CN101261685A - Automatic input device for cloth sample image based on image vector technology - Google Patents

Abstract

The invention relates to an automatic costume sample-piece recording device based on the image vectorization technique. The device comprises a platform used for placing the costume sample-piece, a digital camera used for shooting pictures of the costume sample-piece and a compute used for processing images. A bracket is arranged on the platform, the digital camera is arranged on the bracket, the platform is arranged in the visual scope of the digital camera, and the digital camera is connected with the computer. The digital camera is marked, the images of the costume sample-piece on the platform is collected by the digital camera, the dot-matrix pictures of the costume sample-piece shot by the digital camera are input into the computer, then the dot-matrix pictures of the costume sample-piece are converted into vectorgraph files which can be identified by various costume CAD software by arithmetic of image pre-processing and image vectorization, etc., thus being convenient for the later grading, material-layout and cutting of the costume sample-piece. The automatic costume sample-piece recording device has easy operation, high input efficiency and low cost.

Description

Automatic input device for cloth sample image based on image vector technology

Technical field

The present invention relates to a kind of automatic input device for cloth sample image.

Background technology

Clothing CAD technology has produced far-reaching influence in international rag trade since producing, it makes garment enterprise improve ability to react to market conditions and competitive power.China introduces, digests in the process of advanced foreign technology in eighties of last century the eighties, has researched and developed suitable China clothes practitioner's CAD system, and through the application and the development in twenty or thirty year, the economic benefit of garment enterprise has obtained significant raising.

But existing Garment CAD system generally adopts the method for digitizer input to read in print when putting with the benchmark fashion plate yard.Digitizer is a kind of graphic input device, form by electromagnetic induction plate, vernier and corresponding electronic circuit, when the user on the electromagnetic induction plate moving cursor to assigned address, and when the intersection point of crosshair aimed at digitized, the button of pressing, the order symbol that digitizer is then corresponding with this moment and the position coordinate value of this point are arranged in one group of orderly information, are sent to principal computer by interface (using serial line interface) then more.

If with digitizer typing fashion plate, need experienced making sheet teacher to describe according to the pointwise of print edge, can only import separately and revise a print at every turn, the time-consuming consumption power of manual operations, inefficiency, and precision is not high, and the raising that these have all restricted the garment enterprise production efficiency and the quality of production greatly is unfavorable for the robotization of enterprise version room plate-making.At present digitizer holds at high price, and according to the size of breadth, price does not wait 10,000 to 30,000, and its expensive machine price increase enterprise's production cost of knowing clearly has influenced the popularity of digitizer.

Summary of the invention

In order to overcome limited, the more high deficiency of price of troublesome poeration, inefficiency, precision of existing existing digitizer, the invention provides a kind of simple to operate, input efficiency and the higher low-cost automatic input device for cloth sample image of precision based on image vector technology.

The technical solution adopted for the present invention to solve the technical problems is:

A kind of automatic input device for cloth sample image based on image vector technology, comprise the platform of placing fashion plate, the computing machine that is used to obtain the digital camera of fashion plate picture and is used to handle image, mounting bracket on the described platform, described digital camera is installed on the described support, described platform is positioned at the visual range of digital camera, described digital camera is connected with computing machine, and described computing machine comprises:

Camera calibration module is used for the radial distortion of correcting digital video camera, sets up the corresponding relation between the world coordinate system of pixel coordinate system on the plane of delineation and spatial point;

Image binaryzation is cut apart and the denoising module, is used to receive the image that digital camera obtains, and is binary map with the colored bitmap-converted of being gathered, and binary map is carried out denoising;

Image outline extracts tracking module, is used for the binary map every bit is judged, for black, and its 8 consecutive point with this point deletion, obtain the profile of binary map when all being black as this point; And from first frontier point, the profile of tracking image, record delimitation point coordinate value has been followed the tracks of behind the closed curve one of deletion, begins the tracking of next bar closed curve, up to the tracking of finishing all closed curves;

The contour feature point extraction module, be used for according to coordinate sequence every closed curve extract minutiae, P (k) is [X (k), Y (k)], the coordinate point sequence of expression closed curve, P (k) is resolved into two independently sequences along directions X and Y direction, obtain two discrete one dimension functions (k, X (k)) and (k, Y (k)), wherein, X (k) is P (k) variation in the horizontal direction, and Y (k) is the variation of P (k) at vertical direction, and unique point is the point that has local curvature's maximum value in the corresponding curve, on both direction, seek unique point respectively, and then the unique point on this both direction is comprehensively obtained feature point set on the P (k);

The curve vector fitting module is used for adopting SPL to carry out the recurrence match segment of curve between per two unique points.

As preferred a kind of scheme: in described curve vector fitting module, adopt the para-curve SPL, by three point: P on same straight line not as the match primitive ₁, P ₂, P ₃, the expression-form of a para-curve SPL of definition is as follows:

P(t)＝a ₁+a ₂t+a ₃ t ² 0≤t≤1 (1)

Wherein, a ₁, a ₂, a ₃Be the parameter of SPL, the para-curve batten is crossed P ₁, P ₂, P ₃Three points, and have:

1), parabolic segment is with P ₁Be starting point, promptly when parameter t=0, curve is crossed P ₁The point;

2), parabolic segment is with P ₃Be terminal point, promptly when parameter t=1, curve is crossed P ₃The point;

3), when parameter t=0.5, curve is crossed P ₂Point, and its vector equals P ₃-P ₁

It is as follows to obtain system of equations by above three conditions:

$\left\{\begin{array}{c}{a}_1=P_1\\ a_1+a_2+a_3=P_3\\ a_1+0.5a_2+0.25a_3=P_2\end{array}\right.---\left(2\right)$

Solve:

$\left\{\begin{array}{c}{a}_1=P_1\\ a_2=4P_2-P_3-{3P}_1\\ a_3={2\mathrm{<figure-callout\; id="1"\; label="P"\; filenames="A20081005956900271.png,A20081005956900292.png"\; state="\{\{state\}\}">P</figure-callout>}}_1+{2P}_3-4P_2\end{array}\right.---\left(3\right)$

Formula (13) substitution formula (12) is got:

P(t)＝(2t ²-3t+1)P ₁+(4t-4t ²)P ₂+(2t ²-t)P ₃ 0≤t≤1 (4)

Can get by formula (14), work as P ₁, P ₂, P ₃In the time of on same straight line, i.e. 2P ₂=P ₁+ P ₃The time, a ₃=0, P (t) is a some vector, and it has comprised two coordinate figures [x (t), y (t)] on two dimensional surface, and above-mentioned (14) obtain not three point: P point-blank ₁(x ₁, y ₁), P ₂(x ₂, y ₂) and P ₃(x ₃, y ₃) parabolic equation; According to the value of parameter t, calculate the data point that is positioned on the curve one by one, line is drawn figure in turn then.

Further, in curve fitting process,, the current curves section is carried out match and carried out Error Calculation, if the error of this section match is then returned less than given threshold value with starting point, mid point and parabolic SPL of end point structure of current discrete curve section; Otherwise being divided into two parts with the curve of naming a person for a particular job in the current curves section uses the same method and carries out match respectively.

As preferred another kind of scheme: described image binaryzation cut apart and the denoising module in, adopt the threshold processing image segmentation algorithm, the mathematical model of threshold processing is as follows:

If original image be f (x, y), x and y are the coordinate of pixel in image, f (x, y) and g (x y) is pixel value, through thresholding be after the dividing processing of T image for a g (x, y), g (x y) is bianry image, then has:

$g(x,y)=\left\{\begin{array}{cc}1,& f(x,y)\&GreaterEqual;\mathrm{<figure-callout\; id="0"\; label="T"\; filenames="A20081005956900292.png"\; state="\{\{state\}\}">T</figure-callout>}\\ 0,& f(x,y)<T\end{array}\right.---\left(5\right)$

Wherein, (x y) is the gray-scale map of gray-scale value between 0 to 255 to f, and (x y) for pixel value is 0 or 1 binary map to g;

Thresholding T chooses by the method for automatic Iterative, and the step of choosing is as follows:

1), selects the initial estimate of a T;

2), the T split image, generates two groups of pixel: G1 and form greater than the pixel of T, and G2 is made up of less than the pixel of T all gray-scale values by all gray-scale values;

3), to all the calculated for pixel values average gray value μ among regional G1 and the G2 ₁And μ ₂

4), calculate new threshold value:

$T=\frac{1}{2}({\mathrm{\&mu;}}_1+{\mathrm{\&mu;}}_2)---\left(6\right)$

5), repeating step (2) is to (4), up to the difference of the T of the successive iteration gained value parameter T less than predefined ₀

With the T value split image that obtains, obtain binary map again.

As preferred another scheme: extract in the tracking module at described image outline, the process that profile is followed the tracks of is: according to from left to right, sequential search from top to bottom finds the frontier point of first lower left; Then, initial from first frontier point, defining the initial direction of search is along the upper left side; If upper left point is a white point, then be frontier point, 45 degree otherwise the direction of search turns clockwise; Like this until find till first white point; Then this white point as new frontier point, on the basis of current search direction, be rotated counterclockwise 90 the degree, continue to use the same method the search next frontier point, till returning initial frontier point; While record delimitation point coordinate in the process of search.

Further, in described contour feature point extraction module, adopt threshold value q to represent amount of curvature, be for the q value computation process of directions X:

For discrete function X=X (k):

k	0	1	2	...	n-1
						X	X(0)	X(1)	X(2)	...	X(n-1)

The second derivative that can be drawn discrete function X=X (k) by the interpolation principle is:

$X^{\&prime;\&prime;}\left(k\right)\&ap;\frac{1}{h^2}[X(k-h)-2X\left(k\right)+X(k+h)]---\left(7\right)$

Be approximately equal to the curvature of this point again because of the second derivative on the curve, so, the curvature that k is ordered:

${\mathrm{cvt}}_k\&ap;\frac{1}{h^2}[X(k-h)-2X\left(k\right)+X(k+h)]---\left(8\right)$

Owing to have noise in the image, so the step-length h in the algorithm gets 10, and q is the integral multiple of curvature cvt in addition, gets

q＝|h ²×cvt| (9)

Get by (7) and (8)

q _k≈X(k-10)-2X(k)+X(k+10) (10)

The q value computation process of Y direction is identical with directions X.

Further again, in described camera calibration module, adopt the OpenCV calibration algorithm, camera model is a pin-hole model, the world coordinate system conversion process of pixel coordinate system and spatial point may further comprise the steps:

1), three dimensions rigid body translation: with the coordinate figure P in the world coordinate system _w(x _w, y _w, z _w), be transformed to the coordinate figure P in the camera coordinate system _c=(x _c, y _c, z _c) as follows:

$\left[\begin{array}{c}{x}_c\\ y_c\\ z_c\end{array}\right]=R\left[\begin{array}{c}{x}_w\\ y_w\\ z_w\end{array}\right]+t$ $R=\left[\begin{array}{ccc}{r}_1& r_2& r_3\\ r_4& r_5& r_6\\ r_7& r_8& r_9\end{array}\right]$ $t=\left[\begin{array}{c}{t}_x\\ t_y\\ t_z\end{array}\right]---\left(11\right)$

In the formula, R represents one 3 * 3 rotation real number matrix, r ₁..., r ₉Be real constant; T represents a translation vector, wherein, and t ₁..., t ₃Be real constant;

2), with coordinate figure P _c=(x _c, y _c, z _c) projection that in pin-hole model, standardizes, obtain normalized coordinate figure P _n(x y) is:

$P_n=\left[\begin{array}{c}{x}_c/z_c\\ y_c/z_c\end{array}\right]=\left[\begin{array}{c}x\\ y\end{array}\right]---\left(12\right)$

3), introduce the distortion of lens, the standardization coordinate figure P after the distortion _d(x _d, y _d) be expressed as:

$\left[\begin{array}{c}{x}_d\\ y_d\end{array}\right]=(1+k_1{\mathrm{<figure-callout\; id="2"\; label="r"\; filenames="A20081005956900271.png,A20081005956900292.png"\; state="\{\{state\}\}">r</figure-callout>}}^2+k_2{r}^4)\left[\begin{array}{c}x\\ y\end{array}\right]+\left[\begin{array}{c}2{\mathrm{<figure-callout\; id="3"\; label="k"\; filenames="A20081005956900271.png,A20081005956900292.png"\; state="\{\{state\}\}">k</figure-callout>}}_3\mathrm{xy}+k_4({\mathrm{<figure-callout\; id="2"\; label="r"\; filenames="A20081005956900271.png,A20081005956900292.png"\; state="\{\{state\}\}">r</figure-callout>}}^2+2x^2)\\ k_3({\mathrm{<figure-callout\; id="2"\; label="r"\; filenames="A20081005956900271.png,A20081005956900292.png"\; state="\{\{state\}\}">r</figure-callout>}}^2+2y^2)+2{\mathrm{<figure-callout\; id="4"\; label="k"\; filenames="A20081005956900271.png,A20081005956900292.png"\; state="\{\{state\}\}">k</figure-callout>}}_4\mathrm{xy}\end{array}\right]---\left(13\right)$

In the formula: r ²=x ²+ y ²k ₁, k ₂The expression coefficient of radial distortion; k ₃, k ₄Expression tangential distortion coefficient;

With P _d(x _d, y _d) be converted into the coordinate figure P that pixel coordinate is fastened on the image _p(u, v):

$\left\{\begin{array}{c}u=f_x{x}_d+u_0\\ v=f_y{y}_d+v_0\end{array}\right.---\left(14\right)$

Wherein

$\left\{\begin{array}{c}{f}_x=f\&CenterDot;\mathrm{sx}/\mathrm{dpx}\\ f_y=f/\mathrm{dpy}\end{array}\right.---\left(15\right)$

In the following formula, f is the effective focal length of video camera; f _xAnd f _yFor being the focal length of unit with the pixel.Sx is a scale factor, is used for adapting in computer picture horizontal direction (x direction) going up all uncertain factors that sampling brings; Dpx is the computer picture coverage (mm/pixel) between (x direction) last adjacent two pixels in the horizontal direction; Dpy is the coverage (mm/pixel) of computer picture between last adjacent two pixels of vertical direction (y direction); u ₀Intersection point O for the camera lens optical axis Z and the plane of delineation _iAt pixel coordinate is coordinate figure on the u axle; O _iFor at pixel coordinate being the coordinate figure on the v axle.

Technical conceive of the present invention is: adopt the digital camera typing fashion plate figure of USB interface, and be input to computing machine and do vectorized process.Whole hardware unit is installed the liftable support of digital camera by digital camera, and the desk lamp of illumination is provided, and places the platform of fashion plate, and forms in order to the computing machine of handling image, as shown in Figure 1.Entire image vectored system processing procedure comprises the demarcation of imagination machine, gathers fashion plate figure, and image pre-service, image vector, vector map file are preserved.

Further specify the key issue that the present invention relates to below in implementation process:

(1) because optical lens can produce geometric distortion, and exist inclination angle and corner between video camera imaging plane and the fashion plate plane, thereby there is nonlinear geometric distortion in the image that the optical lens video camera obtains, before image is carried out subsequent treatment, should eliminate these distortion, than higher, the camera lens distortion is the error component that must consider to the accuracy requirement of last image in the present invention.Lens distortion is divided into radial distortion and tangential distortion, and the former is bigger to the image deformation influence, and the latter is less, often can ignore.Camera calibration is meant sets up the video camera imaging geometric model, describe in the space coordinates object point with it in the process of corresponding relation between the picture point on the plane of delineation.The purpose of camera calibration determines that exactly the geometric model parameter is a camera parameters.

Camera calibration at first will be selected suitable camera model, determines internal and external parameter.Camera model in the OpenCV calibration algorithm is based on pin-hole model (pin-hole model), introduce the radial distortion and the tangential distortion of lens, this model has reflected the distortion situation of lens reality more truly than the Tasi model of only introducing the single order radial distortion and pin-hole model.In this model, with the coordinate figure of spatial point in world coordinate system, the process that is transformed to coordinate figure in the system of pixel coordinate on the plane of delineation can be decomposed into 4 following step conversion:

1), three dimensions rigid body translation: with the coordinate figure P in the world coordinate system _w(x _w, y _w, z _w), be transformed to the coordinate figure P in the camera coordinate system _c=(x _c, y _c, z _c) as follows

$\left[\begin{array}{c}{x}_c\\ y_c\\ z_c\end{array}\right]=R\left[\begin{array}{c}{x}_w\\ y_w\\ z_w\end{array}\right]+t$ $R=\left[\begin{array}{ccc}{r}_1& r_2& r_3\\ r_4& r_5& r_6\\ r_7& r_8& r_9\end{array}\right]$ $t=\left[\begin{array}{c}{t}_x\\ t_y\\ t_z\end{array}\right]---\left(11\right)$

R in the formula---one 3 * 3 rotation matrix; T---translation vector.

2), with coordinate figure P _c=(x _c, y _c, z _c) projection that in pin-hole model, standardizes, obtain normalized coordinate figure P _n(x y) is

$P_n=\left[\begin{array}{c}{x}_c/z_c\\ y_c/z_c\end{array}\right]=\left[\begin{array}{c}x\\ y\end{array}\right]---\left(12\right)$

3), introduce the distortion of lens, the standardization coordinate figure after the distortion can be used P _d(x _d, y _d) be expressed as:

$\left[\begin{array}{c}{x}_d\\ y_d\end{array}\right]=(1+k_1{\mathrm{<figure-callout\; id="2"\; label="r"\; filenames="A20081005956900271.png,A20081005956900292.png"\; state="\{\{state\}\}">r</figure-callout>}}^2+k_2{r}^4)\left[\begin{array}{c}x\\ y\end{array}\right]+\left[\begin{array}{c}2{\mathrm{<figure-callout\; id="3"\; label="k"\; filenames="A20081005956900271.png,A20081005956900292.png"\; state="\{\{state\}\}">k</figure-callout>}}_3\mathrm{xy}+k_4({\mathrm{<figure-callout\; id="2"\; label="r"\; filenames="A20081005956900271.png,A20081005956900292.png"\; state="\{\{state\}\}">r</figure-callout>}}^2+2x^2)\\ k_3({\mathrm{<figure-callout\; id="2"\; label="r"\; filenames="A20081005956900271.png,A20081005956900292.png"\; state="\{\{state\}\}">r</figure-callout>}}^2+2y^2)+2{\mathrm{<figure-callout\; id="4"\; label="k"\; filenames="A20081005956900271.png,A20081005956900292.png"\; state="\{\{state\}\}">k</figure-callout>}}_4\mathrm{xy}\end{array}\right]---\left(13\right)$

In the formula: r ²=x ²+ y ²k ₁, k ₂---coefficient of radial distortion; k ₃, k ₄---the tangential distortion coefficient.

With P _d(x _d, y _d) be converted into the coordinate figure P that pixel coordinate is fastened on the image _p(u, v)

$\left\{\begin{array}{c}u=f_x{x}_d+u_0\\ v=f_y{y}_d+v_0\end{array}\right.---\left(14\right)$

Wherein

$\left\{\begin{array}{c}{f}_x=f\&CenterDot;\mathrm{sx}/\mathrm{dpx}\\ f_y=f/\mathrm{dpy}\end{array}\right.---\left(15\right)$

The meaning of each parameter: f is the effective focal length of video camera; Sx is a scale factor, is used for adapting in computer picture (frame is deposited) horizontal direction (x direction) going up all uncertain factors that sampling brings; Dpx is the computer picture coverage (mm/pixel) between (x direction) last adjacent two pixels in the horizontal direction; Dpy is the coverage (mm/pixel) of computer picture between last adjacent two pixels of vertical direction (y direction); u ₀Intersection point O for the camera lens optical axis Z and the plane of delineation _iAt pixel coordinate is coordinate figure on the u axle; O _iFor at pixel coordinate being the coordinate figure on the v axle.

(2) vector quantization of image is based on that the two-value figure carries out, and binary map is to be that pixel value is 0 or 1 figure, and 0 represents black, and 1 represent white.Digital camera obtains is 24 colored bitmap, needs the image pre-service to carry out that image threshold processing binaryzation is cut apart and denoising.

The threshold processing image segmentation algorithm is because its intuitive and be easy to realization property, is in middle cardiac status in speed factor accounts for the application of critical role.In the print photo, the gray-scale value of target and background differs bigger, therefore adopts single global threshold T just the target of needs can be separated basically.Thresholding T can choose by the method for automatic Iterative, and the step of choosing is as follows:

1), selects the initial estimate of a T

2), T split image.Do like this and can form greater than the pixel of T by all gray-scale values by two groups of pixel: G1 of generation, and G2 is made up of less than the pixel of T all gray-scale values.

3), to all the calculated for pixel values average gray value μ among regional G1 and the G2 ₁And μ ₂

4), calculate new threshold value:

$T=\frac{1}{2}\left({\mathrm{\&mu;}}_1{+\mathrm{\&mu;}}_2\right)---\left(6\right)$

5), repeating step (2) is to (4), up to the difference of the T of the successive iteration gained value parameter T less than predefined ₀

As shown in Figure 4, through after the threshold processing, obtain the binary map of print:

There is a spot of isolated point noise in binary map, can remove by corrosion in the two-value morphology and dilation operation.

The concrete steps of erosion operation are as follows:

1), scan former figure, finding first pixel value is 1 point;

2), the initial point with the structural element of pre-set shape and origin position moves to this point;

3), judge whether the pixel value in this structural element institute coverage all is 1, if then the pixel value on the same position of corrosion back image is changed to 1, is 0 if having the value of a pixel at least, then the pixel value on the same position of corrosion back image is 0;

4) be that 1 point repeats 2, to all pixel values of former figure) and 3);

With erosion removal granule noise is very effective, but simultaneously also with the fashion plate edge corrosion pixel, therefore need expand and recover by the dual operations of corrosion.

The concrete steps of dilation operation are as follows:

1), the initial point with the structural element of pre-set shape and origin position moves to all pixels that can contain in the image;

2), judge that whether pixel value in this structure institute coverage exists at least one is 1 point, if exist, then expand in the image of back with structural element initial point same position on pixel value be changed to 1, if the value of all pixels is 0 in this coverage, the pixel value on the same position of the back image that then expands is changed to 0.

3) be that 1 point repeats 1, to all pixel values of former figure) and 2);

(3) purpose that profile extracts and profile is followed the tracks of all is the outer contoured features that obtains image, for the unique point of extraction profile and follow-up match are prepared.

The thought of the algorithm that the bianry image profile extracts is emptied internal point exactly, if having among the former figure a bit for black, and its 8 consecutive point are when all being black (this moment, this point was an internal point), then with this point deletion.Method with mathematical morphology is handled, and is equivalent to the structural element of nine points former figure be corroded, and is deducting the corrosion diagram picture with original image.

The basic skills that profile is followed the tracks of is: basis " visiting criterion " is found out the pixel on the print profile earlier, finds out other pixels on the print profile according to some feature of these pixels with " tracking criterion " again.

At first find first boundary pixel " visiting criterion " to be: according to from left to right, sequential search from top to bottom finds the frontier point of first lower left.

Then, initial from first frontier point, defining the initial direction of search is along the upper left side; If upper left point is white point (pixel value is 1), then is frontier point, 45 degree otherwise the direction of search turns clockwise.Like this until find till first white point.Then this white point as new frontier point, on the basis of current search direction, be rotated counterclockwise 90 the degree, continue to use the same method the search next frontier point, till returning initial frontier point.While record delimitation point coordinate in the process of search.Owing among the width of cloth figure a plurality of fashion plates are arranged, therefore adopt the method for deleting current point while following the tracks of, followed the tracks of a closed curve like this and just deleted one, so that first frontier point search of next bar closed curve.

(4) extract print profile coordinate after, next just according to coordinate sequence to every closed contour curve extract minutiae.The key of extract minutiae is the calculating that X (k) and Y (k) go up each point curvature.Because unique point is extracted according to the degree of crook of curve, the local degree of crook that the curvature absolute value is big more is big more, therefore, the big point of local curvature is exactly a unique point, do not need accurately to draw curvature value, can be used for judging that the size of curvature just can be worth extract minutiae according to this as long as draw a certain numerical value.If this numerical value is q, in actual computation, can get the integral multiple that q equals approximate curvature absolute value.

Be the calculating of example explanation q below with X (k).In numerical analysis, for discrete function X=X (k):

k	0	1	2	...	n-1
						X	X(0)	X(1)	X(2)	...	X(n-1)

The second derivative that can be drawn discrete function X=X (k) by the interpolation principle is:

$X^{\&prime;\&prime;}\left(k\right)\&ap;\frac{1}{h^2}[X(k-h)-2X\left(k\right)+X(k+h)]---\left(7\right)$

Again because the second derivative on the curve is approximately equal to the curvature of this point ^[5]So,, the curvature that k is ordered

${\mathrm{cvt}}_k\&ap;\frac{1}{h^2}[X(k-h)-2X\left(k\right)+X(k+h)]---\left(8\right)$

Owing to have noise in the image, so the step-length h in the algorithm gets 10, and q is the integral multiple of curvature cvt in addition, gets

q＝|h ²×cvt| (9)

Get by (7) and (8)

q _k≈X(k-10)-2X(k)+X(k+10) (10)

The same among the computing method of q value and the X (k) among the Y (k).

(5) on the basis of feature point extraction, be waypoint with the unique point, adopt SPL to carry out the recurrence match to the segment of curve between per two unique points, till precision satisfies certain condition.Because the variation diversity at fashion plate edge, thus this project selection shape more flexibly the para-curve SPL as the match primitive.By three point: P on same straight line not ₁, P ₂, P ₃, the expression-form of a para-curve SPL of definition is as follows:

P(t)＝a ₁+a ₂t+a ₃ t ² 0≤t≤1 (1)

The para-curve batten is crossed P ₁, P ₂, P ₃Three points, and have:

1), parabolic segment is with P ₁Be starting point.Promptly when parameter t=0, curve is crossed P ₁The point.

2), parabolic segment is with P ₃Be terminal point, promptly when parameter t=1, curve is crossed P ₃The point.

3), when parameter t=0.5, curve is crossed P ₂Point, and its vector equals P ₃-P ₁

It is as follows to obtain system of equations by above three conditions:

$\left\{\begin{array}{c}{a}_1=P_1\\ a_1+a_2+a_3=P_3\\ a_1+0.5a_2+0.25a_3=P_2\end{array}\right.---\left(2\right)$

Solve:

$\left\{\begin{array}{c}{a}_1=P_1\\ a_2=4P_2-P_3-{3P}_1\\ a_3={2\mathrm{<figure-callout\; id="1"\; label="P"\; filenames="A20081005956900271.png,A20081005956900292.png"\; state="\{\{state\}\}">P</figure-callout>}}_1+{2P}_3-4P_2\end{array}\right.---\left(3\right)$

Formula (13) substitution formula (12) is got

P(t)＝(2t ²-3t+1)P ₁+(4t-4t ²)P ₂+(2t ²-t)P ₃ 0≤t≤1 (4)

Can get by formula (14), work as P ₁, P ₂, P ₃In the time of on same straight line, i.e. 2P ₂=P ₁+ P ₃The time, a ₃=0, curvilinear equation becomes straight line.Therefore match primitive in this article is actually straight line and para-curve.P (t) is a some vector, and it has comprised two coordinate figures [x (t), y (t)] on two dimensional surface.The formula of more than deriving and obtaining is not point-blank three point: P of our desired mistake ₁(x ₁, y ₁), P ₂(x ₂, y ₂) and P ₃(x ₃, y ₃) parabolic equation.When this para-curve is stored and edited, only need P ₁, P ₂And P ₃Operate.Need not as digital curve, each point all to be operated.When showing this curve on screen, can calculate the data point that is positioned on the curve one by one according to the value of parameter t, line is drawn figure in turn then.

Beneficial effect of the present invention mainly shows: simple to operate, input efficiency is high, precision, minimizing cost.

Description of drawings

Fig. 1 is the device of fashion plate digital camera typing.

Fig. 2 is a fashion plate vector quantization process flow diagram.

Fig. 3 is a camera calibration plate synoptic diagram.

Fig. 4 is the fashion plate binary picture.

Fig. 5 is the fashion plate profile diagram.

Fig. 6 is that 8 directions that center pixel is followed the tracks of are numbered and the skew spirogram.

Fig. 7 is the process flow diagram of image processing system.

Embodiment

Below in conjunction with accompanying drawing the present invention is further described.

With reference to Fig. 1～Fig. 7, a kind of fashion plate input device based on digital camera, described input device comprises digital camera 1, the liftable support 2 of digital camera is installed, the desk lamp 3 of illumination is provided, places the platform 4 of fashion plate, and form in order to the computing machine 5 of handling image.Described intelligent image vectored system comprises camera calibration module 1, and image binaryzation is cut apart and denoising module 2, and image outline extracts tracking module 3, contour feature point extraction module 4, curve vector fitting module 5.

Video camera is installed in the part from platform 1.5-1.8 rice in the invention process, and according to the size adjustment of images acquired height.The breadth size of table top is: wide 1 meter, and long 1.2 meters.Two desk lamps are installed in the both sides of table top, and being used for gathering is the usefulness of illumination, can be according to the needs adjustment height of illumination, as shown in Figure 1.

Described camera calibration module is used to correct the radial distortion that optical lens produces.Camera calibration based on OpenCV adopts plane gridiron pattern calibrating template in order to improve the success ratio that angle point extracts, and in the periphery of demarcating square, also requires to keep the white white space that square is wide, as shown in Figure 3.Video camera only needs to grasp in different angles the picture of a few sheet of planar calibrating templates, just can realize Camera calibration.Obviously, owing to adopt least square method, grab to such an extent that figure is many more, the result of demarcation is just accurate more.

Though it is very high to seek the success ratio of angle point function extraction angle point among the OpenCV automatically, but if situation such as run into that light is blocked, make and will cause calibrating block on the calibrating template unintelligible or the angle point number and the situation that the number of setting does not conform to that extract on image angle point to extract and fail.So must consider the situation that angle point can not be extracted in design during calibration algorithm: very few if angle point extracts successful figure on the one hand, then the result that comes out of demarcation just differs and satisfies the requirement of precision surely, need adopt figure again; On the other hand, because the number of video camera external parameter is associated with the number of uncalibrated image, so when in the end calculating calibration result, should give up with extracting the image that angle point fail, again according to the number that is left image, the dynamically storage area of allocation of parameters in internal memory.So just can guarantee the accuracy requirement of demarcating.

After demarcating, video camera just can collect the fashion plate figure behind the distortion correction, pass through USB interface, it is 24 true color bitmaps that computing machine obtains, it is dot chart of discretize all in space and the brightness, and we at first will obtain is the digitizing boundary curve of fashion plate, therefore can the print zone be separated by threshold processing, and the outward flange that obtains print is followed the tracks of at the edge then.

The threshold processing image segmentation algorithm is because its intuitive and be easy to realization property, is in middle cardiac status in speed factor accounts for the application of critical role.In the print photo, the gray-scale value of target and background differs bigger, therefore adopts single global threshold T just the target of needs can be separated basically.Thresholding T can choose by the method for automatic Iterative, and the step of choosing is as follows:

1), selects the initial estimate of a T

2), T split image.Do like this and can form greater than the pixel of T by all gray-scale values by two groups of pixel: G1 of generation, and G2 is made up of less than the pixel of T all gray-scale values.

3), to all the calculated for pixel values average gray value μ among regional G1 and the G2 ₁And μ ₂

4), calculate new threshold value:

$T=\frac{1}{2}\left({\mathrm{\&mu;}}_1{+\mathrm{\&mu;}}_2\right)---\left(6\right)$

5), repeating step 2) to 4), up to the difference of the T of successive iteration gained value parameter T less than predefined ₀

As shown in Figure 1, through after the threshold processing, obtain the binary map of print, as shown in Figure 4.

Described image outline extracts tracking module in order to extract fashion plate profile coordinate sequence.Owing to will handle a plurality of fashion plates among the width of cloth figure, therefore to extract the profile of bianry image earlier earlier, this paper extracts profile by morphology operations.Morphology operations is the image processing method that grows up according to the method for mathematical morphology set theory at bianry image.The extraction of profile is equivalent to the structural element of nine somes original image be corroded, and makes the border of object reduce by a pixel along periphery, deducts the corrosion diagram picture with original image again, just obtains the continuous fashion plate profile of single pixel, as shown in Figure 5.

Obtain after the profile diagram, carry out pixel again and follow the tracks of, write down the frontier point coordinate of every curve.As seen from Figure 6, the direction that center pixel can be followed the tracks of has 8, each direction direction encoding and side-play amount have been formulated, since image file read the order be from left to right, from top to bottom, therefore choose the most upper left pixel of image as starting point, search for starting point then, after finding starting point, this point is noted, defining initial tracking direction is upper left side 0 direction, judge whether this point is impact point, be then to preserve initial point and, be rotated counterclockwise 90 degree, continue to detect the point on this new tracking direction as new tracking direction the initial starting point of this impact point as tracking; If not impact point is then along 45 degree that turn clockwise, until find impact point.After finding impact point, on the basis of current tracking direction, be rotated counterclockwise 90 degree as new tracking direction, find new frontier point after, old border is preserved, with new detected point as new initial point.Use the same method and follow the tracks of next frontier point, till getting back to starting point.The note pixel follow the tracks of the curved boundary point coordinate sequence draw be P (k)=[X (k), Y (k)] (k=0,1,2...n-1), n is a closed curve frontier point number.

Described feature point extraction module is in order to extracting the bigger point of curvature in the fashion plate closed curve profile, in order to avoid smoothedly in follow-up match fall.The key of edge extracting unique point is the calculating that X (k) and Y (k) go up each point curvature.Because unique point is extracted according to the degree of crook of curve, the local degree of crook that the curvature absolute value is big more is big more, therefore, the big point of local curvature is exactly a unique point, do not need accurately to draw curvature value, can be used for judging that the size of curvature just can be worth extract minutiae according to this as long as draw a certain numerical value.If this numerical value is q, in actual computation, can get the integral multiple that q equals approximate curvature absolute value.

Be the calculating of example explanation q below with X (k).In numerical analysis, for discrete function X=X (k):

k	0	1	2	...	n-1
						X	X(0)	X(1)	X(2)	...	X(n-1)

The second derivative that can be drawn discrete function X=X (k) by the interpolation principle is:

$X^{\&prime;\&prime;}\left(k\right)\&ap;\frac{1}{h^2}[X(k-h)-2X\left(k\right)+X(k+h)]---\left(7\right)$

Again because the second derivative on the curve is approximately equal to the curvature of this point ^[5]So,, the curvature that k is ordered

${\mathrm{cvt}}_k\&ap;\frac{1}{h^2}[X(k-h)-2X\left(k\right)+X(k+h)]---\left(8\right)$

Owing to have noise in the image, so the step-length h in the algorithm gets 10, and q is the integral multiple of curvature cvt in addition, gets

q＝|h ²×cvt| (9)

Get by (7) and (8)

q _k≈X(k-10)-2X(k)+X(k+10) (10)

The same among the computing method of q value and the X (k) among the Y (k).

According to the q value of each point on (19) calculated curve X (k) border, by above analysis as can be known, the local ratio of curvature that the q value is big is bigger.Set a threshold epsilon, judge according to the size of q value:

If in a certain interval, all points all there is q≤ε, then should there be unique point in the interval.At this moment there are two kinds of situations: the one, the q value of interval most of point all equals 0, has only the point of a small amount of q≤ε can think noise spot, the corresponding straight-line segment that has noise in this interval; The 2nd, interval in the q value of most of point be not 0 and less than ε, the curve that this interval degree of crook of correspondence is very little (q when ε gets 2≤2, $\mathrm{cvt}=\frac{\mathrm{<figure-callout\; id="2"\; label="q\; h"\; filenames="A20081005956900271.png,A20081005956900292.png"\; state="\{\{state\}\}">q</figure-callout>}}{h^{}}\&le;0.02$ Be approximately 0), this curve approximation is in straight line.Do not establish unique point under two kinds of situations in the interval.

If in a certain interval, all points all there is q＞ε, then should the corresponding curve that intensity of variation is bigger in interval.For approaching primary curve better, set suitable step-length, the point of finding out q value maximum in each step-length is as unique point.

Specific design is as follows, and wherein max is used for writing down the maximum q value of present segment, and i is used for writing down the sequence number of the point that has maximum q value in the present segment, and T is a threshold epsilon.Going up certain a bit corresponding q (k) in the hope of X (k) is worth and judges; Extract minutiae.Use the same method and to extract unique point on the Y (k).

At last, the unique point on comprehensive X (k) and the Y (k) is asked the feature point set on the P (k), just obtains all unique points.Integrated approach is as follows:

For the arbitrary characteristics point P on the X (k) _xAnd the arbitrary characteristics point P on the Y (k) _y, the sequence number of establishing its point is respectively i, j.

If | i-j|≤3, then P _xAnd P _ySame unique point on the corresponding P (k).

If | i-j|＞3, then P _xAnd P _yLast two the different unique points of corresponding P (k).

Described vector quantization fitting module is in order to fit to curved edge the curve that math equation is represented.Match is meant in the design process of curve, curved surface, makes curve, the curved surface of generation reach some designing requirement with the interpolation or the method for approaching, as pressing close to original point or reference mark sequence etc. in allowed limits.What adopt in this problem is the original point sequence of method match of parabolic spline interpolation, and purpose is to make the more smooth nature in fashion plate edge, is convenient to editor and cutting.On the basis of feature point extraction, be waypoint with the unique point, adopt SPL to carry out the recurrence match to the segment of curve between per two unique points, till precision satisfies certain condition.Because the variation diversity at fashion plate edge, thus this project selection shape more flexibly the para-curve SPL as the match primitive.By three point: P on same straight line not ₁, P ₂, P ₃, the expression-form of a para-curve SPL of definition is as follows:

P(t)＝a ₁+a ₂t+a ₃ t ² 0≤t≤1 (1)

The para-curve batten is crossed P ₁, P ₂, P ₃Three points, and have:

1), parabolic segment is with P ₁Be starting point.Promptly when parameter t=0, curve is crossed P ₁The point.

2), parabolic segment is with P ₃Be terminal point, promptly when parameter t=1, curve is crossed P ₃The point.

3), when parameter t=0.5, curve is crossed P ₂Point, and its vector equals P ₃-P ₁

It is as follows to obtain system of equations by above three conditions:

$\left\{\begin{array}{c}{a}_1=P_1\\ a_1+a_2+a_3=P_3\\ a_1+0.5a_2+0.25a_3=P_2\end{array}\right.---\left(2\right)$

Solve:

$\left\{\begin{array}{c}{a}_1=P_1\\ a_2=4P_2-P_3-{3P}_1\\ a_3={2\mathrm{<figure-callout\; id="1"\; label="P"\; filenames="A20081005956900271.png,A20081005956900292.png"\; state="\{\{state\}\}">P</figure-callout>}}_1+{2P}_3-4P_2\end{array}\right.---\left(3\right)$

Formula (22) substitution formula (21) is got

P(t)＝(2t ²-3t+1)P ₁+(4t-4t ²)P ₂+(2t ²-t)P ₃ 0≤t≤1 (4)

Can get by formula (23), work as P ₁, P ₂, P ₃In the time of on same straight line, i.e. 2P ₂=P ₁+ P ₃The time, a ₃=0, curvilinear equation becomes straight line.Therefore match primitive in this article is actually straight line and para-curve.P (t) is a some vector, and it has comprised two coordinate figures [x (t), y (t)] on two dimensional surface.The formula of more than deriving and obtaining is not point-blank three point: P of our desired mistake ₁(x ₁, y ₁), P ₂(x ₂, y ₂) and P ₃(x ₃, y ₃) parabolic equation.When this para-curve is stored and edited, only need P ₁, P ₂And P ₃Operate.Need not as digital curve, each point all to be operated.When showing this curve on screen, can calculate the data point that is positioned on the curve one by one according to the value of parameter t, line is drawn figure in turn then.

Curve fitting algorithm is of a great variety, though can satisfy the requirement of precision and compressibility, most algorithm complexity, calculated amount is too big.Therefore, this patent approximating method that proposed a kind of recurrence carries out match respectively to the segment of curve between each cusp.This algorithm is thought as follows: starting point, mid point and end parabolic SPL of some structure with current discrete curve section are carried out match and are carried out Error Calculation the current curves section, if the error of this section match is then returned less than the given value of closing; Otherwise being divided into two parts with the curve of naming a person for a particular job in the current curves section uses the same method and carries out match respectively.

Segment of curve between each cusp of match successively just can obtain total fitting result in this way.This algorithm only need just can be realized match with two simple recurrence statements to every section curve as can be seen, and is with respect to the operations such as heredity, variation and intersection of various complexity in the genetic algorithm, very simple and be easy to realize.

Obtaining polar plot after the recurrence match in fact is made up of parabolic segment, and parabolic segment can be made up of starting point, terminal point coordinate and parabolical three parameters, therefore polar plot greatly reduces than data volume with raster image, but because image is often little a lot of than actual print, therefore to enlargement factor be set according to the size of print, and because such polar plot also is not the software that CAD of Garment software and cutting machine can be discerned, so also need the process of process and form conversion.

In example of the present invention, need data are changed into the PLT file that cutting machine can be discerned with the HPGL language, the PLT file is made up of a series of coordinate points, in order to reduce the length of PLT file, get a little density degree by the decision of the size of parabolic segment curvature during conversion, export finally that the PLT file can be cut machine and CAD of Garment software is discerned.

Claims (8)

1, a kind of automatic input device for cloth sample image based on image vector technology, it is characterized in that: described automatic input device for cloth sample image comprises the platform of placing fashion plate, the computing machine that is used to obtain the digital camera of fashion plate picture and is used to handle image, mounting bracket on the described platform, described digital camera is installed on the described support, described platform is positioned at the visual range of digital camera, described digital camera is connected with computing machine, and described computing machine comprises:

Camera calibration module is used for the radial distortion of correcting digital video camera, sets up the corresponding relation between the world coordinate system of pixel coordinate system on the plane of delineation and spatial point;

Image binaryzation is cut apart and the denoising module, is used to receive the image that digital camera obtains, and is binary map with the colored bitmap-converted of being gathered, and binary map is carried out denoising;

Image outline extracts tracking module, is used for the binary map every bit is judged, for black, and its 8 consecutive point with this point deletion, obtain the profile of binary map when all being black as this point;

And from first frontier point, the profile of tracking image, record delimitation point coordinate value has been followed the tracks of behind the closed curve one of deletion, begins the tracking of next bar closed curve, up to the tracking of finishing all closed curves;

The contour feature point extraction module, be used for according to coordinate sequence every closed curve extract minutiae, be [X (k), Y (k)], the coordinate point sequence of expression closed curve, P (k) is resolved into two independently sequences along directions X and Y direction, obtain two discrete one dimension functions (k, X (k)) and (k, Y (k)), wherein, X (k) is P (k) variation in the horizontal direction, and Y (k) is the variation of P (k) at vertical direction, and unique point is the point that has local curvature's maximum value in the corresponding curve, on both direction, seek unique point respectively, and then the unique point on this both direction is comprehensively obtained feature point set on the P (k);

The curve vector fitting module is used for adopting SPL to carry out the recurrence match segment of curve between per two unique points.

2, the automatic input device for cloth sample image based on image vector technology as claimed in claim 1 is characterized in that: in described curve vector fitting module, adopt the para-curve SPL as the match primitive, by three point: P on same straight line not ₁, P ₂, P ₃, the expression-form of a para-curve SPL of definition is as follows:

P(t)＝a ₁+a ₂t+a ₃t ² 0≤t≤1 (1)

Wherein, a ₁, a ₂, a ₃Be the parameter of SPL, the para-curve batten is crossed P ₁, P ₂, P ₃Three points, and have:

1), parabolic segment is with P ₁Be starting point, promptly when parameter t=0, curve is crossed P ₁The point;

2), parabolic segment is with P ₃Be terminal point, promptly when parameter t=1, curve is crossed P ₃The point;

3), when parameter t=0.5, curve is crossed P ₂Point, and its vector equals P ₃-P ₁

It is as follows to obtain system of equations by above three conditions:

$\left\{\begin{array}{c}{a}_1=P_1\\ a_1+a_2+a_3=P_3\\ a_1+0.5a_2+0.25a_3=P_2\end{array}\right.---\left(2\right)$

Solve:

$\left\{\begin{array}{c}{a}_1=P_1\\ a_2=4P_2-P_3-{3P}_1\\ a_3={2P}_1+{2P}_3-4P_2\end{array}\right.---\left(3\right)$

Formula (13) substitution formula (12) is got:

P(t)＝(2t ²-3t+1)P ₁+(4t-4t ²)P ₂+(2t ²-t)P ₃ 0≤t≤1 (4)

Can get by formula (14), work as P ₁, P ₂, P ₃In the time of on same straight line, i.e. 2P ₂=P ₁+ P ₃The time, a ₃=0, P (t) is a some vector, and it has comprised two coordinate figures [x (t), y (t)] on two dimensional surface, and above-mentioned (14) obtain not three point: P point-blank ₁(x ₁, y ₁), P ₂(x ₂, y ₂) and P ₃(x ₃, y ₃) parabolic equation; According to the value of parameter t, calculate the data point that is positioned on the curve one by one, line is drawn figure in turn then.

3, the automatic input device for cloth sample image based on image vector technology as claimed in claim 2, it is characterized in that: in curve fitting process, starting point, mid point and parabolic SPL of end point structure with current discrete curve section, the current curves section is carried out match and carried out Error Calculation, if the error of this section match is then returned less than given threshold value; Otherwise being divided into two parts with the curve of naming a person for a particular job in the current curves section uses the same method and carries out match respectively.

4, the automatic input device for cloth sample image based on image vector technology as claimed in claim 1 is characterized in that: described image binaryzation cut apart and the denoising module in, adopt the threshold processing image segmentation algorithm, the mathematical model of threshold processing is as follows:

If original image be f (x, y), x and y are the coordinate of pixel in image, f (x, y) and g (x y) is pixel value, through thresholding be after the dividing processing of T image for a g (x, y), g (x y) is bianry image, then has:

$g(x,y)=\left\{\begin{array}{cc}1,& f(x,y)\&GreaterEqual;T\\ 0,& f(x,y)<T\end{array}\right.---\left(5\right)$

Wherein, (x y) is the gray-scale map of gray-scale value between 0 to 255 to f, and (x y) for pixel value is 0 or 1 binary map to g;

Thresholding T chooses by the method for automatic Iterative, and the step of choosing is as follows:

1), selects the initial estimate of a T;

2), the T split image, generates two groups of pixel: G1 and form greater than the pixel of T, and G2 is made up of less than the pixel of T all gray-scale values by all gray-scale values;

3), to all the calculated for pixel values average gray value μ among regional G1 and the G2 ₁And μ ₂

4), calculate new threshold value:

$T=\frac{1}{2}({\mathrm{\&mu;}}_1+{\mathrm{\&mu;}}_2)---\left(6\right)$

5), repeating step (2) is to (4), up to the difference of the T of the successive iteration gained value parameter T less than predefined ₀

With the T value split image that obtains, obtain binary map again.

5, the automatic input device for cloth sample image based on image vector technology as claimed in claim 2 is characterized in that: described image binaryzation cut apart and the denoising module in, adopt the threshold processing image segmentation algorithm, the mathematical model of threshold processing is as follows:

If original image be f (x, y), x and y are the coordinate of pixel in image, f (x, y) and g (x y) is pixel value, through thresholding be after the dividing processing of T image for a g (x, y), g (x y) is bianry image, then has:

$g(x,y)=\left\{\begin{array}{cc}1,& f(x,y)\&GreaterEqual;T\\ 0,& f(x,y)<T\end{array}\right.---\left(5\right)$

Wherein, (x y) is the gray-scale map of gray-scale value between 0 to 255 to f, and (x y) for pixel value is 0 or 1 binary map to g;

Thresholding T chooses by the method for automatic Iterative, and the step of choosing is as follows:

1), selects the initial estimate of a T;

2), the T split image, generates two groups of pixel: G1 and form greater than the pixel of T, and G2 is made up of less than the pixel of T all gray-scale values by all gray-scale values;

3), to all the calculated for pixel values average gray value μ among regional G1 and the G2 ₁And μ ₂

4), calculate new threshold value:

$T=\frac{1}{2}({\mathrm{\&mu;}}_1+{\mathrm{\&mu;}}_2)---\left(6\right)$

5), repeating step (2) is to (4), up to the difference of the T of the successive iteration gained value parameter T less than predefined ₀

With the T value split image that obtains, obtain binary map again.

6, as the described automatic input device for cloth sample image of one of claim 1-5 based on image vector technology, it is characterized in that: extract in the tracking module at described image outline, the process that profile is followed the tracks of is: according to from left to right, sequential search from top to bottom finds the frontier point of first lower left; Then, initial from first frontier point, defining the initial direction of search is along the upper left side; If upper left point is a white point, then be frontier point, 45 degree otherwise the direction of search turns clockwise; Like this until find till first white point; Then this white point as new frontier point, on the basis of current search direction, be rotated counterclockwise 90 the degree, continue to use the same method the search next frontier point, till returning initial frontier point; While record delimitation point coordinate in the process of search.

7, as the described automatic input device for cloth sample image of one of claim 1-5, it is characterized in that: in described contour feature point extraction module, adopt threshold value q to represent amount of curvature, be for the q value computation process of directions X based on image vector technology:

For discrete function X=X (k):

k 0 1 2 ... n-1 X X(0) X(1) X(2) ... X(n-1)

The second derivative that can be drawn discrete function X=X (k) by the interpolation principle is:

$X^{\&prime;\&prime;}\left(k\right)\&ap;\frac{1}{h^2}[X(k-h)-2X\left(k\right)+X(k+h)]---\left(7\right)$

Be approximately equal to the curvature of this point again because of the second derivative on the curve, so, the curvature that k is ordered:

${\mathrm{cvt}}_k\&ap;\frac{1}{h^2}[X(k-h)-2X\left(k\right)+X(k+h)]---\left(8\right)$

Owing to have noise in the image, so the step-length h in the algorithm gets 10, and q is the integral multiple of curvature cvt in addition, gets

q＝|h ²×cvt| (9)

Get by (7) and (8)

q _k≈X(k-10)-2X(k)+X(k+10) (10)

The q value computation process of Y direction is identical with directions X.

8, the automatic input device for cloth sample image based on image vector technology as claimed in claim 7, it is characterized in that: in described camera calibration module, adopt the OpenCV calibration algorithm, camera model is a pin-hole model, and the world coordinate system conversion process of pixel coordinate system and spatial point may further comprise the steps:

1), three dimensions rigid body translation: with the coordinate figure P in the world coordinate system _w(x _w, y _w, z _w), be transformed to the coordinate figure P in the camera coordinate system _c=(x _c, y _c, z _c) as follows:

$\left[\begin{array}{c}{x}_c\\ y_c\\ z_c\end{array}\right]=R\left[\begin{array}{c}{x}_w\\ y_w\\ z_w\end{array}\right]+t$ $R=\left[\begin{array}{ccc}{r}_1& r_2& r_3\\ r_4& r_5& r_6\\ r_7& r_8& r_9\end{array}\right]$ $t=\left[\begin{array}{c}{t}_x\\ t_y\\ t_z\end{array}\right]---\left(11\right)$

In the formula, R represents one 3 * 3 rotation real number matrix, r ₁..., r ₉Be real constant; T represents a translation vector, wherein, and t ₁..., t ₃Be real constant;

2), with coordinate figure P _c=(x _c, y _c, z _c) projection that in pin-hole model, standardizes, obtain normalized coordinate figure P _n(x y) is:

$P_n=\left[\begin{array}{c}{x}_c/z_c\\ y_c/z_c\end{array}\right]=\left[\begin{array}{c}x\\ y\end{array}\right]---\left(12\right)$

3), introduce the distortion of lens, the standardization coordinate figure P after the distortion _d(x _d, y _d) be expressed as:

$\left[\begin{array}{c}{x}_d\\ y_d\end{array}\right]=(1+k_1{r}^2+k_2{r}^4)\left[\begin{array}{c}x\\ y\end{array}\right]+\left[\begin{array}{c}2k_3\mathrm{xy}+k_4(r^2+2x^2)\\ k_3(r^2+2y^2)+2k_4\mathrm{xy}\end{array}\right]---\left(13\right)$

In the formula: r ²=x ²+ y ²k ₁, k ₂The expression coefficient of radial distortion; k ₃, k ₄Expression tangential distortion coefficient;

With P _d(x _d, y _d) be converted into the coordinate figure P that pixel coordinate is fastened on the image _p(u, v):

$\left\{\begin{array}{c}u=f_x{x}_d+u_0\\ v=f_y{y}_d+v_0\end{array}\right.---\left(14\right)$

Wherein:

$\left\{\begin{array}{c}{f}_x=f\&CenterDot;\mathrm{sx}/\mathrm{dpx}\\ f_y=f/\mathrm{dpy}\end{array}\right.---\left(15\right)$

In the following formula, f is the effective focal length of video camera; Sx is a scale factor, is used for adapting to all uncertain factors that sampling brings on computer picture x direction; Dpx is that computer picture is at the coverage between adjacent two pixels on the x direction; Dpy is that computer picture is at the coverage between adjacent two pixels on the y direction; u ₀Intersection point O for the camera lens optical axis Z and the plane of delineation _iAt pixel coordinate is coordinate figure on the u axle; O _iFor at pixel coordinate being the coordinate figure on the v axle.

Images