CN106323337A - Stereoscopic-vision relative-measurement-system error analysis method - Google Patents

Abstract

The invention discloses a stereoscopic-vision relative-measurement-system error analysis method. The stereoscopic-vision relative-measurement-system error analysis method includes the following steps that S1, a space three-dimensional target point measuring model of a stereoscopic-vision relative measurement system is established; S2, the accuracy of the image characteristics of the stereoscopic-vision relative measurement system is extracted and analyzed; S3, the focal-distance calibration accuracy of the stereoscopic-vision relative measurement system is analyzed; S4, the rotating-matrix calibration accuracy of the stereoscopic-vision relative measurement system is analyzed; S5, the translation vector calibration accuracy of the stereoscopic-vision relative measurement system is analyzed; S6, space three-dimensional target-point-measuring composition errors of the stereoscopic-vision relative measurement system are analyzed. By means of the stereoscopic-vision relative-measurement-system error analysis method, the focal-distance calibration accuracy, the rotating matrixes between stereoscopic-vision camera pairs and the translation vector calibration accuracy can be analyzed.

Description

A kind of stereoscopic vision relative measurement system error analysis method

Technical field

The present invention relates to spacecraft super close distance relative measurement technique, particularly belong to a kind of stereoscopic vision relative measurement system Error analysis method.

Background technology

By servicing in-orbit, it is possible to use the servicing aircraft in-orbit of advantage of lower cost carries out fault unit to inert satellite The replacing of device, recovers the function of satellite, or carries out fuel make up, extend its service life.And to complete service role in-orbit, Primary premise is can to find target at super close distance viability, and the relative status information obtaining target is supplied to control system Unite corresponding control task.Therefore, spacecraft super close distance relative measurement technique is by the success of service role in-orbit Whether key technology.

Vertical vision relative measurement is to utilize two cameras to obtain target image, based on image characteristic point the most simultaneously Parallax, according to trigonometry principle obtain target characteristic three-dimensional geometric information, including azimuth information and range information.Stereopsis Feel that the performance of relative measurement system is main with camera properties with parameter, locus and measured point and measurement system between two cameras The position of system is relevant.In actual application, once system calibrating, it is necessary for holding and is relatively fixed, the focal length of camera, camera Between baseline and orientation all can not change, so, the optimization of system structure parameter is designed concern it can maximum journey Degree plays the key of measurement performance.

Summary of the invention

It is an object of the invention to provide a kind of stereoscopic vision relative measurement system error analysis method, not only give figure As the impact on certainty of measurement of the feature extraction precision, the most respectively from focal length stated accuracy, vertical vision camera between spin matrix Analyzed with translation vector stated accuracy, and finally given the survey of stereoscopic vision relative measurement system space three-dimensional impact point The analytical expression of amount error analysis.

In order to realize object above, the present invention is achieved by the following technical solutions:

A kind of stereoscopic vision relative measurement system error analysis method, it is characterised in that comprise the steps of:

S1, sets up the space three-dimensional impact point measurement model of stereoscopic vision relative measurement system；

S2, is analyzed the image characteristics extraction precision of stereoscopic vision relative measurement system；

S3, is analyzed the focal length stated accuracy of stereoscopic vision relative measurement system；

S4, is analyzed the spin matrix stated accuracy of stereoscopic vision relative measurement system；

S5, is analyzed the translation vector stated accuracy of stereoscopic vision relative measurement system；

S6, measures synthetical error analysis to the space three-dimensional impact point of stereoscopic vision relative measurement system.

Described stereoscopic vision relative measurement system is the first stereo vision camera and second solid of horizontal symmetry setting Vision camera.

In described step S1, if the first stereo vision camera o-xyz is positioned at the initial point of world coordinate system and without spin, Image coordinate system is O_l-X_lY_l, effective focal length is f₁, the second described stereo vision camera coordinate system is o_r-x_ry_rz_r, image is sat Mark system is O_r-X_rY_r, effective focal length is f_r, camera Perspective transformation model have:

$S_l\left[\begin{array}{c}{X}_l\\ Y_l\\ 1\end{array}\right]=\left[\begin{array}{cccc}{f}_l& 0& 0& 0\\ 0& f_l& 0& 0\\ 0& 0& 1& 0\end{array}\right]\left[\begin{array}{c}{x}_l\\ y_l\\ z_l\end{array}\right]$

$s_r\left[\begin{array}{c}{X}_r\\ Y_r\\ 1\end{array}\right]=\left[\begin{array}{cccc}{f}_r& 0& 0& 0\\ 0& f_r& 0& 0\\ 0& 0& 1& 0\end{array}\right]\left[\begin{array}{c}{x}_r\\ y_r\\ z_r\end{array}\right]$

O-xyz coordinate system and o_r-x_ry_rz_rMutual alignment relation between coordinate system can pass through space conversion matrix M_lrTable It is shown as:

$\left[\begin{array}{c}{x}_r\\ y_r\\ z_r\end{array}\right]=M_{lr}\left[\begin{array}{c}x\\ y\\ z\\ 1\end{array}\right]=\left[\begin{array}{cccc}{r}_1& r_2& r_3& t_x\\ r_4& r_5& r_6& t_y\\ r_7& r_8& r_9& l_z\end{array}\right]\left[\begin{array}{c}x\\ y\\ z\\ 1\end{array}\right]$

M_lr=[R T]

In formula,It is respectively o-xyz coordinate system and o_r-x_ry_rz_rRotation between coordinate system Translation transformation vector between matrix and initial point thereof, it is known that, for the spatial point in o-xyz coordinate system, 2 camera image planes points it Between corresponding relation be:

${\mathrm{\ρ}}_r\left[\begin{array}{c}{X}_r\\ Y_r\\ 1\end{array}\right]=\left[\begin{array}{cccc}{f}_r{r}_1& f_r{r}_2& f_r{r}_3& f_r{t}_x\\ f_r{r}_4& f_r{r}_5& f_r{r}_6& f_r{r}_y\\ r_7& r_8& r_9& t_z\end{array}\right]\left[\begin{array}{c}z{X}_l/f_l\\ {\mathrm{zY}}_l/f_l\\ z\\ 1\end{array}\right];$

ρ_rIt is expressed as a nondimensional scale factor.

Then having, in step S1, objective point coordinates in space is:

$\left\{\begin{array}{c}x={\mathrm{zX}}_l/f_l\\ y={\mathrm{zY}}_l/f_l\\ z=\frac{f_l(f_r{t}_x-X_r{t}_z)}{X_r(r_7{X}_l+r_8{Y}_1+f_1{r}_9)-f_r(r_1{X}_1+r_2{Y}_l+f_l{r}_3)}\\ =\frac{f_l(f_r{t}_y-Y_r{t}_z)}{Y_r(r_7{X}_l+r_8{Y}_l+f_l{r}_9)-f_r(r_4{X}_l+r_5{Y}_l+f_l{r}_6)}\end{array}\right.$

Stereoscopic vision relative measurement system is simplified, it is assumed that f_l=f_r=f, simultaneously the first stereo vision camera and Two stereo vision camera horizontal symmetry are placed, it may be assumed that

$R=\left[\begin{array}{ccc}1& 0& 0\\ 0& 1& 0\\ 0& 0& 1\end{array}\right],T=\left[\begin{array}{c}{t}_x\\ 0\\ 0\end{array}\right],$

Obtain following abbreviation equation:

$z=\frac{f*t_x}{X_r-X_l}.$

The picture point difference that spatial point in described step S2 projects at a pair stereo vision camera determines forward error.

In described step S2, the measurement error of objective point is by image characteristics extraction precision:

${\mathrm{\Δz}}_2=-\frac{d^2}{f*t_x}*\mathrm{\Δ}x$

Wherein d is actual range, and f is camera focus, and Δ x is picture point changes in coordinates small quantity.

The measurement error of objective point is by described step S3 mid-focal length algnment accuracy:

Wherein Δ f is focal length variations small quantity.

Described step S4, particularly as follows: according to the relation between Eulerian angles and spin matrix, describe spin matrix, analyzes rotation The stated accuracy of the corner measurement error to objective point, the measurement to objective point of the stated accuracy of the described anglec of rotation Error is:

Wherein Δ r is that attitude angle changes small quantity, g_maxIntermediate function derivation maximum.

In described step S5, the measurement error of objective point is by translation vector stated accuracy: Wherein Δ t_xSmall quantity is changed for horizontal direction translation vector.

Described step S6, particularly as follows: set up synthetic error TRANSFER MODEL, analyzes stereoscopic vision relative measurement system error Synthetic error, described synthetic error is:

${\mathrm{\Δz}}_6=\frac{d}{f*t_x}*\sqrt{d^2{\mathrm{\Δx}}^2+{t_x}^2{\mathrm{\Δf}}^2+t_x{g}_{max}{\mathrm{\Δr}}^2+{\mathrm{f\Δt}}^2}.$

The present invention compared with prior art, has the advantage that

The present invention is relative to traditional error analysis method, and the present invention not only gives image characteristics extraction precision to survey The impact of accuracy of measurement, the most respectively from focal length stated accuracy, vertical vision camera between spin matrix and translation vector stated accuracy Analyzed, and finally given the resolution table of stereoscopic vision relative measurement system space three-dimensional impact point analysis of measurement errors Reach formula.This achievement has important directive significance to the design of stereoscopic vision relative measurement system system with engineer applied.

Accompanying drawing explanation

Fig. 1 is the schematic diagram of the present invention a kind of stereoscopic vision relative measurement system；

Fig. 2 is the distance change schematic diagram between stereoscopic vision relative measurement system of the present invention and measured target.

Detailed description of the invention

Below in conjunction with accompanying drawing, by describing a preferably specific embodiment in detail, the present invention is further elaborated.

A kind of stereoscopic vision relative measurement system error analysis method, comprises the steps of:

S1, sets up the space three-dimensional impact point measurement model of stereoscopic vision relative measurement system；

S2, is analyzed the image characteristics extraction precision of stereoscopic vision relative measurement system；

S3, is analyzed the focal length stated accuracy of stereoscopic vision relative measurement system；

S4, is analyzed the spin matrix stated accuracy of stereoscopic vision relative measurement system；

S5, is analyzed the translation vector stated accuracy of stereoscopic vision relative measurement system；

S6, measures synthetical error analysis to the space three-dimensional impact point of stereoscopic vision relative measurement system.

Above-mentioned stereoscopic vision relative measurement system is the stereo vision camera of a pair horizontal symmetry setting.

It is positioned at the initial point of world coordinate system and without spin as it is shown in figure 1, set the first stereo vision camera o-xyz, image Coordinate system is O_l-X_lY_l, effective focal length is f₁, second camera coordinate system is o_r-x_ry_rz_r, image coordinate system is O_r-X_rY_r, effectively burnt Away from for f_r, camera Perspective transformation model have:

$S_l\left[\begin{array}{c}{X}_l\\ Y_l\\ 1\end{array}\right]=\left[\begin{array}{cccc}{f}_l& 0& 0& 0\\ 0& f_l& 0& 0\\ 0& 0& 1& 0\end{array}\right]\left[\begin{array}{c}{x}_l\\ y_l\\ z_l\end{array}\right]$

$s_r\left[\begin{array}{c}{X}_r\\ Y_r\\ 1\end{array}\right]=\left[\begin{array}{cccc}{f}_r& 0& 0& 0\\ 0& f_r& 0& 0\\ 0& 0& 1& 0\end{array}\right]\left[\begin{array}{c}{x}_r\\ y_r\\ z_r\end{array}\right]$

O-xyz coordinate system and o_r-x_ry_rz_rMutual alignment relation between coordinate system can pass through space conversion matrix M_lrTable It is shown as:

$\left[\begin{array}{c}{x}_r\\ y_r\\ z_r\end{array}\right]=M_{lr}\left[\begin{array}{c}x\\ y\\ z\\ 1\end{array}\right]=\left[\begin{array}{cccc}{r}_1& r_2& r_3& t_x\\ r_4& r_5& r_6& t_y\\ r_7& r_8& r_9& l_z\end{array}\right]\left[\begin{array}{c}x\\ y\\ z\\ 1\end{array}\right]$

M_lr=[R T]

In formula,It is respectively o-xyz coordinate system and o_r-x_ry_rz_rRotation between coordinate system Translation transformation vector between matrix and initial point thereof.Understand, for the spatial point in o-xyz coordinate system, 2 camera image planes points it Between corresponding relation be:

${\mathrm{\ρ}}_r\left[\begin{array}{c}{X}_r\\ Y_r\\ 1\end{array}\right]=\left[\begin{array}{cccc}{f}_r{r}_1& f_r{r}_2& f_r{r}_3& f_r{t}_x\\ f_r{r}_4& f_r{r}_5& f_r{r}_6& f_r{r}_y\\ r_7& r_8& r_9& t_z\end{array}\right]\left[\begin{array}{c}z{X}_l/f_l\\ {\mathrm{zY}}_l/f_l\\ z\\ 1\end{array}\right].$

r₁ r₂ r₃…r₉Represent the element of spin matrix, t_x t_y t_zIt is expressed as the element of translation vector, ρ_rIt is expressed as a nothing The scale factor of dimension.

According to above-mentioned, having, in step S1, objective point coordinates in space is:

$\left\{\begin{array}{c}x={\mathrm{zX}}_l/f_l\\ y={\mathrm{zY}}_l/f_l\\ z=\frac{f_l(f_r{t}_x-X_r{t}_z)}{X_r(r_7{X}_l+r_8{Y}_1+f_1{r}_9)-f_r(r_1{X}_1+r_2{Y}_l+f_l{r}_3)}\\ =\frac{f_l(f_r{t}_y-Y_r{t}_z)}{Y_r(r_7{X}_l+r_8{Y}_l+f_l{r}_9)-f_r(r_4{X}_l+r_5{Y}_l+f_l{r}_6)}\end{array}\right.$

It is thus known that focal distance f_l,f_rWith spatial point in left and right magazine image coordinate, as long as obtain spin matrix R peace Move vector T and can be obtained by the three dimensional space coordinate of testee point.

Stereoscopic vision relative measurement system is carried out Rational Simplification, it is assumed that f_l=f_r=f, simultaneously two stereo vision camera Horizontal symmetry is placed, it may be assumed that

$R=\left[\begin{array}{ccc}1& 0& 0\\ 0& 1& 0\\ 0& 0& 1\end{array}\right],T=\left[\begin{array}{c}{t}_x\\ 0\\ 0\end{array}\right],$

For Stereo Vision Measurement System, z to error compared to x and y to for be main error, thus the present invention The impact that z is produced by each factor of selective analysis to error, obtains following abbreviation equation:

$z=\frac{f*t_x}{X_r-X_l}.$

The picture point difference that spatial point in above-mentioned steps S2 projects at a pair stereo vision camera determines forward error, When measured target distance camera is the most remote, two picture point coordinate differences of projection are the least, and the error of resolving is the biggest, meanwhile, and two phases Image difference between machine is relevant with distance, and distance is the most remote, and image difference is the least.

Z to the picture point difference that projected at two cameras by spatial point of error determine, get at measured target distance camera Time remote, two picture point coordinate difference (X of projection_r-X_l) the least, the error of resolving is the biggest.Assume X '=X_r-X_l, then solution formula Reform into:

$z=\frac{f*t_x}{X}$

Upper formula derivation is analyzed the precision change of the z that Δ X slight error causes.

$\mathrm{\Δ}z=-\frac{f*t_x}{X^{\′2}}*\mathrm{\Δ}X$

Here the image difference between two cameras is relevant with distance, it is assumed that d is actual range, then exist:

$\frac{X^{\′}}{f}=-\frac{t_x}{d}$

$X^{\′}=-\frac{f*t_x}{d}$

So having, in above-mentioned step S2, the measurement error of objective point is by image characteristics extraction precision:

Wherein d is actual range, and f is camera focus, and Δ x is that picture point changes in coordinates is small Amount.

In described step S3, the focal length value of demarcation independent to camera optical system can not be used directly to carry out stereoscopic vision Measure and use, because not in view of the alignment error that may introduce after assembling with detector, accordingly, it would be desirable to specific device Re-scale the focal length value after optical system is assembled with imaging detector with algorithm, analyze focal length stated accuracy to objective The impact that some measurement error is brought.

The focal length value of demarcation independent to camera lens can not be used directly to carry out the calculating of vision measurement, because not considering The alignment error that may introduce after assembling with detector, accordingly, it would be desirable to specific device and algorithm re-scale Jiao of camera lens Away from value, but calibration result still comprises certain calibration error.Next focal length algnment accuracy will be analyzed to measurement error Impact.

FormulaTo f derivation, can obtain:

By formula:

$X_r-X_1=-\frac{f*t_x}{d}$

Thus, formula (2) is readily modified as

Wherein Δ f is focal length variations small quantity.

Formula (3) is the focal length algnment accuracy measurement error to objective point

In described step S4, during stereo-visiuon measurement, vertical vision camera between rotation relationship as known Condition.The anglec of rotation between two measurement cameras and initial point relative coordinate values is demarcated by some standard feature points.But There is certain error in actual calibration process, the anglec of rotation spin matrix between two cameras represents there are 9 unknown numbers, But only three degree of freedom, the angle of pitch, yaw angle, roll angle between i.e. two cameras, according to Eulerian angles and spin matrix it Between relation, describe spin matrix, analyze the stated accuracy measurement error to objective point of the anglec of rotation.

During stereoscopic vision relative measurement, between two measurement cameras, attitude is as known conditions.Marked by some Quasi-characteristic point (known distance, it is known that null point coordinate figure) demarcates the anglec of rotation between two measurement cameras and initial point relative coordinate Value.But in actual calibration process, there is certain error, the anglec of rotation demarcated and translation accuracy of measurement will be analyzed to survey below The impact that amount error is brought.Anglec of rotation spin matrix between two cameras represents, has 9 unknown numbers, but only three Degree of freedom, the angle of pitch, yaw angle, roll angle between i.e. two cameras.For simplifying the analysis, it is assumed here that cause due to demarcation Three angle errors identical.Then according to the relation between Eulerian angles and spin matrix, spin matrix is:

$\begin{array}{cc}R=\left[\begin{array}{ccc}1& \mathrm{\Δ}r& \mathrm{\Δ}r\\ \mathrm{\Δ}r& 1& \mathrm{\Δ}r\\ \mathrm{\Δ}r& \mathrm{\Δ}r& 1\end{array}\right]& T=\left[\begin{array}{c}{t}_x\\ 0\\ 0\end{array}\right]\end{array}$

R, T are substituted into formulaThen can obtain:

$z=\frac{f^2{t}_x}{X_{r}f-X_{l}f+(X_r{X}_l+X_r{Y}_l-{\mathrm{fY}}_l-f^2)\mathrm{\Δ}r}$

When spin matrix does not has error time, actual value is:

$d=\frac{f^2{t}_x}{X_{r}f-X_{l}f}$

Then error:

$\begin{array}{c}\mathrm{\Δ}z=z-d\\ =\frac{f^2{t}_x\left(\right(X_r{X}_l+X_r{Y}_l-{\mathrm{fY}}_l+f^2\left)\mathrm{\Δ}r\right)}{(X_{r}f-X_{l}f)*(X_{r}f-X_{l}f+(X_r{X}_l+X_r{Y}_l-{\mathrm{fY}}_l-f^2\left)\mathrm{\Δ}r\right)}\end{array}$

Owing to actual error is smaller, time between two cameras, attitude calibrated error is 0.05 °, appearance between two cameras State calibrated error is less, therefore denominator (X in above formula_rX_l+X_rY_l-fY_l-f²) Δ r is much smaller than X_rf-X_lF, is negligible.

Then formulaRewritable it is:

$\mathrm{\Δ}z=\frac{(X_r{X}_l+X_r{Y}_l-{\mathrm{fY}}_l-f^2)\mathrm{\Δ}r}{X_{r}f-X_{l}f}*d$

Image difference is converted into distance, then formulaRewritable it is:

$\mathrm{\Δ}z=\frac{d^2}{f^2{t}_x}*(X_r{X}_l+X_r{Y}_l-{\mathrm{fY}}_l-f^2)*\mathrm{\Δ}r$

Order:

G=X_rX_l+X_rY_l-fY_l-f²

Ask for the extreme value of function g, it is assumed that the detector resolution that we select is h*w, picture dot a size of pix, then X_r,X_l And Y_r,Y_lSpan is respectively

$(-\frac{w*pix}{2},\frac{w*pix}{2})$

$(-\frac{h*pix}{2},\frac{h*pix}{2})$

By formulaSubstitute into formula g=X_rX_l+X_rY_l-fY_l-f²Then can obtain:

$g=X_l^2+\frac{{\mathrm{ft}}_x}{d}{X}_l+X_l{Y}_l+\frac{{\mathrm{ft}}_x}{d}{Y}_l-{\mathrm{fY}}_l-f^2$

Respectively to X in above formula_lAnd Y_lSeeking local derviation, analytic function is worth most:

$X_1=f-\frac{{\mathrm{ft}}_x}{d}$

$Y_1=\frac{{\mathrm{ft}}_x}{d}-2f$

Understand according to analyzing, X_r,X_lAnd Y_r,Y_lStationary point not in span, therefore, be most worth on two borders, i.e.

So, the measurement error of objective point is by the stated accuracy of the anglec of rotation:Wherein Δ R is that attitude angle changes small quantity, g_maxIntermediate function derivation maximum.

In described S5, during stereo-visiuon measurement, vertical vision camera between position relationship as known conditions, Demarcate the position relationship between two measurement cameras by some standard feature points, but in actual calibration process, have one Fixed error, the position relationship translation vector between two cameras represents, has 3 unknown numbers, but the most only considers The single position relationship of horizontal direction, puts before this and analyzes the stated accuracy of translation vector to objective point measurement error The impact brought.

Owing to two camera baselines are main the most in the x direction, substantially think not distance in y and z direction, in Practical Calculation During x relatively big on measurement error impact to the precision demarcated, thus, our Main Analysis is by t_xStated accuracy is measured final The impact of result, makes t_x=t '_x+Δt。

Wherein t_xComprise certain error, formulaIn to t_xDerivation, can obtain:

$\mathrm{\Δ}z=\frac{f}{X_r-X_l}*{\mathrm{\Δt}}_x$

Equally, picture point difference is converted into distance, then can obtain translation vector stated accuracy to objective point Measurement error:

Wherein Δ t_xSmall quantity is changed for horizontal direction translation vector.

Above-mentioned step S6 is particularly as follows: on affecting in stereoscopic vision relative measurement system space three-dimensional impact point measurement model After each component analyses item by item, set up synthetic error TRANSFER MODEL, analyze the comprehensive of stereoscopic vision relative measurement system error Error,

According to formula of error transmission:

$\mathrm{\Δ}z=\sqrt{{\left(\frac{\∂}{\∂x}\right)}^2{\mathrm{\Δx}}^2+{\left(\frac{\∂}{\∂t}\right)}^2{\mathrm{\Δi}}^2+{\left(\frac{\∂}{\∂r}\right)}^2{\mathrm{\Δr}}^2+{\left(\frac{\∂}{\∂f}\right)}^2{\mathrm{\Δf}}^2}$

Can obtain synthetic error:

${\mathrm{\Δz}}_6=\frac{d}{f*t_x}*\sqrt{d^2{\mathrm{\Δx}}^2+{t_x}^2{\mathrm{\Δf}}^2+t_x{g}_{max}{\mathrm{\Δr}}^2+{\mathrm{f\Δt}}^2}.$

In sum, one stereoscopic vision relative measurement system error analysis method of the present invention, not only give image The impact on certainty of measurement of the feature extraction precision, the most respectively from focal length stated accuracy, vertical vision camera between spin matrix and Translation vector stated accuracy is analyzed, and has finally given the measurement of stereoscopic vision relative measurement system space three-dimensional impact point The analytical expression of error analysis.

Although present disclosure has been made to be discussed in detail by above preferred embodiment, but it should be appreciated that above-mentioned Description is not considered as limitation of the present invention.After those skilled in the art have read foregoing, for the present invention's Multiple amendment and replacement all will be apparent from.Therefore, protection scope of the present invention should be limited to the appended claims.

Claims (9)

1. a stereoscopic vision relative measurement system error analysis method, it is characterised in that comprise the steps of:

S1, sets up the space three-dimensional impact point measurement model of stereoscopic vision relative measurement system；

S2, is analyzed the image characteristics extraction precision of stereoscopic vision relative measurement system；

S3, is analyzed the focal length stated accuracy of stereoscopic vision relative measurement system；

S4, is analyzed the spin matrix stated accuracy of stereoscopic vision relative measurement system；

S5, is analyzed the translation vector stated accuracy of stereoscopic vision relative measurement system；

S6, measures synthetical error analysis to the space three-dimensional impact point of stereoscopic vision relative measurement system.

2. stereoscopic vision relative measurement system error analysis method as claimed in claim 1, it is characterised in that described solid Vision relative measurement system is the first stereo vision camera and second stereo vision camera of horizontal symmetry setting.

3. stereoscopic vision relative measurement system error analysis method as claimed in claim 2, it is characterised in that described step In S1, if the first stereo vision camera o-xyz is positioned at the initial point of world coordinate system and without spin, image coordinate system is O_l- X_lY_l, effective focal length is f₁, the second described stereo vision camera coordinate system is o_r-x_ry_rz_r, image coordinate system is O_r-X_rY_r, have Effect focal length is f_r, camera Perspective transformation model have:

$s_l\left[\begin{array}{c}{X}_l\\ Y_l\\ 1\end{array}\right]=\left[\begin{array}{cccc}{f}_l& 0& 0& 0\\ 0& f_l& 0& 0\\ 0& 0& 1& 0\end{array}\right]\left[\begin{array}{c}{x}_l\\ y_l\\ z_l\end{array}\right]$

$s_r\left[\begin{array}{c}{X}_r\\ Y_r\\ 1\end{array}\right]=\left[\begin{array}{cccc}{f}_r& 0& 0& 0\\ 0& f_r& 0& 0\\ 0& 0& 1& 0\end{array}\right]\left[\begin{array}{c}{x}_r\\ y_r\\ z_r\end{array}\right]$

O-xyz coordinate system and o_r-x_ry_rz_rMutual alignment relation between coordinate system can pass through space conversion matrix M_lrRepresent For:

$\left[\begin{array}{c}{x}_r\\ y_r\\ z_r\end{array}\right]=M_{lr}\left[\begin{array}{c}x\\ y\\ z\\ 1\end{array}\right]=\left[\begin{array}{cccc}{r}_1& r_2& r_3& t_x\\ r_4& r_5& r_6& t_y\\ r_7& r_8& r_9& l_z\end{array}\right]\left[\begin{array}{c}x\\ y\\ z\\ 1\end{array}\right]$

M_lr=[R T]

In formula,It is respectively o-xyz coordinate system and o_r-x_ry_rz_rSpin matrix between coordinate system and Translation transformation vector between its initial point, it is known that, for the spatial point in o-xyz coordinate system, right between 2 camera image planes points Should be related to for:

${\mathrm{\ρ}}_r\left[\begin{array}{c}{X}_r\\ Y_r\\ 1\end{array}\right]=\left[\begin{array}{cccc}{f}_r{r}_1& f_r{r}_2& f_r{r}_3& f_r{t}_x\\ f_r{r}_4& f_r{r}_5& f_r{r}_6& f_r{r}_y\\ r_7& r_8& r_9& t_z\end{array}\right]\left[\begin{array}{c}z{X}_l/f_l\\ {\mathrm{zY}}_l/f_l\\ z\\ 1\end{array}\right];$

ρ_rIt is expressed as a nondimensional scale factor；

Then having, in step S1, objective point coordinates in space is:

$\left\{\begin{array}{c}x={\mathrm{zX}}_l/f_l\\ y={\mathrm{zY}}_l/f_l\\ z=\frac{f_l(f_r{t}_x-X_r{t}_z)}{X_r(r_7{X}_l+r_8{Y}_1+f_1{r}_9)-f_r(r_1{X}_l+r_2{Y}_l+f_l{r}_3)}\\ =\frac{f_l(f_r{t}_y-Y_r{t}_z)}{Y_r(r_7{X}_l+r_8{Y}_l+f_l{r}_9)-f_r(r_4{X}_l+r_5{Y}_l+f_l{r}_6)}\end{array}\right.$

Stereoscopic vision relative measurement system is simplified, it is assumed that f_l=f_r=f, the first stereo vision camera and second stands simultaneously Body vision camera horizontal symmetry is placed, it may be assumed that

$R=\left[\begin{array}{ccc}1& 0& 0\\ 0& 1& 0\\ 0& 0& 1\end{array}\right],T=\left[\begin{array}{c}{t}_x\\ 0\\ 0\end{array}\right],$

Obtain following abbreviation equation:

$z=\frac{f*t_x}{X_r-X_l}.$

4. stereoscopic vision relative measurement system error analysis method as claimed in claim 2, it is characterised in that described step S2 In the picture point difference that projects at a pair stereo vision camera of spatial point determine forward error.

5. stereoscopic vision relative measurement system error analysis method as claimed in claim 2, it is characterised in that described step In S2, the measurement error of objective point is by image characteristics extraction precision:

${\mathrm{\Δz}}_2=-\frac{d^2}{f*t_x}*\mathrm{\Δ}x$

Wherein d is actual range, and f is camera focus, and Δ x is picture point changes in coordinates small quantity.

6. stereoscopic vision relative measurement system error analysis method as claimed in claim 2, it is characterised in that described step S3 The measurement error of objective point is by mid-focal length algnment accuracy:

Wherein Δ f is focal length variations small quantity.

7. stereoscopic vision relative measurement system error analysis method as claimed in claim 2, it is characterised in that described step S4, particularly as follows: according to the relation between Eulerian angles and spin matrix, describe spin matrix, analyzes the stated accuracy of the anglec of rotation to three The measurement error of dimension impact point, the measurement error of objective point is by the stated accuracy of the described anglec of rotation:

Wherein Δ r is that attitude angle changes small quantity, g_maxIntermediate function derivation maximum.

8. stereoscopic vision relative measurement system error analysis method as claimed in claim 2, it is characterised in that described step In S5, the measurement error of objective point is by translation vector stated accuracy:

Wherein Δ t_xSmall quantity is changed for horizontal direction translation vector.

9. stereoscopic vision relative measurement system error analysis method as claimed in claim 2, it is characterised in that described step S6, particularly as follows: set up synthetic error TRANSFER MODEL, analyzes the synthetic error of stereoscopic vision relative measurement system error, and described combines Conjunction error is:

${\mathrm{\Δz}}_6=\frac{d}{f*t_x}*\sqrt{d^2{\mathrm{\Δx}}^2+{t_x}^2{\mathrm{\Δf}}^2+t_x{g}_{max}{\mathrm{\Δr}}^2+{\mathrm{f\Δt}}^2}.$