CN102012217B - Binocular vision-based three-dimensional geometric shape measurement method for large-shape object - Google Patents

Abstract

The method for measuring the three-dimensional geometric shape of the large-shape object based on binocular vision realizes the non-contact measurement of the three-dimensional shape of the on-rail object by using a computer vision method based on structured light measurement, utilizes a line laser controlled by a computer to irradiate the on-rail object, and simultaneously carries out differential shooting on laser images projected on the on-rail object by using a left binocular camera and a right binocular camera; performing three-dimensional matching calculation on the laser light bar image imaged by the binocular camera through a computer to obtain three-dimensional space coordinates of points on all on-orbit objects on the laser light bar imaged by the binocular camera; rotating the angle of the laser to enable the laser light bar to cover all the measured surfaces at intervals when the moving object can be scanned or measured, and repeating the operation to obtain a three-dimensional space model and parameters of the whole measured surface; by comparing the three-dimensional appearance of the in-orbit object with the vertical plane coordinate in the orbit, the computer can automatically make size judgment and overrun judgment, and the measurement of the three-dimensional appearance of the outer surface of the object can be realized.

Description

A three-dimensional geometric shape measurement method for large-shaped objects based on binocular vision

technical field

The invention relates to a measurement technology based on three-dimensional structured light, in particular to a fast non-contact detection method for overrunning objects on orbit, and belongs to the technical field of computer vision measurement.

Background technique

In recent years, the speed of domestic railways has been increased many times, especially the operation of high-speed EMUs, and the speed of trains is getting faster and faster, which also puts forward higher requirements for railway safety. When the train, especially the freight train, exceeds the space limit of normal travel, it is called overrun, which may cause serious safety accidents at this time. Therefore, it is extremely important to measure the overrun of on-orbit objects such as railway freight trains.

Structured light measurement is a non-contact three-dimensional shape measurement technology, which can be applied to the overrun detection of objects in orbit. The previous three-dimensional structured light measurement method used the method of monocular camera plus laser light strip extraction, but this method is not suitable for use in the case of strong natural light interference outdoors, and at the same time, the monocular measurement requires relatively high mechanical precision of the turntable. , It is also easy to produce large errors when measuring large-shaped objects. In order to be able to accurately measure the appearance size of large-shaped objects such as trains in harsh outdoor environments, binocular vision measurement, which requires less mechanical precision, has played a unique advantage.

invention problem

The purpose of the present invention is to provide a method for measuring the three-dimensional geometric shape of large-shaped objects based on binocular vision, especially a non-contact overrun measurement method for on-orbit objects, which is used to automatically measure the appearance dimensions of on-orbit objects.

The technical solution of the present invention is: a method for measuring the three-dimensional geometric shape of large-shaped objects based on binocular vision, that is, a binocular overrun measurement method for on-orbit objects, using a computer vision method based on structured light measurement to realize the measurement of the three-dimensional shape of on-orbit objects. The non-contact measurement of appearance, the specific process is as follows:

Utilize computer-controlled lasers to irradiate on-orbit objects, and use left and right binocular cameras to simultaneously perform differential shooting of laser images projected on on-orbit objects (which can be light bar images or light spot images generated by vertical line lasers); through the computer Perform three-dimensional matching calculation on the laser light strip image imaged by the binocular camera, and obtain the three-dimensional space coordinates of all points on the laser light strip imaged by the binocular camera;

Turn the angle of the laser so that the laser light strip can cover all the measured surface after an interval of time when scanning or measuring the moving object, and repeat the above operations to obtain the three-dimensional space model and parameters of the entire measured surface; by comparing the three-dimensional shape of the object in orbit The computer can automatically make size judgment and overrun judgment;

The computer controls the line laser to generate laser light on the measured surface of the object, and uses two cameras to collect an image of the measured surface with laser light bars; the computer controls the laser to turn off the irradiation, and then uses two cameras to collect an image without For the image of the measured surface of the laser light strip, the images obtained by each camera are respectively differentiated to obtain the image of the laser light strip.

The pixel coordinates obtained by matching are at the sub-pixel level, I1 and I2 are the image planes of the left and right cameras, C1 and C2 are the optical centers of the left and right cameras respectively; P is any point in space, it is on the image plane of the left camera The image point on the right camera plane is m1, and the image point on the right camera plane is m2; according to the geometric constraints of the epipolar line, the epipolar line is the connection between point P on the I1 and I2 plane and the intersection point of the I1 and I2 plane and C1 and C2 The connecting line of line intersections, m2 should be located on the epipolar line lm2 of m1 in the right image plane, and m1 should be located on the epipolar line lm1 of m2 in the left image plane, and they are mutually corresponding points; by C1, C2 , the plane II determined by P is the epipolar plane, the epipolar line lm2 of m1 is the intersection line between the epipolar plane II and the right image plane, the epipolar line lm1 of m2 is the intersection line between the epipolar plane II and the left image plane, and the exopolar line The poles e1 and e2 are the intersections between the line connecting C1 and C2 and the left and right image planes. The outer epipolar line of any point in the left image plane in the right image plane must pass through the outer pole e2, and the points on the left image plane still conform to this relationship; that is, take any point m1 on the laser light strip of the left image plane, which corresponds to a point P covered by the laser line on the object, and according to the constraints of the epipolar line, it can be known that the projection of point P in the right image must appear on the epipolar line Lm2 , and the equation of line Lm2 can be obtained, so that the search for two-dimensional matching points can be narrowed to one-dimensional range; at the same time, since the projection of point P on the right image plane must also be on the laser light strip, the right image is taken The two pixel points D1 and Dr closest to the straight line Lm2 on the laser light strip in the plane are calculated for the straight line equation L where these two pixels are located, then the intersection point m2 of the straight line L and Lm2 is the projection of point P on Figure I2, and The coordinates of m2 have sub-pixel precision.

Eliminate the influence of ambient stray light through two cameras with optical filters, suitable for measurement in harsh environments.

Using the binocular measurement method, it is possible to accurately measure large-scale objects such as trains and ships without requiring high precision in machining and installation.

The beneficial effect of the present invention is: the present invention adopts the three-dimensional detection technology of structured light, and uses the differential method and sub-pixel coordinate matching, and can perform fast, non-contact and high-precision overrunning of objects in orbit in harsh outdoor environments The detection overcomes the defects of the existing over-limit measurement, and greatly improves the measurement efficiency and measurement accuracy.

Description of drawings

Figure 1 is a monocular structured light measurement model. In the figure, ① is the beam, ② is the computer, ③ is the camera, ⑥ is the line-structured laser, ⑤ is the turntable, and the angles α and β are respectively the angles between the optical axis of the camera and the laser line and the center line of the beam. The clamps at both ends of the beam can fix the center of the turntable, the focus of the camera and the center of the beam on a straight line.

Figure 2 is a binocular structured light measurement model. In the figure, ① is the beam, ② is the computer, ③ is the camera 1, ④ is the camera 2, ⑤ is the turntable, ⑥ is the line structure laser. Fixtures at both ends and the center of the beam fix the center of the turntable, the focus of the two cameras and the center of the beam on a straight line.

Fig. 3 is a schematic diagram of epipolar line matching. As shown in the figure, the projection points m1 and m2 of a point P on the two camera image planes in space must exist on the plane formed by the optical centers of the two cameras and the point P.

Figure 4 and Figure 5 are schematic diagrams of binocular matching, Figure 4 is the left picture, Figure 5 is the right picture, m1 is a point on the laser light bar in the left picture, lm2 is the outer polar line corresponding to the point m1 on the right picture, and the outer pole The intersection point m2 of the line and the laser line is the matching point of m1 on the right image.

Figure 6 is a schematic diagram of coordinate matching at the sub-pixel level (right figure), Dl and Dr are respectively the two nearest pixel points on the laser light strip from the outer epipolar line, L is the connecting line between the centers of Dl and Dr, and point m2 is L and Dr The intersection point between the epipolar lines lm2 is the matching point to be found.

Detailed ways

The previous monocular structured light measurement device is shown in Figure 1, and the position of point c needs to be accurately known by using triangulation to measure the distance Lab and the angles α, β. However, due to the limitations of mechanical installation and control accuracy, the motor rotation angle β is not easy to control, which affects the measurement accuracy. In this case, binocular measurement shows unique advantages. In binocular measurement, the position and attitude of the camera are relatively fixed, so the distance Lab can be regarded as a constant value, and the angles α and β can be accurately obtained by image processing algorithms, so the appearance size of large-shaped objects can be accurately measured. The measurement device consists of a base, a high-power line-structure laser, two cameras equipped with optical filters, a driving circuit and a computer, as shown in Figure 2. The camera collects image data, and the computer calculates the appearance size of the object according to the three-dimensional space matching relationship.

The left and right binocular matching in three-dimensional space is shown in Figure 3. I1 and I2 are the image planes of the two cameras, and C1 and C2 are the optical centers of the left and right cameras, respectively. P is any point in space, its image point on the left camera image plane is m1, and its image point on the right camera plane is m2. According to the geometric constraints of the epipolar line (point P on the I1 and I2 planes and the intersection point of the I1 and I2 planes and the intersection point of the line between C1 and C2) geometric constraints, m2 should be located at the epipolar line lm2 of m1 in the right image plane above, m1 should also be located on the epipolar line lm1 of m2 in the left image plane, and they are corresponding points to each other. The plane II determined by C1, C2, P is the epipolar surface, the epipolar line lm2 of m1 is the intersection line between the epipolar surface II and the right image plane, and the epipolar line lm1 of m2 is the intersection line between the epipolar surface II and the left image plane The intersection line, the outer poles e1 and e2 are the intersections of the line connecting C1 and C2 with the left and right image planes (also the intersections with the outer poles lm1 and lm2). According to the perspective relationship of binocular imaging, it can be seen that the outer epipolar line of any point in the left image plane in the right image plane must pass through the outer pole e2, and the points on the left image still conform to this relationship. The specific method of image matching is as follows:

Define the homogeneous coordinates of the space point P as [x, y, z, 1]T, and the homogeneous pixel coordinates of the corresponding point m on the image plane as [u, v, 1], we have

$\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; \&CenterDot;</span>\left[\begin{array}{c}\mathrm{<span\; class="notranslate">\; u</span>}\\ \mathrm{<span\; class="notranslate">\; v</span>}\\ \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&Center\; Dot;</span>\left[\begin{array}{c}\mathrm{<span\; class="notranslate">\; x</span>}\\ \mathrm{<span\; class="notranslate">\; the\; y</span>}\\ \mathrm{<span\; class="notranslate">\; z</span>}\\ \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

Here, s is the scale factor, M is the perspective matrix, M=K·[R T], where K is the internal reference matrix of the camera, R and T are the rotation matrix and translation vector respectively.

Without loss of generality, the camera coordinate system of cameras 1 and 2 is the world coordinate system, R and T are the space transformation rotation matrix and translation vector between camera 1 and camera 2 respectively, s ₁ and s ₂ are scale factors, m ₁ , m ₂ is the perspective matrix, k ₁ , k ₂ are the internal reference matrix of the camera, then we have

${\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span>{\mathrm{<span\; class="notranslate">\; m</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; K</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; \&CenterDot;</span><span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; 10</span>}<span\; class="notranslate">\; ]</span>\left[\begin{array}{c}\mathrm{<span\; class="notranslate">\; x</span>}\\ \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$

${\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span>{\mathrm{<span\; class="notranslate">\; m</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; K</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span><span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; RT</span>}<span\; class="notranslate">\; ]</span>\left[\begin{array}{c}\mathrm{<span\; class="notranslate">\; x</span>}\\ \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>$

Eliminate s1 and s2 from the above formula to get

${\mathrm{<span\; class="notranslate">\; m</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}{\mathrm{<span\; class="notranslate">\; K</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}{<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; ]</span>}_{\mathrm{<span\; class="notranslate">\; x</span>}}{\mathrm{<span\; class="notranslate">\; RK</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; m</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; )</span>$

In the formula, [T] _X is an anti-symmetric matrix, which is determined by the translation vector T (T _x T _y T _z is the translation component)

${<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; ]</span>}_{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; =</span>\left[\begin{array}{ccc}\mathrm{<span\; class="notranslate">\; 0</span>}& {<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; the\; y</span>}}& {\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; z</span>}}\\ {\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; the\; y</span>}}& \mathrm{<span\; class="notranslate">\; 0</span>}& {<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}}\\ {<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; z</span>}}& {\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}}& \mathrm{<span\; class="notranslate">\; 0</span>}\end{array}\right]<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 5</span>}<span\; class="notranslate">\; )</span>$

make

$\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; K</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; T</span>}}{<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; ]</span>}_{\mathrm{<span\; class="notranslate">\; x</span>}}{\mathrm{<span\; class="notranslate">\; RK</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 6</span>}<span\; class="notranslate">\; )</span>$

E=[T] _X R (7)

then there is

${\mathrm{<span\; class="notranslate">\; m</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; T</span>}}{\mathrm{<span\; class="notranslate">\; K</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; T</span>}}{\mathrm{<span\; class="notranslate">\; EK</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}}{\mathrm{<span\; class="notranslate">\; m</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 8</span>}<span\; class="notranslate">\; )</span>$

${\mathrm{<span\; class="notranslate">\; m</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; T</span>}}{\mathrm{<span\; class="notranslate">\; Fm</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 9</span>}<span\; class="notranslate">\; )</span>$

Equation (9) is the description of the geometric relationship of the outer epipolar line in binocular vision, where I _m2 = Fm ₁ is the epipolar line corresponding to m ₁ , I _m1 = FT _m2 is the epipolar line corresponding to m ₂ , and the matrix F is the basic matrix ( Fundamental matrix), which is related to the internal parameters of the camera and the corresponding poses between the cameras. According to the basic matrix F obtained through calibration, the coordinates of corresponding points in the two image planes can be matched.

After obtaining the matched pixel points, calculation is performed according to the internal and external parameters of the camera, and the three-dimensional coordinates of the object contour corresponding to the light bar can be obtained. By scanning the laser, it is possible to measure the three-dimensional topography of the outer surface of objects in orbit. Then compare it with the space coordinates of the vertical plane in the orbit, and the computer can automatically make an overrun judgment.

In the experiment, we use a line laser with a central line width of 635nm and a filter, the baseline length is 1 meter, and the distance between the measured point and the epipolar line is 10 meters. After calculation, it can be known that the angles α and β are both 87.14 degrees, and the F matrix value is

F＝-0.000000485008366 -0.000038768789204 0.007350632235932

-0.000045784962796 -0.000001147668094 -0.717662770440072

0.017100613104312 0.778959169714406 -32.324287848102870

The measurement process is as follows:

1) Irradiate the line-structured laser onto the surface of the object to be measured, and use two cameras to obtain an image containing the laser light strip;

2) Turn off the laser, and then use two cameras to obtain an image without laser light bars;

3) Differential processing is performed on the two images obtained by each camera, which can improve the signal-to-noise ratio of the image. At the same time, due to the function of the optical filter, the interference of stray light in the external environment can be further eliminated, and the image quality can be improved.

4) Using a computer to perform matching operations on two laser light strip images, the three-dimensional coordinates of the object contour corresponding to this laser light strip can be obtained

5) Rotate the angle of the laser, repeat steps 1) to 4), and the three-dimensional coordinates of the object contour corresponding to another line of laser light can be obtained. The laser is scanned continuously so that the laser light strip can cover the measured surface of the whole object as much as possible. Then the lines of these measured surfaces are spliced to obtain the three-dimensional appearance size of the object. Comparing the coordinates of the measured surface with the vertical plane of the orbit, a judgment can be made on whether the object in orbit exceeds the limit.

In step 4) in the three-dimensional image matching operation, this method uses the method based on the intersection of epipolar line and line structured light to obtain the intersection coordinates of sub-pixel level, the specific method is as follows:

For a point m1 on the laser light strip in Figure 4, the corresponding epipolar line equation lm2 in Figure 5 can be obtained according to the epipolar line constraints, as shown in Figures 4 and 5, so that the two-dimensional matching point can be selected narrowed down to one dimension. At the same time, since the point m1 is on the laser light strip, the matching point on the right figure should also be on the laser light strip, so that the coordinates of the matching point can be obtained according to the outer epipolar line constraint and the structured light constraint. In Fig. 6, the laser light bar pixels Dl and Dr closest to the outer polar line are found, and the intersection point m2 of the straight line connecting Dl and Dr with the outer polar line is the feature point corresponding to the feature point m1 in the right figure, and the point The coordinates of m2 have sub-pixel precision.

Claims (1)

1. based on the big pattern object dimensional geometric shape measuring method of binocular vision; Promptly in the binocular of the rail object measuring method that transfinites; It is characterized in that using computer vision methods to realize that detailed process is following to the non-contact measurement at rail object dimensional pattern based on structural light measurement:

Utilize computer-controlled laser line generator to be radiated on the rail object, the laser image that is projected on the rail object is carried out the difference shooting simultaneously with left and right sides binocular camera; Through computing machine the laser optical strip image of binocular camera imaging is carried out three-dimensional coupling and calculate, obtain the three dimensional space coordinate of all points on the rail object on this laser striation that binocular camera forms images;

Rotational line laser instrument angle can scan the laser striation or covers all tested surfaces when measuring moving object through an interval time, and repeats aforesaid operations, thereby obtains the three-dimensional space model and the parameter of whole tested surface; At rail object dimensional pattern and track middle vertical plane coordinate, computing machine can be made size automatically and judge and the judgement of transfiniting through contrast;

The computer control laser line generator produces laser radiation on the object tested surface, gathers an image that has the tested surface of laser striation respectively with two cameras; The computer control laser line generator is closed irradiation, gathers an image that does not have the tested surface of laser striation respectively with two cameras again, and the picture that each video camera obtains is done difference respectively, obtains the laser optical strip image

The pixel coordinate of coupling gained is a sub-pixel, I1 and I2 be about the plane of delineation of two binocular cameras, C1 and C2 be respectively about the photocentre of two binocular cameras; P be in the space more arbitrarily, its picture point on left camera review plane is m1, the picture point on right camera plane is m2; According to outer polar curve geometrical constraint; Outer polar curve lm1, lm2 are respectively the lines on the I1 and the I2 plane of delineation; Be the P point with about two binocular camera photocentre C1 and C2 line respectively with the line of the intersection point of line on I1 and I2 plane of the intersection point on said I1 and I2 plane and two photocentre C1 and C2; M2 should be positioned at m1 on the outer polar curve lm2 of the right plane of delineation, and same m1 should be positioned at m2 on the outer polar curve lm1 of the left plane of delineation, and m1, m2 be corresponding point each other each other; By C1; C2, the Planar Mechanism that P confirms is outer pole-face, the intersection that the outer polar curve lm2 of m1 is the outer pole-face II and the right plane of delineation; The intersection that the outer polar curve lm1 of m2 is the outer pole-face II and the left plane of delineation, outer limit e1, e2 are the line of C1 and C2 and the intersection point of the left and right sides plane of delineation; Limit e2 outside the outer polar curve of any point in the right plane of delineation all will pass through in the plane of delineation of a left side; Point on the right plane of delineation still meets this relation; Promptly take up an official post and get 1 m1 at the laser striation of the left plane of delineation, 1 P that laser rays covered on its corresponding object, according to outer polar curve constraint can know the projection of a P in right image necessarily appear at outside on the polar curve lm2; And can try to achieve the equation of outer polar curve lm2, so just can the searching of two-dimentional match point be narrowed down in the one dimension scope; Simultaneously owing to some P also one fixes on the laser light bar in the projection on the right plane of delineation; So get in the right plane of delineation on the laser striation apart from straight line lm2 nearest two pixel Dl, Dr; Obtain the straight-line equation L at these two pixel places; The intersection point m2 of straight line L and lm2 is exactly the projection of a P on the I2 plane of delineation so, and the coordinate of m2 has sub-pixel precision.

Images