respectively with tracking target M ⁱnonoverlapping circular annular region between two circles that the product that maximal rate in the first two field picture horizontal direction and the product in sampling period are the minimum speed in radius and horizontal direction and sampling period forms for radius; Described gauge point vertical associated circular annular region refer to gauge point centered by, respectively with tracking target M ⁱmaximal rate in the first two field picture vertical direction and the product in sampling period for radius and vertical direction on nonoverlapping circular annular region between two circles forming for radius of minimum speed and the product in sampling period.

Above-mentioned deterministic process can be expressed as follows with formula (2) and (3), simultaneously the gauge point of the second two field picture of formula (2) and (3)

be at the first two field picture gauge point

annular associated region in,

$v_{x\min }T<x_2^i-x_1^i<v_{x\max }T---\left(2\right)$

$v_{y\min }T<y_2^i-y_1^i<v_{y\max }T---\left(3\right)$

In formula (2), v _xminand v _xmaxrepresent respectively i tracking target M ⁱmaximal rate and minimum speed in its first two field picture in horizontal direction, in formula (3), v _yminand v _ymaxrepresent respectively tracking target M ⁱmaximal rate and minimum speed in its first two field picture in vertical direction, T represents the sampling period;

with

for tracking target M ⁱthe measuring value of gauge point in the first two field picture,

with

for tracking target M ⁱthe measuring value of gauge point in the second two field picture.

Step 3: according to the measuring value of each temporary transient flight path the first two point, use state estimation and prediction covariance matrix to predict state estimation and the covariance matrix of the next point of temporary transient flight path, then use elliptical wave door mode judge the gauge point of each tracking target in its next frame image whether with temporary transient track association, if with temporary transient track association; retain the gauge point in its next frame image; If not with temporary transient track association, utilize Kalman Prediction mode to extrapolate a point as the gauge point in its next frame image.

The described measuring value according to each temporary transient flight path the first two point is used state estimation and prediction covariance matrix to predict that the state estimation of next one point of temporary transient flight path and the account form of covariance matrix are respectively as shown in formula (4) and (5),

$Z_{k|k-1}^i=\mathrm{\&Phi;}{\hat{Z}}_{k-1|k-1}^i---\left(4\right)$

$P_{k|k-1}^i=\mathrm{\&Phi;}{P}_{k-1|k-1}^i{\mathrm{\&Phi;}}^T+Q---\left(5\right)$

In formula (4), Φ be transition matrix and $\mathrm{\&Phi;}\left[\begin{array}{cccc}1& T& 0& 0\\ 0& 1& 0& 0\\ 0& 0& 1& T\\ 0& 0& 0& 1\end{array}\right];$

for state estimation, concrete as shown in formula (6),

${\hat{Z}}_{k-1|k-1}^i=\left[\begin{array}{c}{x}_{k-1}^i\\ (x_{k-1}^i-x_{k-2}^i)/T\\ y_{k-1}^i\\ (y_{k-1}^i-y_{k-2}^i)/T\end{array}\right]---\left(6\right)$

In formula (6),

with

for the measuring value of tracking target Mi gauge point in k-1 two field picture,

with

for tracking target M ⁱthe measuring value of gauge point in k-2 two field picture;

In formula (5), Q is system noise covariance; for prediction covariance matrix, concrete as shown in formula (7),

$P_{k-1|k-1}^i=\left[\begin{array}{cccc}{\mathrm{\&sigma;}}_x^2& {\mathrm{\&sigma;}}_x^2/T& 0& 0\\ {\mathrm{\&sigma;}}_x^2/T& {\mathrm{\&sigma;}}_x^2/T^2& 0& 0\\ 0& 0& {\mathrm{\&sigma;}}_y^2& {\mathrm{\&sigma;}}_y^2/T\\ 0& 0& {\mathrm{\&sigma;}}_y^2/T& {\mathrm{\&sigma;}}_y^2/T^2\end{array}\right]---\left(7\right)$

In formula (7),

with

represent the horizontal relative error and the vertically opposite error that in computation process, produce;

Described use elliptical wave door mode judge the gauge point of each tracking target in next frame image whether with the account form of temporary transient track association as shown in Equation (8),

${[X_k^i-{\hat{X}}_{k|k-1}^i]}^T{\left(S_k^i\right)}^{-1}[X_k^i-{\hat{X}}_{k|k-1}^i]\&le;\mathrm{\&gamma;}---\left(8\right)$

In formula (8),

represent tracking target M ⁱin next frame image the measuring value of gauge point and $X_k^i=\left[\begin{array}{c}{x}_k^i\\ y_k^i\end{array}\right],$

represent temporary transient flight path future position predicted value and ${\hat{X}}_{k|k-1}^i=\left[\begin{array}{cccc}1& 0& 0& 0\\ 0& 1& 0& 0\end{array}\right]Z_{k|k-1}^i,$ γ is the threshold value of oval associated region,

for the residual error covariance matrix calculating centered by future position, account form as shown in Equation (9),

$S_k^i=H_k{P}_{k|k-1}^i{H}_k^T+R_k---\left(9\right)$

In formula (9), H _kfor observing matrix, R _kfor the average of measurement noise.

Above-mentioned elliptical wave door method can with reference to what friend, build beautiful and close that glad shown < < radar data is processed and application > > (Electronic Industry Press, 2013).

The described Kalman Prediction mode of utilizing is extrapolated the computing method of a point as shown in formula (10) and formula (11)

${\hat{Z}}_{k|k}^i={\hat{Z}}_{k|k-1}^i+{\mathrm{Kg}}_k^i({\hat{X}}_{k|k-1}^i-H_k{\hat{Z}}_{k|k-1}^i)---\left(10\right)$

$P_{k|k}^i=(1-{\mathrm{Kg}}_k^i{H}_k)P_{k|k-1}^i---\left(11\right)$

In formula (10),

for kalman gain.

Above-mentioned Germania Forecasting Methodology can be put down with reference to tight Zhejiang, moving object forecast [J] the > > mono-literary composition (applicating technology of < < based on Kalman filtering shown at Huang Yu peak, 2008,10).

Step 4: utilize least square fitting calculate each tracking target in the horizontal direction with vertical direction on speed, within gauge point in the next frame image that determining step three the obtains deviation angle scope whether line direction allows before tracking target, if within the deviation angle scope that line direction allows before tracking target, think that the measurement point in this next frame image is effective, set it as next point and join in corresponding temporary transient flight path; Otherwise thinking that the measurement point in this next frame image is invalid, is not the subsequent point of temporary transient flight path; If there are multiple effective gauge points within the scope of a certain tracking target deviation angle that front line direction allows in its next frame image, further according to the scope of the position deviation of tracking target and feature deviation, determine the measurement degree of confidence of each measurement point, get the gauge point of degree of confidence maximum as the optimum subsequent point that adds temporary transient flight path.

This step specifically can be shown multi-target traces start algorithm could research [D] the > > (Harbin Institute of Technology, 2012.6) under < < complex background with reference to Zhang Yanhang.

Step 5: just the correlation coefficient of this temporary transient flight path is accumulated once when the temporary transient flight path of tracking target often adds a gauge point, when the cumulative sum of correlation coefficient is more than or equal to predefined judgment threshold, temporary transient flight path is joined in the queue of interim successful flight path and carries out exporting effective flight path after flight path beta pruning; If when the cumulative sum of current correlation coefficient does not reach predefined judgment threshold, skip to step 3.

Effect of the present invention can be described further by following simulation result.

Simulating scenes is set: have 5 targets of doing linear uniform motion, initial position is (150,350), (200,250), (350,350), (400,500), (550,450), and 5 targets are all with initial velocity v _x=5, v _y=0 moves with uniform velocity, and wherein x represents horizontal direction, and y represents vertical direction.

Fig. 2 is the track plot that the track initiation method described in use background technology document three obtains, Fig. 3 is the track initiation figure that adopts the inventive method to obtain, by two width figure, relatively can be found out, track loss in Fig. 2 the length of one and initial flight path all the same, and in Fig. 35 flight paths all initial the and track initiation length of success according to target difference, there is different length, illustrate that the inventive method has improved computational accuracy and computing velocity.

Claims (3)

1. a track initiation method based on improved logic, is characterized in that, comprises the following steps: Step 1: Perform background update processing on the captured video to obtain an updated image sequence, use the measurement point of each tracking target in its first frame of image as the initial point to register a temporary track for each tracking target, and Calculate the correlation coefficient of each tracking target, and use the correlation coefficient of each tracking target as the correlation coefficient of the corresponding temporary track; the calculation method of the correlation coefficient of each tracking target is shown in formula (1): ${\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\left\{\begin{array}{c}\mathrm{<span\; class="notranslate">\; 0.4</span>}<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; ave</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; ></span>\mathrm{<span\; class="notranslate">\; 100</span>}\\ \mathrm{<span\; class="notranslate">\; 0.350</span>}<span\; class="notranslate">\; <</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; ave</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; 100</span>}\\ \mathrm{<span\; class="notranslate">\; 0.210</span>}<span\; class="notranslate">\; <</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; ave</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; <</span>\text{<span class="notranslate"> 50</span>}\\ \mathrm{<span\; class="notranslate">\; 0.1</span>}<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; ave</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; 10</span>}\end{array}\right.<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$ In formula (1), w _i is the correlation coefficient of the i-th tracking target ^Mi , G _i represents the gray level of the tracking target ^Mi in the current frame image, G _ave represents the average background gray level of the current frame image; Step 2: Determine whether the measurement point of each tracking target in its second frame image is within the circular associated area of the measurement point in its first frame image, if the measurement point of the second frame image is within the first In the circular associated area of the measurement point of the frame image, the measurement point of the tracking target in the second frame image is an effective measurement point and this measurement point is added to the corresponding temporary track as the second point, and then Skip to step five; if the measurement point of the second frame of image is not within the circular associated area of the measurement point of the first frame of image, then the corresponding temporary track is a false track, and the false track will be canceled; Step 3: According to the measurement values of the first two points of each temporary track, use the state estimation and prediction covariance matrix to predict the state estimation and covariance matrix of the next point of the temporary track, and then use the elliptic wave gate method to judge each Whether the measurement point of the tracking target in its next frame image is associated with the temporary track, if it is associated with the temporary track, keep the measurement point in the next frame image; if it is not associated with the temporary track, use Cal The Mann prediction method extrapolates a point as the measurement point in the next frame image; Step 4: Use the least squares method to fit and calculate the speed of each tracking target in the horizontal and vertical directions, and judge whether the measurement point in the next frame image obtained in step 3 is within the allowable deviation in the forward direction of the tracking target If it is within the allowable deviation angle range of the tracking target's forward direction, the measurement point in the next frame of image is considered valid, and it is added as the next point to the corresponding temporary track; otherwise, it is considered The measurement point in the next frame image is invalid, not a follow-up point of the temporary track; if a tracking target has multiple valid measurement points within the allowable deviation angle range of its forward direction in the next frame image, then Further determine the measurement confidence of each measurement point according to the range of the position deviation and characteristic deviation of the tracking target, and take the measurement point with the largest confidence as the optimal follow-up point for adding the temporary track; Step 5: Every time a measurement point is added to the temporary track of the tracking target, the correlation coefficient of the temporary track is accumulated once. When the cumulative sum of the correlation coefficients is greater than or equal to the preset judgment threshold, the temporary track After being added to the queue of temporary successful tracks for track pruning, the valid track is output; if the sum of the current correlation coefficients does not reach the preset judgment threshold, skip to step 3. 2. the track starting method based on improved logic as claimed in claim 1, is characterized in that, whether the measuring point of said judging each tracking target in its second frame image is within its first frame image The method in the circular associated area of the measuring point is to track whether the measuring points of the target in the first frame image and the second frame image satisfy the formulas (2) and (3) at the same time, ${\mathrm{<span\; class="notranslate">\; v</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}\mathrm{<span\; class="notranslate">\; min</span>}}\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; v</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}\mathrm{<span\; class="notranslate">\; max</span>}}\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$ ${\mathrm{<span\; class="notranslate">\; v</span>}}_{\mathrm{<span\; class="notranslate">\; the\; y</span>}\mathrm{<span\; class="notranslate">\; min</span>}}\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; v</span>}}_{\mathrm{<span\; class="notranslate">\; the\; y</span>}\mathrm{<span\; class="notranslate">\; max</span>}}\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>$ In formula (2), v _xmin and v _xmax represent respectively the maximum velocity and minimum velocity of the i-th tracking target M ⁱ in the horizontal direction in its first frame image, and in formula (3), v _ymin and v _ymax represent respectively Track the maximum velocity and minimum velocity of the target M ⁱ in the vertical direction in its first frame image, and T represents the sampling period;

and

In order to track the measurement value of the measurement point in the first frame image of the target M ⁱ ,

and

Measure the measurement value of the point in the second frame image for tracking the target ^Mi. 3. the track initiation method based on improved logic as claimed in claim 1, is characterized in that, in step 3, According to the measurement values of the first two points of each temporary track, the state estimation and prediction covariance matrix are used to predict the state estimation and covariance matrix of the next point of the temporary track. The calculation methods are as formula (4) and ( 5) as shown, ${\mathrm{<span\; class="notranslate">\; Z</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; \&Phi;</span>}{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; Z</span>}}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; )</span>$ ${\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; \&Phi;</span>}{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; \&Phi;</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; Q</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 5</span>}<span\; class="notranslate">\; )</span>$ In formula (4), Φ is the transfer matrix and $\mathrm{\&Phi;}\left[\begin{array}{cccc}1& T& 0& 0\\ 0& 1& 0& 0\\ 0& 0& 1& T\\ 0& 0& 0& 1\end{array}\right];$

is the state estimation, specifically as shown in formula (6), ${\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; Z</span>}}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\left[\begin{array}{c}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}\\ <span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; T</span>}\\ {\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}\\ <span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; T</span>}\end{array}\right]<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 6</span>}<span\; class="notranslate">\; )</span>$ In formula (6),

and

To track the measurement value of the measurement point in the k-1th frame image of the target M ⁱ ,

and is the measurement value of the measurement point in the k-2th frame image of the tracking target ^Mi ; In formula (5), Q is the system noise covariance; To predict the covariance matrix, specifically as shown in formula (7), ${\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\left[\begin{array}{cccc}{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}& {\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; T</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ {\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; T</span>}& {\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; /</span>{\mathrm{<span\; class="notranslate">\; T</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& {\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; the\; y</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}& {\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; the\; y</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; T</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& {\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; the\; y</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; T</span>}& {\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; the\; y</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; /</span>{\mathrm{<span\; class="notranslate">\; T</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}\end{array}\right]<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 7</span>}<span\; class="notranslate">\; )</span>$ In formula (7), and Indicates the horizontal relative error and vertical relative error generated during the calculation; The calculation method of using the elliptical gate method to judge whether the measurement point of each tracking target in the next frame image is associated with the temporary track is shown in formula (8), ${<span\; class="notranslate">\; [</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; ]</span>}^{\mathrm{<span\; class="notranslate">\; T</span>}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; [</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; \&gamma;</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 8</span>}<span\; class="notranslate">\; )</span>$ In formula (8),

Indicates the measurement value of the measurement point of the tracking target M ⁱ in the next frame image and $x_k^i=\left[\begin{array}{c}{x}_k^i\\ {\mathrm{the\; y}}_k^i\end{array}\right],$ represents the predicted value of the predicted point of the tentative track and ${\hat{x}}_{k|k-1}^i=\left[\begin{array}{cccc}1& 0& 0& 0\\ 0& 1& 0& 0\end{array}\right]Z_{k|k-1}^i,$ γ is the threshold value of the ellipse association area,

is the residual covariance matrix calculated centered on the prediction point, and the calculation method is shown in formula (9), ${\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; R</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 9</span>}<span\; class="notranslate">\; )</span>$ In formula (9), H _k is the observation matrix, R _k is the mean value of the measurement noise; The calculation method for extrapolating a point using the Kalman prediction method is shown in formula (10) and formula (11) ${\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; Z</span>}}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; Z</span>}}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; kg</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; (</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; Z</span>}}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 10</span>}<span\; class="notranslate">\; )</span>$ ${\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; kg</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 11</span>}<span\; class="notranslate">\; )</span>$ In formula (10),

Gain for Kalman.

Images