CN105809700A - Detecting and positioning method of air refueling taper sleeve image blocked by fuel receiving plug - Google Patents

Abstract

The invention discloses a detecting and positioning method of an air refueling taper sleeve image blocked by a fuel receiving plug, and belongs to the technical field of computer image processing. A mathematical morphology method is used for quickly ejecting interferences including ribs, flexible pipes and the like on the periphery of a black round refueling port in the taper sleeve image, then, an image area where the refueling port in the taper sleeve is positioned is obtained through outline extraction, occlusion area contact ratio judgment and rejection, semi-round condition judgment and the like, a calculation process is simple, and the position and the area of the taper sleeve image under the situation of the blocking of the fuel receiving plug can be quickly detected and positioned.

Description

A kind of drogue image detection and location method blocked by oil plug

Technical field

The invention discloses a kind of drogue image detection and location method blocked by oil plug, belong to the technical field of Computer Image Processing.

Background technology

Existing most of autonomous air refuelling vision navigation method is arranged on fuel filling taper sleeve frequently with special optical markings, including LED light mark, artificial special color labelling etc..Air refuelling project (AAR) as autonomous in the U.S. just employs a set of VisNav system, this VisNav system is accomplished by installing multiple LED light mark on fuel filling taper sleeve, semiconductor position sensor PSD is then utilized to carry out framing (see Dong new people, Xu Yuejian, Chen Bo, " automatic air refuelling Research progress and key issue ", " air force engineering college journal (natural science edition) ", 2008 (12), 9 (6): 1-5).Additionally, the Wang Xufeng etc. of domestic air force engineering university establishes an independently flight-refueled vision Relative Navigation semi physical ground experiment platform, and red mark endless belt has been installed additional in fuel filling taper sleeve cross sectional area, contrast with Enhanced feature region Yu background image, thus color recognition can be passed through obtain framing information (see Wang Xufeng, Dong new people, Kong Xingwei, " the autonomous air refuelling emulation of plug taper sleeve type unmanned plane of machine vision auxiliary ", " science and technology and engineering ", 2013 (6), 13 (18): 5245-5250).

These methods above-mentioned require that fuel filling taper sleeve installs extra optical markings, improve the risk of air-refueling operation in particular for the LED light mark powered.Therefore a kind of AAR visible sensation method not against additional markers, merely with the feature of fuel filling taper sleeve own just has better versatility, convenience and safety.

By the analysis to fuel filling taper sleeve shape facility, its internal oil filler known is circular, radius is about 13cm, presents obvious black nahlock or subcircular oval block after imaging in the picture, can reduce, hence with this obvious own form feature, the amount of calculation that framing processes.Development along with unmanned plane, autonomous flight-refueled demand is also just more and more urgent, and the autonomous air refuelling vision navigation method not against extra optical labelling has higher versatility and convenience, the computer vision methods feature being based particularly on fuel filling taper sleeve inner circular oil filler is obvious, image processing speed is fast, and often photographic head is mounted in by oil plug side rear, fuel filling taper sleeve formation can be blocked as display foreground object by oil plug, solve to be blocked, by oil plug, the key that the fuel filling taper sleeve image detection and location under interference are autonomous air refuelling vision guided navigations, namely this programme is based on aforementioned thinking and produces.

Summary of the invention

The technical problem to be solved is to be in that to overcome the deficiency needing to install extra optical labelling in existing autonomous air refuelling method, extra Charging system brings higher air refuelling dangerous, from relying only on fuel filling taper sleeve unique characteristics, solve vision guided navigation is set out by the angle of oil plug image occlusion issue, provide a kind of drogue image detection and location method blocked by oil plug, the method utilizes Mathematical Morphology Method quickly to reject the umbrella frame of internal black circle oil filler periphery in tapered sleeve image, flexible pipes etc. disturb, then pass through contours extract, occlusion area registration judges and rejects, subcircular shape conditional judgments etc. obtain the image-region at the internal oil filler place of tapered sleeve.

The present invention adopts the following technical scheme that for achieving the above object

A kind of drogue image detection and location method blocked by oil plug, comprises the steps:

A, to being subject to the oil mask images blocked of plug to carry out reverse binary conversion treatment, mask images after reverse binary conversion treatment is carried out mathematical morphology closed operation and obtains the first binary mask image, the first binary mask image is carried out mathematics dynamics etching operation and obtains the second binary mask image；

B, tapered sleeve gray level image to be detected is carried out reverse binary conversion treatment；

C, carry out or computing to the tapered sleeve gray level image to be detected after reverse binary conversion treatment with by the oil mask images blocked of plug, to or the bianry image that obtains of computing carry out mathematical morphology and open operation to obtain bianry image；

D, the step C bianry image obtained is carried out contours extract obtaining circumference, on each circumference, the set of point constitutes circumference edge point set；

E, the area descending that each profile is surrounded, filter out the circumference less than contour area threshold value, the set of point on the circumference filtered out is concentrated from circumference marginal point and rejects；

F, the circumference that screening is stayed carry out the judgement of block masks registration, enter step F1 time misaligned, enter step F2 during coincidence,

F1, circumference is made directly the nearly circle target detection including regional extent constraint, circularity constraint, least square ellipse fitting condition constraint, retains and meet the objective contour that nearly circle target detection requires,

F2, circumference is blocked nearly circle target detection, retain to meet and blocks the objective contour that nearly circle target detection requires: first reject and circumference blocks interference, then the circumference rejected after blocking interference is included the nearly circle target detection of regional extent constraint, least square ellipse fitting condition constraint；

G, record the ellipse fitting parameter of each candidate target profile and calculate the logarithm circularity of each fitted ellipse, selecting the maximum candidate target profile of logarithm circularity as the final image positioning result of tapered sleeve inside oil filler.

As the further prioritization scheme of a kind of described drogue image detection and location method blocked by oil plug, step B adopts expression formula:Tapered sleeve gray level image to be detected is carried out reverse binary conversion treatment,

Wherein, (x y) represents image pixel coordinates, B₀(x y) represents that the tapered sleeve gray level image to be detected after reverse binary conversion treatment is at (x, y) grey scale pixel value of position, D_origin(x y) represents that tapered sleeve gray level image to be detected is at (x, y) grey scale pixel value of position, T₁Represent the threshold value of reverse binary conversion treatment tapered sleeve gray level image to be detected.

Further, a kind of described drogue image detection and location method blocked by oil plug, the circumference in step F, screening stayed carries out the judgement of block masks registration, method particularly includes:

Mask images after the anterior peripheral contour area bianry image extracted and reverse binary conversion treatment is done with computing to extract subimage interested,

Add up the number of pixels that in described subimage, block masks area pixel value is identical:

When number of pixels in statistics exceedes setting threshold value, it is determined that anterior peripheral contour area and block masks area coincidence,

When the number of pixels of statistics is less than or equal to setting threshold value, it is determined that anterior peripheral contour area is misaligned with block masks region.

Further, described a kind of by the oil drogue image detection and location method blocked of plug, step F2 rejects circumference blocks interference, method particularly includes:

Circumference region bianry image and the first binary mask image to extracting do and computing, to with computing after image carry out mathematical morphology and open operation, extract the profile of bianry image after opening operation, to obtain the point set of anterior peripheral profile, to weed out anterior peripheral profile point and concentrate the point with the second binary mask picture registration.

Further, in the step F2 of a kind of described drogue image detection and location method being subject to oil plug to block,

Described regional extent condition is: (x_d>T₃)&(y_d>T₃)&(y_d>T₄₁·x_d)&(x_d>T₄₁·y_d),

Described least square ellipse fitting condition is: { 2a > T₃}&{2b>T₃}&{a>T₄₁·b}&{b>T₄₁A} and

Wherein, Respectively on kth candidate target profile the 1st, 2 ..., n-th_kThe pixel coordinate of individual point, T₃For regional extent threshold value, T₄₁For regional percentage threshold value, & represents that logical AND operates, (a, b) for the length semiaxis of fitted ellipse,For setMeansigma methods and standard deviation, T₆For error of fitting average threshold value, T₇For fit error criterion difference limen value,(x_c,y_c) for the center point coordinate of fitted ellipse, θ is the anglec of rotation of fitted ellipse.

As described a kind of further prioritization scheme of drogue image detection and location method blocked by oil plug, step G calculates fitted ellipse logarithm circularity f (a, expression formula b) is:

The present invention adopts technique scheme, has the advantages that

(1) optical markings need not be additionally installed on fuel filling taper sleeve, merely with fuel filling taper sleeve unique characteristics, there is better versatility, convenience and safety；

(2) this method detects for the fuel filling taper sleeve image being subject under oil plug circumstance of occlusion, the method that binaryzation rear profile is extracted is adopted quickly to determine target area to be detected, adopt the mode that mask is rejected to eliminate the profile with block masks area coincidence and blocked, by oil plug, the image disruption brought, target place profile is determined whether eventually through contour shape, the method clear thinking, calculating process is simple, quickly detection and location can be subject to the tapered sleeve picture position under oil plug circumstance of occlusion and region.

Accompanying drawing explanation

Fig. 1 is the calculation flow chart of the inventive method；

Fig. 2 is the original tapered sleeve coloured image gathered；

Fig. 3 is reversely by oil plug block masks image；

Fig. 4 be tapered sleeve binary image with mask images or operation after bianry image；

Fig. 5 is the tapered sleeve bianry image after operation is opened in morphology；

Fig. 6 is the contours extract result after rejecting little area profile；

Fig. 7 is the contour area bianry image overlapped with masks area；

Fig. 8 is the contours extract result after Fig. 7 bianry image rejects masks area；

Fig. 9 is the result after Fig. 8 profile weeds out the profile point overlapped with masks area；

The internal oil filler image detection and location result of tapered sleeve that Figure 10 is final.

Detailed description of the invention

Below in conjunction with accompanying drawing, the technical scheme of invention is described in detail.

A kind of drogue image detection and location method blocked by oil plug, merely with fuel filling taper sleeve therein circle oil filler feature, by bianry image is carried out mathematical morphology operation, quickly reject internal black circle oil filler periphery interference in tapered sleeve image, then on the basis of contours extract, utilize the shield portions being judged by oil plug mask images and rejecting in profile, and utilizing subcircular shape condition quickly to obtain the subcircular profile of candidate, the candidate solution that finally selection logarithm circularity is maximum is as tapered sleeve image detection and location result.Specifically include following steps.

The mask images that step 1, preparation are blocked by oil plug: by being subject to oil plug occlusion area pixel value to be set to 0 in the mask images M that oil plug blocks, other region is set to 1；Calculating reverse mask images, namely occlusion area pixel value is 1, and other region is 0, and the mask images after reverse binary conversion treatment is designated as M_i；Employing radius is r₁Disc-shape structural element D_rThe mask images M being subject to oil plug to block is carried out mathematical morphology closed operation, it is thus achieved that the first binary mask image M_c；Employing radius is r₁Disc-shape structural element D_rTo image M_cCarry out mathematical morphology etching operation, it is thus achieved that the second binary mask image M_e, r₁Span is 3～12, and employing radius is r₁Disc-shape structural element D_rImage carries out mathematical morphology operation can enlarged area scope.

Step 2, it is loaded into tapered sleeve gray level image D to be detected_origin, adopt threshold value T₁To image D_originReverse binaryzation, it is thus achieved that the reversely tapered sleeve gray level image B to be detected after binary conversion treatment₀.Concrete conversion process is as follows:

$B_0(x,y)=\left\{\begin{array}{cc}0& D_{origin}(x,y)>T_1\\ 1& D_{origin}(x,y)\≤T_1\end{array}\right.,$

In formula, T₁Representing the threshold value of reverse binary conversion treatment tapered sleeve gray level image to be detected, (x y) represents image pixel coordinates, D_origin(x, y) and B₀(x y) represents tapered sleeve gray level image D to be detected respectively_originWith the tapered sleeve gray level image B to be detected after reverse binary conversion treatment₀At position (x, y) grey scale pixel value at place.

Tapered sleeve gray level image B to be detected after step 3, reverse binary conversion treatment₀Carry out with the mask images M being subject to oil plug to block or operate, obtaining bianry image B₁, employing radius is r₁Disc-shape structural element D_rTo bianry image B₁Carry out mathematical morphology and open operation, it is thus achieved that bianry image B₂。

Step 4, to bianry image B₂Carry out contours extract operation, it is thus achieved that circumference edge point setWherein, N₁Represent profile number, C_k(k=1,2 ..., N₁) represent the set of point on each profile, it is represented byN in formula_kRepresent contour edge point number.

In step 5, statistics contour edge point set C, the area of each profile, remembers profile C_kArea be s_k, to sequenceCarry out descending, and weed out area less than T₂The profile of (contour area threshold value), T₂Selection range be 60～200, the final profile numbering obtaining area from big to small, be designated asMeet following formula:

$s_{d_1}\≥s_{d_2}\≥...\≥s_m\≥T_2>s_{m+1}\≥...\≥s_{d_{N_1}}.$

Step 6, put mark f_mask=0, successively to numbering set d={d₁,d₂,...,d_mIn each profile be handled as follows:

It is d for current number_j=k, j=1,2 ..., the profile of m,

If f_mask=0, then enter step 7, current outline is carried out block masks registration judgement, if judged result is for overlapping, then put f_mask=1, and enter step 8, carry out blocking nearly circle target detection, record testing result；If judged result is misaligned, then enter step 9, directly current outline closely justified target detection operation, records testing result,

If f_mask=1, it is directly entered step 9, current outline is closely justified target detection operation, records testing result,

After all profiles have all processed, enter step 10.

Step 7, block masks registration judge: judge that current number is d_j=k, j=1,2 ..., the profile of m whether be subject to oil plug block masks area coincidence.

Step 701, extraction anterior peripheral contour area bianry image C₀, contour area pixel value is set to 1, and other region is set to 0,

Step 702, anterior peripheral contour area bianry image C₀With the mask images M after reverse binary conversion treatment_iCarry out AND-operation, it is thus achieved that bianry image C₁,

Step 703, statistics bianry image C₁Middle pixel value is the number of pixels of 1, if number of pixels exceedes threshold value T_N, then judge current outline region and be subject to oil plug block masks area coincidence；Otherwise judge misaligned.

Step 8, block nearly circle target detection: the anterior peripheral contour area bianry image meeting block masks registration Rule of judgment is blocked nearly circle target detection and judges.

Step 801, the anterior peripheral contour area bianry image C that will extract₀With the first binary mask image M_cCarry out AND-operation, it is thus achieved that bianry image C₂,

Step 802, employing radius are r₂Disc-shape structural element D_rTo bianry image C₂Carry out mathematical morphology and open operation, it is thus achieved that bianry image C₃, r₂Span is 3～6, and employing radius is r₂Disc-shape structural element image is carried out mathematical morphology open operation and can reject the interference such as umbrella frame.

Step 803, to bianry image C₃Carry out contours extract operation, it is thus achieved that circumference point set, and weed out all profile point concentration and the second binary mask image M_eThe point overlapped,

Step 804, entrance step 9, closely justify target detection operation, wherein adopt new regional percentage threshold value T each profile₄=T₄₁, do not carry out circularity conditional judgment, T₄₁Span is 0.35～0.95.

Step 9, closely round target detection: to profile point setRepresented profile closely justifies object judgement.

Step 901, candidate target profile must are fulfilled for following regional extent condition:

Calculate the scope (x of profile point set image coordinate_d,y_d) as follows:

$x_d=max\{x_{k1},x_{k2},...,x_{{\mathrm{kn}}_k}\}-min\{x_{k1},x_{k2},...,x_{{\mathrm{kn}}_k}\}$

$y_d=max\{y_{k1},y_{k2},...,y_{{\mathrm{kn}}_k}\}-min\{y_{k1},y_{k2},...,y_{{\mathrm{kn}}_k}\},$

Regional extent condition is:

(x_d>T₃)&(y_d>T₃)&(y_d>T₄·x_d)&(x_d>T₄·y_d),

In formula, T₃For regional extent threshold value, T₄For regional percentage threshold value, & represents that logical AND operates, T₃Selection range is 7～15, T₄Selection range is 0.5～0.95.

Step 902, candidate target profile must are fulfilled for following circularity condition:

$\frac{4{\mathrm{\πs}}_k}{l_k^2}>T_5$

In formula, T₅For circularity threshold value, T₅Selection range is 0.60～0.95, l_kGirth for kth candidate target profile.

Step 903, candidate target profile must are fulfilled for following least square ellipse fitting condition:

(1) least square ellipse matching: adopt method of least square to point setCarrying out ellipse fitting, fitting result is designated as (x_c,y_c, a, b, θ), wherein (x_c,y_c) for elliptical center point coordinates, (a, is b) oval length semiaxis, and θ is ELLIPTIC REVOLUTION angle.

(2) oval magnitude range judges: candidate target profile must is fulfilled for following condition:

{2a>T₃}&{2b>T₃}&{a>T₄·b}&{b>T₄·a}

(3) matching range error judges: calculate point setRange error e to fitted ellipse_iAs follows:

$e_i=\sqrt{ab}\·(\sqrt{{e_{xi}}^2+{e_{yi}}^2}-1),i=1,2,...,n_k$

In formula, (e_xi,e_yi) it is standardization elliptical point coordinate, can be calculated as follows:

$\left\{\begin{array}{c}{e}_{xi}=\frac{\cos \mathrm{\θ}}{a}(x_{ki}-x_c)+\frac{\sin \mathrm{\θ}}{a}(y_{ki}-y_c)\\ e_{yi}=-\frac{\sin \mathrm{\θ}}{a}(x_{ki}-x_c)+\frac{\cos \mathrm{\θ}}{a}(y_{ki}-y_c)\end{array}\right.,i=1,2,\mathrm{...},n_k$

Candidate target profile must is fulfilled for following condition:

In formula, T₆For error of fitting average threshold value, T₇For fit error criterion difference limen value,For setMeansigma methods and standard deviation, T₆Selection range is 0.01～0.10, T₇Selection range is 0.01～0.10.

Step 10, record meet the ellipse fitting parameter of the candidate region of above-mentioned condition, and calculate the logarithm circularity of ellipseSelect the maximum candidate solution of logarithm circularity as the final image positioning result of tapered sleeve inside oil filler.

Fig. 2 is the original tapered sleeve coloured image gathered, and number of pixels is 640 × 480.After being converted to gray level image, adopt the inventive method that tapered sleeve gray level image is processed, it is as follows that relevant parameter is set:

T₁=75, r₁=8, r₂=3, T₂=300, T₃=9, T₄=0.707, T₄₁=0.4, T₅=0.65, T₆=0.05, T₇=0.03

Fig. 3 is reverse by oil plug block masks image, and Fig. 4 is the bianry image after tapered sleeve image binaryzation, it can be clearly seen that internal oil filler periphery exists the interference such as umbrella frame, background, and is subject to oil plug to be formed to be blocked；Fig. 4 be tapered sleeve binary image with mask images or operation after bianry image；Fig. 5 is the tapered sleeve bianry image after operation is opened in morphology, and the interference of tapered sleeve oil filler periphery is isolated with oil filler region；Fig. 6 is the contours extract result after rejecting little area profile；Fig. 7 is the contour area bianry image overlapped with masks area；Fig. 8 is the contours extract result after Fig. 7 bianry image rejects masks area；Fig. 9 is after Fig. 8 profile weeds out the profile point overlapped with masks area as a result, it is possible to find out because blocking the interference profile point caused disallowable；The internal oil filler image detection and location of tapered sleeve that Figure 10 is final are as a result, it is possible to find out that detection and location result and tapered sleeve inner circular oil filler essentially coincide.The above results demonstrates the correctness of the inventive method.

Claims (6)

1. the drogue image detection and location method blocked by oil plug, it is characterised in that comprise the steps:

A, to being subject to the oil mask images blocked of plug to carry out reverse binary conversion treatment, mask images after reverse binary conversion treatment is carried out mathematical morphology closed operation and obtains the first binary mask image, the first binary mask image is carried out mathematics dynamics etching operation and obtains the second binary mask image；

B, tapered sleeve gray level image to be detected is carried out reverse binary conversion treatment；

C, carry out or computing to the tapered sleeve gray level image to be detected after reverse binary conversion treatment with by the oil mask images blocked of plug, to or the bianry image that obtains of computing carry out mathematical morphology and open operation to obtain bianry image；

D, the step C bianry image obtained is carried out contours extract obtaining circumference, on each circumference, the set of point constitutes circumference edge point set；

E, the area descending that each profile is surrounded, filter out the circumference less than contour area threshold value, the set of point on the circumference filtered out is concentrated from circumference marginal point and rejects；

F, the circumference that screening is stayed carry out the judgement of block masks registration, enter step F1 time misaligned, enter step F2 during coincidence,

F1, circumference is made directly the nearly circle target detection including regional extent constraint, circularity constraint, least square ellipse fitting condition constraint, retains and meet the objective contour that nearly circle target detection requires,

F2, circumference is blocked nearly circle target detection, retain to meet and blocks the objective contour that nearly circle target detection requires: first reject and circumference blocks interference, then the circumference rejected after blocking interference is included the nearly circle target detection of regional extent constraint, least square ellipse fitting condition constraint；

G, record the ellipse fitting parameter of each candidate target profile and calculate the logarithm circularity of each fitted ellipse, selecting the maximum candidate target profile of logarithm circularity as the final image positioning result of tapered sleeve inside oil filler.

2. a kind of by the oil drogue image detection and location method blocked of plug according to claim 1, it is characterised in that step B adopts expression formula:Tapered sleeve gray level image to be detected is carried out reverse binary conversion treatment,

Wherein, (x y) represents image pixel coordinates, B₀(x y) represents that the tapered sleeve gray level image to be detected after reverse binary conversion treatment is at (x, y) grey scale pixel value of position, D_origin(x y) represents that tapered sleeve gray level image to be detected is at (x, y) grey scale pixel value of position, T₁Represent the threshold value of reverse binary conversion treatment tapered sleeve gray level image to be detected.

It is 3. a kind of by the oil drogue image detection and location method blocked of plug according to claim 2, it is characterised in that the circumference in step F, screening stayed carries out the judgement of block masks registration, method particularly includes:

Mask images after the anterior peripheral contour area bianry image extracted and reverse binary conversion treatment is done with computing to extract subimage interested,

Add up the number of pixels that in described subimage, block masks area pixel value is identical:

When the number of pixels of statistics exceedes setting threshold value, it is determined that anterior peripheral contour area and block masks area coincidence,

The number of pixels of statistics is less than or equal to when setting threshold value, it is determined that anterior peripheral contour area is misaligned with block masks region.

It is 4. a kind of by the oil drogue image detection and location method blocked of plug according to claim 3, it is characterised in that step F2 to reject circumference blocks interference, method particularly includes:

Circumference region bianry image and the first binary mask image to extracting do and computing, to with computing after image carry out mathematical morphology and open operation, extract the profile of bianry image after opening operation, to obtain the point set of anterior peripheral profile, to weed out anterior peripheral profile point and concentrate the point with the second binary mask picture registration.

5. according to a kind of described drogue image detection and location method blocked by oil plug any in Claims 1-4, it is characterised in that in step F2,

Described regional extent condition is: (x_d>T₃)&(y_d>T₃)&(y_d>T₄₁·x_d)&(x_d>T₄₁·y_d),

Described least square ellipse fitting condition is: { 2a > T₃}&{2b>T₃}&{a>T₄₁·b}&{b>T₄₁A} and

Wherein,(x_k1,y_k1)、(x_k2,y_k2)、……、Respectively on kth candidate target profile the 1st, 2 ..., n-th_kThe pixel coordinate of individual point, T₃For regional extent threshold value, T₄₁For regional percentage threshold value, & represents that logical AND operates, (a, b) for the length semiaxis of fitted ellipse,For setMeansigma methods and standard deviation, T₆For error of fitting average threshold value, T₇For fit error criterion difference limen value,(x_c,y_c) for the center point coordinate of fitted ellipse, θ is the anglec of rotation of fitted ellipse.

6. a kind of by the oil drogue image detection and location method blocked of plug according to claim 5, it is characterised in that step G calculates fitted ellipse logarithm circularity f (a, expression formula b) is: