A point-to-point registration method

Abstract

The present invention discloses a point-to-point registration method, which includes Step 1: making a registration plate with a material having good X-ray transparency, and randomly placing n objects P1, P2,..., Pn capable of extracting and determining coordinates. Step 2: using a C-arm machine to perform fluoroscopic imaging, transmitting the captured fluoroscopic image into a PC, preprocessing the image, and then obtaining the center coordinates C1, C2,..., Cn of the projections of the steel balls in the fluoroscopic image. Step 3: measuring and obtaining the three-dimensional positions of the centers of the steel balls on the registration plate with any point on the registration plate as the origin, taking the direction perpendicular to the registration plate as the Z direction, such as P1(x1, y1, z1), P2(x2, y3, z2),..., Pn(xn, yn, zn), and mapping P1(x1, y1, z1), P2(x2, y3, z2),..., Pn(xn, yn, zn) to a two-dimensional plane; through the implementation of the present invention, the registration accuracy between the feature points in the perspective view and the feature points in the model is not affected by factors such as the occlusion of key steel balls and different poses of the model, and both the practicability and applicability are relatively strong, achieving a prominent progress and having certain use value and promotion value.

Description

Technical Field

[0001] The present invention relates to the technical field of feature point registration, and particularly to a point-to-point registration method. Background Art

[0002] During the application of a surgical robot, it is necessary to find the one-to-one correspondence between the feature points in the model and the projections of the feature points in the C-arm perspective view for calculating image distortion correction, focal length measurement, camera pose estimation, and so on.

[0003] When using steel balls with different diameters as feature points in the same plane, the positions of the steel balls are made to correspond one by one by calculating the diameters of the projections of the steel balls in the perspective view. For example, Figure 1 However, in actual use, the occlusion of key steel balls, different poses of the model, etc. can easily affect the registration accuracy, and the applicability is poor. Summary of the Invention

[0004] Aiming at the deficiencies of the prior art, the present invention provides a point-to-point registration method to solve the above problems.

[0005] To achieve the above object, the present invention provides the following technical solution: A point-to-point registration method, including the following steps:

[0006] Step 1: Make a registration plate using a material with good X-ray permeability, and randomly place n objects P1, P2,..., Pn whose coordinates can be extracted and determined.

[0007] Step 2: Use a C-arm machine to perform perspective shooting, transfer the captured perspective image into a PC, preprocess the image, and then obtain the center coordinates C1, C2,..., Cn of the projections of the steel balls in the perspective image;

[0008] Step 3: Measure and obtain the three-dimensional positions of the centers of the steel balls on the registration plate with an arbitrary point on the registration plate as the origin, and the direction perpendicular to the registration plate is the Z direction, such as P1(x1, y1, z1), P2(x2, y3, z2),..., Pn(xn, yn, zn). Map P1(x1, y1, z1), P2(x2, y3, z2),..., Pn(xn, yn, zn) to a two-dimensional plane, and the corresponding two-dimensional point coordinates obtained by mapping each point are M1(X1, Y1), M2(X2, Y2),..., Mn(Xn, Yn). Connect each pair of points among M1, M2,..., Mn to generate a pairwise connection diagram of the spatial positions of the registration plate mapped to the two-dimensional plane. Similarly, connect each pair of the circle center coordinates C1, C2,..., Cn obtained from the perspective image to generate a pairwise connection diagram of the centers of the steel ball projections on the registration plate;

[0009] Step 4: Extract a series of feature points A and feature points B from the pairwise connection diagram of the registration plate mapped to the two-dimensional plane space position and the pairwise connection diagram of the projection centers of the registration plate steel balls respectively. Obtain the best match between the feature points A and the feature points B through the feature point registration algorithm, find the most matching items, extract the corresponding feature point coordinates, and obtain the transformation matrix H between the pairwise connection diagram of the registration plate mapped to the two-dimensional plane space position and the pairwise connection diagram of the projection centers of the registration plate steel balls through homography;

[0010] Step 5: The projection coordinates C1, C2, …, Cn of the objects P1, P2, …, Pn in the perspective image are obtained through the perspective transformation matrix H to obtain the corresponding coordinates M1’, M2’, …, Mn’ in the pairwise connection diagram of the registration plate mapped to the two-dimensional plane space position. Find the position M in the pairwise connection diagram of the registration plate mapped to the two-dimensional plane space position that is closest to the position coordinates of the projection of the registration plate steel balls mapped to the two-dimensional plane. The projection of the object P corresponding to the M coordinate is C, and thus C and the corresponding P are put into one-to-one correspondence.

[0011] Further, the registration plate is a plane or a curved surface.

[0012] Further, in the above Step 1, the object whose coordinates can be extracted and determined is a steel ball, and the steel ball can be placed inside or on the surface of the registration plate.

[0013] Further, in the above Step 2, the perspective image is preprocessed using filtering, and the method for obtaining the projection center coordinates C1, C2, …, Cn of the steel balls in the perspective view is any one of blob detection, edge detection, centroid, and cross feature points.

[0014] Further, in the above Step 4, a series of feature points A and feature points B are extracted from the pairwise connection diagram of the registration plate mapped to the two-dimensional plane space position and the pairwise connection diagram of the projection centers of the registration plate steel balls respectively using feature extraction schemes such as Harris corner detection, sift, suft, brief, and ORB.

[0015] Further, the projection coordinates C1, C2, …, Cn of the steel balls P1, P2, …, Pn are obtained through the perspective transformation matrix H to obtain the corresponding coordinates M1’, M2’, …, Mn’ in the pairwise connection diagram of the registration plate mapped to the two-dimensional plane space position, where M’ = H * C.

[0016] The substantial effect of the present invention: Even when any part of the steel balls is blocked, the steel balls can still be put into one-to-one correspondence with the projections, and the calculation time is short, the accuracy is high, the practicability and applicability are both strong, which represents a prominent progress and has certain use value and promotion value. Description of the Drawings

[0017] Figure 1 It is a schematic diagram of the projection of traditional steel balls.

[0018] Figure 2 It is a schematic diagram of a traditional metal wire grid.

[0019] Figure 3 It is a schematic diagram of the registration plate structure of the present invention.

[0020] Figure 4 It is a schematic diagram of the projection of the registration plate of the present invention.

[0021] Figure 5 It is a schematic diagram of the pairwise connection of the positions where the registration plate of the present invention is mapped to the two-dimensional plane space.

[0022] Figure 6 It is a schematic diagram of the pairwise connection of the projection centers of the steel balls of the registration plate of the present invention.

[0023] Figure 7 It is a schematic diagram of the optimal registration of the feature points of the present invention.

[0024] Figure 8 It is for the steel ball projection of the present invention in Figure 6 the corresponding position diagram. Detailed implementation manners

[0025] Example 1:

[0026] As Figure 3 , 4 , 5, 6, 7, 8 show, a point-to-point registration method includes the following steps:

[0027] Step 1: Make a registration plate using a material with good X-ray transparency. The registration plate is flat or curved, and randomly place n steel balls P1, P2, Pn that can extract and determine coordinates. The steel balls are placed inside or on the surface of the registration plate.

[0028] Step 2: Use a C-arm machine for fluoroscopic imaging, transfer the captured fluoroscopic image into a PC, preprocess the image using filtering, and then obtain the center coordinates C1, C2, Cn of the projection of the steel balls in the fluoroscopic image through any one of spot detection, edge detection, centroid, and cross feature points;

[0029] Step 3: Taking any point on the registration plate as the origin, measure and obtain the three-dimensional positions of the centers of the steel balls on the registration plate. The direction perpendicular to the registration plate is the Z direction, such as P1(x1, y1, z1), p2(x2, y3, z2), Pn(xn, yn, zn). Map P1(x1, y1, z1), p2(x2, y3, z2), Pn(xn, yn, zn) to a two-dimensional plane. The corresponding two-dimensional point coordinates obtained by mapping each point are M1(X1, Y1), M2(X2, Y2), Mn(Xn, Yn). Connect the points M1, M2, Mn pairwise to generate a pairwise connection diagram of the spatial positions of the registration plate mapped to the two-dimensional plane. Similarly, connect the center coordinates C1, C2, Cn of the circles obtained from the perspective image pairwise to generate a pairwise connection diagram of the projection centers of the steel balls on the registration plate;

[0030] Step 4: Use feature extraction schemes such as Harris corner detection, sift, suft, brief, and ORB to extract a series of feature points A and feature points B in the pairwise connection diagram of the spatial positions of the registration plate mapped to the two-dimensional plane and the pairwise connection diagram of the projection centers of the steel balls on the registration plate. Obtain the best match between feature point A and feature point B through the feature point registration algorithm, such as Figure 7 , find the most matching items, extract the corresponding feature point coordinates, and obtain the transformation matrix H between the pairwise connection diagram of the spatial positions of the registration plate mapped to the two-dimensional plane and the pairwise connection diagram of the projection centers of the steel balls on the registration plate through homography;

[0031] Step 5: The projection coordinates C1, C2, Cn of the steel balls P1, P2, Pn in the perspective image are used to obtain the corresponding coordinates M1’, M2’, Mn’ in the pairwise connection diagram of the spatial positions of the registration plate mapped to the two-dimensional plane through the perspective transformation matrix H. M’ = H * C. Find the position M in the pairwise connection diagram of the spatial positions of the registration plate mapped to the two-dimensional plane that is closest to the position coordinates of the projection of the steel balls on the registration plate mapped to the two-dimensional plane. The projection of the object P corresponding to the M coordinate is C. Thus, C and the corresponding P are in one-to-one correspondence.

[0032] Example 2:

[0033] A point-to-point registration method includes the following steps:

[0034] Step 1: Make a registration plate using a material with good X-ray permeability. The registration plate is a plane, and randomly place 3 steel balls P1, P2, P3 that can extract and determine coordinates. The steel balls are placed inside the registration plate.

[0035] Step 2: Use a C-arm machine to take a perspective image, transfer the taken perspective image into a PC, preprocess the image using filtering, and then obtain the projection center coordinates C1, C2, C3 of the steel balls in the perspective image through crosshair feature points;

[0036] Step 3: Take any point on the registration plate as the origin to measure and obtain the three-dimensional positions of the centers of the steel balls on the registration plate. The direction perpendicular to the registration plate is the Z direction, such as P1(x1, y1, z1), p2(x2, y3, z2), P3(xn, yn, zn). Map P1(x1, y1, z1), p2(x2, y3, z2), P3(xn, yn, zn) to the two-dimensional plane. The two-dimensional point coordinates corresponding to each point after mapping are M1(X1, Y1), M2(X2, Y2), M3(Xn, Yn). Connect the points M1, M2, and M3 pairwise to generate a pairwise connection diagram of the spatial positions of the registration plate mapped to the two-dimensional plane. Similarly, connect the center coordinates C1, C2, and C3 of the circles obtained from the perspective image pairwise to generate a pairwise connection diagram of the projection centers of the steel balls on the registration plate.

[0037] Step 4: Use the Harris corner detection feature extraction scheme to extract a series of feature points A and feature points B in the pairwise connection diagram of the spatial positions of the registration plate mapped to the two-dimensional plane and the pairwise connection diagram of the projection centers of the steel balls on the registration plate. Obtain the best match between feature point A and feature point B through the feature point registration algorithm, find the most matching items, extract the corresponding feature point coordinates, and obtain the transformation matrix H between the pairwise connection diagram of the spatial positions of the registration plate mapped to the two-dimensional plane and the pairwise connection diagram of the projection centers of the steel balls on the registration plate through homography.

[0038] Step 5: The projection coordinates C1, C2, and C3 of the steel balls P1, P2, and P3 in the perspective image are used to obtain the corresponding coordinates M1’, M2’, and M3’ in the pairwise connection diagram of the spatial positions of the registration plate mapped to the two-dimensional plane through the perspective transformation matrix H, M’ = H * C. Find the mapping position M closest to the position coordinates of the steel ball on the registration plate mapped to the two-dimensional plane in the pairwise connection diagram of the spatial positions of the registration plate mapped to the two-dimensional plane. The projection of the object P corresponding to the M coordinate is C. Thus, C and the corresponding P are in one-to-one correspondence.

[0039] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, or improvements made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A point-to-point registration method, characterized in that, It includes the following steps: Step 1: Make a registration plate with a material having good X-ray penetrability, and randomly place n objects P1, P2,..., Pn that can extract and determine coordinates. Step 2: Use a C-arm machine for fluoroscopic imaging, transmit the captured fluoroscopic image into a PC, preprocess the image, and then obtain the center coordinates C1, C2,..., Cn of the steel ball projections in the fluoroscopic image. Step 3: Measure and obtain the three-dimensional positions of the centers of the steel balls on the registration plate with an arbitrary point on the registration plate as the origin. The direction perpendicular to the registration plate is the Z direction, such as P1(x1, y1, z1), P2(x2, y3, z2),..., Pn(xn, yn, zn). Map P1(x1, y1, z1), P2(x2, y3, z2),..., Pn(xn, yn, zn) to a two-dimensional plane, and the corresponding two-dimensional point coordinates obtained by mapping each point are M1(X1, Y1), M2(X2, Y2),..., Mn(Xn, Yn). Connect the points M1, M2,..., Mn pairwise to generate a pairwise connection diagram of the spatial positions in the two-dimensional plane. Similarly, connect the center coordinates C1, C2,..., Cn of the circles obtained from the fluoroscopic image pairwise to generate a pairwise connection diagram of the projection centers of the steel balls on the registration plate. Step 4: Respectively extract a series of feature points A and feature points B from the pairwise connection diagram of the spatial positions in the two-dimensional plane and the pairwise connection diagram of the projection centers of the steel balls on the registration plate. Obtain the best match between the feature points A and the feature points B through a feature point registration algorithm, find the most matching items, extract the corresponding feature point coordinates, and obtain the transformation matrix H between the pairwise connection diagram of the spatial positions in the two-dimensional plane and the pairwise connection diagram of the projection centers of the steel balls on the registration plate through homography. Step 5: The projection coordinates C1, C2,..., Cn of the objects P1, P2,..., Pn in the fluoroscopic image obtain the corresponding coordinates M1', M2',..., Mn' in the pairwise connection diagram of the spatial positions in the two-dimensional plane through the perspective transformation matrix H. Find the position M in the pairwise connection diagram of the spatial positions in the two-dimensional plane that is closest to the position coordinates of the mapping of the steel balls on the registration plate to the two-dimensional plane space. The projection of the object P corresponding to the M coordinate is C, and thus C and the corresponding P are in one-to-one correspondence.

2. The point-to-point registration method according to claim 1, characterized in that, The registration plate is a plane or a curved surface.

3. The point-to-point registration method according to claim 1, characterized in that, In the above Step 1, the objects that can extract and determine coordinates are steel balls, and the steel balls are placed inside or on the surface of the registration plate.

4. The point-to-point registration method according to any one of claims 1 or 3, characterized in that, In the above Step 2, use filtering to preprocess the fluoroscopic image. And the method for obtaining the center coordinates C1, C2,..., Cn of the steel ball projections in the fluoroscopic image is any one of blob detection, edge detection, centroid, and cross feature points.

5. The point-to-point registration method according to claim 1, characterized in that, In the above Step 4, use feature extraction schemes such as Harris corner detection, sift, suft, brief, and ORB to respectively extract a series of feature points A and feature points B from the pairwise connection diagram of the spatial positions in the two-dimensional plane and the pairwise connection diagram of the projection centers of the steel balls on the registration plate.

6. The point-to-point registration method according to claim 3, characterized in that, The projection coordinates C1, C2,..., Cn of the steel balls P1, P2,..., Pn obtain the corresponding coordinates M1', M2',..., Mn' in the pairwise connection diagram of the spatial positions in the two-dimensional plane through the perspective transformation matrix H, where M' = H * C.

Images