CN103839274B - A kind of Extended target tracking based on geometric proportion relation - Google Patents

Abstract

The present invention provides a kind of Extended target tracking based on geometric proportion relation, initially with Gaussian smoothing filter, pending image is carried out pretreatment to remove the noise impact on subsequent algorithm, secondly use Fuzzy C-Means Cluster Algorithm FCM that the image after above-mentioned smoothing is split, obtain bianry image, then the bianry image obtained after segmentation is carried out skeletal extraction, head and tail characteristic point is selected from the skeletal point extracted, calculate the length of fuselage in the plane of delineation, ratio relation according to head on known aerocraft real fuselage to fuselage with wing intersection point with fuselage length, when aircraft occurs in time blocking, according to this geometric proportion invariance, obtain the intersection point of fuselage and wing, thus obtain trace point.The present invention solves the problem extracting the skeleton less than wing when certainly blocking occurs in aircraft thus cause trace point to be lost, it is achieved that the aircraft tenacious tracking when certainly blocking.

Description

A kind of Extended target tracking based on geometric proportion relation

Technical field

The present invention relates to a kind of motor-driven Extended target tracking, particularly a kind of foundation geometric proportion relation carrys out tracing machine The method of dynamic Extended target, is mainly used in image procossing, computer vision.Belong to target detection tracing skill in photoeletric measuring system Art field.

Background technology

In photoeletric measuring system, in order to improve tracking accuracy, the visual field of detector is smaller, and target size is the most inclined Greatly.Therefore, in detector, target presents the form of extension.Distant object imaging, due to atmospheric turbulance, thrashing and light The degraded factors such as the aberration of system cause target the fuzzyyest in the imaging of system, poor contrast；Additionally, target texture-free letter Breath, different, without characterizing and identifying clarification of objective information.Target there is also the obvious feature of attitudes vibration, along with target The change of attitude, trace point also can drift about therewith.Choose stable characteristic point and carry out locking tracking, be that Extended target is followed the tracks of The a great problem faced.

At present, the conventional algorithm for Extended target is coupling, including the coupling of the aspect such as gray scale, feature.Due to mesh Target is moved, and target is likely to occur the changes such as size, shape, attitude, adds at the various interference such as background, illumination, and image The precision problem of the minimum measurement unit of reason, matched jamming can not get the most optimal matched position, and this can bring the drift of trace point Move.Due to target texture-free and marked feature information, attitudes vibration is relatively big, and traditional tracking based on gray feature is worked as Easily with losing target during the target bigger attitudes vibration of appearance, for this situation, occur the most again using skeletal extraction characteristic point Follow the tracks of Extended target.Although this method can process the situation of attitudes vibration, but, Extended target this kind of for aircraft and Speech, when target occurs in time blocking (as vertical with imaging surface with wing place plane in fuselage), the simple side relying on skeletal extraction Method can not obtain fuselage or wing place axis, thus causes trace point to be lost, it is impossible to meets and is actually needed.Therefore in the urgent need to Study new method to adapt to the engineer applied demand followed the tracks of.

Summary of the invention

The technology of the present invention solves problem: for the deficiencies in the prior art, it is provided that a kind of extension based on geometric proportion relation Method for tracking target, inherently by abstract out for the geometry information of motor-driven Extended target, simultaneously according to known priori Information, according to certain geometric proportion relation, it is achieved target is at bigger attitudes vibration and the tenacious tracking under circumstance of occlusion.

For realizing such purpose, technical scheme: a kind of Extended target based on geometric proportion relation is followed the tracks of Method, comprises the steps:

Step one, Image semantic classification: use Gaussian smoothing filter that pending image is processed, remove the shadow of noise Ring, obtain filtered smoothed image；

Step 2, use Fuzzy C-Means Cluster Algorithm FCM(Fuzzy C-Means Cluster) step one is obtained Image after Ping Hua is split, it is thus achieved that bianry image；

Step 3, the image utilizing the method for skeletal extraction to obtain step 2 process, and extract the skeleton on aircraft Point；

Step 4, the skeletal point obtained according to step 3, choose frame head and tail characteristic point according to certain rule, be calculated as picture The length of fuselage in face；

In step 5, known reality, head is to the proportionate relationship of fuselage with wing intersection point with fuselage length, according to step 4 The fuselage length obtained is calculated as fuselage and wing intersection point in image planes；

Step 6, intersection point step 5 obtained combine interframe continuity correction trace point position, and last trace point is Trace point in present frame.

Wherein, in described step 2, use Fuzzy C-Means Cluster Algorithm FCM(Fuzzy C-Means Cluster) to step Rapid one obtain smooth after image split, it is thus achieved that the method for bianry image is:

Step (21), initialization: C=2 in the given cluster classification number C(present invention), set iteration stopping threshold epsilon, initialize Fuzzy partition matrix U⁽⁰⁾, iterations l=0, Fuzzy Weighting Exponent m (m=2 in the present invention)；

Step (22), by U^(l)Substitution formula (5), calculates cluster centre matrix V^(l):

$v_i=\frac{1}{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{\left(u_{\mathrm{ik}}\right)}^m}\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{\left(u_{\mathrm{ik}}\right)}^m{x}_k,i=1,...,c---\left(5\right)$

Wherein n is number of pixels to be clustered, and m is FUZZY WEIGHTED index, and c is cluster classification number, u_ikMould when being the l time iteration Stick with paste classification matrix U^(l)In the i-th row kth column element, x_kFor kth pixel value, v in image to be clustered_iGathering when being the l time iteration Class center matrix V^(l)Middle ith cluster central value；

Step (23), according to formula (6), utilize V^(l)Update U^(l), obtain new fuzzy classified matrix U^(l+1):

$u_{\mathrm{ik}}=\frac{1}{\underset{j=1}{\overset{c}{\mathrm{\Σ}}}{\left(\frac{d_{\mathrm{ik}}}{d_{\mathrm{jk}}}\right)}^{\frac{2}{m-1}}},i=1,...,c---\left(6\right)$

Wherein d_ikFor the Euclidean distance of kth element in image to be clustered Yu ith cluster center, d_jkFor figure to be clustered Kth element and the Euclidean distance of jth cluster centre in Xiang；

Step (24) if | | U^(l)-U^l+1| | ＜ ε, iteration stopping.Otherwise, put l=l+1, return step (22)；

Step (25), calculate the cluster centre that in image to be clustered, each pixel distance above-mentioned steps (21)-(24) obtain The Euclidean distance of value, is set to 1 by pixel value nearest for distance cluster centre, is otherwise set to 0, the binary map after thus being split Picture.

Wherein, in described step 3, the image utilizing the method for skeletal extraction to obtain step 2 processes, and extraction flies Skeletal point on machine, the present invention uses the iterative refinement algorithm gradually eliminating boundary point to extract skeleton, and algorithm is as follows:

If known target point is labeled as 1, background dot is labeled as 0, and definition boundary point is itself to be labeled as 1, and its 8-connects In region, at least point is labeled as the point of 0.Algorithm considers the 8-neighborhood centered by boundary point, and note central point is p₁, its 8 some clockwise around central points of neighborhood are designated as p respectively₂,p₃,...,p₉, wherein p₂At p₁Top；

Including border point being carried out two step operations:

(3.1) labelling meets the boundary point of following condition simultaneously:

(3.1.1) 2≤N (p₁)≤6;

(3.1.2) S (p₁)=1;

(3.1.3) p₂·p₄·p₆=0;

(3.1.4) p₄·p₆·p₈=0;

Wherein N (p₁) it is p₁Non-zero adjoint point number, S (p₁) it is with p₂,p₃,...,p₉,p₂For the value of these somes during sequence from 0 The number of → 1.After all boundary point is checked, all points that marked are removed.(42) labelling meets following simultaneously The boundary point of condition:

(3.2) labelling meets the boundary point of following condition simultaneously:

(3.2.1) 1≤N (p₁)≤6

(3.2.2) S (p₁)=1;

(3.2.3) p₂·p₄·p₈=0;

(3.2.4) p₂·p₆·p₈=0;

Above two step operations constitute an iteration, and algorithm iterates until point does not meets flag condition again, at this moment remains Under some composition skeletal point.

Wherein, in described step 4, the skeletal point obtained according to step 3, choose frame head and tail feature according to certain rule Point, is calculated as the length of fuselage in image planes, and its method is:

Step (41), the skeleton end points judged in the skeletal point that step 3 is obtained, calculate its position；

Step (42), the skeleton end points obtained for step (41), be considered as head by the end points of abscissa minimum and maximum With tail characteristic point (head is on a left side)；

Step (43), the head obtained according to step (42) and tail characteristic point, according to Euclidean distance calculate these 2 Distance in imaging surface, namely the fuselage length in imaging surface；

Wherein, in described step 5, before whole word change into: in described step 5, it is assumed that in body axis system, head Being labeled as an A at place, fuselage and wing place crossing point of axes are labeled as B, and tail is labeled as C at place, the length of the most known AB Degree and the ratio of AC length, be calculated as fuselage and wing intersection point in image planes according to the fuselage length of step 4 acquisition, and its method is:

The length of known above-mentioned AB and the ratio of AC length are R, step (42) head and tail in the imaging surface obtained Characteristic point coordinate is respectively (x_h,y_h),(x_t,y_t), required fuselage and wing cross point coordinate are (x_c,y_c), exist with head As a example by a left side, can obtain according to the geometric proportion relation between line segment:

$\frac{x_c-x_h}{x_t-x_h}=\frac{y_c-y_h}{y_t-y_h}=R---\left(7\right)$

Thus can obtain, fuselage and wing cross point transverse and longitudinal coordinate are respectively as shown in formula (8) and formula (9):

x_c=x_h+R·(x_t-x_h) (8)

y_c=y_h+R·(y_t-y_h) (9)

Wherein, in described step 6, intersection point step 5 obtained combines interframe continuity correction trace point position, finally Trace point be trace point in present frame.The trace point position that previous frame obtains is P_o, present frame obtains according to step 5 Position of intersecting point is P_n, as follows according to trace point method in interframe continuity correction present frame:

P_c=(1-α) P_o+α·P_n(10)

Wherein, α is modifying factor, and the present invention takes 0.95, P_cFor the trace point position through revised present frame.

The present invention having the beneficial effects that compared with prior art:

(1) image after segmentation is used the method for skeletal extraction to obtain the architectural feature point on Aircraft Targets by the present invention, Solving tradition relies on gray value to carry out the problem that feature point extraction causes poor robustness.

(2) present invention is according to certain Rule Extraction head and wing characteristic point from the skeletal point extracted, and solves tradition Gray scale angle point grid operator is to illumination and attitudes vibration sensitive issue.

(3) present invention provides a kind of motor-driven Extended target tracking utilizing geometric proportion relation, carries with simple dependence The skeletal point taken is compared to the method carrying out aircraft tracking, and the present invention has incorporated the aircraft geometry information of priori, depends on Obtain aircraft trace point under attitudes vibration blocks more greatly and certainly according to this geometry constant rate, solve and fly The row device problem that trace point is lost when occurring certainly to block (as vertical with imaging surface with wing place plane in fuselage), it is achieved that fly Row device tenacious tracking under attitudes vibration blocks more greatly and certainly.

Accompanying drawing explanation

Fig. 1 is the inventive method flowchart；

Fig. 2 is the result that the present invention is tracked location to the 1st two field picture of sequence used；

Fig. 3 is the result that the present invention is tracked location to the 86th two field picture of sequence used；

Fig. 4 is the result that the present invention is tracked location to the 215th two field picture of sequence used；

Fig. 5 is the result that the present invention is tracked location to the 945th two field picture of sequence used；

Fig. 6 is each trace point geometric locus that sequence used is tracked obtaining by the present invention.

Detailed description of the invention

Below in conjunction with the accompanying drawings embodiments of the invention are elaborated.The present embodiment is being front with technical solution of the present invention Put and implement, give detailed embodiment and concrete operating process, but protection scope of the present invention be not limited to Under embodiment.

The present invention is based on a kind of Extended target tracking based on geometric proportion relation, and input picture is single station flash ranging boat Mould image sequence.

As it is shown in figure 1, the invention provides a kind of Extended target tracking based on geometric proportion relation, comprise following Step:

Step one, Image semantic classification.Due to illumination or the defect of imaging system, the pending image of acquisition can be by noise Impact, thus affect follow-up process.Therefore, before Processing Algorithm continuous after execution, pending image is carried out pre-place Reason.This method uses Gaussian smoothing filter to remove effect of noise, obtains filtered smoothed image.

Step 2, use Fuzzy C-Means Cluster Algorithm FCM(Fuzzy C-Means Cluster) step one is obtained Image after Ping Hua is split, it is thus achieved that bianry image.In essence, image segmentation be one based on certain attribute to pixel Carry out the process classified.The property complicated and changeable of natural image determines many pixels Uncertain, thus consider that the image ration of division is more reasonable from the angle of fuzzy clustering.Fuzzy C-Means Cluster Algorithm (FCM, Fuzzy C-Means Cluster) it is to develop from hard C mean algorithm (HCM, Hard C-Means Cluster), its Essence is a kind of nonlinear iteration optimization method based on object function, and object function uses each pixel in image poly-with each Weighted Similarity between class center is estimated.The task of FCM algorithm is through iteration, selects rational fuzzy membership matrix Individual cluster centre, makes object function minimize, thus obtains optimal segmentation result.

Fuzzy C-Means Cluster Algorithm divides by the iteration optimization of object function is realized set, and it can represent publishes picture The degree belonged to a different category as each pixel.If n is pixel count to be clustered, c be classification number (c=2 in the present invention), m be fuzzy Weighted Index (takes m=2) in the present invention, it controls degree of membership all kinds of shared degree.The value of object function is each in image Pixel, to the weighted sum of squares of C cluster centre, is represented by:

$J_m(U,V)=\underset{i=1}{\overset{c}{\mathrm{\Σ}}}\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{u}_{\mathrm{ik}}^m{\left(d_{\mathrm{ik}}\right)}^2---\left(11\right)$

Wherein, u_ikFor the kth pixel degree of membership to the i-th class, d_ikFor the distance of kth pixel to the i-th class, U is fuzzy Classification matrix, V is cluster centre set.

Clustering criteria seeks to seek optimal group to (U, V) so that J_m(U, V) is minimum.J_mMinimization can by following repeatedly Realize for algorithm:

(2.1) initialize: C=2 in the given cluster classification number C(present invention), set iteration stopping threshold epsilon, initialize fuzzy Matrix dividing U⁽⁰⁾, iterations l=0, Fuzzy Weighting Exponent m (m=2 in the present invention)；

(2.2) by U^(l)Substitution formula (12), calculates cluster centre matrix V^(l):

$v_i=\frac{1}{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{\left(u_{\mathrm{ik}}\right)}^m}\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{\left(u_{\mathrm{ik}}\right)}^m{x}_k,i=1,...,c---\left(12\right)$

Wherein n is number of pixels to be clustered, and m is FUZZY WEIGHTED index, and c is cluster classification number, u_ikMould when being the l time iteration Stick with paste classification matrix U^(l)In the i-th row kth column element, x_kFor kth pixel value, v in image to be clustered_iGathering when being the l time iteration Class center matrix V^(l)Middle ith cluster central value；

(2.3) according to formula (13), V is utilized^(l)Update U^(l), obtain new fuzzy classified matrix U^(l+1):

$u_{\mathrm{ik}}=\frac{1}{\underset{j=1}{\overset{c}{\mathrm{\Σ}}}{\left(\frac{d_{\mathrm{ik}}}{d_{\mathrm{jk}}}\right)}^{\frac{2}{m-1}}},i=1,...,c---\left(13\right)$

Wherein, d_ikFor the Euclidean distance of kth element in image to be clustered Yu ith cluster center, similarly, d_jkFor Kth element and the Euclidean distance of jth cluster centre in image to be clustered；

(2.4) if | | U^(l)-U^l+1| | ＜ ε, iteration stopping.Otherwise put l=l+1, return step (2.2)；

(2.5) the cluster centre value that in image to be clustered, each pixel distance above-mentioned steps (2.1)-(2.4) obtain is calculated Euclidean distance, the distance nearest pixel value of cluster centre is set to 1, is otherwise set to 0, the binary map after thus being split Picture.

Experiment finds, uses fuzzy C-means clustering method to split image and can obtain than the side using Threshold segmentation The more preferable effect of method, is especially apparent natural image.This is owing to natural image is complicated and changeable, and level is complicated, each pixel Belong to the boundary which kind of neither one determines.Which kind of degree is each pixel is belonged to probability by fuzzy C-means clustering Form shows, and unlike hard C mean cluster (HCM) method directly thinks what each pixel determined belongs to which kind of, therefore Fuzzy C-means clustering method is used for image segmentation and can preferably embody the property complicated and changeable of natural image.

Step 3, the image utilizing the method for skeletal extraction to obtain step 2 process, and extract the skeleton on aircraft Point.Skeleton has the topological sum shape information identical with the original, it is possible to effectively describe object, is the several of a kind of function admirable What feature.The method realizing skeletal extraction has multiple thinking, Medial-Axis Transformation (medial axis transform, MAT) to be a kind of Relatively effective technology.But the method needs the distance that calculates all boundary points to all intra-zone points, amount of calculation is very Greatly.Therefore, the present invention uses and gradually eliminates the iterative refinement algorithm of boundary point to extract skeleton.

If known target point is labeled as 1, background dot is labeled as 0.Definition boundary point is itself to be labeled as 1, and its 8-connects In region, at least point is labeled as the point of 0.Algorithm considers the 8-neighborhood centered by boundary point, and note central point is p₁, its 8 some clockwise around central points of neighborhood are designated as p respectively₂,p₃,...,p₉, wherein p₂At p₁Top.

Algorithm includes carrying out border point two step operations:

(3.1) labelling meets the boundary point of following condition simultaneously:

(3.1.1) 2≤N (p₁)≤6;

(3.1.2) S (p₁)=1;

(3.1.3) p₂·p₄·p₆=0;

(3.1.4) p₄·p₆·p₈=0;

Wherein N (p₁) it is p₁Non-zero adjoint point number, S (p₁) it is with p₂,p₃,...,p₉,p₂For the value of these somes during sequence from 0 The number of → 1.After all boundary point is checked, all points that marked are removed.

(3.2) labelling meets the boundary point of following condition simultaneously:

(3.2.1) 1≤N (p₁)≤6

(3.2.2) S (p₁)=1;

(3.2.3) p₂·p₄·p₈=0;

(3.2.4) p₂·p₆·p₈=0;

Above two step operations constitute an iteration, and algorithm iterates until point does not meets flag condition again, at this moment remains Under some composition skeletal point.

Step 4, the skeletal point obtained according to step 3, choose frame head and tail characteristic point according to certain rule, be calculated as picture The length of fuselage in face, method is as follows:

(4.1), the skeleton end points that judges in the skeletal point that step 3 is obtained, calculate its position.Its rule is, for step The rapid three all skeletal point obtained, if only one of which skeletal point in its eight neighborhood, it is judged that this point would be skeleton end points；

(4.2) the skeleton end points, for (4.1) obtained, is considered as head and tail by the end points of abscissa minimum and maximum Characteristic point (head is on a left side)；

(4.3) head, according to (4.2) obtained and tail characteristic point, calculate at these 2 at imaging surface according to Euclidean distance In distance, namely the fuselage length in imaging surface；

Step 5, assume in body axis system, at head place, be labeled as an A, fuselage and wing place crossing point of axes Being labeled as B, tail is labeled as C at place, it is known that the length of AB and the ratio of AC length, the fuselage length obtained according to step 4 It is calculated as fuselage and wing intersection point in image planes.For the different aircraft of specific model, the geometry of itself is phase As.No matter, this specific geometric relationship is constant if there is which kind of attitudes vibration in aircraft.The invention reside in and probe into Such a invariance, thus solve the aircraft problem that trace point is lost under bigger attitudes vibration and circumstance of occlusion.Depend on Method according to these geometric proportion Relation acquisition tenacious tracking points is as follows:

The length of known above-mentioned AB and the ratio of AC length are R, step (4.2) head and tail in the imaging surface obtained Characteristic point coordinate is respectively (x_h,y_h),(x_t,y_t), required fuselage and wing cross point coordinate are (x_c,y_c), exist with head As a example by a left side, can obtain according to the geometric proportion relation between line segment:

$\frac{x_c-x_h}{x_t-x_h}=\frac{y_c-y_h}{y_t-y_h}=R---\left(14\right)$

Thus can obtain, fuselage and wing cross point transverse and longitudinal coordinate are respectively as shown in formula (15) and formula (16):

x_c=x_h+R·(x_t-x_h) (15)

y_c=y_h+R·(y_t-y_h) (16)

Wherein, in described step 6, intersection point step 5 obtained combines interframe continuity correction trace point position, finally Trace point be trace point in present frame.The trace point position that previous frame obtains is P_o, present frame obtains according to step 5 Position of intersecting point is P_n, as follows according to trace point method in interframe continuity correction present frame:

P_c=(1-α) P_o+α·P_n(17)

Wherein, α is modifying factor, and the present invention takes 0.95, P_cFor the trace point position through revised present frame.

In order to verify the robustness of the inventive method, experiment uses model plane image sequence, totally 1321 frame, intercepts wherein the 1st The tracking result that frame, the 86th frame, the 215th frame and the 945th two field picture obtain is respectively as shown in Fig. 2,3,4,5.Three ashes in figure Color rectangle frame is respectively tracking box centered by head, fuselage and wing intersection point and tail characteristic point, rectangle frame center Lycoperdon polymorphum Vitt cross represents the head of last extraction, fuselage and wing intersection point and tail characteristic point.It can be seen that when target is sent out (driftage, pitching and rolling, time as shown in Figure 2 and Figure 5), utilize the method for skeletal extraction accurately to obtain to raw attitudes vibration Stable head and tail characteristic point, utilize geometric proportion relation can obtain the intersection point of fuselage and wing simultaneously；When target is sent out It is conigenous and blocks that (fuselage is vertical with imaging surface with wing place plane, time as shown in Figure 3 and Figure 4), as long as fuselage institute can be extracted At axis, remain able to the constant rate according to head to fuselage with wing intersection point with fuselage length, it is thus achieved that stable tracking Point position, thus realize that aircraft is relatively big at attitudes vibration and tenacious tracking under circumstance of occlusion.

In order to verify the stability of the inventive method, the experiment front 400 frames to above-mentioned model plane image sequence (totally 1321 frame) Test, Fig. 6 be fuselage with wing intersection point in the position error of adjacent interframe respectively along X and the curve of Y-direction, stdx be with Track point interframe error standard deviation in X direction, stdy is trace point interframe error standard deviation in X direction.Can from figure Going out, trace point is respectively less than a pixel along the error to standard deviation of X and Y-direction respectively, when target occurs from blocking (fuselage and wing Place plane is vertical with imaging surface, time as shown in Figure 3 and Figure 4), still can obtain the intersection point of fuselage and wing, thus realize certainly Tenacious tracking under blocking.

Non-elaborated part of the present invention belongs to the known technology of those skilled in the art.

Those of ordinary skill in the art it should be appreciated that above embodiment be intended merely to illustrate the present invention, And it is not used as limitation of the invention, as long as in the spirit of the present invention, embodiment described above is changed, Modification all will fall in the range of claims of the present invention.

Claims (3)

1. an Extended target tracking based on geometric proportion relation, its feature comprises the steps:

Step one, Image semantic classification: use Gaussian smoothing filter that pending image is processed, remove effect of noise, Obtain filtered smoothed image；

It is smooth that step one is obtained by step 2, use Fuzzy C-Means Cluster Algorithm FCM (Fuzzy C-Means Cluster) After image split, it is thus achieved that bianry image；

Step 3, the image utilizing the method for skeletal extraction to obtain step 2 process, and extract the skeletal point on aircraft；

Step 4, the skeletal point obtained according to step 3, choose head and tail characteristic point according to certain rule, be calculated as image planes The length of middle fuselage；

Step 5, assume in body axis system, at head place, be labeled as an A, fuselage and wing place crossing point of axes labelling For B, tail is labeled as C at place, the length of the most known AB and the ratio of AC length, the fuselage length meter obtained according to step 4 It is counted as fuselage and wing intersection point in image planes；

Step 6, fuselage step 5 obtained are combined interframe continuity correction trace point position with wing intersection point, last with Track point is trace point in present frame；

Described step 4 chooses head and tail characteristic point according to certain rule, is calculated as the length of fuselage in image planes specifically real As follows:

Step (41), the skeleton end points judged in the skeletal point that step 3 is obtained, calculate its position；

Step (42), the skeleton end points obtained for step (41), be considered as head and machine by the end points of abscissa minimum and maximum Tail characteristic point；

Step (43), the head obtained according to step (42) and tail characteristic point, utilize Euclidean distance to calculate in imaging at these 2 Distance in face, namely the fuselage length in imaging surface.

Extended target tracking based on geometric proportion relation the most according to claim 1, it is characterised in that: described step It is calculated as fuselage in image planes in rapid five to be implemented as follows with wing intersection point:

Assuming in body axis system, be labeled as an A at head place, fuselage and wing place crossing point of axes are labeled as B, tail Being labeled as C at place, the length of the most known AB is R with the ratio of AC length, and required fuselage with wing cross point coordinate is (x_c,y_c), obtain according to the geometric proportion relation between line segment:

$\frac{x_c-x_h}{x_t-x_h}=\frac{y_c-y_h}{y_t-y_h}=R---\left(1\right)$

Thus, fuselage and wing cross point transverse and longitudinal coordinate are respectively as shown in formula (2) and formula (3):

x_c=x_h+R·(x_t-x_h) (2)

y_c=y_h+R·(y_t-y_h) (3)

(x_h,y_h),(x_t,y_t) it is respectively head and tail characteristic point coordinate in imaging surface.

Extended target tracking based on geometric proportion relation the most according to claim 1, it is characterised in that: described step In rapid six present frames, trace point is defined below:

P_c=(1-α) P_o+α·P_n (4)

Wherein, P_oThe trace point position obtained for previous frame, P_nFor fuselage and wing position of intersecting point, α is modifying factor, takes 0.95, P_cFor the trace point position through revised present frame.