CN102096899A - Correction method of linear array camera image - Google Patents

Abstract

The invention discloses a correction method of a linear array camera image, comprising the following steps of, firstly, setting a calibration object in an object image to be processed; secondly, detecting a boundary point in the calibration object; thirdly, solving a motion parameter of the calibration object; and fourthly, carrying out affine transformation on a corresponding point in the object image to be processed according to a calculation result of the third step, wherein the image acquired after transformation is a correction image. The image correction method realizes geometric correction of a deformation image photographed by a linear array camera through mathematical transformation, is simple and easy to operate, has good correction effect and is particularly suitable for being combined with a method for processing the image by automatically searching the boundary point.

Description

A kind of bearing calibration of line-scan digital camera image

Technical field

The present invention relates to a kind of method for correcting image, relate in particular to a kind of method of carrying out geometry correction, belong to the machine vision technique field at the captured deformation pattern of line-scan digital camera (linecamera).

Background technology

In field of machine vision, line-scan digital camera is the special visual machine of a class.(area camera) compares with normally used area array cameras, and its imageing sensor has only delegation's photosensitive pixel, and this design makes high sweep frequency and high resolving power become possibility.Therefore, line-scan digital camera is mainly used in the Flame Image Process in fields such as industry, medical treatment, scientific research and safety.

The typical field of application of line-scan digital camera is to detect the object of continuous motion, especially lays particular emphasis on the detection to successive logistics, as wire rod, coiled material, tobacco, cotton, paper etc.Detected object is made uniform motion usually, utilizes one or more line-scan digital camera to its continuous sweep line by line, to reach the purpose that body surface is evenly detected.Under the situation of the very big visual field of needs or very high degree of precision, often use repeatedly excitation line array camera of excitation apparatus, repeatedly take pictures, again several captured " bar " shape images are merged, thereby generate complete subject image.

Because the image of line-scan digital camera on each constantly can only a certain line of shot object.In the practice,, be combined into the complete image of object often by continuous shot object each image constantly through camera.But if movement velocity and the design rate of object when the line-scan digital camera is inconsistent, then being out of shape can appear in the captured image of line-scan digital camera, as Fig. 1～shown in Figure 4.

In order to address this problem, people have successively attempted the multiple technologies solution.For example Gao Junchai, Lei Zhiyong and Wang Zemin are in " laser and optoelectronics progress " paper " image rectification of line-scan digital camera " that the 47th the 9th phase of volume delivered in 2010, at in the high precision dimensional measurement of part, the image acquisition of linear array CCD camera and the stationary problem of part movement are studied.The method for synchronous that adopts servomotor and motion control card is analyzed, proposed to adopt the control method of common asynchronous moter and scrambler combination; Studied kinetic control system and camera and gathered relation between the line rate, designed synchronous control mode based on step-length such as non-.According to the characteristics of line-scan digital camera image acquisition, as spotting, extracted the diameter parameters of chip direction and direction of motion with the standard circular target, realized the demarcation of object-image relation and correction interpolation coefficient; Designed linear interpolation algorithm, can proofread and correct the target imaging of arbitrary shape.Experimental result shows that this method is simple, and cost is low, can proofread and correct line array CCD effectively because the stretcher strain that motion causes.

Summary of the invention

Technical matters to be solved by this invention is to provide a kind of deformation pattern geometric correction method at line-scan digital camera.

For realizing above-mentioned goal of the invention, the present invention adopts following technical scheme:

A kind of bearing calibration of line-scan digital camera image is characterized in that comprising the steps:

(1) in pending subject image, sets the demarcation thing;

(2) detect the frontier point of demarcating in the thing;

(3) find the solution the kinematic parameter of demarcating thing;

(4) according to the result of calculation of step (3), the corresponding point in the pending subject image are carried out affined transformation, the image that is obtained after the conversion is correcting image.

Wherein more preferably, in described step (1), described demarcation thing is the circular object of determining in pending subject image.

Described demarcation thing is positioned at the central area of described subject image.

In the described step (2), the quantity of described frontier point is no less than 5.

Described step (3) comprises following substep:

(31) by (x of all frontier points _py _py _p ²x _py _p1) and x _p ²Form system of linear equations, under the situation of equation quantity, use least square method to find the solution more than unknown number quantity:

$\left(\begin{array}{ccccc}{x}_{p1}{y}_{p1}& {y_{p1}}^2& x_{p1}& y_{p1}& 1\\ ...& ...& ...& ...& ...\\ x_{\mathrm{pn}}{y}_{\mathrm{pn}}& {y_{\mathrm{pn}}}^2& x_{\mathrm{pn}}& y_{\mathrm{pn}}& 1\end{array}\right)\left(\begin{array}{c}{a}_1\\ a_2\\ a_3\\ a_4\\ a_5\end{array}\right)=\left(\begin{array}{c}-{x_{p1}}^2\\ ...\\ -{x_{\mathrm{pn}}}^2\end{array}\right)$

Wherein, (x _p, y _p) be the coordinate of frontier point, 1,2 ... n represents the sequence number of each frontier point;

(32) calculate the kinematic parameter of demarcating thing:

$c_1=\frac{a_1}{2}$

$c_2=\sqrt{a^2-{c_1}^2}$

$d_1=\frac{a_3}{2}+x_c$

$d_2=\frac{(a_4-c_1{a}_3)}{2c_2}+y_c.$

Wherein, c ₁Be x direction speed, c ₂Be y direction speed, d ₁Be the central point of x direction, d ₂Central point for the y direction.

In the described step (4), the c that calculates according to described step (3) ₁, c ₂, d ₁And d ₂Numerical value, carry out affined transformation to the corresponding point in the image:

$\left(\begin{array}{c}{x}_q\\ y_q\end{array}\right)=\left(\begin{array}{cc}1& c_1\\ 0& c_2\end{array}\right)\left(\begin{array}{c}{x}_p\\ y_p\end{array}\right)+\left(\begin{array}{c}{d}_1\\ d_2\end{array}\right)$

Wherein, (x _q, y _q) be the corresponding point coordinate after the conversion.

Method for correcting image provided by the present invention has realized that by mathematic(al) manipulation operation is simple to the geometry correction of the captured deformation pattern of line-scan digital camera, and calibration result is good; Especially be fit to the use that combines of the image processing method of search frontier point automatically.

Description of drawings

The present invention is described in further detail below in conjunction with the drawings and specific embodiments.

Fig. 1 is as the circular object synoptic diagram of demarcating thing;

Fig. 2 is the demarcation thing synoptic diagram that exists x direction mistake to cut;

Fig. 3 is the demarcation thing synoptic diagram that exists the y direction to stretch;

Fig. 4 cuts the demarcation thing synoptic diagram that stretches with the y direction for there being x direction mistake simultaneously;

Fig. 5 (a) and Fig. 5 (b) are respectively and adopt this method for correcting image to proofread and correct an example of the coin image of front and back;

Fig. 6 (a) and Fig. 6 (b) are respectively and adopt this method for correcting image to proofread and correct another example of the coin image of front and back.

Embodiment

Method for correcting image provided by the present invention can carry out geometry correction to the captured subject image of line-scan digital camera, thereby recovers the original true picture of object.Basic implementation step to this method for correcting image is described in detail below.

Step 1: in pending subject image, set and demarcate thing

Geometry deformation takes place in line-scan digital camera captured subject image easily, and the circular geometry deformation degree that can embody all angles, therefore, the present invention at first determines a circular object as the demarcation thing in pending subject image, and hypothesis (x _c, y _c) be the ideal position (as the central area of image) of employed demarcation thing in pending subject image, radius is r, the point (x on the then desirable circle _q, y _q) satisfy:

(x _q-x _c) ²+(y _q-y _c) ²＝r ² (1)

Step 2: detect the frontier point of demarcating in the thing

In this step, search for the frontier point of demarcating thing automatically or manually.Generally speaking, the quantity that should guarantee frontier point is no less than 5.

For detected each frontier point, suppose that its coordinate is (x _p, y _p), captured subject image has respectively that x direction mistake is cut, the y direction stretches and the conversion of translation, as Fig. 2, Fig. 3 and shown in Figure 4.Like this, each frontier point (x on the demarcation thing _p, y _p) all with ideal circle (ideal image after promptly proofreading and correct) on a frontier point (x _q, y _q) corresponding, so the establishment of formula (2) is arranged:

$\left(\begin{array}{c}{x}_q\\ y_q\end{array}\right)=\left(\begin{array}{cc}1& c_1\\ 0& c_2\end{array}\right)\left(\begin{array}{c}{x}_p\\ y_p\end{array}\right)+\left(\begin{array}{c}{d}_1\\ d_2\end{array}\right)---\left(2\right)$

Wherein, c ₁Be x direction speed, c ₂Be y direction speed, d ₁Be the central point of x direction, d ₂Central point for the y direction.Bring formula (2) into formula (1), obtain:

(x _p+c ₁y _p+d ₁-x _c) ²+(c ₂y _p+d _p-y _c) ²＝r ²

That is:

$\left(\begin{array}{ccccc}{x}_p{y}_p& {y_p}^2& x_p& y_p& 1\end{array}\right)\left(\begin{array}{c}{2c}_1\\ {c_1}^2+{c_2}^2\\ 2(d_1-x_c)\\ 2c_1(d_1-x_c)+2c_2(d_2-y_c)\\ {(d_1-x_c)}^2+{(d_2-y_c)}^2-r^2\end{array}\right)=-{x_p}^2---\left(3\right)$

Order:

$\left(\begin{array}{c}{a}_1\\ a_2\\ a_3\\ a_4\\ a_5\end{array}\right)=\left\{\begin{array}{c}{2c}_1\\ {c_1}^2+{c_2}^2\\ 2(d_1-x_c)\\ 2c_1(d_1-x_c)+2c_2(d_2-y_c)\\ {(d_1-x_c)}^2+{(d_2-y_c)}^2-r^2\end{array}\right\}---\left(4\right)$

Get (the x of all frontier points _py _py _p ²x _py _p1) and x _p ²Form system of linear equations, can obtain least square solution

Step 3: find the solution the kinematic parameter of demarcating thing

This step comprises following substep:

(1) finds the solution following system of equations, under the situation of equation quantity, use least square method to find the solution, so that make the result more stable more than unknown number quantity.

$\left(\begin{array}{ccccc}{x}_{p1}{y}_{p1}& {y_{p1}}^2& x_{p1}& y_{p1}& 1\\ ...& ...& ...& ...& ...\\ x_{\mathrm{pn}}{y}_{\mathrm{pn}}& {y_{\mathrm{pn}}}^2& x_{\mathrm{pn}}& y_{\mathrm{pn}}& 1\end{array}\right)\left(\begin{array}{c}{a}_1\\ a_2\\ a_3\\ a_4\\ a_5\end{array}\right)=\left(\begin{array}{c}-{x_{p1}}^2\\ ...\\ -{x_{\mathrm{pn}}}^2\end{array}\right)---\left(5\right)$

Wherein, 1,2 ... n represents the sequence number of the corresponding point that search obtains from pending subject image.

(2) calculate the kinematic parameter of demarcating thing.These kinematic parameters comprise x direction speed c ₁, y direction speed c ₂, x direction central point d ₁With y direction central point d ₂, wherein:

$c_1=\frac{a_1}{2}$

$c_2=\sqrt{a^2-{c_1}^2}$

$d_1=\frac{a_3}{2}+x_c---\left(6\right)$

$d_2=\frac{(a_4-c_1{a}_3)}{2c_2}+y_c$

Step 4:, the corresponding point in the pending subject image are carried out affined transformation according to the result of calculation of step 3.

Before address, demarcate each the frontier point (x on the thing _p, y _p) all with proofread and correct after ideal image on a frontier point (x _q, y _q) corresponding, the geometric transformation between them concerns as the formula (2):

$\left(\begin{array}{c}{x}_q\\ y_q\end{array}\right)=\left(\begin{array}{cc}1& c_1\\ 0& c_2\end{array}\right)\left(\begin{array}{c}{x}_p\\ y_p\end{array}\right)+\left(\begin{array}{c}{d}_1\\ d_2\end{array}\right)---\left(2\right)$

Obtaining under many situations the line-scan digital camera image after can obtaining to proofread and correct by affined transformation to the relation of the geometric transformation between the frontier point.Wherein, can adopt similar step to obtain corresponding geometric transformation relation, thereby realize affined transformation for other point of non-frontier point in the image.After each corresponding point in the image were carried out affined transformation respectively, the image of being made up of these corresponding point was the due true picture of the captured object of line-scan digital camera.

Fig. 5 (a) and Fig. 5 (b) are respectively and adopt this method for correcting image to proofread and correct an example of the coin image of front and back; Fig. 6 (a) and Fig. 6 (b) are respectively and adopt this method for correcting image to proofread and correct another example of the coin image of front and back.From these two groups correction results as can be seen, no matter be the coin image that any mode is twisted, adopt this method for correcting image all can obtain preferable geometry correction effect.

It is emphasized that method for correcting image provided by the present invention proofreaies and correct as benchmark with a circular frontier point of demarcating thing, 5 frontier points of minimum needs just can be proofreaied and correct, but the quantity of frontier point is many more, and the effect of correction is good more.Under a fairly large number of situation of frontier point, even there is error in the search to frontier point, even the minority frontier point be marked with mistake, can the result who proofread and correct not produced a very large impact, so this method for correcting image is well suited for the use that combines of the image processing method of search frontier point automatically yet.

More than the bearing calibration of line-scan digital camera image provided by the present invention is had been described in detail.To those skilled in the art, any conspicuous change of under the prerequisite that does not deviate from connotation of the present invention it being done all will constitute to infringement of patent right of the present invention, with corresponding legal responsibilities.

Claims (6)

1. the bearing calibration of a line-scan digital camera image is characterized in that comprising the steps:

(1) in pending subject image, sets the demarcation thing;

(2) detect the frontier point of demarcating in the thing;

(3) find the solution the kinematic parameter of demarcating thing;

(4) according to the result of calculation of step (3), the corresponding point in the pending subject image are carried out affined transformation, the image that is obtained after the conversion is correcting image.

2. the bearing calibration of line-scan digital camera image as claimed in claim 1 is characterized in that:

In the described step (1), described demarcation thing is the circular object of determining in pending subject image.

3. the bearing calibration of line-scan digital camera image as claimed in claim 1 is characterized in that:

In the described step (1), described demarcation thing is positioned at the central area of described subject image.

4. the bearing calibration of line-scan digital camera image as claimed in claim 1 is characterized in that:

In the described step (2), the quantity of described frontier point is no less than 5.

5. the bearing calibration of line-scan digital camera image as claimed in claim 1 is characterized in that:

Described step (3) comprises following substep:

(31) by (x of all frontier points _py _py _p ²x _py _p1) and x _p ²Form system of linear equations, under the situation of equation quantity, use least square method to find the solution more than unknown number quantity:

$\left(\begin{array}{ccccc}{x}_{p1}{y}_{p1}& {y_{p1}}^2& x_{p1}& y_{p1}& 1\\ ...& ...& ...& ...& ...\\ x_{\mathrm{pn}}{y}_{\mathrm{pn}}& {y_{\mathrm{pn}}}^2& x_{\mathrm{pn}}& y_{\mathrm{pn}}& 1\end{array}\right)\left(\begin{array}{c}{a}_1\\ a_2\\ a_3\\ a_4\\ a_5\end{array}\right)=\left(\begin{array}{c}-{x_{p1}}^2\\ ...\\ -{x_{\mathrm{pn}}}^2\end{array}\right)$

Wherein, (x _p, y _p) be the coordinate of frontier point, 1,2 ... n represents the sequence number of each frontier point;

(32) calculate the kinematic parameter of demarcating thing:

$c_1=\frac{a_1}{2}$

$c_2=\sqrt{a^2-{c_1}^2}$

$d_1=\frac{a_3}{2}+x_c$

$d_2=\frac{(a_4-c_1{a}_3)}{2c_2}+y_c.$

Wherein, c ₁Be x direction speed, c ₂Be y direction speed, d ₁Be the central point of x direction, d ₂Central point for the y direction.

6. the bearing calibration of line-scan digital camera image as claimed in claim 5 is characterized in that:

In the described step (4), the c that calculates according to described step (3) ₁, c ₂, d ₁And d ₂Numerical value, carry out affined transformation to the corresponding point in the image:

$\left(\begin{array}{c}{x}_q\\ y_q\end{array}\right)=\left(\begin{array}{cc}1& c_1\\ 0& c_2\end{array}\right)\left(\begin{array}{c}{x}_p\\ y_p\end{array}\right)+\left(\begin{array}{c}{d}_1\\ d_2\end{array}\right)$

Wherein, (x _q, y _q) be the corresponding point coordinate after the conversion.

Images