CN1758020A - Stereo vision detection system based on adaptive sine streak projection - Google Patents

Abstract

A stereoscopic visional detecting system based on the adaptive sine fringe projection is composed of computer system, 5D scanning frame, sine fringe projector based on acousto-optic deflector, electric unit, visional sensor and image acquisition unit. When multiple sine rasters are projected onto surface of object, a series of images with different phases is captured by image sensor, then sent to image acquisition unit, and finally processed by computer to obtain the phase diagram about the 3D information of object.

Description

Stereo vision detection system based on the adaptive sine streak projection

Technical field

The present invention relates to a kind of stereo vision detection system based on the adaptive sine streak projection.

Background technology

Because the pattern and the distance of zone of ignorance and target can faster, more accurate, more automatically be surveyed and locate to real-time three-dimensional stereoscopic visual detector fast, make Vision Builder for Automated Inspection have stronger function, therefore aspect automatic target identification, guiding and location, caused common concern both domestic and external, be widely used in the autonomous space articulation automatically between product quality detection, reverse-engineering and spacecraft.

The outstanding advantage of stereoscopic vision is less demanding to measurement environment; Only need calibrating camera, so have higher stated accuracy.Three-dimensional coupling is the key that guarantees the stereoscopic vision precision, and the high contrast characteristics of existing matching process such as dependency graph pictures such as the constraint of outer polar curve rule, sequence constraint rule and area of space class rule, but general objectives does not have such feature, therefore cause the matching precision of stereoscopic vision lower, be difficult to be suitable in a lot of applications.

The projection gate phase matching then is a kind of novel three-dimensional matching way, stereoscopic vision detecting method based on the projection gate phase matching adopts projection striated structure measured surface, thereby obtain three-dimensional match point, this method can obtain the measuring accuracy higher than passive vision method.

Because the pattern of measured target is different, the curved transition scope is bigger, therefore when throwing measured surface with a kind of striped of fixed width, resolution can not meet the demands, phase hit appears in the zone big in curved transition, make phase place go parcel mistake to occur, can't find three-dimensional match point in this zone, also just can't obtain the data of relevant range, therefore can't solve the high-acruracy survey problem of measured target, so the width of projection striped need be determined according to the curved transition size of measured surface.When the curved transition of measured surface is big, need the wide striped of projection, prevent phase hit; When the curved transition of measured surface hour, need the projection stria, improve systemic resolution; Simultaneously, because the projection of this striped is unidirectional, little when the requirement ratio of the intensity of light source is added corner reflector.

The present stereoscopic vision detecting method based on the sine streak projective technique has following several: (1) utilizes the projection of Luo Qi grating out of focus, (2) utilize double grating to form Moire fringe, (3) utilize interferometer to form interference fringe, (4) utilize projector projects numeral striped.In this several method, the sine streak that first three kind can obtain to simulate, but the change of fringe phase and width must realize by the structural parameters that change projection system, want to realize accurate phase shifts, must be attached with complicated accurate mechanical movement means, complex structure is regulated difficulty, system's poor anti jamming capability.The 4th kind of method is to utilize commercial product as the projection projectoscope projection striped based on LCD or DLP.Because the striped of projection is the Computer Simulation image, therefore can accurately obtain desired phase shift and width of fringe by program, satisfy the characteristics of striped self-adaptation projection, but because this projector projects is digital phase, so separating the phase precision is affected, projector volume, weight are big in addition, are unfavorable for integrated and dirigibility on-line testing.

Summary of the invention

Technology of the present invention is dealt with problems and is: overcome the deficiencies in the prior art, provide a kind of under the condition that does not change structural parameters, be applied to signal frequency and phase place on the striped grenade instrumentation by change, accurately change the phase place and the width of striped, increase the stereo vision detection system based on the adaptive sine streak projection of space application reliability, this system has enlarged the scope of application of 3D vision target identification system, adaptivity is strong, improved the system identification accuracy, solved workpiece sensing identification problem with complex surface.

One of technical solution of the present invention: based on the stereo vision detection system of adaptive sine streak projection, its characteristics are to comprise:

Based on the adaptive sine streak grenade instrumentation of acoustooptic deflector, be used for joining with computer system, by the phase place and the frequency of sine streak that computer system control is throwed to several sinusoidal gratings of measured object surface projection;

Binocular sensor based on visual theory, be positioned at two ends based on the adaptive sine streak grenade instrumentation of acoustooptic deflector, by the relation of binocular sensor and the formation triangulation of testee surface, obtain the phase modulation image of measured surface, send into computing machine via image pick-up card;

Five dimension scanning supports, vision sensor is installed and based on the adaptive sine streak grenade instrumentation of acoustooptic deflector on it, the binocular vision sensor that is used to drive based on visual theory arrives different spatial and with different attitudes testee is taken measurement, and the motor for five dimension scanning supports provides driving power simultaneously;

Electric part, input end and computer system output terminal join altogether, output is connected to five dimension scanning supports, is used for the control command of computer system output is converted to drive control signal, moves and rotates along surface level and vertical plane along three directions of XYZ to drive five dimension scanning supports;

The image acquisition part, the phase image signal of the measured surface that will take based on the vision sensor of visual theory amplifies, sends into computer system after the A/D conversion process via image pick-up card filtering;

Computer system, as the control center of whole detection system, control five dimension scanning supports and move, export the phase place of the sine streak that control signal control throwed based on the adaptive sine streak grenade instrumentation of acoustooptic deflector and phase image data that frequency, control collection are obtained based on the binocular sensor of visual theory, the phase image data of being gathered are carried out the phase diagram that Flame Image Process obtains having object dimensional information thereby be used to output a control signal to electrical system.

Two of technical solution of the present invention: the stereo vision detection system based on the adaptive sine streak projection is characterized in that comprising following parts:

Based on the adaptive sine streak grenade instrumentation of acoustooptic deflector, be used for joining with computer system, by the phase place and the frequency of sine streak that computer system control is throwed to several sinusoidal gratings of measured object surface projection;

Based on the monocular sensor of visual theory, the monocular sensor is positioned at a side of adaptive sine streak grenade instrumentation, with the relation of testee surface formation triangulation, obtains the phase modulation image of measured surface, sends into computing machine via image pick-up card;

Five dimension scanning supports, the monocular sensor is installed and based on the adaptive sine streak grenade instrumentation of acoustooptic deflector on it, the monocular sensor that is used to drive based on visual theory arrives different spatial and with different attitudes testee is taken measurement, and the motor for five dimension scanning supports provides driving power simultaneously;

Electric part, input end and computer system output terminal join altogether, output is connected to five dimension scanning supports, is used for the control command of computer system output is converted to drive control signal, moves and rotates along surface level and vertical plane along three directions of XYZ to drive five dimension scanning supports;

The image acquisition part, the phase image signal of the measured surface that will take based on the monocular sensor of visual theory amplifies, sends into computer system after the A/D conversion process via image pick-up card filtering;

Computer system, as the control center of whole detection system, control phase place that five dimension scanning supports moved, exported the sine streak that control signal control throwed based on the adaptive sine streak grenade instrumentation of acoustooptic deflector and frequency, acquisition phase view data, the phase image data of being gathered are carried out phase place separate and handle the phase diagram that obtains having object dimensional information mutually thereby be used to output a control signal to electrical system.

The above-mentioned adaptive sine streak grenade instrumentation based on acoustooptic deflector is made up of acoustooptic deflector, modulated laser, drive signal circuit and optical system, drive signal circuit produces two paths of signals, one the road drives acoustooptic deflector, another road is used to drive modulated laser, produce the grid that reversed of stress distributes in the acoustooptic deflector, when modulating lasering beam passes through acoustooptic deflector with certain angle, outgoing beam produces diffraction, and two lens focuss of bundle diffraction light in optical system produce the sinusoidal interference ripple.

Above-mentioned image processing section comprises the image pre-service, separates phase, whole audience phase demodulation, the three-dimensional coupling of phase place and three-dimensional appearance calculation processes.

Principle of work of the present invention: move on the measured object surface by computer system output control signal control five dimension scanning supports, and by computer system control based on the sine streak grenade instrumentation of acoustooptic deflector to several sinusoidal gratings of measured object surface projection, obtained the image of a series of outs of phase delivers in the computer system via image pick-up card by vision sensor, the processing such as reaching noise filtering mutually of separating of view data is carried out in view data processing section in the computer system then, finally obtains having the phase diagram of object dimensional information.

Because the present invention compared with prior art has following advantage:

(1) the present invention uses adaptive striped grenade instrumentation, can be according to the curved transition situation on measured target surface, change spacing, phase place and the contrast of throwing striped fast by computer control, can enlarge the scope of application of 3D vision target identification system, adaptivity is strong.

(2) since fringe phase move need not be complicated accurate mechanical movement means, therefore simple in structure, easy to adjust, the measuring accuracy height, antijamming capability is strong.

(3) adopt binocular sensor or monocular sensor and striped grenade instrumentation and testee surface to form the relation of triangulation, can take, improved the system identification accuracy from different visual angles.

(4) can obtain highdensity point cloud data by the three-dimensional coupling of phase place.

Description of drawings

Fig. 1 is a composition structured flowchart of the present invention;

Fig. 2 is the perspective view of one of technical solution of the present invention;

Fig. 3 is two spatial structure structured flowcharts of technical solution of the present invention;

Fig. 4 is the adaptive sine streak grenade instrumentation structural representation based on acoustooptic deflector of the present invention;

The software flow pattern of Fig. 5 computer system of the present invention;

Fig. 6 resolves process flow diagram for phase place in the computer system of the present invention;

Fig. 7 determines the phase place synoptic diagram for three kinds of space sinusoidal grating gray-scale values of the present invention;

The raster phase figure of Fig. 8 service plate model of the present invention;

Fig. 9 accurate phase diagram of the present invention;

Use the phase diagram after the method for lining by line scan is launched fully among Figure 10 the present invention;

The process flow diagram of Figure 11 judgement phase quadrant of the present invention;

The background identification synoptic diagram of Figure 12 phase diagram of the present invention;

Figure 13 is the service plate three-dimensional surface point cloud chart among the present invention;

Figure 14 is the binocular vision sensor three-dimensional coordinates measurement mathematical model among the present invention.

Embodiment

As Fig. 1, shown in Figure 2, the present invention is by computer system 1, five dimension scanning supports 2, sine streak grenade instrumentation 3 based on acoustooptic deflector, electric part 4, vision sensor 5 based on visual theory, image acquisition part 6 is formed, wherein computer system 1 is as the center of whole detection, thereby output a control signal to electrical system 4 control five the dimension scanning supports 2 move, export the phase place and the frequency of the sine streak that control signal control throwed based on the adaptive sine streak grenade instrumentation 3 of acoustooptic deflector simultaneously, gather the phase image data that the vision sensor 5 based on visual theory obtains in addition, the phase image data of being gathered are carried out phase place separate and handle the phase diagram that obtains having object dimensional information mutually; On the five dimension scanning supports 2 vision sensor 5 and adaptive sine streak grenade instrumentation 3 are installed, vision sensor 5 is positioned at the two ends of adaptive sine streak grenade instrumentation 3, vision sensor 5 adopts two ccd video cameras, be left CCD and the right CCD among Fig. 2, testee 7 places on the five dimension scanning supports 2 simultaneously, move five dimension scanning supports 2 and be used to realize that vision sensor 5 arrives different spatial and takes measurement with different attitudes, five dimension scanning supports 2 comprise realizes straight-line leading screw of three dimensions and guide rail, and three stepper motors, wherein three stepper motors are controlled three the axle Xs of vision sensor 5 along quadrature respectively, Y, motion on the Z direction, two universal stages of two step motor control rotate, be separately implemented at the rotation on surface level and the vertical plane, by controlling five stepper motors, realize the scanning support control of five degree of freedom; Electric part 4 connects between five dimension scanning supports 2 and the computer system 1, its effect is to provide power supply for five stepper motors in the five dimension scanning supports 2, rotation by 5 stepper motors of the driver control in it, reception by the data acquisition and processing (DAP) in the computer system 1 partly send to the Stepping Motor Control pulse signal, and the limit switch signal of each guide rail and turntable, and export five dimension scanning supports 2 to; Image acquisition part 6 is an image pick-up card, the OK-M20 of image pick-up card for producing from company in the good perseverance.

As shown in Figure 3, another technical solution of the present invention for a change binocular sensor is the monocular sensor, monocular sensor, striped grenade instrumentation and measured workpiece surface form the triangulation relation, when grenade instrumentation 3 during to several sinusoidal gratings of testee 7 surface projection, obtain phase image by ccd video camera, send into then and carry out Flame Image Process in the computer system 1.Measure based on the trigonometry principle, be to constitute a triangle between video camera and striped grenade instrumentation and the testee 7, position relation between known video camera and the striped grenade instrumentation just can be measured the three-dimensional dimension of object in the camera field of view scope and the three-dimensional coordinate of space object unique point.Before measuring, must demarcate the mutual alignment relation of system and the parameter of each components interior.

As shown in Figure 4, adaptive sine streak grenade instrumentation 3 based on acoustooptic deflector of the present invention is by acoustooptic deflector 31, modulated laser 32, drive signal circuit 33 and optical system 34 are formed, drive signal circuit 33 produces two paths of signals, one the road drives acoustooptic deflector 31, another road is used to drive modulated laser 32, produce the grid that reversed of stress distributes in the acoustooptic deflector 31, when modulating lasering beam passes through acoustooptic deflector with certain angle, outgoing beam produces diffraction, the plano-convex lens of two bundle diffraction lights in optical system 34 focuses on and produces the sinusoidal interference ripple, and the phase place of the sinusoidal interference striped that is produced is relevant with the frequency difference between two drive signals of acoustooptic deflector with the phase differential of acousto-optic drive signal by the laser modulation signal respectively with spacing.After laser beam entered acoustooptic deflector 31, mixed frequency signal was divided into two bundles with incident beam, and along with the increase of front-end driven signal, the angle of two light beams also becomes greatly, and the cycle of resulting interference fringe diminishes, and made that promptly striped attenuates, pitch smaller.When the front end drive signal reduces, the angle of two light beams diminishes, and resulting striped chap, spacing become big.There is phase differential between the control signal of the control signal of modulated laser 32 and acoustooptic deflector 31, make the asynchronism(-nization) of the propagation of two light beams in acoustooptic deflector 31, produce constant phase differential, so just make two luminous points after lens converge produce constant phase differential, therefore the phase place of interference fringe just can change by the phase differential that changes control signal, the automatic control of drive signal be can realize, the phase place and the spacing variation of not moving any parts and finishing the projection striped reached.

Above-mentioned drive signal circuit 33 produces the two-way frequency by the DDS chip circuit and is all f _mThe sinusoidal signal of scalable phase value and frequency values, wherein one road signal and acoustooptic deflector 31 center frequency signal f _cBy the synthetic mixed frequency signal f of high-speed multiplier _c+ f _mAnd f _c-f _m, being used to drive acoustooptic deflector 31, other one road signal generates square-wave signal through after the frequency multiplication by high-speed comparator, is used to drive modulated laser 32.Modulated laser 32 adopts the TTL modulated laser, in the process that produces interference fringe, modulated laser 32 is for total system provides light source, behind modulated laser 32 irradiation acoustooptic deflectors 31, ultrasound wave in the acoustooptic deflector 31 can be projected on the background, produce interference fringe.Converging the two-beam line with lens after, acoustooptic deflector 31 just can produce spacing, interference fringe that phase place is adjustable, the spacing of front-end driven signal deciding striped wherein, the difference on the frequency F of the two-way drive signal by adjusting acoustooptic deflector can realize the adjusting to fringe spacing b.

$b=\frac{v_{s}L}{2F{f}_m}---\left(7\right)$

V in the formula _sBe ultrasonic speed in crystal; L is a striped outgoing distance.

By regulating drive signal and the laser modulation phase difference between signals that drives acoustooptic deflector, can realize phase shift control to striped.

Optical system 34 comprises various optics supports, the turntable that sets up experimental provision, also comprises striped lens combination and laser collimation lens.

As shown in Figure 5, the software section of computer system 1 of the present invention mainly comprise the control that drives five dimension scanning supports 2 and move, to based on the control of the phase place of the adaptive sine streak grenade instrumentation 3 projection sine streaks of acoustooptic deflector and frequency with the phase image data of being gathered are carried out phase place and separate the view data processing section of handling mutually.When detection system of the present invention started, at first detection system hardware guaranteed that the various piece of system all normally is connected with system, and working properly, and the parameter of initialization system various piece, guarantees that each subsystem is in original state; Second step, judge image capturing system whether be in workpiece directly over, whether operating distance meets the demands and the angle of systematic optical axis and surface of the work normal whether less than 15 degree, otherwise drive controlling five dimension scanning supports 2 drive acquisition systems arrive correct position; In the 3rd step, judge whether the width and the brightness of projection striped is suitable, according to the parameter regulation width of fringe and the brightness such as reflectivity, angle of inclination and surface curvature of surface of the work; In the 4th step, order changes the projection fringe phase, gathers by the image after the surface of the work modulation; In the 5th step, use image phase as shown in Figure 6 to resolve the D coordinates value that software flow calculates surface of the work.

As shown in Figure 6, the flow process of Flame Image Process of the present invention comprises the image pre-service, resolves phase diagram, resolves the steps such as the quadrant that launches, judges phase place fully, the three-dimensional coupling of identification background, phase place and three-dimensional appearance calculating of precise phase figure, phase image.

The image pre-service mainly is the image that is obtained by sensor acquisition to be carried out filtering and noise reduction handle, and the present invention adopts general filtering and noise reduction algorithm.

It is the raster image through the testee modulation to be carried out phase place resolve mutually that image is separated, and its algorithm principle is as follows:

When grating projects testee 7 surface and goes up, on body surface, produce distortion, form deformed grating, write carrier phase and modulated phase place as one and be:

I(x，y)＝γ(x，y){I _DC+I _mcos[Φ(x，y)]} (1)

Wherein (x y) is the reflection coefficient of projection light gate region, I to γ _DCBe the DC component of grating sine streak, I _mCos[Φ (x, y)] be that the Grating Modulation light intensity changes.

Using fringe phase φ is three kinds of sinusoidal projection raster images of { φ=0, φ=pi/2, φ=π }, carries out phase place and resolves.Phase place Φ can differ the gray-scale value I of the CCD picture point of pi/2 according to three width of cloth ₀, I _Pi/2, I _πDetermine that formula is as follows:

I ₀＝γ(I _DC+I _msinΦ)

I _π/2＝γ(I _DC+I _mcosΦ) (2)

I _π＝γ(I _DC-I _msinΦ)

$\mathrm{\Φ}={\tan }^{-1}\left[\frac{I_0-I_{\mathrm{\π}}}{2I_{\mathrm{\π}/2}-(I_0+I_{\mathrm{\π}})}\right]---\left(3\right)$

As shown in Figure 7, pass through I ₀, I _πDetermine the DC component of sine streak, just can determine phase value according to formula (2) again.The phase diagram that uses the service plate model that this method obtains as shown in Figure 8.

Be to improve the adaptive capacity to environment of phase place calculation accuracy and system, the present invention adopts the method for thickness striped combination, cord is easy to phase unwrapping and the many advantages of stria phase information fully combine.The cord period ratio is bigger, and the quantity that phase step occurs is fewer, but is difficult to obtain satisfied phase place calculation accuracy; Stria calculation accuracy height falls but occur the phase place volume easily, if the morphology change of object is excessive, stria is easy to generate the disappearance in whole cycle, brings a lot of troubles to phase unwrapping.Adopt the method for thickness striped combination to remedy mutually, can effectively avoid the defective of single striped method.Can get by formula (2).The phase value of each picture point is:

$\mathrm{\Φ}={\tan }^{-1}\left[\frac{I_0-I_{\mathrm{\π}}}{2I_{\mathrm{\π}/2}-(I_0+I_{\mathrm{\π}})}\right]$

Fringe position number for the phase place correspondence of this picture point can be determined by following formula:

${\mathrm{<figure-callout\; id="1"\; label="N"\; filenames="A20051008692800221.png"\; state="\{\{state\}\}">N</figure-callout>}}_1=\frac{\mathrm{\&Phi;}}{2\mathrm{\&pi;}}\&le;1,(\mathrm{\&Phi;}\&le;2\mathrm{\&pi;},{\mathrm{<figure-callout\; id="1"\; label="N"\; filenames="A20051008692800221.png"\; state="\{\{state\}\}">N</figure-callout>}}_1\&le;1)---\left(4\right)$

When the striped that adopts different thicknesses carries out phase place when resolving, can determine accurate fringe position number by following formula.

$N_i=\frac{{\mathrm{\&Phi;}}_i}{2\mathrm{\&pi;}}+\mathrm{round}(X{N}_{i-1}-\frac{{\mathrm{\&Phi;}}_i}{2\mathrm{\&pi;}})---\left(5\right)$

X is the ratio in cord cycle and stria cycle in the formula, and round is the rounding operation symbol.Formula (5) is the recursion formula of a phase place, and upper level striped (cord) is determined the integral part of phase value, and next stage striped (stria) is determined the fraction part of phase place, can effectively avoid the phase place disappearance of stria like this.By two kinds of stripeds of thickness, can accurately determine the phase value of each picture point.The present invention adopts the phase image of three kinds of width, uses recursion formula (5), makes phase image more accurate.The result compares with the phase image (Fig. 8) of cord as shown in Figure 8, and the image that Fig. 9 obtains has not been seen the influence of plate shadow, and has fully kept the phase information of stria simultaneously.

The purpose of whole audience phase demodulation is exactly that reference point when each row is launched is all unified on the point.The present invention adopts the method for lining by line scan, at first noise spot is carried out filtration treatment, judge the phase step point then, in deterministic process, might run into some noises, it is the phase step point that these noise spots are easy to be mistaken as, launch to make troubles for like this whole audience of phase place, so the method that the present invention uses continuous detecting is eliminated the influence of noise spot to phase unwrapping.Phase diagram after the whole audience that the application method of lining by line scan obtains launches as shown in figure 10.

Because the span of Φ is-pi/2～pi/2 in the formula (2), it need be expanded to 0～2 π, so just need judge the quadrant at Φ place.As shown in figure 11, I ₀, I _Pi/2, I _πBe three gray-scale values of certain picture point, just can judge the quadrant at Φ place by the positive and negative relation of Φ and the magnitude relationship of these three amounts, thereby determine Φ is launched in 0～2 π scope.

In resolving the phase place process, because background equally can the reflective projection grating, therefore be easy to together be resolved out together with image phase information, bring very burden to three-dimensional imaging.Automatically the identification background is very important in a three-dimensional imaging technology.

The method of the identification background that the present invention adopts is finished according to gray-value variation, and in general, the ability of object surfaces and background surface reflection ray is different, i.e. reflection coefficient γ (x, y) difference in the formula (1).Identification background concrete grammar is as follows: suppose I ₀I _Pi/2I _πBe three gray-scale values of certain picture point, if

|I ₀-I _π/2|+|I ₀-I _π|＞threshold (6)

Prove that this picture point is a testee, otherwise this picture point is a background.In application, according to actual conditions threshold value threshold can be set suitably background is effectively discerned.The phase image of identification phase image of background and unidentified background as shown in figure 12, left figure is the phase image of unidentified background, right figure is the phase image of discerning behind the background image, and background is made as same gray-scale value.

The single eye stereo vision system is by single camera and striped grenade instrumentation and the surperficial triangle relation that forms of testee, system is behind the phase image that obtains the testee surface, obtain the conversion coefficient of phase image and three-dimensional coordinate point according to the parameter of the mutual alignment of single camera and striped grenade instrumentation relation and video camera and striped grenade instrumentation, phase image can be converted to three-dimensional appearance surface point cloud by these conversion coefficients.

One of three-dimensional gordian technique of mating the stereo vision detection system that is based on the binocular vision principle of phase place, it mainly is according to the phase image of separating the left and right sides sensor that obtains mutually, employing is based on the projection gate phase method of phase place, the coupling of two width of cloth image corresponding point about carrying out.

Based on the coupling of phase place is with the phase place of each pixel foundation as coupling.The space a bit because they are the same by the situation of Grating Modulation, so the phase value in left and right image should equate in theory.When the projection travers in measured object when surface, the every bit on the object is through calculating the horizontal phase value that all can obtain this point, but because carrier wave and modulated, the phase value of point that may be on a certain striped is identical.So, need throw longitudinal stripe again, the every bit on the object can obtain vertical phase value of this point.Like this, the every bit on the object has two phase values, has the coordinate figure of x, y both direction just as the every bit in the rectangular coordinate system.The PHASE DISTRIBUTION of supposing left figure point p is (Φ _Lu, Φ _Lv), then point by point scanning in right figure is up to finding 1 q (Φ _Ru, Φ _Rv), make | Φ _Lu-Φ _Ru|＜α, | Φ _Rv-Φ _Rv|＜β wherein, α, β are the phase deviation threshold values of setting, and illustrates that some p are a pair of match points with putting q.

According to this principle, can carry out the coupling of being had a few in the public domain.The method of phase matching need not be sought unique point, and is easy to operate, and operand is also little, and the ratio of precision of coupling is higher, and the mistake matching rate is low.

Intrinsic parameter according to relation of the mutual alignment between two video cameras of system calibrating gained and camera chain itself, above-mentioned match point is transformed into world coordinate system, obtaining object dimensional pattern surface point cloud, is the three-dimensional table millet cake cloud of service plate gained after aforementioned a series of data processing as shown in figure 13.

As Fig. 2 and shown in Figure 14, binocular sensor 3 based on visual theory of the present invention by about two ccd video cameras constitute, when grenade instrumentation 2 during to several sinusoidal gratings of testee 7 surface projection, obtain phase image by two ccd video cameras, send into then and carry out Flame Image Process in the computer system 1.Binocular sensor 6 based on visual theory adopts perspective imaging model, measure based on the trigonometry principle, promptly constitute a triangle between the plane of delineation of two video cameras and the testee 7, position relation between known two video cameras just can be measured the three-dimensional dimension of object in the two video camera public view field and the three-dimensional coordinate of space object unique point.Before measuring, must demarcate the mutual alignment relation of system and the parameter of each components interior.

Corresponding relation between spatial point that the sensor measurement coordinate system is represented and the 2 video camera image planes points is:

In the formula

Be desirable computer picture coordinate;

Be world coordinates; P is 3 * 4 matrixes, becomes projection matrix, and it has characterized the fundamental relation between two dimensional image coordinate and the three-dimensional world coordinate.A is a upper triangular matrix, is called the intrinsic parameters of the camera matrix; R and T are respectively rotation matrix between world coordinate system and the camera coordinate system and the translation transformation vector between the initial point.

Because the camera optics system exists mismachining tolerance and rigging error, consider that radially proofreading and correct with tangential distortion of imaging system is a kind of expansion to desirable perspective model.Radial distortion cause the real image coordinate at the plane of delineation along the diametric(al) displacement, tangential distortion is caused by the centrifugal distortion of camera lens.

But the radial distortion approximate representation is:

$\left[\begin{array}{c}{\mathrm{\&delta;u}}^r\\ {\mathrm{\&delta;v}}^r\end{array}\right]=\left[\begin{array}{c}{X}_u(k_1{\mathrm{<figure-callout\; id="2"\; label="r"\; filenames="A20051008692800242.png"\; state="\{\{state\}\}">r</figure-callout>}}^2+k_2{r}^4)\\ Y_u(k_1{\mathrm{<figure-callout\; id="2"\; label="r"\; filenames="A20051008692800242.png"\; state="\{\{state\}\}">r</figure-callout>}}^2+k_2{r}^4)\end{array}\right]---\left(8\right)$

But the tangential distortion approximate representation is:

$\left[\begin{array}{c}{\mathrm{\&delta;u}}^t\\ \mathrm{\&delta;}{v}^t\end{array}\right]=\left[\begin{array}{c}{2p}_1{X}_u{Y}_u+p_2({\mathrm{<figure-callout\; id="2"\; label="r"\; filenames="A20051008692800242.png"\; state="\{\{state\}\}">r</figure-callout>}}^2+2X_u^2)\\ p_1({\mathrm{<figure-callout\; id="2"\; label="r"\; filenames="A20051008692800242.png"\; state="\{\{state\}\}">r</figure-callout>}}^2+2Y_u^2)+2p_2{X}_u{Y}_u\end{array}\right]---\left(9\right)$

Wherein, X _u, Y _uBe the ideal image coordinate of initial point at the figure principal point, ${\mathrm{<figure-callout\; id="2"\; label="r"\; filenames="A20051008692800242.png"\; state="\{\{state\}\}">r</figure-callout>}}^2={\mathrm{<figure-callout\; id="2"\; label="X\; u"\; filenames="A20051008692800242.png"\; state="\{\{state\}\}">X</figure-callout>}}_u^{}+{\mathrm{<figure-callout\; id="2"\; label="Y\; u"\; filenames="A20051008692800242.png"\; state="\{\{state\}\}">Y</figure-callout>}}_u^{};$ k ₂, k ₁, p ₁, p ₂Be distortion factor.Then Shi Ji frame deposit coordinate (u v) is:

$\left[\begin{array}{c}u\\ v\end{array}\right]=\left[\begin{array}{c}{X}_u+{\mathrm{\&delta;u}}^r+{\mathrm{\&delta;u}}^t\\ Y_u+{\mathrm{\&delta;v}}^r+{\mathrm{\&delta;v}}^t\end{array}\right]+\left[\begin{array}{c}{u}_0\\ v_0\end{array}\right]---\left(10\right)$

The objective function of nonlinear optimization is expressed as with analytical expression:

$F\left(x\right)=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}({(U_i-u_i)}^2+{(V_i-v_i)}^2)---\left(11\right)$

(U wherein _i, V _i) be model image coordinate according to the model projection of the video camera unique point to the plane of delineation; (u _i, v _i) image coordinate that arrives for actual detection.The process of nonlinear optimization is exactly to make deviation between model image coordinate and the real image coordinate reach minimum, is the estimated value of camera parameters thereby obtain deviation parameter hour.

Formula is rewritten as:

$\mathrm{\&rho;}\left[\begin{array}{c}{u}_i\\ v_i\\ 1\end{array}\right]=\left[\begin{array}{cccc}{a}_{11}& a_{12}& a_{13}& a_{14}\\ a_{21}& a_{22}& a_{23}& a_{24}\\ a_{31}& a_{32}& a_{33}& a_{34}\end{array}\right]\left[\begin{array}{c}{X}_i\\ Y_i\\ Z_i\\ 1\end{array}\right]---\left(12\right)$

In order to solve parameter a ₁₁... a ₃₄, following formula is rewritten as:

LA＝0 (13)

0 is the zero vector of 2n * 1 in the formula

A＝[a ₁₁ a ₁₂ a ₁₃ a ₁₄ a ₂₁ a ₂₂ a ₂₃ a ₂₄ a ₃₁ a ₃₂ a ₃₃ a ₃₄] ^T

$L=\left[\begin{array}{cccccccccccc}{X}_1& Y_1& {\mathrm{<figure-callout\; id="1"\; label="Z"\; filenames="A20051008692800221.png"\; state="\{\{state\}\}">Z</figure-callout>}}_1& 1& 0& 0& 0& 0& {-X}_1{u}_1& -Y_1{u}_1& -Z_1{u}_1& -{\mathrm{<figure-callout\; id="1"\; label="u"\; filenames="A20051008692800221.png"\; state="\{\{state\}\}">u</figure-callout>}}_1\\ 0& 0& 0& 0& X_1& Y_1& {\mathrm{<figure-callout\; id="1"\; label="Z"\; filenames="A20051008692800221.png"\; state="\{\{state\}\}">Z</figure-callout>}}_1& 1& -X_1{v}_1& -Y_1{v}_1& -Z_1{v}_1& -v_1\\ M& M& M& M& M& M& M& M& M& M& M& M\\ X_i& Y_i& {\mathrm{<figure-callout\; id="1"\; label="Z\; i"\; filenames="A20051008692800221.png"\; state="\{\{state\}\}">Z</figure-callout>}}_i& 1& 0& 0& 0& 0& -X_i{u}_i& -Y_i{u}_i& -Z_i{u}_i& -{\mathrm{<figure-callout\; id="0"\; label="u\; i"\; filenames="A20051008692800242.png"\; state="\{\{state\}\}">u</figure-callout>}}_i\\ 0& 0& 0& 0& X_i& Y_i& Z_i& 1& -X_i{v}_i& -Y_i{v}_i& -Z_i{v}_i& -v_i\\ M& M& M& M& M& M& M& M& M& M& M& M\\ X_n& Y_n& {\mathrm{<figure-callout\; id="1"\; label="Z\; n"\; filenames="A20051008692800221.png"\; state="\{\{state\}\}">Z</figure-callout>}}_n& 1& 0& 0& 0& 0& -X_n{u}_n& -Y_n{u}_n& -Z_n{u}_n& -{\mathrm{<figure-callout\; id="0"\; label="u\; n"\; filenames="A20051008692800242.png"\; state="\{\{state\}\}">u</figure-callout>}}_n\\ 0& 0& 0& 0& X_n& Y_n& Z_n& 1& -X_n{v}_n& -Y_n{v}_n& -Z_n{v}_n& -v_n\end{array}\right]---\left(14\right)$

Separating of equation, promptly calculate the constrained optimization problem that A is actually a following objective function:

J＝A ^TL ^TLA (15)

Satisfy constraint: || A||=1, corresponding Lagrangian function is

1＝A ^TL ^TLA-λ(A ^TA-1) (16)

Differential is: 1/ A=2L ^TLA-2 λ A=0,

Promptly

(L ^TL)A＝λA (17)

Formula (17) shows matrix L ^TThe eigenvector of L is separating of A, matrix L ^TThe eigenwert of L is: λ=A ^TL ^TLA.By the definition of singular values of a matrix as can be known, matrix L ^TTherefore the eigenvector of L is actually the unusual vector of matrix L, by matrix L is carried out svd, and need not calculate L ^TL just can solve A, thereby increases data stability.Matrix L ^TThe least square eigenvector of L is corresponding to the eigenwert of minimum, i.e. the pairing unusual vector of minimum singular value.

After obtaining projection matrix A, just can calculate whole parameters of video camera.Decomposable process is as follows:

$\left[\begin{array}{cc}{a}_1^T& a_{14}\\ a_2^T& a_{24}\\ a_3^T& a_{34}\end{array}\right]=\left[\begin{array}{cccc}{f}_x& 0& {\mathrm{<figure-callout\; id="0"\; label="u"\; filenames="A20051008692800242.png"\; state="\{\{state\}\}">u</figure-callout>}}_0& 0\\ 0& {\mathrm{<figure-callout\; id="0"\; label="f\; y\; v"\; filenames="A20051008692800242.png"\; state="\{\{state\}\}">f</figure-callout>}}_y& v_{}{}_0& 0\\ 0& 0& 1& 0\end{array}\right]\left[\begin{array}{cc}{r}_1^T& t_x\\ r_2^T& t_y\\ r_3^T& t_z\\ 0^{\mathrm{<figure-callout\; id="1"\; label="T"\; filenames="A20051008692800221.png"\; state="\{\{state\}\}">T</figure-callout>}}& 1\end{array}\right]=\left[\begin{array}{cc}{f}_x{r}_1^T+u_0{r}_3^T& f_x{t}_x+u_0{t}_z\\ f_y{r}_2^T+v_0{r}_3^T& f_y{t}_y+v_0{t}_z\\ r_3^T& t_z\end{array}\right]---\left(18\right)$

A in the formula _i ^TThe row that (i=1～3) are formed for capable first three element of i of the projection matrix A that tries to achieve is vectorial; a _I4(i=1～3) are capable the 4th column element of i of matrix A; r _i ^T(i=1～3) are that the i of rotation matrix R is capable; t _x, t _y, t _zElement for translation vector.By the orthogonality of formula (18) and R as can be known:

r ₃＝p ₃ r ₁＝(p ₁-u ₀p ₃)/f _x

${\mathrm{<figure-callout\; id="0"\; label="u"\; filenames="A20051008692800242.png"\; state="\{\{state\}\}">u</figure-callout>}}_0=p_1^T{p}_3$ r ₂＝(p ₂-v ₀p ₃)/f _y

${\mathrm{<figure-callout\; id="0"\; label="v"\; filenames="A20051008692800242.png"\; state="\{\{state\}\}">v</figure-callout>}}_0=p_2^T{p}_3$ t _x＝p ₃₄ (19)

f _x＝|p ₁×p ₃| t _y＝(p ₁₄-u ₀)/f _x

f _y＝|p ₂×p ₃| t _z＝(p ₂₄-v ₀)/f _y

The camera parameters that formula (19) is obtained is as the initial value of camera parameters, and suppose that all distortion factor initial values are 0, be optimized with the objective function of Levenberg-Marquardt nonlinear optimization method, just can obtain optimum whole camera parameters formulate.

The vision process of reconstruction is exactly utilize to demarcate the corresponding 2D picture point coordinate that the model parameter that obtains and matching process obtain to find the solution the 3d space point, it is the inverse process of perspective imaging, be finding the solution of a kind of uncertain problem, reliability of separating and precision are influenced by several factors.After having demarcated the internal and external parameter of video camera, just set up the relation of having got well between image coordinate and the world coordinate system, after two video cameras are as the coordinate on the plane, just can obtain the world coordinates of this point in known a certain spatial point.Detailed process is as follows:

Two camera parameters and corresponding image coordinate about known can be rewritten as formula (12):

(a ₃₁u _li-a ₁₁)X _i+(a ₃₂u _li-a ₁₂)Y _i+(a ₃₃u _li-a ₁₃)Z _i＝a ₁₄-a ₃₄u _li

(a ₃₁v _li-a ₂₁)X _i+(a ₃₂v _li-a ₂₂)Y _i+(a ₃₃v _li-a ₂₃)Z _i＝a ₂₄-a ₃₄v _li (20)

(b ₃₁u _ri-b ₁₁)X _i+(b ₃₂u _ri-b ₁₂)Y _i+(b ₃₃u _ri-b ₁₃)Z _i＝b ₁₄-b ₃₄u _ri

(b ₃₁V _ri-b ₂₁)X _i+(b ₃₂v _ri-b ₂₂)Y _i+(b ₃₃v _ri-b ₂₃)Z _i＝b ₂₄-b ₃₄v _ri

Three unknown numbers, four equations just can be obtained the world coordinates of this point with least square method.

Claims (6)

1, based on the stereo vision detection system of adaptive sine streak projection, it is characterized in that comprising following parts:

Based on the adaptive sine streak grenade instrumentation of acoustooptic deflector, be used for joining with computer system, by the phase place and the frequency of sine streak that computer system control is throwed to several sinusoidal gratings of measured object surface projection;

Binocular sensor based on visual theory, be positioned at two ends based on the adaptive sine streak grenade instrumentation of acoustooptic deflector, by the relation of binocular sensor and the formation triangulation of testee surface, obtain the phase modulation image of measured surface, send into computing machine via image pick-up card;

Five dimension scanning supports, vision sensor is installed and based on the adaptive sine streak grenade instrumentation of acoustooptic deflector on it, the binocular vision sensor that is used to drive based on visual theory arrives different spatial and with different attitudes testee is taken measurement, and the motor for five dimension scanning supports provides driving power simultaneously;

Electric part, input end and computer system output terminal join altogether, output is connected to five dimension scanning supports, is used for the control command of computer system output is converted to drive control signal, moves and rotates along surface level and vertical plane along three directions of XYZ to drive five dimension scanning supports;

The image acquisition part, the phase image signal of the measured surface that will take based on the vision sensor of visual theory amplifies, sends into computer system after the A/D conversion process via image pick-up card filtering;

Computer system, as the control center of whole detection system, control five dimension scanning supports and move, export the phase place of the sine streak that control signal control throwed based on the adaptive sine streak grenade instrumentation of acoustooptic deflector and phase image data that frequency, control collection are obtained based on the binocular sensor of visual theory, the phase image data of being gathered are carried out the phase diagram that Flame Image Process obtains having object dimensional information thereby be used to output a control signal to electrical system.

2, stereo vision detection system based on the adaptive sine streak projection according to claim 1, it is characterized in that: described adaptive sine streak grenade instrumentation based on acoustooptic deflector is by acoustooptic deflector, modulated laser, drive signal circuit and optical system are formed, drive signal circuit produces two paths of signals, one the road drives acoustooptic deflector, another road is used to drive modulated laser, produce the grid that reversed of stress distributes in the acoustooptic deflector, when modulating lasering beam passes through acoustooptic deflector with certain angle, outgoing beam produces diffraction, and two lens focuss of bundle diffraction light in optical system produce the sinusoidal interference ripple.

3, the stereo vision detection system based on adaptive sine streak projection according to claim 1 is characterized in that: described image processing section comprises the image pre-service, separates phase, whole audience phase demodulation, the three-dimensional coupling of phase place and three-dimensional appearance calculation processes.

4, based on the stereo vision detection system of adaptive sine streak projection, it is characterized in that comprising following parts:

Based on the adaptive sine streak grenade instrumentation of acoustooptic deflector, be used for joining with computer system, by the phase place and the frequency of sine streak that computer system control is throwed to several sinusoidal gratings of measured object surface projection;

Based on the monocular sensor of visual theory, the monocular sensor is positioned at a side of adaptive sine streak grenade instrumentation, with the relation of testee surface formation triangulation, obtains the phase modulation image of measured surface, sends into computing machine via image pick-up card;

Five dimension scanning supports, the monocular sensor is installed and based on the adaptive sine streak grenade instrumentation of acoustooptic deflector on it, the monocular sensor that is used to drive based on visual theory arrives different spatial and with different attitudes testee is taken measurement, and the motor for five dimension scanning supports provides driving power simultaneously;

Electric part, input end and computer system output terminal join altogether, output is connected to five dimension scanning supports, is used for the control command of computer system output is converted to drive control signal, moves and rotates along surface level and vertical plane along three directions of XYZ to drive five dimension scanning supports;

The image acquisition part, the phase image signal of the measured surface that will take based on the monocular sensor of visual theory amplifies, sends into computer system after the A/D conversion process via image pick-up card filtering;

Computer system, as the control center of whole detection system, control phase place that five dimension scanning supports moved, exported the sine streak that control signal control throwed based on the adaptive sine streak grenade instrumentation of acoustooptic deflector and frequency, acquisition phase view data, the phase image data of being gathered are carried out phase place separate and handle the phase diagram that obtains having object dimensional information mutually thereby be used to output a control signal to electrical system.

5, stereo vision detection system based on the adaptive sine streak projection according to claim 4, it is characterized in that: described adaptive sine streak grenade instrumentation based on acoustooptic deflector is by acoustooptic deflector, modulated laser, drive signal circuit and optical system are formed, drive signal circuit produces two paths of signals, one the road drives acoustooptic deflector, another road is used to drive modulated laser, produce the grid that reversed of stress distributes in the acoustooptic deflector, when modulating lasering beam passes through acoustooptic deflector with certain angle, outgoing beam produces diffraction, and two lens focuss of bundle diffraction light in optical system produce the sinusoidal interference ripple.

6, the stereo vision detection system based on adaptive sine streak projection according to claim 4 is characterized in that: described Flame Image Process comprises the image pre-service, separates phase, whole audience phase demodulation and three-dimensional appearance calculation processes.

Images