CN104820973A - Image correction method for distortion curve radian detection template - Google Patents

Abstract

The invention discloses an image correction method for a distortion curve radian detection template. First, the coordinates of the distortion center of a calibration plate image are calculated based on the radian of a distortion curve; then, on the basis, the area near the distortion center is initially corrected, and the image is reconstructed with use of the corrected area; next, the distortion coefficient is solved based on the distorted image, a distortion model and the image after reconstruction; and finally, an acquired scene image is corrected based on the distortion coefficient. Compared with the prior art, by independently solving the distortion center and the distortion coefficient, the coupling error of parameter solving is reduced effectively, and the correction precision is improved.

Description

The method for correcting image of distortion curve radian detection template

Technical field

The invention belongs to technical field of optical instrument, be specifically related to a kind of method for correcting image of distortion curve radian detection template.

Background technology

In recent years, along with deepening continuously of optical technology research, the application of imaging system is increasingly extensive.But because optical imaging system exists nonlinear geometry distortion, lens surface radian and the factor such as lens center and imaging plane dislocation, make imaging system camera lens existence distortion in various degree (as radial, tangential, thin prism distortion) problem.Under water (as fishing ground cultivation) in monitoring and measuring application, install bucker additional and use wide-angle lens, more exacerbating lens distortion, considerable influence is produced to successive image result.

Try to achieve systematical distortion coefficient by scaling method, utilize distortion factor to correct image, thus obtain the major technique that orthoscopic image is solution aberration problems.At present, the scaling method that optical imaging system mainly adopts has Tsai method, Zhang Zhengyou method etc.Tsai method utilizes radial constraint to solve video camera external parameter (translation matrix, rotation matrix and the anglec of rotation), based on the initial value of pin-hole model acquisition algorithm iteration, then obtains distortion of camera coefficient through nonlinear optimization search.Tsai method can reach good stated accuracy, but needs the non-co-planar three-dimensional scaling point obtained to be subject to the restriction of calibrating block precision, and timing is using picture centre as center of distortion, often cause the image scaled after correcting and ideal image inconsistent.Zhang Zhengyou method improves Tsai, the three-dimensional coordinate that scaling board in Tsai method is put is solved and is transformed into the two-dimensional coordinate that scaling board is put and solves, (i.e. scaling board used Z=0 in world coordinate system), camera parameters optimization solution is drawn by pin-hole model analytical calculation, then carry out nonlinear operation with based on maximum likelihood method [7], solve distortion factor.ZHANG method has good robustness, and can replace with plane reference plate the calibrating block that precision is higher, and practicality is stronger, but it is not independent to there is parametric solution, the large problem of coupling error.In recent years, propose based on above-mentioned two kinds of template standardizations that some demarcation speed are fast, succinct the improving one's methods of solution procedure, as the widely used Matlab tool box of one that the people such as Bouguet utilize Zhang Zhengyou algorithm to propose, input several different angles images, identify the angle point on scaling board, carry out distortion factor based on Zhang Zhengyou algorithm and solve, the method is applicable to the less demanding occasion of stated accuracy, reduce demarcation complexity, but need manual identification angle point; The camera self-calibration new method based on square templates that Chen Xi proposes, in conjunction with the double ratio relation in scaling board and projective geometry and harmonic conjugates character, solve distortion factor linearly, the method is applicable to the scene that there is geometrical property, dirigibility is higher, but poor robustness.But these methods above-mentioned do not solve the problem causing coupling error in parametric solution process.

In this context, pin image processing field distortion correction application demand of the present invention, a kind of method for correcting image of distortion curve radian detection template is provided, uncalibrated image and scene image is obtained by image collecting device, based on the method for distortion curve radian, calculate the image after correction efficiently.

Summary of the invention

Goal of the invention: in view of this, the present invention is directed to the above-mentioned or other defect existed in prior art, a kind of method for correcting image of distortion curve radian detection template is provided, solved by the calibrating template image gathered and obtain distortion factor, the target fault image gathered is processed, thus the correcting image of target can be obtained.

Technical scheme: for achieving the above object, the present invention adopts following technical scheme:

A method for correcting image for distortion curve radian detection template, solved by the scaling board image gathered and obtain distortion factor, then correct the scene image obtained based on distortion factor, wherein solving of distortion factor comprises the steps:

(1) radian of all curves in X-axis and Y direction calculating that black and white lattice intersection on scaling board image forms;

(2) in X-axis and Y direction, the minimum curve of two radians is respectively chosen respectively;

(3) least square line matching is carried out to two curves selected in step (2), obtain the undistorted straight line of ideal in X-axis and Y direction respectively;

(4) two the undistorted straight lines obtained in step (3) are carried out system of equations computing and obtain center of distortion coordinate;

(5) point centered by center of distortion, selected N × n-quadrant is that initial calibration is carried out in prime area, and obtain the N+1 bar straight line after the correction of X-axis and Y direction respectively, wherein N is default prime area black and white grid number;

(6) with the positive straight line of prime area lieutenant colonel for bus carries out scaling board image reconstruction;

(7) in conjunction with distortion model, distortional point coordinate and reconstruction point coordinate, solve and obtain distortion factor.

Further, the method solving the radian of a curve in X-axis or Y direction in step (1) for: on i-th curve coordinate a little, carry out least square line Fitting Analysis, the straight line expression formula of matching is y=a _ix+b _i, then utilize minimum mean-square error to solve crooked radian r ⁱ, wherein, (y _i,k, x _i,k) be the coordinate of kth on an i-th curve point.

Further, carrying out to the curve that two radians are minimum the method that fitting a straight line obtains desirable undistorted straight line in step (3) is: according to formula

$\left\{\begin{array}{c}{x}_{3,j}=x_{1,j}+(x_{2,j}-x_{1,j})\frac{\max (r_1,r_2)}{r_1+r_2}\\ y_{3,j}=y_{1,j}+(y_{2,j}-y_{1,j})\frac{\max (r_1,r_2)}{r_1+r_2}\end{array}\right.$

Ask for the curve L that two radians are minimum ₁and L ₂, crooked radian is designated as r respectively ₁and r ₂, between the coordinate figure of each point of Article 3 curve L3, carry out least square line matching, calculate crooked radian r ₃if, r ₃< r ₁, then by curve L ₃be designated as L ₁, and r ₃=r ₁; Otherwise r ₃< r ₂, then by curve L ₃be designated as L ₂, and r ₃=r ₂; Successive ignition calculate until | r ₃| < ε, wherein ε is the threshold value of setting, the curve L obtained ₃for undistorted straight line.

Further, in step (5), prime area bearing calibration specifically comprises:

(5.1) point centered by center of distortion O (U, V) obtained with step (4), selected 3 × 3 regions, as prime area, obtain 17, prime area coordinate points;

(5.2) three characteristics that the image after being corrected prime area meets are expressed as formula:

$\left\{\begin{array}{c}{p}_1={\mathrm{\&Sigma;}}_{l=1}^4{\mathrm{\&Sigma;}}_{i=1}^4{(y_i-a_1{x}_i-b_1)}^2\\ p_2={\mathrm{\&Sigma;}}_{r=1}^4{\mathrm{\&Sigma;}}_{i=1}^4{(y_i-a_r{x}_i-b_r)}^2\\ p_3=|{\mathrm{\&Sigma;}}_1^3{\mathrm{\&Sigma;}}_{i=1}^3\sqrt{{(x_i-x_{i+1})}^2+{(y_i-y_{i+1})}^2}+\sqrt{{\left(x_i{-x}_{i+4}\right)}^2+{(y_i-y_{i+4})}^3}-2L|\\ p_4={\mathrm{\&Sigma;}}_{r=1}^4{\mathrm{\&Sigma;}}_{l=1}^4{(k_l{k}_r+1)}^2\end{array}\right.$

Wherein a _r, b _rfor by the straight line coefficient that some matching is tried to achieve of four on straight line, (x _i, y _i) coordinate for scaling board image is put, L is the foursquare pixel length of side of gridiron pattern on scaling board image, k _land k _rfor the slope of straight line, p ₁and p ₂for curve linear characteristic is estimated, be worth less, curve is more close to ideal value; p ₃for adjacent two dot characteristics are estimated, be worth less, arbitrary neighborhood 2 distances are more close to ideal value; p ₄for the mutual vertical property of transverse and longitudinal axle straight line is estimated, be worth less, any transverse and longitudinal axle is close to ideal just state of value;

(5.3) p=p is asked for by least-squares iteration ₁+ p ₂+ p ₃+ p ₄minimum value, when p gets minimum value, prime area has corrected.

Further, in step (6), specifically comprise:

(6.1) extended to two ends by the straight line of prime area, with initial calibration edges of regions point for starting point, spacing distance L gets a little, tries to achieve the coordinate of remaining point on the straight line in all prime areas;

(6.2) four points arranged with remaining n-4 simulate the straight line of n-4 bar Y direction, and n is total columns of intersection point on scaling board image;

(6.3) simulate the straight line of m-4 bar X-direction with four points that remaining m-4 is capable, m is total line number of intersection point on scaling board image;

(6.4) coordinate of the point after namely the intersection point calculating all X-axis and Y direction straight line goes out to correct.

Further, the distortion model in step (7) is:

$\left[\begin{array}{cccc}{X}_p{r}^2& X_p{r}^4& 2X_p^2+r^2& 2X_p{Y}_P\\ Y_p{r}^2& Y_p{r}^4& 2X_p{Y}_P& 2Y_p^2+r^2\end{array}\right]\left[\begin{array}{c}{k}_1\\ k_2\\ k_3\\ p_1\\ p_2\end{array}\right]=\left[\begin{array}{c}{X}_d-X_p\\ Y_d-Y_p\end{array}\right]$

Wherein (U, V) is center of distortion point coordinate, (X _p, Y _p) be reconstruction point coordinate, (X _d, Y _d) be distortional point coordinate, k ₁, k ₂, k ₃, p ₁, p ₂for distortion factor.

Further, described based on distortion factor in the method that corrects of scene image obtained according to system of equations $\left\{\begin{array}{c}{X}_d=X_P+X_P(k_1{r}^2+k_2{r}^4)+p_1({2X}_p^2+r^2)+{2p}_2{X}_P{Y}_P\\ Y_d=Y_P+Y_P(k_1{r}^2+k_2{r}^4)+2p_1{X}_P{Y}_P+p_2({2Y}_p^2+r^2)\end{array}\right.$ Utilize distortional point coordinate figure (X _d, Y _d) try to achieve the ideal coordinates (X of all pixels in image _p, Y _p), thus obtain the image after correcting.

Beneficial effect: provided by the invention a kind of in the bearing calibration of fault image based on calibrating template distortion curve radian detect carry out solving of distortion factor.First, distortion curve radian is utilized to calculate center of distortion coordinate; On this basis, initial calibration is carried out to center of distortion near zone, then, utilize correcting area to rebuild image; Finally, carry out distortion factor in conjunction with image after fault image, distortion model and reconstruction and solve, then to the correct image obtained.Compared with prior art, the inventive method independently solves center of distortion and distortion factor, effectively reduces parametric solution coupling error, improves correction accuracy.

Accompanying drawing explanation

In order to make content of the present invention be more likely to be clearly understood, below basis specific embodiment and by reference to the accompanying drawings, the present invention is further detailed explanation, wherein:

Fig. 1 is the scaling board image schematic diagram gathered in the embodiment of the present invention;

Fig. 2 is center of distortion and the undistorted linear relation schematic diagram of ideal in the embodiment of the present invention;

Fig. 3 is initial calibration regional choice schematic diagram in the embodiment of the present invention;

Fig. 4 adopts the image rectification schematic flow sheet in the system of embodiment of the present invention method.

Embodiment

Disclosed in the invention process, a kind of method for correcting image of distortion curve radian detection template, first, utilizes Harris Corner Detection Algorithm, detects the coordinate of the joining of all black and white lattice on scaling board; Then based on the above-mentioned joining coordinate asked for, distortion curve radian is utilized to calculate center of distortion coordinate; On this basis, initial calibration is carried out to center of distortion near zone, then, utilize correcting area to rebuild image; Finally, carry out distortion factor in conjunction with image after fault image, distortion model and reconstruction and solve, then to the correct image obtained.Specifically comprise the steps:

(1) radian of every a line curve in scaling board image is calculated.As shown in Figure 1, gridiron pattern plane reference plate is the black and white lattice composition that the capable n of m arranges.Harris Corner Detection Algorithm is utilized to detect the pixel coordinate of all black and white lattice intersection on shooting image.With the upper left corner of image for initial point, image left edge is x-axis, and image coboundary is y-axis, terminate to the lower right corner from the upper left corner, the pixel coordinate of point is [i, j] (0≤i≤m-1,0≤j≤n-1), calculate the 0th row as follows to the method for the capable degree of crook of m-1.First to the coordinate of the n above every a line point, carry out least square line Fitting Analysis, the straight line expression formula of matching is y=a _ix+b _i, then utilize minimum mean-square error to solve crooked radian r _i,

$r_i=\underset{k=1}{\overset{j}{\mathrm{\&Sigma;}}}{(y_{i,k}-a_i{x}_{i,k}-b_i)}^2---\left(1\right)$

Wherein, (y _i,k, x _i,k) be the coordinate of kth on an i-th curve point.

(2) in the m bar curve required by step (1), two the curve i looking for crooked radian r minimum and curve i+1.Because curved direction, both sides, center of distortion is contrary, and adjacent, namely center of distortion is between curve i and i+1.Curve is designated as L ₁, L ₂, crooked radian is designated as r ₁and r ₂.

(3) according to formula

$\left\{\begin{array}{c}{x}_{3,j}=x_{1,j}+(x_{2,j}-x_{1,j})\frac{\max (r_1,r_2)}{r_1+r_2}\\ y_{3,j}=y_{1,j}+(y_{2,j}-y_{1,j})\frac{\max (r_1,r_2)}{r_1+r_2}\end{array}\right.---\left(2\right)$

Ask for Article 3 curve L between two curves ₃the coordinate figure (as Fig. 2) of each point, carry out least square line matching, calculate crooked radian r ₃.If r ₃< r ₁, then by curve L ₃be designated as L ₁, and r ₃=r ₁; Otherwise r ₃< r ₂, then by curve L ₃be designated as L ₂, and r ₃=r ₂.

(4) step (3) is repeated, until | r ₃| < ε, herein ε=0.001, now curve L ₃be similar to very much the X-axis straight line by photocentre, be designated as l ₁: y=a ₁x+b ₁.

(5) adopt the method for step (1) to (4) to calculate to each row in test template, the Y-axis straight line be similar to by photocentre can be tried to achieve, be designated as l ₂: y=a ₂x+b ₂.

Ask for the intersecting point coordinate of above-mentioned two straight lines, simultaneous two straight line, solving equation group

$\left\{\begin{array}{c}y=a_{1}x+b_1\\ y=a_{2}x+b_2\end{array}\right.---\left(3\right)$

Namely center of distortion coordinate is:

$\left[\begin{array}{c}y\\ x\end{array}\right]={\left[\begin{array}{cc}1& -a\\ 1& -a_2\end{array}\right]}^{-1}\left[\begin{array}{c}{b}_1\\ b_2\end{array}\right]---\left(4\right)$

(6) with the above-mentioned center of distortion obtained for initial point, selected 3*3 region shown in Fig. 3 is as prime area.Prime area selected in Fig. 3 comprises 17 coordinate points, is designated as q respectively ₁(x _i, y _j), q ₂(x _i, y _j+1), q ₃(x _i, y _j+2), q ₄(x _i, y _j+3), q ₅(x _i+1, y _j), q ₆(x _i+1, y _j+1), q ₇(x _i+1, y _j+2), q ₈(x _i+1, y _j+3), q ₉(x _i+2, y _j), q ₁₀(x _i+2, y _j+1), q ₁₁(x _i+2, y _j+2), q ₁₂(x _i+2, y _j+3), q ₁₃(x _i+3, y _j), q ₁₄(x _i+3, y _j+1), q ₁₅(x _i+3, y _j+2), q ₁₆(x _i+3, y _j+3), O (U, V).

(7) essence of distorted image correction is image three characteristic: a after correcting: the straight line during line of arbitrary curve point; B: the distance between arbitrary neighborhood 2 is equal; C: transverse and longitudinal axle straight line is mutually vertical; Representation formula is as shown in (5).

$\left\{\begin{array}{c}{p}_1={\mathrm{\&Sigma;}}_{l=1}^4{\mathrm{\&Sigma;}}_{i=1}^4{(y_i-a_1{x}_i-b_1)}^2\\ p_2={\mathrm{\&Sigma;}}_{r=1}^4{\mathrm{\&Sigma;}}_{i=1}^4{(y_i-a_r{x}_i-b_r)}^2\\ p_3=|{\mathrm{\&Sigma;}}_1^3{\mathrm{\&Sigma;}}_{i=1}^3\sqrt{{(x_i-x_{i+1})}^2+{(y_i-y_{i+1})}^2}+\sqrt{{\left(x_i{-x}_{i+4}\right)}^2+{(y_i-y_{i+4})}^3}-2L|\\ p_4={\mathrm{\&Sigma;}}_{r=1}^4{\mathrm{\&Sigma;}}_{l=1}^4{(k_l{k}_r+1)}^2\end{array}\right.---\left(5\right)$

Wherein k _land k _rbe respectively the slope of row and column place straight line, L is the square pixel length of side in the picture in gridiron pattern.

(8) distortion degree due to the region near center of distortion is little relative to other regional compare, so to choose near center of distortion 3*3 region as initial calibration region, using this initial calibration region as image reconstruction basis, general image reconstruction error is substituted, integral image error after correction for reduction with regional area correction error.Formula (5) is transformed to formula (6).So far the problem that distortional point corrects just transfers the formula of asking for (6) minimum problems to, when the distance of picture point all on corresponding straight line and between adjacent 2 all equals a fixed value (length of side of little square shaped cells grid), formula (6) will obtain minimum value, so by least square (LM) iteration when formula (6) obtains minimum value, prime area has corrected.

p＝p ₁+p ₂+p ₃+p ₄(6)

By using LM iterative algorithm, with the length of side L of template square cell, distortional point coordinate set q _iwith initial straight parameter (a, b) as input value, carry out interative computation, make the coordinate q of distortional point _i,dmove closer to ideal dot position q _i,p, correcting straight line and move closer to desirable undistorted straight line, when the distance D of consecutive point meets | D-L| < 0.001, iteration terminates, and formula (6) obtains minimum value, obtains the prime area after correcting.

(9) with the X of prime area, Y direction four straight lines for bus, three features that after correcting, on scaling board, straight line meets are utilized to seek out the coordinate that in X-direction, on every bar straight line, remaining n-4 is put, and the coordinate that in Y direction, on every bar straight line, remaining m-4 is put.

After the correction of prime area, this regional correction point puts three features after meeting correction, the point on same direction is utilized to be in collinear characteristic, 8 of prime area straight lines are extended to two ends, then utilize adjacent 2 distance properties equivalent, with initial calibration edges of regions point for starting point, spacing distance L gets a little, try to achieve the coordinate of remaining n-4 point on every bar straight line in X-direction, and the coordinate that in Y direction, on every bar straight line, remaining m-4 is put.

(10) 4 somes matching on same y direction on 4 horizontal linears is obtained n-4 bar longitudinal axis straight line.

(11) 4 somes matching on same level direction on 4 longitudinal axis straight lines is obtained m-4 bar horizontal linear.

(12) coordinate of the point after namely the intersection point calculating all X-axis and Y direction straight line goes out to correct.

(13) above-mentioned check point coordinate of trying to achieve and original distortional point coordinate are substituted into distortion model formula (7)

$\left[\begin{array}{cccc}{X}_p{r}^2& X_p{r}^4& 2X_p^2+r^2& 2X_p{Y}_P\\ Y_p{r}^2& Y_p{r}^4& 2X_p{Y}_P& 2Y_p^2+r^2\end{array}\right]\left[\begin{array}{c}{k}_1\\ k_2\\ k_3\\ p_1\\ p_2\end{array}\right]=\left[\begin{array}{c}{X}_d-X_p\\ Y_d-Y_p\end{array}\right]---\left(7\right)$

Wherein (X _p, Y _p) be ideal point coordinate, (X _d, Y _d) be distortional point coordinate.Obtain distortion factor (k ₁, k ₂, k ₃, p ₁, p ₂).

(14) distortion factor is utilized to carry out distortion correction to scene image.Scene image distortion correction process is as follows: distortion model formula (7) is carried out conversion and obtains updating formula (8):

$\left\{\begin{array}{c}{X}_d=X_P+X_P(k_1{r}^2+k_2{r}^4)+p_1({2X}_p^2+r^2)+{2p}_2{X}_P{Y}_P\\ Y_d=Y_P+Y_P(k_1{r}^2+k_2{r}^4)+2p_1{X}_P{Y}_P+p_2({2Y}_p^2+r^2)\end{array}\right.---\left(8\right)$

According to above-mentioned system of equations, utilize distortional point coordinate figure (X _d, Y _d) try to achieve the ideal coordinates (X of all pixels in image _p, Y _p), the image after correcting can be obtained.

As shown in Figure 4, for applying the image correction process process flow diagram of the intelligent distorted image correction system of method for correcting image of the present invention, first scaling board image and scene image is obtained by image collecting device, try to achieve distortion factor be stored in storer if uncalibrated image is then asked for by desirable undistorted straight line, center of distortion solves, prime area corrects, image reconstruction and distortion factor solve etc., when camera gets scene image, utilize the distortion factor stored to carry out image rectification in conjunction with distortion model to scene image, obtain orthoscopic image.Specific works flow process in system between each functional module is as follows:

1) camera is aimed at scaling board image and scene image respectively and is gathered scaling board image and scene image

2), during acquisition module end-of-job, acquisition control module sends a signal to storage control module, storage control module by gather view data through I ²c bus is stored into the raw data district of memory module.

3) in advance calibration coefficient acquisition algorithm and aberration correction algorithm are stored in the algorithm memory block of memory module.

4) after calibration coefficient acquisition module receives feedback signal, the calibration coefficient acquisition algorithm program of memory module algorithm memory block and the uncalibrated image in raw data district are loaded in calibration coefficient acquisition module, the calibration coefficient that the uncalibrated image loaded carries out detecting based on distortion curve radian is calculated.

5), after the process of calibration coefficient acquisition module terminates, calibration coefficient obtains control module will send feedback signal to storage control module and image procossing control module simultaneously.

6) calibration coefficient in storage area and scene fault image can be loaded into image processing module by memory module, after process terminates, image procossing control module can send and feed back signal to storage control module and display module, for storage and the display of correcting image.

Obviously, above-described embodiment is only for example of the present invention is clearly described, and is not the restriction to embodiments of the present invention.For those of ordinary skill in the field, can also make other changes in different forms on the basis of the above description.Here exhaustive without the need to also giving all embodiments.And these belong to spirit institute's apparent change of extending out of the present invention or change and are still within protection scope of the present invention.

Claims (7)

1. the method for correcting image of distortion curve radian detection template, solved by the scaling board image gathered and obtain distortion factor, then correct the scene image obtained based on distortion factor, it is characterized in that, wherein solving of distortion factor comprises the steps:

(1) radian of all curves in X-axis and Y direction calculating that black and white lattice intersection on scaling board image forms;

(2) in X-axis and Y direction, the minimum curve of two radians is respectively chosen respectively;

(3) least square line matching is carried out to two curves selected in step (2), obtain the undistorted straight line of ideal in X-axis and Y direction respectively;

(4) two the undistorted straight lines obtained in step (3) are carried out system of equations computing and obtain center of distortion coordinate;

(5) point centered by center of distortion, selected N × n-quadrant is that initial calibration is carried out in prime area, and obtain the N+1 bar straight line after the correction of X-axis and Y direction respectively, wherein N is default prime area black and white grid number;

(6) with the positive straight line of prime area lieutenant colonel for bus carries out scaling board image reconstruction;

(7) in conjunction with distortion model, distortional point coordinate and reconstruction point coordinate, solve and obtain distortion factor.

2. the method for correcting image of distortion curve radian detection template according to claim 1, it is characterized in that, the method solving the radian of a curve in X-axis or Y direction in step (1) for: on i-th curve coordinate a little, carry out least square line Fitting Analysis, the straight line expression formula of matching is y=a _ix+b _i, then utilize minimum mean-square error to solve crooked radian r _i, wherein, (y _i,k, x _i,k) be the coordinate of kth on an i-th curve point.

3. the method for correcting image of distortion curve radian detection template according to claim 1, is characterized in that, carries out the method that fitting a straight line obtains desirable undistorted straight line to be in step (3) to the curve that two radians are minimum: according to formula

$\left\{\begin{array}{c}{x}_{3,j}=x_{1,j}+(x_{2,j}-x_{1,j})\frac{\max (r_1,r_2)}{r_1+r_2}\\ y_{3,j}=y_{1,j}+(y_{2,j}-y_{1,j})\frac{\max (r_1,r_2)}{r_1+r_2}\end{array}\right.$

Ask for the curve L that two radians are minimum ₁and L ₂, crooked radian is designated as r respectively ₁and r ₂, between Article 3 curve L ₃the coordinate figure of each point, carry out least square line matching, calculate crooked radian r ₃if, r ₃< r ₁, then by curve L ₃be designated as L ₁, and r ₃=r ₁; Otherwise r ₃< r ₂, then by curve L ₃be designated as L ₂, and r ₃=r ₂; Successive ignition calculate until | r ₃| < ε, wherein ε is the threshold value of setting, the curve L obtained ₃for undistorted straight line.

4. the method for correcting image of distortion curve radian detection template according to claim 1, is characterized in that, in step (5), prime area bearing calibration specifically comprises:

(5.1) point centered by center of distortion O (U, V) obtained with step (4), selected 3 × 3 regions, as prime area, obtain 17, prime area coordinate points;

(5.2) three characteristics that the image after being corrected prime area meets are expressed as formula:

$\left\{\begin{array}{c}{p}_1={\mathrm{\&Sigma;}}_{l=1}^4{\mathrm{\&Sigma;}}_{i=1}^4{(y_i-a_1{x}_i-b_1)}^2\\ p_2={\mathrm{\&Sigma;}}_{r=1}^4{\mathrm{\&Sigma;}}_{i=1}^4{(y_i-a_r{x}_i-b_r)}^2\\ p_3=|{\mathrm{\&Sigma;}}_1^3{\mathrm{\&Sigma;}}_{i=1}^3\sqrt{{(x_i-x_{i+1})}^2+{(y_i-y_{i+1})}^2}+\sqrt{{(x_i-x_{i+4})}^2+{(y_i-y_{i+4})}^2}-2L|\\ p_4={\mathrm{\&Sigma;}}_{r=1}^4{\mathrm{\&Sigma;}}_{l=1}^4{(k_l{k}_r+1)}^2\end{array}\right.$

Wherein a _r, b _rfor by the straight line coefficient that some matching is tried to achieve of four on straight line, (x _i, y _i) coordinate for scaling board image is put, L is the foursquare pixel length of side of gridiron pattern on scaling board image, k _land k _rfor the slope of straight line, p ₁and p ₂for curve linear characteristic is estimated, be worth less, curve is more close to ideal value; p ₃for adjacent two dot characteristics are estimated, be worth less, arbitrary neighborhood 2 distances are more close to ideal value; p ₄for the mutual vertical property of transverse and longitudinal axle straight line is estimated, be worth less, any transverse and longitudinal axle is close to ideal just state of value;

(5.3) p=p is asked for by least-squares iteration ₁+ p ₂+ p ₃+ p ₄minimum value, when p gets minimum value, prime area has corrected.

5. the method for correcting image of distortion curve radian detection template according to claim 4, is characterized in that, in step (6), specifically comprises:

(6.1) extended to two ends by the straight line of prime area, with initial calibration edges of regions point for starting point, spacing distance L gets a little, tries to achieve the coordinate of remaining point on the straight line in all prime areas;

(6.2) four points arranged with remaining n-4 simulate the straight line of n-4 bar Y direction, and n is total columns of intersection point on scaling board image;

(6.3) simulate the straight line of m-4 bar X-direction with four points that remaining m-4 is capable, m is total line number of intersection point on scaling board image;

(6.4) coordinate of the point after namely the intersection point calculating all X-axis and Y direction straight line goes out to correct.

6. the method for correcting image of distortion curve radian detection template according to claim 1, its feature exists

In, the distortion model in step (7) is:

$\left[\begin{array}{cccc}{X}_p{r}^2& X_p{r}^4& {2X}_p^2+r^2& {2X}_p{Y}_p\\ Y_p{r}^2& Y_p{r}^4& {2X}_p{Y}_p& {2Y}_p^2+r^2\end{array}\right]\left[\begin{array}{c}{k}_1\\ k_2\\ k_3\\ p_1\\ p_2\end{array}\right]=\left[\begin{array}{c}{X}_d-X_p\\ Y_d-Y_p\end{array}\right]$

Wherein (U, V) is center of distortion point coordinate, (X _p, Y _p) be reconstruction point

Coordinate, (X _d, Y _d) be distortional point coordinate, k ₁, k ₂, k ₃, p ₁, p ₂for distortion factor.

7. the method for correcting image of distortion curve radian detection template according to claim 7, is characterized in that, described based on distortion factor in the method that corrects of scene image obtained according to system of equations $\left\{\begin{array}{c}{X}_d=X_P+X_P(k_1{r}^2+k_2{r}^4)+p_1({2X}_p^2+r^2)+{2p}_2{X}_P{X}_P\\ Y_d=Y_P+Y_P(k_1{r}^2+k_2{r}^4)+{2p}_1{X}_P{Y}_P+p_2({2Y}_p^2+r^2)\end{array}\right.$ Utilize distortional point coordinate figure (X _d, Y _d) try to achieve the ideal coordinates (X of all pixels in image _p, Y _p), thus obtain the image after correcting.