CN105666274A - Dinner plate edging method based on vision control - Google Patents

Abstract

The invention discloses a dinner plate edging method based on vision control. The method comprises the following steps: clamping and locating a dinner plate to be edged, and acquiring images of the dinner plate by virtue of two cameras which are symmetrically arranged at the two sides of the dinner plate; integrating the images through pre-treatment, image registration and fusion processing, and then extracting contour features to generate a complete dinner plate contour drawing; comparing the dinner plate contour drawing with a theoretical CAD model of the dinner plate, and obtaining a central position and a deviation angle of the current clamping and locating; and carrying out coordinate offset in a numerical control system according to the central position and the deviation angle, and controlling an edging unit to realize edging for the dinner plate by the numerical control system after the coordinate offset according to G code corresponding to the theoretical CAD model. According to the method disclosed by the invention, the two cameras which are symmetrically distributed at the two sides of the dinner plate are used for carrying out contour acquisition on the edge of the dinner plate, at a machining station, comprehensive acquisition and control for the contour of the dinner plate can be realized, and the G code is finally used for machining; and the method has the advantages of high edging quality, high machining efficiency and the like.

Description

The service plate edging method that a kind of view-based access control model controls

Technical field

The invention belongs to automatization's edging technical field, more particularly, to the service plate edging method that a kind of view-based access control model controls.

Background technology

Service plate is as the common tool in life, and it needs to carry out edging process in process of production. At present, service plate edging processing method is broadly divided into two classes: edging and automatization's edging manually, wherein automatization's edging processes owing to automaticity is high, efficiency advantages of higher is widely applied, and automatization's edging is most commonly used that the edging technology that view-based access control model controls in processing.

Existing techniques below scheme: the CN201520216408.2 of edging technology controlled about view-based access control model in prior art discloses the glass edging machining center of a kind of automatization, it launches infrared ray to workpiece by RF transmitter, and utilize the infrared ray anchor point on scanning device sensing workpiece profile, generating machining profile with this and obtain machining locus, mainshaft mechanism is processed according to machining locus; CN201420470616.0 discloses a kind of display glass substrate edge polisher, it seizes, by vision imaging, the anchor point that camera lens is seized on glass substrate, work out the reference route of processing edging according to the coordinate seizing anchor point, emery wheel and bearing fixture carry out edging work along machining benchmark route.

But, the edging technology that above-mentioned view-based access control model controls there is problems in that it adopts the scheme of single-station single camera, the problems such as inaccurate or Set and Positioning inaccuracy are transmitted due to drive mechanism, although edging efficiency is high, but edging weak effect, for the edging of the plate-rack product of high-grade environmental protection, such as service plates such as all kinds of sponges, edging required precision is higher, it is impossible to meet instructions for use.

Summary of the invention

Disadvantages described above or Improvement requirement for prior art, the invention provides the service plate edging method that a kind of view-based access control model controls, the camera of two plate-rack both sides being symmetrically distributed in Set and Positioning is wherein adopted at processing stations, the edge of plate-rack is carried out profile acquisition, and it is preprocessed, image is integrated by registration and fusion, the plate-rack profile diagram complete to generate a width, contrasted by the theoretical cad model with plate-rack, obtain center and the angle of deviation of current Set and Positioning, coordinate biasing is carried out according to center, the edging of plate-rack uniform high-efficiency is realized according to the G code programme-control that theoretical model is corresponding, the present invention can not only light all kinds of special-edge plate-rack of grinding, also there is edging quality good, edging precision is high, working (machining) efficiency advantages of higher.

For achieving the above object, the present invention proposes the service plate edging method that a kind of view-based access control model controls, and the method comprises the steps:

The service plate Set and Positioning of edging will be treated, utilize the contour images of two collected by camera service plates being symmetricly set on service plate both sides;

Preprocessed for the contour images collected, image registration and fusion treatment are integrated, then extracts the service plate profile diagram that contour feature is complete to generate a width;

The theoretical cad model of the service plate profile diagram of generation Yu service plate is contrasted, obtains center and the angle of deviation of current Set and Positioning;

Digital control system is carried out coordinate biasing by the center according to current Set and Positioning, then according to G code corresponding to theoretical cad model controls edging unit and realize the uniform high-efficiency edging of service plate.

As it is further preferred that described pretreatment is mean filter process, specifically processing in the following way:

If there being n pixel in image, the gray scale of each pixel is f_i(x, y), with g (x, y) representative process after the pixel value of this image, then:

$g(x,y)=\frac{1}{n}{\Σ}_{i=1}^n{f}_i(x,y).$

As it is further preferred that described image registration specifically processes in the following way:

Select piece image as with reference to figure, another piece image is as search graph, with reference to the image subblock chosen on figure centered by a certain impact point registration template as image, allow the movement that registration template is orderly on search graph, often move on to a position, registration template is carried out correlation ratio relatively with the corresponding part in search graph, until finding registration position.

As it is further preferred that described registration template carries out correlation ratio relatively with the corresponding part in search graph, until finding registration position, particularly as follows:

The coordinate of establishing quasi-mode plate upper left angle point is (x, y), the region being registered template covering on search graph is search subgraph, the coordinate of this search subgraph upper left angle point be (x ', y '), relatively (x, y) with the gray value acquiring size optimal registration point at (x ', y ') place.

As it is further preferred that compare (x, y) with the gray value acquiring size optimal registration point at (x ', y ') place particularly as follows: calculate the minima of difference of two squares D in region of search (x ', y '), it is thus achieved that optimal registration point, described inWherein, (m, n), (m, n) respectively (m, n) the search subgraph at pixel coordinate place and the gray value of actual subgraph, c and d represents length and the width of registration template for N for M.

As it is preferred that, described fusion treatment is specially and moves line by line in the overlapping region of two images with a window, using pixel minimum for every a line grey value difference as image mosaic point, it is sequentially connected with each splice point and obtains best splicing seams, then smoothing processing it is weighted, image overlapping region is carried out transition, it is achieved fusion naturally smooth between adjacent image.

As it is further preferred that described extraction contour feature specifically includes following steps with the service plate profile diagram generating a width complete:

Divide an image into background and two regions of profile target, obtain gray value when gray variance is maximum between the class making background and profile target as optimal segmenting threshold t;

Make gray value ω in image_iThe new gray value of the pixel less than or equal to t is 0, and the new gray value of the pixel more than t is 1, and segmentation obtains two-valued function image: $F\left({\mathrm{\ω}}_1\right)=\left\{\begin{array}{cc}0,& {\mathrm{\ω}}_i\≤t\\ 1,& {\mathrm{\ω}}_i>t\end{array}\right.;$ Utilize two-valued function image zooming-out bianry image profile.

As it is further preferred that obtain and make gray value that between the class of background and profile target, gray variance is maximum as optimal segmenting threshold t, specifically including:

If the gray level of image is L level, the probability that background occurs is P_A, the probability that profile target occurs is P_B, the average gray value of background area is ω_A, the average gray value of profile target area is ω_B, then:

$P_A={\mathrm{\Σ}}_{i=0}^t{P}_i,P_B={\mathrm{\Σ}}_{i=t+1}^{L-1}{P}_i=1-P_A;$

${\mathrm{\ω}}_A={\mathrm{\Σ}}_{i=0}^t{\mathrm{iP}}_i/P_A,{\mathrm{\ω}}_B={\mathrm{\Σ}}_{i=t+1}^{L-1}{\mathrm{iP}}_i/P_B;$

The meansigma methods ω of gradation of image₀It is then:

${\mathrm{\ω}}_0=P_A{\mathrm{\ω}}_A+P_B{\mathrm{\ω}}_B={\mathrm{\Σ}}_{i=0}^{L-1}{\mathrm{iP}}_i;$

Gray variance σ between the class of background area and profile target area²For:

σ²=P_A(ω_A-ω₀)²+P_B(ω_B-ω₀)²;

Try to achieve and image makes gray variance σ between above-mentioned class²It is optimal segmenting threshold t for gray value time maximum.

In general, by the contemplated above technical scheme of the present invention compared with prior art, mainly possess following technological merit:

1. in the present invention, adopt two cameras of single-station, and two cameras are symmetrically distributed in the both sides of the plate-rack of Set and Positioning, can realize at clamping position, the whole of plate-rack edge being gathered, eliminate because of trembling and the error that causes when conveyer belt transmits inaccurate or Set and Positioning, make edging precision higher, complete the profile acquisition that plate-rack is complete and control at grinding station.

2. in the present invention, after plate-rack profile of two collected by cameras, once gather again after plate-rack 90-degree rotation, four profile diagrams collected twice merge, and obtain the profile of whole plate-rack, and the plate-rack profile got and existing cad model are carried out registration, to find the position at plate-rack center under current workpiece coordinate system, then carry out coordinate biasing, can further reduce error, improve edging precision.

3. in the present invention, preprocessed for the image collected, image registration and fusion treatment are integrated, then the plate-rack profile diagram that contour feature is complete to generate a width is extracted, carry out studying to pretreatment, image registration, fusion treatment and contours extract mode simultaneously and set, obtain and be applicable to the best pretreatment of tableware edging, image registration, fusion treatment and contours extract mode, improve picture quality, fusion accuracy and edging precision and efficiency further.

Accompanying drawing explanation

Fig. 1 is based on the front view of the edging device of visual spatial attention;

Fig. 2 is based on the side view of the edging device of visual spatial attention;

Fig. 3 (a), (b) are the schematic diagrams of desirable clip position and actual clip position.

Detailed description of the invention

In order to make the purpose of the present invention, technical scheme and advantage clearly understand, below in conjunction with drawings and Examples, the present invention is further elaborated. Should be appreciated that specific embodiment described herein is only in order to explain the present invention, is not intended to limit the present invention. As long as just can be mutually combined additionally, technical characteristic involved in each embodiment of invention described below does not constitute conflict each other.

The service plate edging method that a kind of view-based access control model of the embodiment of the present invention controls, its ultimate principle is: with edging device, the edge of service plate is carried out automatization's edging, grinding station adopts two cameras, two cameras are symmetrically distributed in the both sides of the service plate of Set and Positioning, the edge of service plate is carried out profile acquisition by two cameras, the image collected is integrated, through pretreatment, registration and fusion, generate service plate profile diagram, then contrast with the theoretical cad model of service plate, thus getting center and the angle of deviation of current Set and Positioning, digital control system carries out coordinate biasing, the G code corresponding according to theoretical model is controlled, finally realize the edging of the uniform high-efficiency of service plate. as shown in Figure 1-2, edging device includes the cantilever beam 1 for fixed camera 5, drives the rotating servo motor 2 that plate-rack 7 rotates, for clamping the newel 3 of plate-rack, for carrying out the bistrique 4 of edging and for transmitting the synchrome conveying belt 6 of plate-rack.

The method specifically includes following steps:

(1) the service plate Set and Positioning of edging will be treated, utilize the image of two collected by camera service plates being symmetricly set on service plate both sides, the multiple image of the Same Scene of service plate different visual angles, different time can be obtained by symmetrically arranged two cameras, acquisition mode is after service plate clamps, service plate profile of two collected by cameras, then once gather again after service plate 90-degree rotation, obtain four service plate profile diagrams.

(2) preprocessed for the image collected, image registration and fusion treatment are integrated, then extract the service plate profile diagram that contour feature is complete to generate a width, below each is processed step and be described in detail:

Pretreatment: pretreatment is to improve processing accuracy and speed, after obtaining multiple such as four images to be spliced, first noise reduction pretreatment it is filtered, to avoid the output that the noise on image that the deviation because of local is bigger splices to impact, the present embodiment adopts the mode of mean filter to process, and specifically carries out in the following way:

If there being n pixel in image near current pixel, (x, y) for the two dimensional surface position at pixel place, the gray scale of each pixel is f_i(x, y), with g (x, y) representative process after the gray value of this image current pixel, then:

$g(x,y)=\frac{1}{n}{\mathrm{\Σ}}_{i=1}^n{f}_i(x,y).$

Image registration: the four width images that image registration is the Same Scene in order to different visual angles, different time be obtained after pre-processing transform under the same coordinate system, then the match information extracted in image finds best match position, the present embodiment adopts the method compared one by one to carry out registration, specifically carries out in the following way:

Taking piece image therein as joining reference of reference (namely with reference to figure), another piece image is as search graph; With reference to figure chooses the image subblock centered by the impact point registration template N as image successively from the upper left corner, then movement orderly on search graph for registration template N is allowed, often move on to a position, registration template N is carried out correlation ratio relatively with the corresponding part in search graph, until finding registration position.

Concrete, if search graph is sized to the rectangle of a × b, registration masterplate N is sized to the rectangle of c × d, the coordinate of establishing quasi-mode plate N upper left angle point is (x, y), search graph intercepts one and the equirotal block image of template N, the total individual block image of (a-c+1) × (b-d+1), the target of registration finds out an image most like with registration template exactly in the individual block image of (a-c+1) × (b-d+1), and the datum mark that this image is corresponding is optimal registration point.

More specifically, registration template N moves on search graph, the region covered by template on search graph is search subgraph M, if the coordinate of this search subgraph M upper left angle point is (x ', y '), with as a reference point, then compare that (x, y) with the size of (x ', y ') place's gray value, if both are consistent, then N (x, y) and the difference of M (x ', y ') be zero, and in real image registration process, two width images are difficult to completely the same, only need to ensure N (x, y) with M (x ', y ') difference minimum, to obtain optimal registration point.

Further, the present embodiment adopts difference of two squares method to obtain optimal registration point, calculates the minima of difference of two squares D (x ', y '), it is thus achieved that optimal registration point, concrete:

Wherein, M (m, n), N (m, n) respectively (m, n) the search subgraph at pixel coordinate place and the gray value of actual subgraph, due toIt is a constant definite value for searching for the half-tone information of subgraph;

OrderAccording to Cauchy-Schwarz inequality known 0 < R < 1, and and if only if:

$\frac{M(1,1)}{N(1,1)}=\frac{M(1,2)}{N(1,2)}=\frac{M(1,3)}{N(1,3)}=...=\frac{M(m,n)}{N(m,n)}$ Time, R (x ', y ') takes maximum, and now D (x ', y ') is minimum, then this search subgraph and actual subgraph are most like, and (x ', y ') some place cooperation at this moment is the most accurate.

Fusion treatment: significantly connect vestige by image overlapping region can be made after image registration to occur, cause the fuzzy of image border and profile to a certain extent, fusion treatment is precisely in order to eliminate the impact of stitching thread and luminosity and colourity, eliminate the vestige of splicing, making image vision reach gratifying effect, the present embodiment adopts linear weighted function algorithm to merge, for four profile diagrams, merge between two respectively, particularly as follows:

Move line by line in the overlapping region of two images with a window, by pixel minimum for every a line grey value difference, namely the most similar between neighbor during splicing pixel is as image mosaic point, it is sequentially connected with each splice point composition splicing line using as best splicing seams, then smoothing processing it is weighted, image overlapping region is carried out transition, it is achieved fusion naturally smooth between adjacent image.In like manner, all the other two profile diagrams can merge to constitute the whole figure of a width service plate successively.

More specifically, if two adjacent width image lap pixel value respectively P₁(x, y) and P₂(x, y), then be weighted the pixel value P after smoothing processing₃(x, y) has:

P₃(x, y)=α P₁(x,y)+(1-α)P₂(x,y)

Wherein α is variable factor, when lap position before overlapping widths in the overlapping region of 1/3 and the overlapping region of rear 1/3 time, P₃(x, y) is the pixel value in two width image each regions, when the middle part at overlapping widths, the lap position, after being weighted smoothing processing, and now P₃(x, y)=α P₁(x,y)+(1-α)P₂(x, y). If the pixel value after weighting smoothing processing is P₃(x, y), namely through the pixel value of fusion treatment.

$\left\{\begin{array}{c}{P}_3(x,y)=P_1(x,y)(0<l<\frac{L}{3})\\ P_3(x,y)=\mathrm{\&alpha;}{P}_1(x,y)+(1-\mathrm{\&alpha;})P_2(x,y)(\frac{L}{3}<l<\frac{2L}{3})\\ P_3(x,y)=P_2(x,y)(\frac{2L}{3}<l<L)\end{array}\right.$

Wherein, l is pixel position in horizontal direction in overlapping region, and L is the width that horizontal direction epigraph is overlapping, takes α=(3l-L)/L in the present invention.

Contours extract: in order to find and demonstrate the obvious pixel of brightness flop in image, contours extract need to be carried out, the present embodiment adopts grey relevant dynamic matrix to carry out image segmentation, selected optimal segmenting threshold, then each pixel of pending image is carried out standalone processes, compare by its gray value and set threshold value, thus obtaining the artwork master of binaryzation. Particularly as follows:

A) adopting maximum between-cluster variance algorithm, divide an image into background and two parts of profile target according to gamma characteristic, partitioning standards is for choosing threshold value so that the variance between background and target is maximum.

Being provided with L gray level, the probability that each gray level occurs is P, then obtain following image grey level histogram:

$O={\mathrm{\&Sigma;}}_{i=0}^{L-1}{n}_i,P_i=n_i/O;$

Wherein: O is the sum of image pixel, n_iFor the number of pixel that gray value is i, P_iBeing the i probability occurred for gray value, i is gray value.

If t is optimal segmenting threshold, P_ARepresent the probability that background occurs, P_BRepresenting the probability that profile target occurs, tonal range is 0-N, then the probability P that background occurs_AFor:

$P_A={\mathrm{\&Sigma;}}_{i=0}^t{P}_i;$

The probability P that target occurs_BFor:

$P_B={\mathrm{\&Sigma;}}_{i=t+1}^{L-1}{P}_i=1-P_A;$

The average gray value ω of background area_AFor:

${\mathrm{\&omega;}}_A={\mathrm{\&Sigma;}}_{i=0}^t{\mathrm{iP}}_i/P_A;$

The average gray value ω of profile target area_BFor:

${\mathrm{\&omega;}}_B={\mathrm{\&Sigma;}}_{i=t+1}^{L-1}{\mathrm{iP}}_i/P_B;$

The meansigma methods ω of gradation of image₀For:

${\mathrm{\&omega;}}_0=P_A{\mathrm{\&omega;}}_A+P_B{\mathrm{\&omega;}}_B={\mathrm{\&Sigma;}}_{i=0}^{L-1}{\mathrm{iP}}_i;$

The inter-class variance σ of background area and profile target area²:

σ²=P_A(ω_A-ω₀)²+P_B(ω_B-ω₀)²

The optimal segmenting threshold t tried to achieve enables to gray variance σ between the class of background area and profile target area in image²For gray value time maximum.

B) all gray value ω in image are made_iThe new gray value of the pixel less than or equal to t is all 0, and the new gray value of the pixel more than t is all 1, and segmentation obtains two-valued function image:

$F\left({\mathrm{\&omega;}}_i\right)=\left\{\begin{array}{cc}0,& {\mathrm{\&omega;}}_i\&le;t\\ 1,& {\mathrm{\&omega;}}_i>t\end{array}\right.;$

Utilize two-valued function image zooming-out bianry image profile.

Concrete, utilize the two-valued function image zooming-out bianry image profile to be:

By two-valued function image F (ω_i) be 1 point be extracted as black color dots, black color dots is stitched together and is linked to be curve, can obtain the circumference of workpiece.

(3) the theoretical cad model of the service plate profile diagram of generation Yu service plate is contrasted, obtain the center of current Set and Positioning, specifically in order to eliminate the location deflection and angular deviation produced because transmission and clamping are inaccurate, and calculate the size of the skew with the central point of the service plate profile diagram of generation of the center of current Set and Positioning and angle of deviation α. When plate-rack clamping station due to transmit inaccurate, tremble the deviations caused, as Fig. 3 (a), (b), center and angle all there occurs deviation, the service plate profile diagram of Nature creating also creates position and angular deviation.Such as Fig. 3 (a), offset deviation length is the distance between O1, O2 2; Such as Fig. 3 (b), when there is not angular deflection error in theory, grinding position is A1, and after there is angular deflection, A1 rotates to the position of A2, currently grinding position should arrive B point, it is therefore desirable to use rotation transformation function, it is possible to specified angle α will be rotated according to the machining locus of theoretical CAD profile establishment around center of rotation.

(4) according to the center of current Set and Positioning, digital control system carried out coordinate biasing (i.e. displacement bias and angle offset), then according to G code corresponding to theoretical cad model controls edging unit and realizes the uniform high-efficiency edging of service plate, particularly as follows: the overall some position in known G code program is translated and rotation process by digital control system. First carry out coordinate translation biasing, the profile initial point collected is overlapped with the profile initial point of cad model; Then carry out coordinate again and rotate biasing, by the G code of the cad model point position cosine value divided by the angle of deviation, bias through twice coordinate, processing can be completed according to the G code of cad model.

Detailed process is as follows:

First, as Fig. 3 (a) profile diagram central point O1 generated offsets to the center O2 of plate-rack theory CAD profile, to eliminate offset deviation; Then, rotation alpha angle as counterclockwise in Fig. 3 (b), to eliminate angular deviation.

Concrete, utilize G code coordinate offset instructions can realize location bias and angle offset, successfully elimination transmission is inaccurate, tremble the deviations caused, it is achieved efficient G code processing.

Specifically, location bias can by the machining locus translation designated displacement according to theoretical CAD profile establishment, G code format sample:

G90G01X[100+(#50001)]Y[200+(#50001)]F800

X[150+(#50001)]Y[240+(#50001)]F700

X[180+(#50001)]Y[260+(#50001)]F900

…

Wherein #50001 is the user's macro-variable that can customize, and here by side-play amount, namely the distance of O1, O2 2 is assigned to #50001, can realize displacement bias.

Specifically, angle offset, in G code, use rotation transformation function, it is possible to specified angle α will be rotated according to the machining locus of theoretical CAD profile establishment around center of rotation. G code format sample:

G17

G68X0Y0P[#50002]

G69

…

Wherein, G17 selects XOY Plane of rotation; G68X0Y0P [#50002] _ set up rotation transformation, P [#50002] refer to rotate [#50002] degree, user's macro-variable that #50002 also can customize, and here offset angle are assigned to #50002, and namely all positions have rotated α degree; G69 cancels rotation transformation.

Those skilled in the art will readily understand; the foregoing is only presently preferred embodiments of the present invention; not in order to limit the present invention, all any amendment, equivalent replacement and improvement etc. made within the spirit and principles in the present invention, should be included within protection scope of the present invention.

Claims (8)

1. the service plate edging method that a view-based access control model controls, it is characterised in that the method comprises the steps:

The service plate Set and Positioning of edging will be treated, utilize the contour images of two collected by camera service plates being symmetricly set on service plate both sides;

Preprocessed for the contour images collected, image registration and fusion treatment are integrated, then extracts the service plate profile diagram that contour feature is complete to generate a width;

The theoretical cad model of the service plate profile diagram of generation Yu service plate is contrasted, obtains center and the angle of deviation of current Set and Positioning;

Digital control system is carried out coordinate biasing by center and the angle of deviation according to current Set and Positioning, then according to G code corresponding to theoretical cad model controls edging unit and realize the uniform high-efficiency edging of service plate.

2. the service plate edging method that view-based access control model as claimed in claim 1 controls, it is characterised in that described pretreatment is that mean filter processes, and specifically processes in the following way:

If there being n pixel in image, the gray scale of each pixel is f_i(x, y), with g (x, y) representative process after the pixel value of this image, then:

$g(x,y)=\frac{1}{n}{\mathrm{\&Sigma;}}_{i=1}^n{f}_i(x,y).$

3. the service plate edging method that view-based access control model as claimed in claim 1 or 2 controls, it is characterised in that described image registration specifically processes in the following way:

Select piece image as with reference to figure, another piece image is as search graph, with reference to the image subblock chosen on figure centered by a certain impact point registration template as image, allow the movement that registration template is orderly on search graph, often move on to a position, registration template is carried out correlation ratio relatively with the corresponding part in search graph, until finding registration position.

4. the service plate edging method that view-based access control model as claimed in claim 3 controls, it is characterised in that described registration template carries out correlation ratio relatively with the corresponding part in search graph, until finding registration position, particularly as follows:

The coordinate of establishing quasi-mode plate upper left angle point is (x, y), the region being registered template covering on search graph is search subgraph, the coordinate of this search subgraph upper left angle point be (x ', y '), relatively (x, y) with the gray value acquiring size optimal registration point at (x ', y ') place.

5. the service plate edging method that view-based access control model as claimed in claim 4 controls, it is characterized in that, relatively (x, y) with (x ', y ') place gray value acquiring size optimal registration point particularly as follows: calculate difference of two squares D in region of search (x ', y ') minima, it is thus achieved that optimal registration point, wherein:

In formula, (m, n), (m, n) respectively (m, n) the search subgraph at pixel coordinate place and the gray value of actual subgraph, c and d represents length and the width of registration template for N for M.

6. the service plate edging method that view-based access control model as claimed in claim 1 or 2 controls, it is characterized in that, described fusion treatment is specially and moves line by line in the overlapping region of two images with a window, using pixel minimum for every a line grey value difference as image mosaic point, it is sequentially connected with each splice point and obtains best splicing seams, then it is weighted smoothing processing, image overlapping region is carried out transition, it is achieved fusion naturally smooth between adjacent image.

7. the service plate edging method that view-based access control model as claimed in claim 1 or 2 controls, it is characterised in that described extraction contour feature specifically includes following steps with the service plate profile diagram generating a width complete:

Divide an image into background and two regions of profile target, obtain gray value when gray variance is maximum between the class making background and profile target as optimal segmenting threshold t;

Make gray value ω in image_iThe new gray value of the pixel less than or equal to t is 0, and the new gray value of the pixel more than t is 1, and segmentation obtains two-valued function image: $F\left({\mathrm{\&omega;}}_i\right)=\left\{\begin{array}{cc}0,& {\mathrm{\&omega;}}_i\&le;t\\ 1,& {\mathrm{\&omega;}}_i>t\end{array}\right.;$ Utilize two-valued function image zooming-out bianry image profile.

8. the edging method that view-based access control model as claimed in claim 7 controls, it is characterised in that obtain gray value that between the class making background and profile target, gray variance is maximum as optimal segmenting threshold t, specifically include:

If the gray level of image is L level, the probability that background occurs is P_A, the probability that profile target occurs is P_B, the average gray value of background area is ω_A, the average gray value of profile target area is ω_B, then:

$P_A={\mathrm{\&Sigma;}}_{i=0}^t{P}_i,P_B={\mathrm{\&Sigma;}}_{i=t+1}^{L-1}{P}_i=1-P_A;$

${\mathrm{\&omega;}}_A={\mathrm{\&Sigma;}}_{i=0}^t{\mathrm{iP}}_i/P_A,{\mathrm{\&omega;}}_B={\mathrm{\&Sigma;}}_{i=t+1}^{L-1}{\mathrm{iP}}_1/P_B;$

The meansigma methods ω of gradation of image₀It is then:

${\mathrm{\&omega;}}_0=P_A{\mathrm{\&omega;}}_A+P_B{\mathrm{\&omega;}}_B={\mathrm{\&Sigma;}}_{i=0}^{L-1}{\mathrm{iP}}_i;$

Gray variance σ between the class of background area and profile target area²For:

σ²=P_A(ω_A-ω₀)²+P_B(ω_B-ω₀)²;

Try to achieve and image makes gray variance σ between above-mentioned class²It is optimal segmenting threshold t for gray value time maximum.