A real-time geometric correction method for perspective imaging device

Abstract

The present invention discloses a real-time geometric correction method for a perspective imaging device, comprising the following steps: constructing a real-time calibration cover with a high-precision gold mark implanted therein; fixing the perspective imaging device to a shooting table, shooting a perspective view containing the real-time calibration cover; extracting the position of the gold mark in the perspective view, and making a one-to-one correspondence with the three-dimensional position of the gold mark; and calculating and obtaining the geometric parameters of the real-time perspective device. The present invention realizes real-time geometric calibration of the perspective device during shooting, overcomes the real-time error caused by the instability of the perspective device during use, and can improve the geometric accuracy of the perspective image.

Description

Technical Field

[0001] The present invention relates to a geometric correction technology for perspective imaging, and in particular to a real-time geometric correction method for a perspective imaging device. Background Art

[0002] In the applications of fluoroscopic images such as CBCT and fluoroscopic image-guided positioning, there are precision requirements for the geometric parameters of the fluoroscopic imaging device. When acquiring the fluoroscopic image, it is necessary to obtain the geometric parameters of the projection fluoroscopic imaging device that forms the perspective view. These geometric parameters refer to the perspective projection matrix or equivalent parameters such as the perspective source coordinates, imaging plate coordinates and direction.

[0003] For example, in the application of cone beam computed tomography (CBCT), the inaccuracy of geometric parameters may cause artifacts and other defects in the CBCT three-dimensional images reconstructed from two-dimensional and three-dimensional images during the back-projection reconstruction process, thus affecting their subsequent use.

[0004] In current applications in this field, geometric parameters are generally obtained through an independent calibration process in advance, and pre-stored geometric parameters are used during perspective imaging. However, such methods are heavily dependent on the repeatability of the device and are difficult to apply to devices with insufficient repeatability of the mechanical structure. Summary of the invention

[0005] Purpose of the invention: The purpose of the present invention is to provide a real-time geometric correction method for a perspective imaging device, which realizes real-time geometric calibration of the perspective device during the shooting process. Only rough initial geometric parameters are needed to obtain accurate geometric parameters for perspective image synchronization, and can be conveniently applied to imaging devices with different mechanical precision and general repeatability.

[0006] Technical solution: The real-time geometric correction method of the perspective imaging device of the present invention comprises the following steps:

[0007] (1) Constructing a real-time calibration cover with a high-precision gold standard implant;

[0008] (2) Fixing the real-time calibration cover to the shooting table and shooting a perspective view including the real-time calibration cover;

[0009] (3) extracting the gold mark position in the perspective image and making a one-to-one correspondence with the gold mark's three-dimensional position;

[0010] (4) Calculate and obtain the geometric parameters of the real-time perspective device.

[0011] In step (1), a double helix is ​​provided on the outer surface of the real-time calibration cover, and a plurality of gold labels are embedded in the double helix.

[0012] In step (1), the gold mark is a spherical metal marker, and the real-time calibration cover is made of a material with high transmittance to radiographic rays.

[0013] In step (2), after the real-time calibration cover is fixed to the shooting table, the displacement information of the real-time calibration cover relative to the reference coordinate system of the shooting table is obtained, so as to calculate the three-dimensional coordinate sequence T = {xt1, xt2, ..., xt n}.

[0014] In step (2), when taking a perspective view including the real-time calibration mask, the projection matrix M of the perspective imaging device measured in advance is recorded. 0 , the calculation formula is as follows:

[0015]

[0016] In step (3), the two-dimensional coordinates D = {xd1, xd2, ..., xd m}, xd i ∈R 2 .

[0017] In step (3), extracting the two-dimensional coordinates of the gold mark in the real-time calibration cover in the perspective image includes the following steps:

[0018] (3.1) According to the projection matrix M of the initial perspective imaging device 0 And the three-dimensional coordinate sequence T of the gold mark, calculate the expected projection sequence D of each gold mark point in the image 0 ={xd 0 1, xd 0 2, ..., xd 0 m},xd 0 i ∈R 2 ;

[0019] (3.2) For each expected coordinate, the image of the gold mark pixels is filtered out by the threshold method. A point that meets the gold mark projection characteristics is searched on the image of the gold mark pixels. The pixel coordinates around the point are weighted and summed with the pixel value as the weight for the points whose threshold is higher than the lower limit of the gold mark projection. The accurate two-dimensional coordinates of the center of the gold mark on the projection image are obtained, and the two-dimensional coordinate sequence D is obtained.

[0020] In step (4), the specific process of calculating the geometric parameters of the real-time perspective device is as follows: minimizing the sum of the distances between the projection points calculated by the projection matrix and the corresponding gold mark points on the projection image as the optimization target, iteratively optimizing and updating the projection matrix and the three-dimensional and two-dimensional coordinate mapping relationship of the gold mark, and finally obtaining the optimized projection matrix M.

[0021] Beneficial effect: Compared with the prior art, the present invention has the following beneficial effects: it realizes real-time geometric calibration of the perspective device during shooting, overcomes the real-time error caused by the instability of the perspective device during use, and improves the geometric accuracy of the perspective image. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a schematic diagram of the process of the present invention;

[0023] Figure 2 A real-time calibration cover for implanting a gold standard in the present invention;

[0024] Figure 3 This is a state diagram of the real-time calibration cover in the present invention when it is placed on the shooting table;

[0025] Figure 4 is a two-dimensional perspective view of the gold-containing mark obtained in the present invention;

[0026] Figure 5 This is a flowchart of the optimization of the projection matrix in the present invention. DETAILED DESCRIPTION

[0027] The technical solution of the present invention is further described in detail below in conjunction with specific implementation methods and the accompanying drawings.

[0028] like Figure 1 As shown, the real-time geometric correction method of the perspective imaging device of the present invention comprises the following steps:

[0029] (1) Figure 2 and Figure 3 As shown, a real-time calibration cover 1 with a high-precision embedded gold mark 2 is constructed, and the size of the real-time calibration cover 1 is estimated according to the distance between the source, the object, and the imaging plate of the fluoroscopic imaging device to ensure that when the real-time calibration cover 1 is placed on the shooting table 3, there are enough gold marks 2 (more than 12) included in the radiation field formed by the X-ray source 5 and the imaging plate 4; in this embodiment, the real-time calibration cover 1 adopts an organic glass cover with a thickness of 1 cm, and a double helix is ​​provided on the outer surface of the organic glass cover, and a plurality of gold marks are embedded in the double helix; specifically, the gold mark 2 is a spherical metal marker, and the gold mark 2 has a strong ability to block transmitted projection rays. In this embodiment, the gold mark 2 adopts a stainless steel bead, and the real-time calibration cover 1 adopts a material with high transmittance to fluoroscopic rays, and has a weak ability to block transmitted projection rays; in addition, the three-dimensional coordinate sequence of the steel bead relative to the real-time calibration cover 1 is also recorded.

[0030] (2) Figure 3As shown, the X-ray source 5 and the imaging flat panel 4 together constitute a perspective imaging device. Fix the real-time calibration cover 1 to the shooting table 3 to ensure that the real-time calibration cover 1 has sufficient rigidity and will not deform. Adjust the position of the real-time calibration cover 1 to ensure that the perspective imaging can capture enough gold marks 2 on the real-time calibration cover 1, generally more than 20. After the real-time calibration cover 1 is fixed to the shooting table 3, the displacement information of the position of the real-time calibration cover 1 relative to the reference coordinate system of the shooting table 3 is used to calculate the three-dimensional coordinate sequence T = {xt1, xt2,…, xt n}, xt i ∈R 3 ; Take a perspective view of the real-time calibration cover 1;

[0031] In addition, when taking a perspective view containing the real-time calibration cover 1, the geometric parameters of the rough perspective imaging device measured in advance are recorded, and the projection matrix M of the perspective imaging device is calculated. 0 , which is used to obtain the two-dimensional coordinates projected from the three-dimensional space coordinates onto the plate. The calculation formula is as follows:

[0032]

[0033] (3) Figure 4 As shown, the position of the gold mark 2 in the perspective image is extracted and corresponds to the three-dimensional position of the gold mark 2 one by one; specifically, the two-dimensional coordinates D = {xd1, xd2, ..., xd m}, xd i ∈R 2 , including the following steps:

[0034] (3.1) According to the projection matrix M of the initial perspective imaging device 0 And the three-dimensional coordinate sequence T of the gold mark, calculate the expected projection sequence D of each gold mark point in the image 0 ={xd 0 1, xd 0 2,…,xd 0 m},xd 0 i ∈R 2 ;

[0035] (3.2) For each expected coordinate, the image of the gold mark pixels is filtered out by the threshold method, and each gold mark image area is obtained by template matching and other methods. The pixel values ​​in the area all exceed the threshold, the area shape is the gold mark projection circle, and the size is consistent with the theoretical size of the gold mark projection; search for points that meet the gold mark projection characteristics on the image of the filtered gold mark pixels, and weighted sum the pixel point coordinates around the point with a threshold higher than the lower limit of the gold mark projection with the pixel value as the weight to obtain the accurate two-dimensional coordinates of the center of the gold mark on the projection image, and obtain the two-dimensional coordinate sequence D.

[0036] (4) Figure 5 As shown, the geometric parameters of the real-time perspective imaging device, that is, the optimized projection matrix M, are calculated.

[0037]

[0038] The specific process is: through the optimization algorithm, the mapping relationship between the gold mark position in the image and the three-dimensional gold mark position and the projection matrix of the perspective projection device are optimized. The optimization algorithm here can be a common optimization algorithm, and the optimization input is the sequence number I of the gold mark two-dimensional coordinates associated with the three-dimensional coordinates 0 ={i 0 1,i 0 2,…,i 0 m},i 0 i ∈{1,2,…,m}, the optimization output is the updated geometric parameter M j And the updated three-dimensional and two-dimensional mapping relationship I k , where k refers to the number of iterations to solve the projection matrix, and the optimization goal is to obtain the theoretical two-dimensional projection coordinates based on M and I k The corresponding two-dimensional coordinates have the smallest average distance.

Claims

1. A real-time geometric correction method for a perspective imaging device, characterized in that: The following steps are involved: (1) Constructing a real-time calibration cover with a high-precision gold standard implant; (2) Fixing the real-time calibration cover to the shooting table and shooting a perspective view including the real-time calibration cover; (3) extracting the gold mark position in the perspective image and making a one-to-one correspondence with the gold mark's three-dimensional position; (4) Calculate and obtain the geometric parameters of the real-time perspective device; In step (2), after the real-time calibration cover is fixed to the shooting table, the displacement information of the perspective imaging device relative to the reference coordinate system of the shooting table is obtained, so as to calculate the three-dimensional coordinate sequence T = {xt1, xt2, ..., xt n }, xt n ∈R 3 ; In step (2), when taking a perspective view including the real-time calibration mask, the projection matrix M of the perspective imaging device measured in advance is recorded. 0 , the calculation formula is as follows: In step (3), the two-dimensional coordinates D = {xd1, xd2, ..., xd m },xd m ∈R 2 ; In step (3), extracting the two-dimensional coordinates of the gold mark in the real-time calibration cover in the perspective image includes the following steps: (3.1) According to the projection matrix M of the initial perspective imaging device 0 And the three-dimensional coordinate sequence T of the gold mark in the real-time calibration mask, calculate the expected projection sequence D of each gold mark point in the image 0 ={xd 0 1,xd 0 2,…,xd 0 m },xd 0 m ∈R 2 ; (3.2) For each expected coordinate, the image of the gold mark pixels is filtered out by the threshold method. Points that meet the gold mark projection characteristics are searched on the image of the gold mark pixels. The pixel coordinates around the point whose threshold is higher than the lower limit of the gold mark projection are weighted and summed with the pixel value as the weight to obtain the two-dimensional coordinates of the gold mark on the projection image, and obtain the two-dimensional coordinate sequence D.

2. The real-time geometric correction method for a perspective imaging device according to claim 1, characterized in that: In step (1), a double helix is ​​provided on the outer surface of the real-time calibration cover, and a plurality of gold labels are embedded in the double helix.

3. The real-time geometric correction method for a perspective imaging device according to claim 2, characterized in that: In step (1), the gold mark is a spherical metal marker, and the real-time calibration cover is made of a material with high transmittance to radiographic rays.

4. The real-time geometric correction method for a perspective imaging device according to claim 1, characterized in that: In step (4), the specific process of calculating the geometric parameters of the real-time perspective device is as follows: minimizing the sum of the distances between the projection points calculated by the projection matrix and the corresponding gold mark points on the projection image as the optimization target, iteratively optimizing and updating the projection matrix and the three-dimensional and two-dimensional coordinate mapping relationship of the gold mark, and finally obtaining the optimized projection matrix M.

Images