Low-curvature closed contour image registration error evaluation method based on circular domain error area ratio

Abstract

The invention discloses a low-curvature closed contour image registration error evaluation method based on a circular domain error area ratio, and relates to the technical field of image registration error evaluation. Performing error evaluation by using a relative spatial position relation between a two-dimensional image edge contour curve and a three-dimensional model projection edge contour curve at the same geometrical shape position after image registration; the technical key points are as follows: an error evaluation circular domain division method capable of customizing an evaluation scale according to an actual application condition is provided; dividing an error evaluation circular domain by taking a two-dimensional image edge contour curve as a reference curve, judging whether the error evaluation circular domain can be enveloped to a three-dimensional model projection edge contour curve or not by utilizing geometrical characteristics of the error evaluation circular domain, and further performing initial judgment on the overall distance deviation of the contour curve in the region; the method comprises the following steps of: introducing a concept of a circular domain error area ratio by utilizing the characteristic that an area enclosed between local parts of two profile curves after image registration can reflect the registration condition of the region, judging whether the circular domain error area ratio meets the requirement or not, and further judging whether the overall distance deviation of the two profile curves in the region meets the requirement or not; according to the method, the relative spatial position relation between the two-dimensional image edge contour curve and the three-dimensional model projection edge contour curve at the same geometrical shape position after image registration is taken as a judgment condition, so that the error evaluation efficiency after image registration can be improved.

Description

Technical Field

[0001] The patent of this invention relates to a low-curvature closed contour image registration error evaluation method based on the circular domain error area ratio, which belongs to the technical field of image registration error evaluation. Background Art

[0002] Image registration technology is widely used in fields such as medical imaging, computer vision, and augmented reality. The accuracy and efficiency of image registration have become crucial. Current image registration error evaluation methods mainly focus on the image registration between two-dimensional images and the image registration between three-dimensional models. In addition, traditional image registration error evaluation methods tend to ignore the characteristics and structural information of the image itself. Since the registration of two-dimensional images and three-dimensional models involves different dimensions, there is no suitable image registration error evaluation method. This makes it impossible to guarantee the registration effect of two-dimensional images and three-dimensional models, which hinders the application of two-dimensional image and three-dimensional model registration in medical imaging, computer vision, and augmented reality.

[0003] In order to evaluate the registration error between a two-dimensional image and a three-dimensional model image, the problem of dimensional unification needs to be addressed. A common method is to "upgrade" the two-dimensional image, but this method can only be achieved when the camera can capture multiple two-dimensional images at different angles and directions, which is often not possible in reality. Therefore, in order to evaluate the registration error between a two-dimensional image and a three-dimensional model image, it is necessary to project the three-dimensional model perpendicular to the two-dimensional image plane, obtain the edge contour of the three-dimensional model projection, achieve "dimensionality reduction", and then use the edge contour points of the two-dimensional image and the edge contour points of the three-dimensional model projection for error evaluation. Current error evaluation methods mostly focus on the statistical characteristics of the overall contour. Local errors of low-curvature closed contours are often masked due to the gentle curvature, resulting in significant deviations between the evaluation results and the actual registration quality. In addition, current bounding domain error evaluation methods are prone to unreasonable error judgment area division and discontinuous judgment areas when dealing with closed contours with gentle curvature, resulting in errors in the calculation of the error area. In summary, the current field of image registration error evaluation technology urgently needs a method that can accurately and quantitatively evaluate the image registration effect of low-curvature closed contours. Summary of the Invention

[0004] In response to the above problems, the present invention proposes a low-curvature closed contour image registration error evaluation method based on the circular domain error area ratio, which solves the problem in the current field of image registration error evaluation technology that there is a lack of image registration error evaluation methods that can intuitively reflect the local area registration status after low-curvature closed contour image registration, and realizes the quantitative description of the registration effect of two-dimensional images and three-dimensional model images.

[0005] A method for evaluating registration errors of low-curvature closed contour images based on circle error area ratio is characterized in that the specific implementation process of the method is as follows:

[0006] Step 1: Import the registered 2D image contour point set and 3D model projection contour point set:

[0007] The lower left corner of the two-dimensional image is used as the coordinate origin, the right is the positive direction of the x-axis, the upward is the positive direction of the y-axis, and according to the right-hand rule, the vertical screen outward is the positive direction of the z-axis to establish a three-dimensional image registration error evaluation coordinate system w, obtain the contour curve information of the registered two-dimensional image, and input the contour point set of the registered two-dimensional image contour curve is the position information of the i-th contour point of the registered 2D image contour curve in the o-xy plane of the 3D image registration error evaluation coordinate system w. The value range of i is 1≤i≤n, where, is the x-axis coordinate of the i-th contour point of the registered 2D image contour curve located in the o-xy plane of the 3D image registration error evaluation coordinate system w, The y-axis coordinate of the i-th contour point of the registered two-dimensional image contour curve is located in the o-xy plane of the three-dimensional image registration error evaluation coordinate system w; the registered three-dimensional model is parallel projected onto the o-xy plane along the negative direction of the z-axis of the three-dimensional image registration error evaluation coordinate system w to obtain the projection contour curve information, and the contour point set of the registered three-dimensional model projection contour curve is input. is the position information of the jth contour point of the projection contour curve of the registered 3D model in the o-xy plane of the 3D image registration error evaluation coordinate system w. The value range of j is 1≤j≤m, where The jth contour point of the projection contour curve of the registered 3D model is located in the x-axis coordinate of the o-xy plane of the 3D image registration error evaluation coordinate system w. The jth contour point of the projection contour curve of the registered 3D model is located on the y-axis coordinate of the o-xy plane of the 3D image registration error evaluation coordinate system w;

[0008] Step 2: Error assessment circle division:

[0009] The i-th contour point and the i+1-th contour point of the registered two-dimensional image contour curve are connected by a straight line segment. Connect, calculate straight line segments The length of a i When i=n, the nth contour point of the registered 2D image contour curve is connected to the first contour point by a straight line segment. Connect, calculate straight line segments The length of a n; Define the error assessment circle division preset radius r, error assessment circle division judgment coefficient ε, error assessment circle division starting coefficient k, error assessment circle division span coefficient t, error assessment circle division compensation coefficient λ, error assessment circle division compensation span coefficient j, error assessment circle division retrieval coefficient α, error assessment circle division radius judgment distance L α , The actual radius r of the error assessment circle α , with r α As the radius, point The error evaluation circle area divided by the center of the circle is the error evaluation circle area C α , where the error assessment circle domain division preset radius r and the error assessment circle domain division determination coefficient ε are constant values; the error assessment circle domain division process is specifically as follows:

[0010] a) Set k = 1, λ = 0, j = 0, α = 1, and jump to step 2b);

[0011] b) Set t = 0 and jump to step 2 c);

[0012] c) Calculate the error assessment circle division radius judgment distance Jump to step 2d);

[0013] d) Determine 2r-L α <ε is established, specifically:

[0014] If 2r-L α <ε holds true, jump to step 2f);

[0015] If 2r-L α <ε does not hold, set t = t + 1, and jump to step 2 e);

[0016] e) Determine whether k+t≤n holds, specifically:

[0017] If k+t≤n holds, jump to step 2c);

[0018] If k + t ≤ n does not hold, set λ = 1, j = j + 1, t = t - 1, and jump to step 2 c);

[0019] f) Determine 2r-L α <0 is true, specifically:

[0020] If 2r-L α <0 holds, let r α =L α / 2, jump to step 2g);

[0021] If 2r-L α <0 does not hold, let r α= r, jump to step 2g);

[0022] g) Determine whether k+t=n is true, specifically:

[0023] If k+t=n holds, jump to step 2h);

[0024] If k+t=n does not hold, jump to step 2i);

[0025] h) Point is the center of the circle, r α Establish error evaluation circle C for radius α , the error assessment circle area division is completed, jump to step 3;

[0026] i) Point is the center of the circle, r α Establish error evaluation circle C for radius α , let k = k + t + 1, jump to step 2 j);

[0027] j) Determine whether k≤n is true, specifically:

[0028] If k≤n holds, set α=α+1 and jump to step 2b);

[0029] If k≤n does not hold, the error assessment circle division is completed and jump to step 3;

[0030] Step 3: Calculate the circular error area ratio

[0031] All error assessment circles are regarded as a collection C = {C1, C2, ..., C β ,…,C α}, C β For point is the center of the circle, r β The error evaluation circle domain is established with the radius, and the value range of β is 1≤β≤α; the contour point set P of the projection contour curve of the 3D model after registration is calculated by using the K nearest neighbor method. Th Find the point The nearest point Specifically: For the query point, set the parameter K value to 1. The value of parameter K represents the number of neighboring points to be found near the query point. Setting the parameter K value to 1 means that the contour point set P of the projection contour curve of the 3D model after registration is Th Finding and querying points The closest point to the Euclidean distance Euclidean distance to the nearest point That is the point The closest point, the Euclidean distance between the two points is d β ; Define the error evaluation circle Cβ The two intersection points of the contour curve of the registered two-dimensional image are close to the error evaluation circle C β-1 The intersection point on one side is point Approach error assessment circle C β+1 The intersection point on one side is point When d β <r β When the error evaluation circle C is defined β The two intersection points of the projection contour curve of the registered 3D model and the point The closest intersection point is point with dot The closest intersection point is point When d β =r β When the error evaluation circle C is defined β An intersection point with the projection contour curve of the registered 3D model is When d β <r β When the calculation is done from point Arrive The contour curve segment of the two-dimensional image after registration, point Arrive The projection contour curve segment of the 3D model after registration, as well as The area of the enclosed region is S β ; When d β =r β When the calculation is done from point Arrive The contour curve segment of the two-dimensional image after registration, as well as The area of the enclosed region is S β ; Define the area S of the closed region β and error evaluation circle C β The area ratio of the circular error is the area ratio of the circular error, and the symbol is ρ β express, ρ β The upper limit value is specified as ρ max ; Ratio of the circle error area calculated each time to ρ β Stored in the circle domain error area ratio data set ρ, ρ={ρ1,ρ2,…,ρ β ,…,ρ α};

[0032] a) Judge d β ≤r β Whether it is established, specifically:

[0033] If d β ≤r βIf established, jump to step 3b);

[0034] If d β ≤r β If it does not hold, then the two curves after the output image registration are in the error evaluation circle C. β There are large differences in the image, the evaluation is unqualified, and the image registration error evaluation is completed;

[0035] b) Judge d β <r β Whether it is established, specifically:

[0036] If d β <r β If established, jump to step 3c);

[0037] If d β <r β If not, jump to step 3d);

[0038] c) Calculate the points Arrive The contour curve segment of the two-dimensional image after registration, point Arrive The projection contour curve segment of the 3D model after registration, as well as The area of the enclosed region is S β , jump to step 3e);

[0039] d) Calculate the value of the point Arrive The contour curve segment of the two-dimensional image after registration, as well as The area of the enclosed region is S β , jump to step 3e);

[0040] e) Calculate the circular error area ratio Determine ρ β ≤ρ max Whether it is established, specifically:

[0041] If ρ β ≤ρ max If the circle error area ratio ρ β Store it in the circle error area ratio data set ρ and jump to step f);

[0042] If ρ β ≤ρ max If it is not true, the output is in the error assessment circle C β The inner circle error area ratio is unqualified, and the output is ρ β The image registration error evaluation is completed;

[0043] f) Whether all circular regions have been judged:

[0044] Determine whether β+1>α is true, specifically:

[0045] If β+1>α holds, jump to step 3g);

[0046] If β+1>α does not hold, set β=β+1 and jump to step 3a);

[0047] g) Obtain the maximum value max{ρ1,ρ2,…,ρ β ,…,ρ α}, and the β value corresponding to the maximum value, the output image registration is qualified, and the area with the largest registration error is the error evaluation circle C β , the circle error area ratio is max{ρ1,ρ2,…,ρ β ,…,ρ α}, the image registration error evaluation is completed.

[0048] The beneficial effects of the present invention are:

[0049] 1. When evaluating image registration errors, the present invention achieves an overall evaluation of the characteristic contour points in the area by dividing the error evaluation circle, solving the problem that the number of collected edge contour points of the two-dimensional image and the edge contour points of the three-dimensional model projection is different, which may lead to the inability to find matching characteristic contour points; and the method can also control the scale of the evaluation by adjusting the size of the divided error evaluation circle according to actual usage.

[0050] 2. Based on the characteristics that the registration error of low-curvature closed contour images is mainly reflected in the overall overlap and symmetry deviation, the present invention proposes the concept of circular domain error area ratio and sets an upper limit value of the circular domain error area ratio. By judging the overall distance deviation between the edge contour curve of the two-dimensional image and the edge contour curve of the three-dimensional model projection after image registration within the error evaluation circular domain, the registration image error is evaluated.

[0051] 3. Compared with the invention patent "A method for evaluating the registration error of high-curvature contour images based on an enclosing domain threshold" applied by the inventor on the same day, although both methods are used for the evaluation of the registration error of two-dimensional images and three-dimensional model images, the method mentioned in "A method for evaluating the registration error of high-curvature contour images based on an enclosing domain threshold" is based on the premise that the registered image contour has high curvature. By calculating the neighboring distances of high-curvature contour points, the proximity of the projected contour, and the area of the error evaluation enclosing domain, it focuses on the error evaluation of the local curvature mutation area; the premise focused on by this method is that the registered image contour is closed and has low curvature, and the closed area is flexibly covered by the center and radius parameters. The error evaluation circle can be dynamically divided according to the evaluation scale, and the overall distance deviation of the registered image within the error evaluation circle is quantified by calculating the circle error area ratio, thereby realizing the evaluation of the registered image error; therefore, the proposed method compensates for the other method, thereby improving the series of methods for evaluating the registration error of two-dimensional images and three-dimensional model images.

[0052] 4. Compared with the invention patent "A method for evaluating the registration error of two-dimensional images and three-dimensional model images based on contour curve shape features" applied by the inventor on the same day, although both methods are used for evaluating the registration error of two-dimensional images and three-dimensional model images, the method mentioned in "A method for evaluating the registration error of two-dimensional images and three-dimensional model images based on contour curve shape features" comprehensively considers the reliability of curvature on neighboring point matching and the overall shape characteristics of the contour curve by calculating the maximum mismatch and center offset; the premise emphasized by this method is that the contour of the registered image is closed and has low curvature, and the error evaluation circle domain can be dynamically divided according to the evaluation scale. By calculating the circle domain error area ratio, the overall overlap and symmetry deviation of the registered image are quantitatively evaluated; therefore, the proposal of this method and the other method compensate each other, thereby improving the series of methods for evaluating the registration error of two-dimensional images and three-dimensional model images. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] For ease of explanation, the present invention is described in detail with reference to the following specific implementations and accompanying drawings.

[0054] Figure 1 This is a flow chart of a method for evaluating the registration error of low-curvature closed contour images based on the circle error area ratio.

[0055] Figure 2 This is a flowchart of error assessment circle division for a low-curvature closed contour image registration error evaluation method based on circle error area ratio;

[0056] Figure 3 Flowchart for calculating circle error area ratio;

[0057] Figure 4 This is a schematic diagram of the registration between the actual two-dimensional image and the three-dimensional model image;

[0058] Figure 5 Schematic diagram of an implementation example of a method for evaluating low-curvature closed contour image registration error based on circle error area ratio; DETAILED DESCRIPTION

[0059] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is described below through the specific embodiments shown in the accompanying drawings. However, it should be understood that these descriptions are only exemplary and are not intended to limit the scope of the present invention. In addition, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present invention.

[0060] Example 1: Figure 1 、 Figure 2 、 Figure 3 、 Figure 4 、 Figure 5 As shown, this specific embodiment adopts the following technical solution: a low-curvature closed contour image registration error evaluation method based on the circular domain error area ratio, and the specific implementation process of the method is as follows:

[0061] Step 1: Import the registered 2D image contour point set and 3D model projection contour point set:

[0062] The lower left corner of the two-dimensional image is used as the coordinate origin, the right is the positive direction of the x-axis, the upward is the positive direction of the y-axis, and according to the right-hand rule, the vertical screen outward is the positive direction of the z-axis to establish a three-dimensional image registration error evaluation coordinate system w, obtain the contour curve information of the registered two-dimensional image, and input the contour point set of the registered two-dimensional image contour curve is the position information of the i-th contour point of the registered 2D image contour curve in the o-xy plane of the 3D image registration error evaluation coordinate system w. The value range of i is 1≤i≤n, where, is the x-axis coordinate of the i-th contour point of the registered 2D image contour curve located in the o-xy plane of the 3D image registration error evaluation coordinate system w, The y-axis coordinate of the i-th contour point of the registered two-dimensional image contour curve is located in the o-xy plane of the three-dimensional image registration error evaluation coordinate system w; the registered three-dimensional model is parallel projected onto the o-xy plane along the negative direction of the z-axis of the three-dimensional image registration error evaluation coordinate system w to obtain the projection contour curve information, and the contour point set of the registered three-dimensional model projection contour curve is input. is the position information of the jth contour point of the projection contour curve of the registered 3D model in the o-xy plane of the 3D image registration error evaluation coordinate system w. The value range of j is 1≤j≤m, where The jth contour point of the projection contour curve of the registered 3D model is located in the x-axis coordinate of the o-xy plane of the 3D image registration error evaluation coordinate system w. The jth contour point of the projection contour curve of the registered 3D model is located on the y-axis coordinate of the o-xy plane of the 3D image registration error evaluation coordinate system w;

[0063] Step 2: Error assessment circle division:

[0064] The i-th contour point and the i+1-th contour point of the registered two-dimensional image contour curve are connected by a straight line segment l i Tw Connect and calculate the straight line segment l i Tw The length of a i , when i=n, the nth contour point of the registered 2D image contour curve is connected to the first contour point by a straight line segment l n Tw Connect and calculate the straight line segment l n Tw The length of a n ; Define the error assessment circle division preset radius r, error assessment circle division judgment coefficient ε, error assessment circle division starting coefficient k, error assessment circle division span coefficient t, error assessment circle division compensation coefficient λ, error assessment circle division compensation span coefficient j, error assessment circle division retrieval coefficient α, error assessment circle division radius judgment distance L α , The actual radius r of the error assessment circle α , with r α As the radius, point The error evaluation circle area divided by the center of the circle is the error evaluation circle area C α , where the error assessment circle domain division preset radius r and the error assessment circle domain division determination coefficient ε are constant values; the error assessment circle domain division process is specifically as follows:

[0065] a) Set k = 1, λ = 0, j = 0, α = 1, and jump to step 2b);

[0066] b) Set t = 0 and jump to step 2 c);

[0067] c) Calculate the error assessment circle division radius judgment distance Jump to step 2d);

[0068] d) Determine 2r-L α <ε is established, specifically:

[0069] If 2r-L α <ε holds true, jump to step 2f);

[0070] If 2r-L α <ε does not hold, set t = t + 1, and jump to step 2 e);

[0071] e) Determine whether k+t≤n holds, specifically:

[0072] If k+t≤n holds, jump to step 2c);

[0073] If k + t ≤ n does not hold, set λ = 1, j = j + 1, t = t - 1, and jump to step 2 c);

[0074] f) Determine 2r-L α <0 is true, specifically:

[0075] If 2r-L α <0 holds, let r α =L α / 2, jump to step 2g);

[0076] If 2r-L α <0 does not hold, let r α = r, jump to step 2g);

[0077] g) Determine whether k+t=n is true, specifically:

[0078] If k+t=n holds, jump to step 2h);

[0079] If k+t=n does not hold, jump to step 2i);

[0080] h) Point is the center of the circle, r α Establish error evaluation circle C for radius α , the error assessment circle area division is completed, jump to step 3;

[0081] i) Point is the center of the circle, r α Establish error evaluation circle C for radius α , let k = k + t + 1, jump to step 2 j);

[0082] j) Determine whether k≤n is true, specifically:

[0083] If k≤n holds, set α=α+1 and jump to step 2b);

[0084] If k≤n does not hold, the error assessment circle division is completed and jump to step 3;

[0085] Step 3: Calculate the circular error area ratio

[0086] All error assessment circles are regarded as a collection C = {C1, C2, ..., C β ,…,C α}, C β For point is the center of the circle, r β The error evaluation circle domain is established with the radius, and the value range of β is 1≤β≤α; the contour point set P of the projection contour curve of the 3D model after registration is calculated by using the K nearest neighbor method. Th Find the point The nearest point Specifically: For the query point, set the parameter K value to 1. The value of parameter K represents the number of neighboring points to be found near the query point. Setting the parameter K value to 1 means that the contour point set P of the projection contour curve of the 3D model after registration is Th Finding and querying points The closest point to the Euclidean distance Euclidean distance to the nearest point That is the point The closest point, the Euclidean distance between the two points is d β ; Define the error evaluation circle C β The two intersection points of the contour curve of the registered two-dimensional image are close to the error evaluation circle C β-1 The intersection point on one side is point Approach error assessment circle C β+1 The intersection point on one side is point When d β <r β When the error evaluation circle C is defined β The two intersection points of the projection contour curve of the registered 3D model and the point The closest intersection point is point with dot The closest intersection point is point When d β =r β When the error evaluation circle C is defined β An intersection point with the projection contour curve of the registered 3D model is When d β <r β When the calculation is done from point Arrive The contour curve segment of the two-dimensional image after registration, point Arrive The projection contour curve segment of the 3D model after registration, as well as The area of the enclosed region is S β ; When d β =r β When the calculation is done from point Arrive The contour curve segment of the two-dimensional image after registration, as well as The area of the enclosed region is S β ; Define the area S of the closed region β and error evaluation circle C β The area ratio of the circular error is the area ratio of the circular error, and the symbol is ρ β express, ρ β The upper limit value is specified as ρ max ; Ratio of the circle error area calculated each time to ρ β Stored in the circle domain error area ratio data set ρ, ρ={ρ1,ρ2,…,ρ β ,…,ρ α};

[0087] a) Judge d β ≤r β Whether it is established, specifically:

[0088] If d β ≤r β If established, jump to step 3b);

[0089] If d β ≤r β If it does not hold, then the two curves after the output image registration are in the error evaluation circle C. β There are large differences in the image, the evaluation is unqualified, and the image registration error evaluation is completed;

[0090] b) Judge d β <r β Whether it is established, specifically:

[0091] If d β <r β If established, jump to step 3c);

[0092] If d β <r β If not, jump to step 3d);

[0093] c) Calculate the points Arrive The contour curve segment of the two-dimensional image after registration, point Arrive The projection contour curve segment of the 3D model after registration, as well as The area of the enclosed region is S β , jump to step 3e);

[0094] d) Calculate the value of the point Arrive The contour curve segment of the two-dimensional image after registration, as well as The area of the enclosed region is S β , jump to step 3e);

[0095] e) Calculate the circular error area ratio Determine ρ β ≤ρ max Whether it is established, specifically:

[0096] If ρ β ≤ρ max If the circle error area ratio ρ β Store it in the circle error area ratio data set ρ and jump to step f);

[0097] If ρ β ≤ρ max If it is not true, the output is in the error assessment circle C β The inner circle error area ratio is unqualified, and the output is ρ β The image registration error evaluation is completed;

[0098] f) Whether all circular regions have been judged:

[0099] Determine whether β+1>α is true, specifically:

[0100] If β+1>α holds, jump to step 3g);

[0101] If β+1>α does not hold, set β=β+1 and jump to step 3a);

[0102] g) Obtain the maximum value max{ρ1,ρ2,…,ρ β ,…,ρ α}, and the β value corresponding to the maximum value, the output image registration is qualified, and the area with the largest registration error is the error evaluation circle C β , the circle error area ratio is max{ρ1,ρ2,…,ρ β ,…,ρ α}, the image registration error evaluation is completed.

[0103] Implementation Example 2: Figure 5 As shown, the contour point set P of the two-dimensional image contour curve after input registration Tw , point set P Tw Contains n = 1034 contour points; Input the contour point set P of the 3D model projection contour curve after registration Th , point set P ThContains m = 1172 contour points; defines the error assessment circle division preset radius r = 0.5, the error assessment circle division determination coefficient ε = 0.05; sets the error assessment circle division starting coefficient k = 1, the error assessment circle division compensation coefficient λ = 0, the error assessment circle division compensation span coefficient j = 0, and the error assessment circle division retrieval coefficient α = 1;

[0104] Let the error assessment circle domain division span coefficient t=0, calculate The judgment that 2r-L1<0.05 is not true; let t=1, the judgment that k+t≤1034 is true, calculate The judgment that 2r-L1<0.05 is not true; let t=2, the judgment that k+t≤1034 is true, calculate The judgment that 2r-L1<0.05 is not true; let t=3, the judgment that k+t≤1034 is true; calculate The judgment that 2r-L1<0.05 is not true; let t=4, and the judgment that k+t≤1034 is true; calculate The judgment that 2r-L1<0.05 is not true; let t=5, and the judgment that k+t≤1034 is true; calculate The judgment that 2r-L1<0.05 is not true; let t=6, and the judgment that k+t≤1034 is true; calculate The judgment that 2r-L1<0.05 is not true; let t=7, the judgment that k+t≤1034 is true; calculate The judgment that 2r-L1<0.05 is not true; let t=8, the judgment that k+t≤1034 is true; calculate The judgment that 2r-L1<0.05 is not true; let t=9, the judgment that k+t≤1034 is true; calculate The judgment 2r-L1<0.05 is true; the judgment 2r-L1<0 is true, r1=0.555; the judgment k+t=1034 is not true, point by point With t as the center and r1 = 0.555 as the radius, establish the error evaluation circle C1; set k = k + t + 1, determine if k ≤ 1034, and set α = α + 1; repeat the above steps until k ≤ 1034 is no longer true, then the error evaluation circle division is complete and jump to step 3;

[0105] Regulation ρ max = 0.4; Use the K nearest neighbor method to project the contour curve contour point set P of the 3D model after registration Th Find the point The nearest point The Euclidean distance between the two points is d1=0.472, so d1<r1 is established. Arrive The contour curve segment of the two-dimensional image after registration, point Arrive The projection contour curve segment of the 3D model after registration, as well as The area of the enclosed region is S1 = 0.312, and the calculated circular domain error area ratio is ρ1 = 0.322; judge ρ1≤ρ max If the circular domain error area ratio ρ1 is established, store it in the circular domain error area ratio data set ρ; determine whether β+1>α is not established, set β=β+1; repeat the above steps until β+1>α is established; obtain the maximum value ρ in the circular domain error area ratio data set ρ. 25 =0.389, the output image registration is qualified, and the area with the largest registration error is the error assessment circle C 25 , the circular error area ratio is ρ 25 =0.389, the image registration error evaluation is completed.

Claims

1. A method for evaluating registration error of low-curvature closed contour images based on circular error area ratio, characterized by: The specific implementation process of the method is as follows: Step 1: Import the registered 2D image contour point set and 3D model projection contour point set: The lower left corner of the two-dimensional image is used as the coordinate origin, the right is the positive direction of the x-axis, the upward is the positive direction of the y-axis, and according to the right-hand rule, the vertical screen outward is the positive direction of the z-axis to establish a three-dimensional image registration error evaluation coordinate system w, obtain the contour curve information of the registered two-dimensional image, and input the contour point set of the registered two-dimensional image contour curve is the position information of the i-th contour point of the registered 2D image contour curve in the o-xy plane of the 3D image registration error evaluation coordinate system w. The value range of i is 1≤i≤n, where, is the x-axis coordinate of the i-th contour point of the registered 2D image contour curve located in the o-xy plane of the 3D image registration error evaluation coordinate system w, The y-axis coordinate of the i-th contour point of the registered two-dimensional image contour curve is located in the o-xy plane of the three-dimensional image registration error evaluation coordinate system w; the registered three-dimensional model is parallel projected onto the o-xy plane along the negative direction of the z-axis of the three-dimensional image registration error evaluation coordinate system w to obtain the projection contour curve information, and the contour point set of the registered three-dimensional model projection contour curve is input. is the position information of the jth contour point of the projection contour curve of the registered 3D model in the o-xy plane of the 3D image registration error evaluation coordinate system w. The value range of j is 1≤j≤m, where The jth contour point of the projection contour curve of the registered 3D model is located in the x-axis coordinate of the o-xy plane of the 3D image registration error evaluation coordinate system w. The jth contour point of the projection contour curve of the registered 3D model is located on the y-axis coordinate of the o-xy plane of the 3D image registration error evaluation coordinate system w; Step 2: Error assessment circle division: The i-th contour point and the i+1-th contour point of the registered two-dimensional image contour curve are connected by a straight line segment l i Tw Connect and calculate the straight line segment l i Tw The length of a i When i=n, the nth contour point of the registered 2D image contour curve is connected to the first contour point by a straight line segment l n Tw Connect and calculate the straight line segment l n Tw The length of a n ; Define the error assessment circle division preset radius r, error assessment circle division judgment coefficient ε, error assessment circle division starting coefficient k, error assessment circle division span coefficient t, error assessment circle division compensation coefficient λ, error assessment circle division compensation span coefficient j, error assessment circle division retrieval coefficient α, error assessment circle division radius judgment distance L α , The actual radius r of the error assessment circle α , with r α As the radius, point The error evaluation circle area divided by the center of the circle is the error evaluation circle area C α , where the error assessment circle domain division preset radius r and the error assessment circle domain division determination coefficient ε are constant values; the error assessment circle domain division process is specifically as follows: a) Set k = 1, λ = 0, j = 0, α = 1, and jump to step 2b); b) Set t = 0 and jump to step 2 c); c) Calculate the error assessment circle division radius judgment distance Jump to step 2d); d) Determine 2r-L α <ε is established, specifically: If 2r-L α <ε holds true, jump to step 2f); If 2r-L α <ε does not hold, set t = t + 1, and jump to step 2 e); e) Determine whether k+t≤n holds, specifically: If k+t≤n holds, jump to step 2c); If k + t ≤ n does not hold, set λ = 1, j = j + 1, t = t - 1, and jump to step 2 c); f) Determine 2r-L α <0 is true, specifically: If 2r-L α <0 holds, let r α =L α / 2, jump to step 2g); If 2r-L α <0 does not hold, let r α = r, jump to step 2g); g) Determine whether k+t=n is true, specifically: If k+t=n holds, jump to step 2h); If k+t=n does not hold, jump to step 2i); h) Point is the center of the circle, r α Establish error evaluation circle C for radius α , the error assessment circle area division is completed, jump to step 3; i) Point is the center of the circle, r α Establish error evaluation circle C for radius α , let k = k + t + 1, jump to step 2 j); j) Determine whether k≤n is true, specifically: If k≤n holds, set α=α+1 and jump to step 2b); If k≤n does not hold, the error assessment circle division is completed and jump to step 3; Step 3: Calculate the circular error area ratio All error assessment circles are regarded as a collection C = {C1, C2, ..., C β ,…,C α }, C β For point is the center of the circle, r β The error evaluation circle domain is established with the radius, and the value range of β is 1≤β≤α; the contour point set P of the projection contour curve of the 3D model after registration is calculated by using the K nearest neighbor method. Th Find the point The nearest point Specifically: For the query point, set the parameter K value to 1. The value of parameter K represents the number of neighboring points to be found near the query point. Setting the parameter K value to 1 means that the contour point set P of the projection contour curve of the 3D model after registration is Th Finding and querying points The closest point to the Euclidean distance Euclidean distance to the nearest point That is the point The closest point, the Euclidean distance between the two points is d β ; Define the error evaluation circle C β The two intersection points of the contour curve of the registered two-dimensional image are close to the error evaluation circle C β-1 The intersection point on one side is point Approach error assessment circle C β+1 The intersection point on one side is point When d β <r β When the error evaluation circle C is defined β The two intersection points of the projection contour curve of the registered 3D model and the point The closest intersection point is point with dot The closest intersection point is point When d β =r β When the error evaluation circle C is defined β An intersection point with the projection contour curve of the registered 3D model is When d β <r β When the calculation is done from point Arrive The contour curve segment of the two-dimensional image after registration, point Arrive The projection contour curve segment of the 3D model after registration, as well as The area of the enclosed region is S β ; When d β =r β When the calculation is done from point Arrive The contour curve segment of the two-dimensional image after registration, as well as The area of the enclosed region is S β ; Define the area S of the closed region β and error evaluation circle C β The area ratio of the circular domain error is the ratio of the circular domain error area, and the symbol is ρ β express, ρ β The upper limit value is specified as ρ max ; Ratio of the circle error area calculated each time to ρ β Stored in the circle domain error area ratio data set ρ, ρ={ρ1,ρ2,…,ρ β ,…,ρ α }; a) Judge d β ≤r β Whether it is established, specifically: If d β ≤r β If established, jump to step 3b); If d β ≤r β If it does not hold, then the two curves after the output image registration are in the error evaluation circle C. β There are large differences in the image, the evaluation is unqualified, and the image registration error evaluation is completed; b) Judge d β <r β Whether it is established, specifically: If d β <r β If established, jump to step 3c); If d β <r β If not, jump to step 3d); c) Calculate the points Arrive The contour curve segment of the two-dimensional image after registration, point Arrive The projection contour curve segment of the 3D model after registration, as well as The area of the enclosed region is S β , jump to step 3e); d) Calculate the value of the point Arrive The contour curve segment of the two-dimensional image after registration, as well as The area of the enclosed region is S β , jump to step 3e); e) Calculate the circular error area ratio Determine ρ β ≤ρ max Whether it is established, specifically: If ρ β ≤ρ max If the circle error area ratio ρ β Store it in the circle error area ratio data set ρ and jump to step f); If ρ β ≤ρ max If it is not true, the output is in the error assessment circle C β The inner circle error area ratio is unqualified, and the output is ρ β The image registration error evaluation is completed; f) Whether all circular regions have been judged: Determine whether β+1>α is true, specifically: If β+1>α holds, jump to step 3g); If β+1>α does not hold, set β=β+1 and jump to step 3a); g) Obtain the maximum value max{ρ1,ρ2,…,ρ β ,…,ρ α }, and the β value corresponding to the maximum value, the output image registration is qualified, and the area with the largest registration error is the error evaluation circle C β , the circle error area ratio is max{ρ1,ρ2,…,ρ β ,…,ρ α }, the image registration error evaluation is completed.

Images