A method for calibrating the geometric parameters of an X-ray imaging system

Abstract

This invention discloses a geometric parameter calibration method for an X-ray imaging system, relating to the field of medical imaging equipment technology. This invention significantly improves the adaptability of portable X-ray equipment in scenarios involving rapid switching between pets of various sizes through an anatomical feature-driven parameter self-calibration mechanism. Utilizing the inherent proportional relationships of mammalian skeletons, such as the ratio of intervertebral space to rib width, geometric parameter calibration is completed in a single exposure, eliminating reliance on traditional physical calibration modules and avoiding proportional distortion of reconstructed images due to size differences. For live respiratory motion, respiratory frequency components are separated in the frequency domain of the projection data, and displacement errors are dynamically corrected through reverse compensation, effectively suppressing motion artifacts without the need for external control equipment. For the edge distortion problem of low-cost detectors, a polynomial distortion model is constructed based on the symmetry of 180° projection, combined with a robust iterative optimization algorithm, to automatically complete calibration during initial power-on or periodic scans.

Description

Technical Field

[0001] This invention relates to the field of medical imaging equipment technology, and in particular to a method for calibrating the geometric parameters of an X-ray imaging system. Background Technology

[0002] Portable X-ray imaging equipment in veterinary medicine needs to be adapted to rapid switching of images for animals with significant size differences, such as continuous examinations of small dogs to large dogs; the geometric parameters of the equipment, such as the distance between the X-ray source and detector and the center of rotation, directly affect the accuracy of the imaging scale; the current equipment relies on the initial calibration parameters, and when the size of the object being photographed changes beyond the preset range, the reconstructed image is prone to distortion of the anatomical structure scale, requiring manual recalibration and interrupting the workflow.

[0003] Current common adaptive calibration schemes match animal weight ranges by pre-setting multiple sets of parameter templates or using deep learning body shape recognition algorithms to predict parameter adjustment amounts. However, template switching requires operators to manually confirm the animal body shape classification, increasing the number of operation steps. Vision-based body shape recognition is easily affected by animal position and hair thickness, leading to incorrect parameter matching. For non-standard positions, such as a curled-up state, existing methods still rely on manual intervention to correct geometric parameters.

[0004] In addition, some solutions introduce dual-view rapid scanning to assist calibration, estimating body shape parameters through two orthogonal projections; however, this method requires additional radiation exposure, which does not conform to the dose control principles of veterinary medicine; and it is difficult for moving animals to maintain consistent posture in both views; another type of solution uses miniature markers on the edge of the detector to assist calibration, but the markers are easily obscured by limbs and fail in complex shooting positions; therefore, there is an urgent need for a geometric parameter calibration method for X-ray imaging systems to solve these problems. Summary of the Invention

[0005] In view of the aforementioned existing problems, the present invention is proposed.

[0006] This invention provides a method for calibrating the geometric parameters of an X-ray imaging system, which solves the problem of low efficiency and image distortion caused by traditional geometric calibration methods that rely on manual intervention and physical calibration objects when rapidly switching between multiple body types in pet medical treatment.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0008] This invention provides a method for calibrating the geometric parameters of an X-ray imaging system, comprising:

[0009] Step S1: Acquire a single projection image of the object under test;

[0010] Step S2: Identify the positions of at least two anatomical feature points in the projected image, wherein the anatomical feature points include paired feature parts in mammalian skeletons that have a fixed proportional relationship;

[0011] The identification of anatomical feature points in step S2 includes:

[0012] Perform bone edge enhancement processing on the projected image;

[0013] Connectivity analysis was used to extract the connection points between the spine and ribs.

[0014] Select paired feature parts with a spacing greater than 10mm;

[0015] Step S3: Calculate the actual distance ratio of the paired feature parts;

[0016] Step S4: When the actual distance ratio value deviates from the preset ratio threshold range, the distance parameter from the X-ray source to the detector is dynamically adjusted based on the deviation.

[0017] As a preferred embodiment of the geometric parameter calibration method for an X-ray imaging system according to the present invention, the paired feature regions include a combination of intervertebral space and the width of adjacent ribs.

[0018] As a preferred embodiment of the geometric parameter calibration method for an X-ray imaging system according to the present invention, the dynamic adjustment step in step S4 includes:

[0019] Divide the actual distance ratio by the preset standard ratio to obtain the ratio correction factor;

[0020] Multiply the distance from the current X-ray source to the detector by the aforementioned scaling factor to output the corrected geometric parameters.

[0021] As a preferred embodiment of the geometric parameter calibration method for an X-ray imaging system according to the present invention, in the dynamic adjustment process of step S4, the distance dynamic correction of the X-ray source detector based on the proportional deviation is performed, and the steps include:

[0022] Deviation calculation and judgment are performed. When the ratio value R of paired feature distance is detected, act Deviation from calibration standard R std And when |ΔR|>ε, calculate ΔR=R act -R std ,

[0023] Among them, R act R represents the ratio of paired feature distances measured from the projected image, dimensionless. std The standard proportional value obtained during the calibration phase is dimensionless; ΔR is the proportional deviation, dimensionless; ε is the upper limit of the allowable proportional deviation, dimensionless.

[0024] Constructing sensitivity coefficients related to uncertainty to suppress ranging noise:

[0025]

[0026] Where, k sd is the proportional distance sensitivity coefficient, dimensionless; c is the inverse variance scaling coefficient, in mm. -2 , σ D The variance of the distance measurement to the X-ray source detector in the previous cycle, in mm. 2 σ0 is the variance control threshold, in mm. 2 ;

[0027] Map the proportional deviation to a distance correction:

[0028]

[0029] Where η is the distance adjustment coefficient, which is dimensionless, and ζ is the near-zero denominator stability constant, which is taken as 10. -3 ;

[0030] Obtain the candidate correction distance:

[0031] D new =Dη,

[0032] Where D is the current distance to the X-ray source detector, in mm. new This is a one-time correction result, in mm;

[0033] Exponential smoothing iterations are performed to reduce oscillations caused by mechanical hysteresis. Exponential smoothing is introduced to suppress mechanical hysteresis.

[0034] D t+1 =D t +α(D new -D t ),

[0035] Among them, D t Distance of the X-ray source detector used in the current frame, in mm, D t+1 The distance after the update is in mm, and α is the smoothing factor.

[0036] As a preferred embodiment of the geometric parameter calibration method for an X-ray imaging system according to the present invention, it further includes a respiratory motion compensation step.

[0037] (a) Obtain the projection sequence in a continuous rotating scan;

[0038] (b) Perform frequency domain analysis on the projection sequence to extract the projection offset corresponding to the respiratory motion frequency band;

[0039] (c) Before image reconstruction, inject a compensation amount opposite to the projection offset into each projection angle;

[0040] In steps (a)-(c), the frequency domain analysis in step (b) includes:

[0041] The projection sequence is divided into overlapping segments according to the time window. Motion frequency band energy is detected independently for each segment, and the offsets of multiple segments are fused to generate a continuous compensation curve.

[0042] As a preferred embodiment of the geometric parameter calibration method for an X-ray imaging system according to the present invention, the frequency domain analysis includes performing a Fourier transform on the 180° projection sequence;

[0043] The motion component in the frequency band from 0.1Hz to 0.6Hz is extracted as the projection offset.

[0044] As a preferred embodiment of the geometric parameter calibration method for an X-ray imaging system according to the present invention, it further includes a detector distortion calibration step:

[0045] (d) Select the location of the first set of feature points in the 0° projected image;

[0046] (e) Match the positions of a second set of feature points that are symmetrical to the first set of feature points in the 180° projected image;

[0047] (f) Solve for the detector distortion correction parameters based on the positional deviation between the first group and the second group of feature points;

[0048] In step (d) of the detector distortion calibration, the selection of the first set of feature points must satisfy the following:

[0049] They are distributed in the central and edge regions of the detector;

[0050] The spacing between adjacent feature points shall not be less than 15% of the detector width;

[0051] Exclude occlusion points in the overlapping projection area.

[0052] As a preferred embodiment of the geometric parameter calibration method for an X-ray imaging system according to the present invention, the solution step includes:

[0053] Establish the mapping relationship between the positional deviation and the polynomial distortion model;

[0054] The coefficients of the polynomial distortion model are optimized through least squares iteration.

[0055] As a preferred embodiment of the geometric parameter calibration method for an X-ray imaging system according to the present invention, the method for solving the distortion model coefficients during the detector distortion calibration process is as follows:

[0056] Take the coordinates (u) of the Mth feature point in the 0° projected image. i ,v i Coordinates of the point symmetrical to 180° Calculate the ideal symmetric coordinates:

[0057]

[0058] Among them, u i ,v i The x and y coordinates of the i-th original pixel, in pixels (px). Let u' be the x and y coordinates of the i-th symmetric pixel, in pixels (px). i ,v′ i The coordinates of the ideal center of symmetry are in pixels (px).

[0059] With the detector center (u c ,v c Based on ), construct an nth-order bivariate polynomial:

[0060] u′ i =u i +∑ p+q≤n A pq (u i -u c ) p (v i -v c ) q ,

[0061] v′ i =v i +∑ p+q≤n B pq (u i -u c ) p (v i -v c ) q ,

[0062] Among them, A pq B pq The distortion coefficient is to be determined, and the unit is pixels (px). 1-p-q p and q are non-negative integer orders, and n is the highest order of the model;

[0063] Define the matrix form and the residual vector, denoted as:

[0064]

[0065] Where G is a 2M×2P design matrix, The transpose operator is 2P, which means that each of the two polynomials u and v has P coefficients, for a total of 2P unknowns. θ is the dimensionless column vector of distortion coefficients, Δ is the column vector of observation residuals, and px and P are the total number of coefficients.

[0066] Perform weighted least squares iterations, setting the robust weight matrix as follows:

[0067]

[0068] Iterative solution:

[0069]

[0070] r (k) =Δ-Gθ (k) ,

[0071] in, Let r be the weight of the j-th observation in the k-th round, dimensionless. j (k) For the corresponding residual, τ is the residual scaling constant; iterate to ||θ (k) -θ (k-1) The iteration stops when 2 < δ, where δ represents the convergence threshold and is dimensionless.

[0072] Output Then, anti-distortion mapping is performed on the entire projected coordinate system, and the root mean square of the corrected residuals is calculated:

[0073]

[0074] in, The coordinates are the corrected coordinates, in pixels (px). RMS is the root mean square error. If RMS ≤ β, the distortion correction is considered to have converged. β represents the upper limit of RMS, in pixels (px).

[0075] As a preferred embodiment of the geometric parameter calibration method for an X-ray imaging system according to the present invention, it further includes a triggering mechanism:

[0076] When the difference in weight between two consecutive photos of the subject exceeds a set threshold, steps S2-S4 are executed automatically.

[0077] The method for determining the set threshold includes: calculating the proportional deviation distribution of animals of different body sizes based on historical shooting data, using the distribution boundary value covering the preset probability interval as the trigger threshold, and dynamically updating the distribution boundary value as new shooting data is added;

[0078] If there is no weight data, the proportional deviation value in step S3 will be used as the trigger.

[0079] The beneficial effects of this invention are as follows: This invention significantly improves the adaptability of portable X-ray equipment in scenarios of rapid switching between pets of different body sizes through a parameter self-calibration mechanism driven by anatomical features; by utilizing the inherent proportional relationship of mammalian skeletons, such as the ratio of intervertebral space to rib width, geometric parameter calibration can be completed in a single exposure, eliminating the dependence on traditional physical calibration modules and avoiding proportional distortion of reconstructed images caused by differences in body size.

[0080] This invention addresses the issue of in vivo respiratory motion by innovatively separating respiratory frequency components in the frequency domain of projected data. It dynamically corrects displacement errors through reverse compensation, effectively suppressing motion artifacts without the need for external control equipment. For edge distortion issues in low-cost detectors, a polynomial distortion model is constructed based on the symmetry of 180° projection, combined with a robust iterative optimization algorithm, automatically completing calibration during initial startup or periodic scans. Furthermore, an intelligent triggering mechanism is introduced, autonomously initiating the correction process when body shape switching or historical deviations exceed a threshold, reducing the need for manual intervention. The overall solution simplifies the operation process while maintaining accuracy, making it particularly suitable for scenarios requiring high-frequency, multi-type imaging, such as veterinary hospitals. Attached Figure Description

[0081] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0082] Figure 1 This is a flowchart illustrating the geometric parameter calibration method for the X-ray imaging system in Example 1. Detailed Implementation

[0083] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0084] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0085] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0086] Example 1, referring to Figure 1 This embodiment provides a method for calibrating the geometric parameters of an X-ray imaging system, including the following steps:

[0087] Step S1: Acquire a single projection image of the object under test;

[0088] Step S2: Identify the location of at least two anatomical feature points in the projected image. The anatomical feature points include paired feature parts in the mammalian skeleton that have a fixed proportional relationship. The paired feature parts include the combination of the intervertebral space and the width of the adjacent rib.

[0089] Step S2 involves identifying anatomical feature points, including:

[0090] Perform bone edge enhancement processing on the projected image;

[0091] Connectivity analysis was used to extract the connection points between the spine and ribs.

[0092] Select paired feature parts with a spacing greater than 10mm;

[0093] Step S3: Calculate the actual distance ratio of paired feature parts;

[0094] Step S4: When there is a deviation between the actual distance ratio value and the preset ratio threshold range, the distance parameter from the X-ray source to the detector is dynamically adjusted based on the deviation.

[0095] The dynamic adjustment steps in step S4 include:

[0096] Divide the actual distance ratio by the preset standard ratio to obtain the ratio correction factor;

[0097] Multiply the distance from the current X-ray source to the detector by the scaling factor to output the corrected geometric parameters;

[0098] During the dynamic adjustment process in step S4, the X-ray source detector performs dynamic distance correction based on the proportional deviation. The steps include:

[0099] Deviation calculation and judgment are performed. When the ratio value R of paired feature distance is detected, act Deviation from calibration standard R std And when |ΔR|>ε, calculate ΔR=R act -R std ,

[0100] Among them, R act R represents the ratio of paired feature distances measured from the projected image, dimensionless. std The standard proportional value obtained during the calibration phase is dimensionless; ΔR is the proportional deviation, dimensionless; ε is the upper limit of the allowable proportional deviation, dimensionless.

[0101] Constructing sensitivity coefficients related to uncertainty to suppress ranging noise:

[0102]

[0103] Where, k sd is the proportional distance sensitivity coefficient, dimensionless; c is the inverse variance scaling coefficient, in mm. -2 , σ D The variance of the distance measurement to the X-ray source detector in the previous cycle, in mm. 2 σ0 is the variance control threshold, in mm. 2 ;

[0104] Map the proportional deviation to a distance correction:

[0105]

[0106] Where η is the distance adjustment coefficient, which is dimensionless, and ζ is the near-zero denominator stability constant, which is taken as 10. -3 ;

[0107] Obtain the candidate correction distance:

[0108] D new =Dη,

[0109] Where D is the current distance to the X-ray source detector, in mm. new This is a one-time correction result, in mm;

[0110] Exponential smoothing iterations are performed to reduce oscillations caused by mechanical hysteresis. Exponential smoothing is introduced to suppress mechanical hysteresis.

[0111] D t+1 =D t +α(D new -D t ),

[0112] Among them, D t Distance of the X-ray source detector used in the current frame, in mm, D t+1 The distance is the updated distance in mm, and α is the smoothing factor.

[0113] Specifically, the correction process is driven by proportional deviation and uses the logistic curve to adaptively adjust the sensitivity, minimizing the risk of overcorrection in high-noise scenarios. Exponential smoothing compensates for the slow start and stop characteristics of mechanical movement, significantly reducing oscillations.

[0114] It also includes respiratory movement compensation steps:

[0115] (a) Obtain the projection sequence in a continuous rotating scan;

[0116] (b) Perform frequency domain analysis on the projection sequence to extract the projection offset corresponding to the respiratory motion frequency band;

[0117] (c) Before image reconstruction, inject a compensation amount opposite to the projection offset into each projection angle;

[0118] In steps (a)-(c), the frequency domain analysis in step (b) includes:

[0119] The projection sequence is divided into overlapping segments according to the time window. Motion frequency band energy is detected independently for each segment. The offsets of multiple segments are fused to generate a continuous compensation curve.

[0120] Frequency domain analysis includes:

[0121] Perform a Fourier transform on the 180° projection sequence;

[0122] Motion components in the 0.1Hz to 0.6Hz frequency band are extracted as projection offsets, preferably 0.2Hz-0.4Hz. The frequency can be adjusted as needed; here it is set to the average breathing frequency of the example cat and dog.

[0123] It also includes the detector distortion calibration step:

[0124] (d) Select the location of the first set of feature points in the 0° projected image;

[0125] (e) Match the positions of the second set of feature points that are symmetrical to the first set of feature points in the 180° projected image;

[0126] (f) Solve for the detector distortion correction parameters based on the positional deviation between the first and second groups of feature points;

[0127] In step (d) of detector distortion calibration, the selection of the first set of feature points must satisfy the following:

[0128] They are distributed in the central and edge regions of the detector;

[0129] The spacing between adjacent feature points shall not be less than 15% of the detector width;

[0130] Exclude occlusion points in the overlapping projection area;

[0131] The solution steps include:

[0132] Establish the mapping relationship between positional deviation and polynomial distortion model;

[0133] The coefficients of the polynomial distortion model are optimized through least squares iteration;

[0134] The method for solving the distortion model coefficients during detector distortion calibration is as follows:

[0135] Take the coordinates (u) of the Mth feature point in the 0° projected image.i ,v i Coordinates of the point symmetrical to 180° Calculate the ideal symmetric coordinates:

[0136]

[0137] Among them, u i ,v i The x and y coordinates of the i-th original pixel, in pixels (px). Let u' be the x and y coordinates of the i-th symmetric pixel, in pixels (px). i ,v′ i The coordinates of the ideal center of symmetry are in pixels (px).

[0138] With the detector center (u c ,v c Based on ), construct an nth-order bivariate polynomial:

[0139] u′ i =u i +∑ p+q≤n A pq (u i -u c ) p (v i -v c ) q ,

[0140] v′ i =v i +∑ p+q≤n B pq (u i -u c ) p (v i -v c ) q ,

[0141] Among them, A pq B pq The distortion coefficient is to be determined, and the unit is pixels (px). 1-p-q p and q are non-negative integer orders, and n is the highest order of the model;

[0142] Define the matrix form and the residual vector, denoted as:

[0143]

[0144] Where G is a 2M×2P design matrix, The transpose operator is 2P, which means that each of the two polynomials u and v has P coefficients, for a total of 2P unknowns. θ is the dimensionless column vector of distortion coefficients, Δ is the column vector of observation residuals, and px and P are the total number of coefficients.

[0145] Perform weighted least squares iterations, setting the robust weight matrix as follows:

[0146]

[0147] Iterative solution:

[0148]

[0149] r (k) =Δ-Gθ (k) ,

[0150] in, Let r be the weight of the j-th observation in the k-th round, dimensionless. j (k) For the corresponding residual, τ is the residual scaling constant; iterate to ||θ (k) -θ (k-1) The iteration stops when 2 < δ, where δ represents the convergence threshold and is dimensionless.

[0151] Output Then, anti-distortion mapping is performed on the entire projected coordinate system, and the root mean square of the corrected residuals is calculated:

[0152]

[0153] in, The coordinates are the corrected coordinates, in pixels (px), and RMS is the root mean square error. If RMS ≤ β, the distortion correction is considered to have converged, where β represents the upper limit of RMS, in pixels (px).

[0154] Specifically, this method transforms the detector distortion problem into a bivariate multinomial regression through symmetric constraints. The robust weight function adaptively suppresses abnormal matching caused by occlusion and artifacts, avoiding extreme value bias. The iteration matrix is ​​reweighted in each round, and the residual decreases rapidly, stabilizing within three rounds.

[0155] It also includes a triggering mechanism: when the difference in weight between two consecutive photos of the subject exceeds a set threshold, steps S2-S4 are automatically executed;

[0156] The method for determining the threshold includes: calculating the proportional deviation distribution of animals of different body sizes based on historical shooting data, using the distribution boundary value covering the preset probability interval as the trigger threshold, and dynamically updating the distribution boundary value as new shooting data is added;

[0157] If there is no weight data, the actual measurement error in step S3 will be used as the trigger.

[0158] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for calibrating geometric parameters of an X-ray imaging system, characterized in that, The method comprises the following steps: Step S1, acquiring a single projection image of a subject to be measured; Step S2, identifying the positions of at least two anatomical feature points in the projection image, the anatomical feature points including a pair of feature parts in mammalian bones having a fixed proportional relationship; The step S2 of identifying the anatomical feature points comprises: Performing a bone edge enhancement process on the projection image; Extracting the connection points of the spine and the ribs through connected domain analysis; Screening the pair of feature parts with a distance greater than 10 mm; Step S3, calculating the actual distance proportional value of the pair of feature parts; Step S4, when the actual distance proportional value deviates from a preset proportional threshold range, dynamically adjusting the distance parameter of the ray source to the detector based on the deviation.

2. The method of calibrating geometric parameters of an X-ray imaging system of claim 1, wherein, The pair of feature parts includes the combination of the spinal gap and the width of the adjacent ribs.

3. The method of claim 1, wherein the step of determining the geometric parameters of the X-ray imaging system is performed by using a calibration phantom. The step of dynamic adjustment in step S4 comprises: Dividing the actual distance proportional value by a preset standard proportional value to obtain a proportional correction factor; Multiplying the distance of the current ray source to the detector by the proportional correction factor to output the corrected geometric parameter.

4. The method of claim 3, wherein the step of determining the geometric parameters of the X-ray imaging system is performed by using a calibration phantom. In the dynamic adjustment process of step S4, the ray source detector is dynamically corrected based on the proportional deviation, and the steps comprise: The deviation calculation and determination are performed, and when the distance ratio value of the pair of features is detected deviates from the calibration standard and , the following is calculated , wherein, is a pair-wise feature distance scale value measured for the projected image, dimensionless, is a standard scale value obtained in the calibration phase, dimensionless, is a scale deviation, dimensionless, is an upper limit of scale deviation allowed, dimensionless; Constructing a sensitivity coefficient related to the uncertainty to suppress the ranging noise: ， wherein, is a proportional distance sensitivity coefficient, dimensionless, is an inverse-variance scaling coefficient, dimensionless , is a previous cycle ray source detector distance measurement variance, dimensionless , is a variance enable threshold, dimensionless ; Mapping the proportional deviation to the distance correction amount: ， wherein, is a distance adjustment factor, dimensionless, is a near-zero denominator stabilizing constant, taken as ; Obtaining a candidate correction distance: ， wherein, is the current ray source detector distance in millimeters, mm, is the one-time correction result in millimeters, mm; Performing exponential smoothing iteration to reduce the oscillation caused by mechanical hysteresis, and introducing exponential smoothing to suppress mechanical hysteresis: ， wherein, is the ray source detector distance used for the current frame in millimeters, mm, is the updated distance in millimeters, mm, is a smoothing factor.

5. The method of claim 1, wherein the step of determining the geometric parameters of the X-ray imaging system is performed by using a calibration phantom. Further comprising a respiratory motion compensation step: (a) acquiring a projection sequence in a continuous rotation scan; (b) performing frequency domain analysis on the projection sequence to extract a projection offset corresponding to a respiratory motion frequency band; (c) injecting a compensation amount opposite to the projection offset to each projection angle before image reconstruction; In steps (a)-(c), the frequency domain analysis in step (b) comprises: Dividing the projection sequence into overlapping sub-sections according to the time window, independently performing motion frequency band energy detection on each sub-section, and fusing the offset amounts of multiple sub-sections to generate a continuous compensation curve.

6. The method of calibrating the geometry of an x-ray imaging system of claim 5, wherein, The frequency domain analysis comprises: performing Fourier transform on the 180° projection sequence; Extracting the motion component in the 0.1 Hz to 0.6 Hz frequency band as the projection offset.

7. The method of calibrating the geometry of an x-ray imaging system of claim 1, wherein, Further comprising a detector distortion calibration step: (d) selecting a first group of feature point positions in the 0° projection image; (e) matching a second group of feature point positions symmetrical to the first group in the 180° projection image; (f) solving a detector distortion correction parameter according to the position deviation of the first group and the second group of feature points; The selection of the first group of feature points in step (d) of the detector distortion calibration needs to meet: Distributed in the central and edge regions of the detector; The distance between adjacent feature points is not less than 15% of the width of the detector; Excluding the occlusion points in the projection overlap region.

8. The method of claim 7, wherein the step of determining the geometric parameters of the X-ray imaging system is performed by using a calibration phantom having a known shape and size. The solving step comprises: Establishing a mapping relationship between the position deviation and a polynomial distortion model; Optimizing the coefficients of the polynomial distortion model through least squares iteration.

9. The geometric parameter calibration method for an X-ray imaging system as described in claim 8, characterized in that, In the detector distortion calibration process, the way of solving the distortion model coefficients is: Take the 0° projection image of the first feature point coordinates and 180° symmetry point coordinates , calculate the ideal symmetry coordinates: ， wherein, is the horizontal coordinate of the first original pixel, is the vertical coordinate of the first original pixel, , is the horizontal coordinate of the first symmetric pixel, is the vertical coordinate of the first symmetric pixel, , is the ideal symmetric center coordinate, ; with the center of the detector constructed order bivariate polynomial: ， ， wherein, is the distortion coefficient to be determined, in units of , is a non-negative integer order, is the highest order of the model; Define the matrix form and the residual vector as follows: ， ， wherein, is design matrix, is the transpose symbol, u, v two polynomials each containing P coefficients, a total of 2P unknowns, is a distortion coefficient column vector, dimensionless, is an observation residual column vector, , is the total number of coefficients ; Perform weighted least squares iteration, and set the robust weight matrix as follows: ， ， Iteratively solve: ， ， wherein, is the weight of the th observation in the th round, dimensionless, is the corresponding residual, is a residual scale constant; iterating until convergence is reached, denotes an iteration convergence threshold, dimensionless; Output After, the inverse distortion mapping is performed on the whole projection coordinates, and the corrected residual root mean square is calculated: ， wherein, is the corrected back coordinate, unit is , is the root mean square error; if convergence of distortion correction is considered, denotes the upper limit of the RMS, unit is px.

10. The method of calibrating the geometry of an x-ray imaging system of claim 1, wherein, Further comprising a triggering mechanism: When the weight difference of the object to be measured exceeds the set threshold value for two consecutive times, steps S2-S4 are automatically executed; The determination method of the set threshold value comprises: calculating the proportion deviation distribution of animals of different body types according to historical shooting data, taking the distribution boundary value covering the preset probability interval as the trigger threshold value, and dynamically updating the distribution boundary value with new shooting data; If there is no weight data, the proportion deviation value in step S3 is used as the trigger.

Images