CN105261047A - Docking ring circle center extraction method based on close-range short-arc image - Google Patents

Abstract

The invention relates to a docking ring circle center extraction method based on a close-range short-arc image. The docking ring circle center extraction method comprises the following steps: carrying out epipolar rectification on a binocular image containing a short-arc docking ring, dividing a threshold by using a standard deviation method, extracting a ring surface by means of regional attributes, and then extracting inner and outer edges; carrying out dense reconstruction on the extracted ring surface region by using a semi-global matching method to obtain three-dimensional point cloud in a camera coordinate system; calculating a ring surface normal vector of the point cloud, and projecting a coordinate in a plane vertical to the normal vector; distinguishing the inner and outer edges in a new plane, and respectively carrying out perfect circle fitting on the inner and outer edges; and finally, obtaining an accurate circle center coordinate of the docking ring by use of concentric ring constraint according to obtained centers and radiuses of the inner and outer edges. According to the docking ring circle center extraction method provided by the invention, from the characteristics of the docking ring, three-dimensional point cloud data are directly projected into a normal incidence spatial plan, ellipse fitting is simplified to circle fitting, and meanwhile, error points are eliminated by constraint conditions of inner and outer rings, and the docking ring circle center extraction method has the advantages of accurate result and small calculated amount.

Description

A kind of butt joint ring center of circle extracting method based on closely short arc segments image

Technical field

The present invention relates to the method that anchor ring extracts and the anchor ring center of circle is estimated, be specifically related to a kind of butt joint ring center of circle extracting method based on closely short arc segments image.

Background technology

In field of aerospace, butt joint ring is target common in spacecraft, only has to carry out pose accurately to butt joint ring target and estimate just can accurately dock, thus carries out the space tasks such as Material Transportation, aircraft repairing.Wherein the detection of anchor ring and the estimation in the anchor ring center of circle are steps crucial during pose is estimated.Recent decades, the pose of butt joint ring estimates it is a hot issue all the time, and scholars propose the method that many poses are estimated.

Current, the pose of spacecraft is estimated to carry out mainly for cooperative target, namely can mark monumented point in cooperative target, carry out to carry out pose estimation according to monumented point position in docking operation with aircraft.Aircraft object pose based on monumented point is estimated to need to carry out careful mark before vehicle launch, makes Flight Vehicle Design work become loaded down with trivial details.Method based on mark point is not suitable for the noncooperative target of the non-labels tokens point in space.In addition, this method only needs the binocular image obtaining target aircraft, and corresponding camera apparatus is portable, cheap.

Summary of the invention

Technology of the present invention is dealt with problems: overcome the deficiencies in the prior art, a kind of butt joint ring center of circle extracting method based on closely short arc segments image is provided, the method can overcome in classic method needs hand labeled monumented point on butt joint ring and the shortcoming consuming a large amount of manpower, avoided simultaneously space annulus project two dimensional surface after distortion of projection, do not take pains the fitting problems solving oval or quafric curve, but by measurement image that binocular camera obtains, on the projection plane that the three-dimensional information being projected recovering its anchor ring and edge is just being penetrated, simplify the positive round being projected as two dimensional surface of space annulus, simplify and calculate, only need just can accurately estimate anchor ring center according to the topography of annulus, under the prerequisite ensureing certain robustness, there is estimation result accurate, low advantage consuming time.

The technical solution used in the present invention is: as shown in Figure 1

(1) for the binocular image comprising short arc segments butt joint ring target closely obtained, carry out polar curve correction, then adopt standard deviation method to carry out binary-state threshold segmentation, obtain the bianry image after splitting.Utilize the area attribute in bianry image to carry out annulus region extraction, annulus region adopts its outer edge of canny operator extraction after extracting;

(2) utilize the matching process of half overall situation to carry out densification to the annulus region extracted to rebuild, obtain a large amount of point on anchor ring (be exactly all on anchor ring can the pixel of registration, determined by the imaging area size of anchor ring and registration rate) parallax value, thus the three-dimensional reconstruction carried out under camera coordinates system, obtain three-dimensional point cloud;

(3) adopt multi-point fitting to the three-dimensional point cloud under camera coordinates system, the method for largest optimization calculates anchor ring normal vector and projects in the plane vertical with normal vector by the three-dimensional point cloud represented under camera coordinates system;

(4) in the plane vertical with normal plane, distinguish inward flange and outward flange, adopt multi-point fitting, the method for result optimum carries out the matching of standard positive round respectively, obtains center and the radius of butt joint ring outer edge;

(5) according to the outer edge center obtained and radius, butt joint ring is adopted to be the constraint condition of concentric ring, judge that the centre distance of outer edge is minimum, the radius difference result suitable with butt joint ring width of outer edge is optimum, finally obtains from comprising the butt joint ring central coordinate of circle accurately calculated the binocular image of local butt joint ring target.

The concrete grammar that above-mentioned steps (1) adopts standard deviation method to carry out binary-state threshold segmentation is:

1) standard deviation image is defined as, the standard deviation value to each pixel of image I is calculated as follows:

$s_k\left(I\right)=\sqrt{|u_k\left(I^2\right)-u_k{\left(I\right)}^2|}---\left(1\right)$

Wherein, for any one pixel, u _k() is the mean value of value in surrounding (2k+1) × (2k+1) neighborhood centered by this point, and k represents window half-breadth;

2) carry out statistics with histogram to standard deviation image, adopt averaging method to calculate binary-state threshold ξ, each pixel of criterion offset images, what pixel value was more than or equal to ξ is designated as 255, and what be less than ξ is designated as 0, obtains the bianry image after splitting.

The concrete grammar that above-mentioned steps (1) utilizes the area attribute in bianry image to carry out annulus region extraction is:

1) according to the constraint condition of area setting, the region excessive or too small for area is rejected;

2) in the image after removing large regions or region in small, broken bits, the solid degree of annulus region is less than other regions, by calculating solid degree, judges that solid degree Minimum Area is the annulus region finally extracted.The computing formula of solid degree is as follows:

The matching process that in above-mentioned steps (2), utilization half is overall carries out the fine and close concrete grammar rebuild to the anchor ring extracted:

Half global registration algorithm along one dimension path r, the cost value L of pixel p to be matched under parallax d condition _r(p, d) recursive definition is:

$\begin{array}{c}{L}_r(p,d)=C(p,d)+\min \left(\begin{array}{c}{L}_r(p-r,d)=L_r(p-r,d-1)+P_1,L_r(p-r,d+1)+P_1,\\ \underset{i}{\min }{L}_r(p-r,i)+P_2\end{array}\right)\\ -\underset{k}{\min }{L}_r(p-r,k)\end{array}---\left(3\right)$

Wherein, min represents and gets minimum value, and pixel represents the Matching power flow of pixel p to be matched under parallax d to matched data item C (p, d), and Matching power flow is selected and calculated pixel intensity difference based on Birchfield and Tomasi insensitive method of sampling.Section 2 represents the minimum cost value of neighbor under current pixel p parallax is d condition in path, and p-r represents that p's under current path is upper.When the change of a pixel cell occurs the parallax of neighborhood pixels, cost value is P ₁, when parallax change is more than a pixel unit, cost value is P ₂, P ₁with P ₂for set-point.Last in above formula, the parallax d chosen with current pixel p has nothing to do, just to preventing L excessive, to meet L≤C _max+ P ₂result of calculation, C _maxrepresent the maximum occurrences of Matching power flow.I and k represents that the traversal of parallax is from d _minto d _max, d _minrepresent parallax minimum value, d _maxrepresent parallax maximal value.

Finally, the cost value of each pixel is all paths _rcost value sum S (p, d)=∑ _rl _r(p, d), the parallax of matching result corresponding to minimum cost value of every bit pixel, thus the three-dimensional coordinate completing each point in anchor ring calculates, i.e. fine and close reconstruction.

The three-dimensional reconstruction process of carrying out in above-mentioned steps (2) under camera coordinates system is: the three-dimensional coordinate calculating each point under camera coordinates system according to parallax result, wherein camera coordinates system is defined as: cross camera photocentre, parallel picture planar horizontal direction to the right be x _caxle, that parallel picture plane orthogonal direction is downward is y _caxle, z _caxle meets the right-hand rule.

Adopt multi-point fitting in above-mentioned steps (3), the method for largest optimization calculates anchor ring normal vector and is:

In the three dimensional point cloud of anchor ring, get arbitrarily three points, carry out plane fitting, the normal vector obtaining plane equation represents, then calculate the distance of other point to plane, normal vector when Distance geometry is minimum, as final output, is exactly the normal vector of anchor ring.

In above-mentioned steps (3), the three-dimensional point cloud represented under camera coordinates system is projected in the plane vertical with normal vector and is:

The normal vector of plane is (p _h, q _h, r _h) (p _h, q _h, r _hrepresent the direction number along current coordinate system XYZ tri-directions respectively) and (p _h ²+ q _h ²+ r _h ²)=1, the coordinate of three-dimensional point cloud is (x _i, y _i, z _i) i=0,1 ... n-1, n are the total number of three-dimensional point cloud.Then following coordinate transform is carried out to three-dimensional point cloud:

$\left[\begin{array}{ccc}\sqrt{1-p_h^2}& 0& p_h\\ -p_h{q}_h& p_h^2-r_h\sqrt{1-p_h^2}& q_h\sqrt{1-p_h^2}\\ p_h& q_h& r_h\end{array}\right]\left[\begin{array}{c}{x}_i\\ y_i\\ z_i\end{array}\right]---\left(4\right)$

Above-mentioned steps distinguishes inward flange and outward flange in (4) in the plane vertical with normal plane, and adopt multi-point fitting, the method for result optimum carries out the matching of standard positive round respectively, obtains center and the radius of butt joint ring outer edge, is specially:

The inward flange obtained according to canny operator and outward flange, adopt the corresponding relation of point to obtain the three-dimensional coordinate of inward flange and outward flange point, calculate the mean distance between inward flange and outward flange two groups of points; Then in inward flange point group or outward flange point group, random selecting three points carry out the matching of positive round center and radius respectively, judge the distance of other point and matching annulus in this group, what distance summation was minimum thinks optimum, exports respectively, obtains center and the radius of butt joint ring outer edge.

The beneficial effect that the present invention compared with prior art has is: instant invention overcomes in classic method and need hand labeled monumented point on butt joint ring and the shortcoming consuming a large amount of manpower, avoided simultaneously space annulus project two dimensional surface after distortion of projection, do not take pains the fitting problems solving oval or quafric curve, but by measurement image that binocular camera obtains, on the projection plane that the three-dimensional information being projected recovering its anchor ring and edge is just being penetrated, simplify the positive round being projected as two dimensional surface of space annulus, simplify and calculate, only need just can accurately estimate anchor ring center according to the topography of annulus, under the prerequisite ensureing certain robustness, there is estimation result accurate, low advantage consuming time.

Accompanying drawing explanation

Fig. 1 is the step schematic diagram of method of the present invention;

Fig. 2 (a)-(i) adopts the butt joint ring center of circle extracting method based on closely short arc segments image to extract the processing procedure in the butt joint ring center of circle for embodiment 1; Wherein (a) is the short arc segments butt joint ring target left images pair of input, b () be rear design sketch for polar curve corrects, c () is standard deviation image, the binaryzation design sketch of (d) standard deviation figure, design sketch e () utilizes the area attribute of solid degree to carry out annulus region extraction after, the design sketch of (f) butt joint ring edge extracting, g the disparity map obtained is mated in the densification of () butt joint ring region, h () annular section Three-dimensional Gravity is laid foundations the design sketch of cloud re-projection, (i) concentric circles matching constraint obtains the schematic three dimensional views in the final center of circle.

Embodiment

Below in conjunction with Fig. 1, Fig. 2 (a)-(i), the present invention is described in detail.

(1) input image to be detected: image to be detected is the binocular gray level image comprising short arc segments butt joint ring, comprise left figure and right figure, as shown in Fig. 2 (a).

(2) correction of image polar curve is carried out, as shown in Fig. 2 (b) to binocular image.Rotation matrix between the camera of known left and right is that (Eulerian angle that R is corresponding are A to R _lr), when translation matrix is T, the step that polar curve corrects is as follows:

1) in order to make two camera coordinates be parallel, first time rotation is carried out to two coordinate systems.Left camera rotates R _l=F (A _lr/ 2), right camera rotates R _r=F (-A _lr/ 2), F represents the transfer function between rotation matrix and Eulerian angle herein, postrotational translation vector T ₂=R _rt;

2) in order to ensure T ₂be not only 0 (i.e. base direction) in X-direction, second time carried out to Two coordinate system and rotates, make T ₂be parallel to vector n=[-100] ^t.

(3) binocular image binary conversion treatment:

1) standard deviation image is defined as, the standard deviation value to each pixel of image I is calculated as follows:

$s_k\left(I\right)=\sqrt{|u_k\left(I^2\right)-u_k{\left(I\right)}^2|}---\left(1\right)$

Wherein, for any one pixel, u _k() is the mean value of value in surrounding (2k+1) × (2k+1) neighborhood centered by this point, and k represents window half-breadth, determines window size, is taken as 2 here.|| represent signed magnitude arithmetic(al).Actual computation process adopts the method pulling speed of integrogram, calculated product component only needs to image traversal once, utilizes integrogram, the pixel value in any one rectangle and can be determined by four summits, calculate and have nothing to do with rectangular dimension, only need four sub-addition computings.Standard deviation figure is as shown in Fig. 2 (c).

2) carry out statistics with histogram to standard deviation image, the method for average calculates binary-state threshold ξ, each pixel of criterion offset images, and what pixel value was more than or equal to ξ is designated as 255, and what be less than ξ is designated as 0.Binary image is as shown in Fig. 2 (d).

(4) area attribute is utilized to extract annulus region:

1) owing to there is region in small, broken bits in binary picture non-circular region, region in small, broken bits can effectively be removed according to size;

2), in the image behind removing region in small, broken bits, the solid degree of annulus region is less than other regions, finally can extract annulus region by calculating solid degree:

The extraction effect of annulus region is as shown in (e) in Fig. 2.

(5) utilize canny operator to carry out the edge extracting of annulus region, obtain outer edge point position.Extraction effect is as shown in Fig. 2 (f).

(6) because the texture of anchor ring is less, conventional matching process easily lost efficacy, and carries out anchor ring densification rebuild by the method for half overall situation.

Half global registration algorithm along one dimension path r, the cost value L of pixel p to be matched under parallax d condition _r(p, d) recursive definition is:

$\begin{array}{c}{L}_r(p,d)=C(p,d)+\min \left(\begin{array}{c}{L}_r(p-r,d)=L_r(p-r,d-1)+P_1,L_r(p-r,d+1)+P_1,\\ \underset{i}{\min }{L}_r(p-r,i)+P_2\end{array}\right)\\ -\underset{k}{\min }{L}_r(p-r,k)\end{array}---\left(3\right)$

Wherein, min represents and gets minimum value, and pixel represents the Matching power flow of pixel p to be matched under parallax d to matched data item C (p, d), and Matching power flow is selected and calculated pixel intensity difference based on Birchfield and Tomasi insensitive method of sampling.Section 2 represents the minimum cost value of neighbor under current pixel p parallax is d condition in path, and p-r represents that p's under current path is upper.When the change of a pixel cell occurs the parallax of neighborhood pixels, cost value is P ₁, when parallax change is more than a pixel unit, cost value is P ₂, P ₁with P ₂for set-point.Last in above formula, the parallax d chosen with current pixel p has nothing to do, just to preventing L excessive, to meet L≤C _max+ P ₂result of calculation, C _maxrepresent the maximum occurrences of Matching power flow.I and k represents that the traversal of parallax is from d _minto d _max, d _minrepresent parallax minimum value, d _maxrepresent parallax maximal value.

Finally, the cost value of each pixel is all paths _rcost value sum S (p, d)=∑ _rl _r(p, d), the parallax of matching result corresponding to minimum cost value of every bit pixel, thus the three-dimensional coordinate completing each point in anchor ring calculates, i.e. fine and close reconstruction.

Fine and close disparity map effect of rebuilding as shown in Fig. 2 (g), for the anchor ring that left image extracts.

(7) plane of a loop equation is solved according to the anchor ring rebuild: any 3 the three-dimensional point X on the anchor ring obtained in Stochastic choice (6) step of the present invention _i, Y _i, Z _ii=1,2,3 parameters calculating one group of plane equation, plane equation is as follows:

aX+bY+cZ+d＝0(4)

Wherein, a, b, c, d are the expression coefficient of plane equation.General closed planar equation can not through coordinate axis center (anchor ring has certain vertical range from camera), and therefore above formula can be reduced to:

AX+BY+CZ＝-1(5)

By X _i, Y _i, Z _ii=1,2,3 bring above formula into, solve A, B, C.Then other three-dimensional point (x is calculated _j, y _j, z _j) to the distance of current plane all distances add up, and the minimum result of Distance geometry exports, and obtains A, B, C, is converted to planar process vector (p _h, q _h, r _h), export in order to improve counting yield and obtain the result restrained, through many experiments statistics, generally after 200 Stochastic choice calculating parameters, one group of result that ideal distance is minimum can be obtained.

(8) projected in the plane vertical with normal vector by three-dimensional point cloud according to the plane equation in step (7), specific practice is:

The normal vector of plane is (p _h, q _h, r _h) and (p _h ²+ q _h ²+ r _h ²)=1, the coordinate of three-dimensional point cloud is (x _i, y _i, z _i) i=0,1 ... n-1, n are the total number of three-dimensional point cloud.Then following coordinate transform is carried out to three-dimensional point cloud, obtain new three-dimensional coordinate (x' _i, y' _i, z' _i):

$\left[\begin{array}{c}{x}_i^{\′}\\ y_i^{\′}\\ z_i^{\′}\end{array}\right]\left[\begin{array}{ccc}\sqrt{1-p_h^2}& 0& p_h\\ -p_h{q}_h& p_h^2-r_h\sqrt{1-p_h^2}& q_h\sqrt{1-p_h^2}\\ p_h& q_h& r_h\end{array}\right]\left[\begin{array}{c}{x}_i\\ y_i\\ z_i\end{array}\right]---\left(7\right)$

In the three-dimensional point cloud effect of normal vector vertical plane inner projection as shown in Fig. 2 (h).

(9) carry out positive round matching to the coordinate after projection, specific practice is:

Equation of a circle is (x' _k-x ₀) ²+ (y' _k-y ₀) ²=R ₀ ², wherein (x' _k, y' _k) be the inner annular edge point chosen in the some cloud that obtains of (8) step, k=1,2,3 three points can solve, x ₀, y ₀and R ₀for amount to be asked.Least square method is utilized to try to achieve x ₀, y ₀and R ₀.Then by other inner annular edge point (x' _l, y' _l, z' _l) substitute into ∑ ((x' _l-x ₀) ²+ (y' _l-y ₀) ²-R ₀ ²), total error is minimum, exports x ₀, y ₀and R ₀.

In like manner, (x' is got _m, y' _m) be the outer shroud marginal point chosen in the some cloud that obtains of (8) step, m=1,2,3 three points can solve, x ₁, y ₁and R ₁for amount to be asked.Least square method is utilized to try to achieve x ₁, y ₁and R ₁.Then by other outer shroud marginal point (x' _n, y' _n, z' _n) substitute into ∑ ((x' _n-x ₁) ²+ (y' _n-y ₁) ²-R ₁ ²), total error is minimum, exports x ₁, y ₁and R ₁.

(10) carry out evaluation to last result to export

Two center of circle x that (9) step obtains ₀, y ₀, x ₁, y ₁and radius R ₀, R ₁, judge the difference of the Distance geometry radius of home position, be less than x and the y coordinate as the output center of circle of setting threshold value, then (x, y) is rotated the Spatial outlier value (x that the plane of getting back to before normal vector projection obtains the center of circle _center, y _center, z _center).Determine export butt joint ring center and anchor ring 3-D effect as shown in Fig. 2 (i).

Experimental example:

Treat (a) in detected image Fig. 2, process according to each step in the present invention successively, by result, visible the present invention can detect circle ring area exactly, and can effectively carry out the reconstruction of plane of a loop and the estimation in the center of circle.

Claims (8)

1., based on a butt joint ring center of circle extracting method for closely short arc segments image, it is characterized in that performing step is as follows:

(1) for the binocular image comprising short arc segments butt joint ring target closely obtained, carry out polar curve correction, then standard deviation method is adopted to carry out binary-state threshold segmentation, obtain the bianry image after splitting, utilize the area attribute in bianry image to carry out annulus region extraction, annulus region adopts its outer edge of canny operator extraction after extracting;

(2) utilize the matching process of half overall situation to carry out densification to the annulus region extracted to rebuild, obtain the parallax value of a large amount of point on anchor ring, thus carry out the three-dimensional reconstruction under camera coordinates system, obtain three-dimensional point cloud; Described a large amount of point be all on anchor ring can the pixel of registration, determined by the imaging area size of anchor ring and registration rate;

(3) adopt multi-point fitting to the three-dimensional point cloud under camera coordinates system, the method for largest optimization calculates anchor ring normal vector and projects in the plane vertical with normal vector by the three-dimensional point cloud represented under camera coordinates system;

(4) in the plane vertical with normal plane, distinguish inward flange and outward flange, adopt multi-point fitting, the method for result optimum carries out the matching of standard positive round respectively, obtains center and the radius of butt joint ring outer edge;

(5) according to the outer edge center obtained and radius, butt joint ring is adopted to be the constraint condition of concentric ring, judge that the centre distance of outer edge is minimum, the radius difference result suitable with butt joint ring width of outer edge is optimum, finally obtains from comprising the butt joint ring central coordinate of circle accurately calculated the binocular image of local short arc segments butt joint ring target.

2. a kind of butt joint ring center of circle extracting method based on closely short arc segments image according to claim 1, is characterized in that: the concrete grammar adopting standard deviation method to carry out binary-state threshold segmentation in described step (1) is:

(11) standard deviation image is defined as, the standard deviation value to each pixel of image I is calculated as follows:

$s_k\left(I\right)=\sqrt{|u_k\left(I^2\right)-u_k{\left(I\right)}^2|}---\left(1\right)$

Wherein, for any one pixel, u _k() is the mean value of value in surrounding (2k+1) × (2k+1) neighborhood centered by this point, and k represents window half-breadth;

(12) carry out statistics with histogram to standard deviation image, adopt averaging method to calculate binary-state threshold ξ, each pixel of criterion offset images, what pixel value was more than or equal to ξ is designated as 255, and what be less than ξ is designated as 0, obtains the bianry image after splitting.

3. a kind of butt joint ring center of circle extracting method based on closely short arc segments image according to claim 1, is characterized in that: the concrete grammar that described step (1) utilizes the area attribute in bianry image to carry out annulus region extraction is:

(21) according to the constraint condition of area setting, the region excessive or too small for area is rejected;

(22) in the image after removing large regions or region in small, broken bits, the solid degree of annulus region is less than other regions, by calculating solid degree, judges that solid degree Minimum Area is the annulus region finally extracted.The computing formula of solid degree is as follows:

4. a kind of butt joint ring center of circle extracting method based on closely short arc segments image according to claim 1, is characterized in that: described step (2) utilizes the matching process of half overall situation to carry out the fine and close concrete grammar rebuild to the anchor ring extracted to be:

Half global registration algorithm along one dimension path r, the cost value L of pixel p to be matched under parallax d condition _r(p, d) recursive definition is:

$\begin{array}{c}{L}_r(p,d)=C(p,d)+\min \left(\begin{array}{c}{L}_r(p-r,d)=L_r(p-r,d-1)+P_1,L_r(p-r,d+1)+P_1,\\ \underset{i}{\min }{L}_r(p-r,i)+P_2\end{array}\right)\\ -\underset{k}{\min }{L}_r(p-r,k)\end{array}---\left(3\right)$

Wherein, min represents and gets minimum value, pixel is to matched data item C (p, d) Matching power flow of pixel p to be matched under parallax d is represented, Matching power flow is selected and is calculated pixel intensity difference based on Birchfield and Tomasi insensitive method of sampling, and Section 2 represents the minimum cost value of neighbor under current pixel p parallax is d condition in path, and p-r represents that p's under current path is upper, when the change of a pixel cell occurs the parallax of neighborhood pixels, cost value is P ₁, when parallax change is more than a pixel unit, cost value is P ₂, P ₁with P ₂for set-point; Last in above formula, the parallax d chosen with current pixel p has nothing to do, just to preventing L excessive, to meet L≤C _max+ P ₂result of calculation, C _maxrepresent the maximum occurrences of Matching power flow, i and k represents that the traversal of parallax is from d _minto d _max, d _minrepresent parallax minimum value, d _maxrepresent parallax maximal value;

Finally, the cost value of each pixel is all paths _rcost value sum S (p, d)=∑ _rl _r(p, d), the parallax of matching result corresponding to minimum cost value of every bit pixel, thus the three-dimensional coordinate completing each point in anchor ring calculates, i.e. fine and close reconstruction.

5. a kind of butt joint ring center of circle extracting method based on closely short arc segments image according to claim 1, it is characterized in that: the three-dimensional reconstruction process of carrying out in described step (2) under camera coordinates system is: the three-dimensional coordinate calculating each point under camera coordinates system according to parallax result, wherein camera coordinates system is defined as: cross camera photocentre, parallel picture planar horizontal direction to the right be x _caxle, that parallel picture plane orthogonal direction is downward is y _caxle, z _caxle meets the right-hand rule.

6. a kind of butt joint ring center of circle extracting method based on closely short arc segments image according to claim 1, is characterized in that: adopt multi-point fitting in described step (3), and the method for largest optimization calculates anchor ring normal vector and is:

In the three dimensional point cloud of anchor ring, get arbitrarily three points, carry out plane fitting, the normal vector obtaining plane equation represents, then calculate the distance of other point to plane, normal vector when Distance geometry is minimum, as final output, is exactly the normal vector of anchor ring.

7. a kind of butt joint ring center of circle extracting method based on closely short arc segments image according to claim 1, is characterized in that: projected in the plane vertical with normal vector by the three-dimensional point cloud represented under camera coordinates system in described step (3) and be:

The normal vector of plane is (p _h, q _h, r _h) (p _h, q _h, r _hrepresent the direction number along current coordinate system XYZ tri-directions respectively) and (p _h ²+ q _h ²+ r _h ²)=1, the coordinate of three-dimensional point cloud is (x _i, y _i, z _i), i=0,1, " n-1, n are the total number of three-dimensional point cloud, then carry out following coordinate transform to three-dimensional point cloud:

$\left[\begin{array}{ccc}\sqrt{1-p_h^2}& 0& p_h\\ -p_h{q}_h& p_h^2-r_h\sqrt{1-p_h^2}& q_h\sqrt{1-p_h^2}\\ p_h& q_h& r_h\end{array}\right]\left[\begin{array}{c}{x}_i\\ y_i\\ z_i\end{array}\right]---\left(4\right)$

8. a kind of butt joint ring center of circle extracting method based on closely short arc segments image according to claim 1, it is characterized in that: described step distinguishes inward flange and outward flange in (4) in the plane vertical with normal plane, adopt multi-point fitting, the method of result optimum carries out the matching of standard positive round respectively, obtain center and the radius of butt joint ring outer edge, be specially:

The inward flange obtained according to canny operator and outward flange, adopt the corresponding relation of point to obtain the three-dimensional coordinate of inward flange and outward flange point, calculate the mean distance between inward flange and outward flange two groups of points; Then in inward flange point group or outward flange point group, random selecting three points carry out the matching of positive round center and radius respectively, judge the distance of other point and matching annulus in this group, what distance summation was minimum thinks optimum, exports respectively, obtains center and the radius of butt joint ring outer edge.