Large heterogeneous component CT imaging method and device based on bidirectional cooperative scanning

Abstract

The invention discloses a large heterogeneous component CT imaging method and device based on bidirectional cooperative scanning. According to the method, a rectangular coordinate system with the long axis of the battery pack as the X axis and the thickness direction as the Z axis is established, scanning parameters are initialized, and bidirectional translation scanning including forward scanning, reverse scanning and dip angle iteration adjustment is executed; monitoring the position deviation of the radiation source / detector in real time, correcting the deviation, and carrying out geometric calibration compensation; and finally, carrying out image reconstruction by adopting an ordered subset combined algebraic reconstruction technology, modeling a scanning process into a linear equation set, dividing projection data into ordered subsets, and realizing image reconstruction through an iterative updating strategy. According to the method, the scanning track planning and motion control strategy is innovated, efficient collection of multi-dimensional projection data is achieved within the limited mechanical stroke, meanwhile, the reconstruction algorithm is optimized in combination with the battery pack material characteristics, and a non-destructive testing solution with the high cost performance is provided for power battery quality control.

Description

Technical Field

[0001] The present invention relates to the field of CT imaging technology, and in particular to a CT imaging method and device for large heterogeneous components based on bidirectional collaborative scanning. Background Art

[0002] Computed Tomography (CT) technology achieves three-dimensional reconstruction of the internal structure of an object through multi-angle projection data, and has important applications in industrial inspection. However, traditional CT needs to meet the conditions of rotation scanning of more than 180°. For rectangular structures with large length, width and thickness (such as new energy vehicle battery packs), their geometric dimensions far exceed the carrying range of conventional CT turntables, resulting in limited sample rotation and inability to obtain complete projection data. In addition, when the rays penetrate the long axis direction of the battery pack, the long path will cause severe attenuation, and the projection signal-to-noise ratio will drop significantly, affecting the accuracy of defect detection.

[0003] As an alternative to limited-angle scanning, Computed Laminography (CL) technology usually uses a non-coaxial motion mode between the X-ray source, detector and sample, and is suitable for layered inspection of plate-like or thin-walled components. Existing CL systems are mostly designed for flat structures (such as circuit boards and wings) whose length and width are much greater than their thickness. Their scanning trajectories (such as linear and C-arm types) reduce the risk of geometric occlusion by constraining the inclination angle between the ray and the sample normal. However, the three-dimensional scale of the rectangular battery pack is large and the internal cells are densely stacked. The existing CL technology faces the following limitations: 1) The scanning trajectory with a fixed inclination angle is difficult to adapt to the composite dimensions of the battery pack in terms of length, width, and thickness, which can easily lead to insufficient projection coverage in local areas; 2) Traditional CL systems rely on high-precision mechanical structures (such as arc guides, multi-axis linkage turntables), which have high manufacturing costs and insufficient flexibility, making it difficult to meet the needs of large field of view and rapid detection at the production line level; 3) The existing layered reconstruction algorithm is optimized for thin-layer structures, and has poor adaptability to the differences in attenuation characteristics of multi-layer heterogeneous materials (metal shell, diaphragm, electrolyte) in the battery pack, and is prone to introducing artifacts. Summary of the invention

[0004] Based on this, the embodiment of the present application provides a large heterogeneous component CT imaging method and device based on bidirectional collaborative scanning. For large rectangular battery packs, innovative scanning trajectory planning and motion control strategies are used to achieve efficient acquisition of multi-dimensional projection data within a limited mechanical stroke. At the same time, the reconstruction algorithm is optimized in combination with the material characteristics of the battery pack, providing a cost-effective non-destructive testing solution for power battery quality control.

[0005] In a first aspect, a method for CT imaging of a large heterogeneous component based on bidirectional collaborative scanning is provided, the method comprising:

[0006] S1. Establish a rectangular coordinate system with the long axis of the battery pack as the X-axis and the thickness direction as the Z-axis, and initialize the scanning parameters. Among them, it includes adjusting the angle between the line connecting the ray source focus and the detector center and the normal of the battery pack, and fixing the length of the line.

[0007] S2. Perform bidirectional translation scanning, including forward scanning and reverse scanning, and inclination angle iterative adjustment.

[0008] S3. Real-time monitor the position deviation of the ray source / detector and perform correction. After each angle adjustment, perform geometric calibration compensation, and calculate the Z-axis step size to ensure the integrity of the interlayer information.

[0009] S4. Use the ordered subset combined algebraic reconstruction technique for image reconstruction. Model the scanning process as a system of linear equations, divide the projection data into ordered subsets, and achieve image reconstruction through an iterative update strategy.

[0010] Optionally, performing bidirectional translation scanning, including forward scanning and reverse scanning, and inclination angle iterative adjustment, specifically includes:

[0011] S21 Forward scanning: The ray source system uniformly translates along the positive X-axis direction from the initial left end position, and at the same time, the detector system synchronously moves right along the mirror image trajectory, and the two maintain a constant speed and relative distance. During the translation, the X-ray beam covers the cross-section of the battery pack with a fan angle, and the detector continuously acquires projection data at the frame rate.

[0012] S22 Reverse scanning: When the ray source reaches the right end position, the system immediately switches the movement direction, and the ray source and the detector translate in the opposite direction to the initial left end position at the same speed. During the reverse scanning, the detector synchronously acquires the reverse projection data set, which forms a complementary projection coverage with the forward data to eliminate the one-way scanning blind area.

[0013] S23 Inclination angle iterative adjustment: After completing the forward / reverse scanning of the current angle, drive the ray source and the detector to synchronously deflect around the center of the battery pack, update the connection inclination angle to a new angle; repeat the above forward / reverse scanning process until the entire preset inclination angle range is covered.

[0014] Optionally, the initialization of the scanning parameters further includes setting the initial positions of the ray source and the detector so that the line connecting the ray source focus and the detector center is perpendicular to the initial scanning plane of the battery pack.

[0015] Optionally, the real-time monitoring of the position deviation of the ray source / detector and performing correction, and performing geometric calibration compensation after each angle adjustment, and calculating the Z-axis step size to ensure the integrity of the interlayer information, includes:

[0016] Real-time monitor the position deviation of the ray source / detector through a grating ruler. If the lateral offset is greater than the threshold, immediately trigger the correction algorithm to adjust the torque of the servo motor.

[0017] After each angle adjustment, insert a calibration phantom, invert the system geometric parameters based on the projection coordinates, and correct the projection data to a unified coordinate system;

[0018] Automatically calculate the Z-axis step size according to the layer spacing of the battery pack to ensure that the overlapping area of adjacent scan layers is ≥ 20%.

[0019] Optionally, use the ordered subset joint algebraic reconstruction technique for image reconstruction, specifically including:

[0020] Divide the total projection data into multiple non-overlapping ordered subsets, each subset contains several rays, and the subset division is optimized according to the projection angle or spatial distribution;

[0021] In each iteration, update each subset and suppress iteration oscillation through a dynamic relaxation factor;

[0022] Perform forward projection calculation, residual calculation, correction accumulation, and voxel update until the convergence condition is met to obtain the reconstructed image.

[0023] Optionally, model the CT scanning process as the following linear equation system:

[0024] AX = b

[0025] Where:

[0026] b = (b1, b2, …, b M ) ∈ R M Is the projection data vector, and M is the total number of rays;

[0027] X = (X1, X2, …, X N ) ∈ R N Is the image vector to be reconstructed, and N is the total number of voxels;

[0028] A = (a mn ) ∈ R M×N Is the system matrix, and the element a mn Represents the path length of the m-th ray passing through the n-th voxel.

[0029] Optionally, divide the total projection data into multiple non-overlapping ordered subsets, including:

[0030] Divide the total projection data into T non-overlapping subsets S1, S2, …, S T Satisfy:

[0031]

[0032] Each subset S T Contains The rays, and the subset division is optimized according to the projection angle or spatial distribution, including allocating adjacent angular intervals to different subsets.

[0033] Optionally, each subset is updated to suppress iterative oscillation by a dynamic relaxation factor, including:

[0034] In the k-th iteration, the subset S [k] ([k]=(k mod T)+1) is updated as follows:

[0035]

[0036] where λ k ∈(0,2) is the dynamic relaxation factor for suppressing iterative oscillation; the denominator term is the total weighted contribution of the subset S [k] to the voxel n.

[0037] Optionally, forward projection calculation, residual calculation, correction accumulation, and voxel update are performed until the convergence condition is met to obtain the reconstructed image, specifically including:

[0038] Perform forward projection calculation through the formula ;

[0039] Perform residual calculation through the formula ;

[0040] Perform correction accumulation calculation through the formula ;

[0041] Perform voxel update calculation through the formula ;

[0042] In a second aspect, a CT imaging device for large heterogeneous components based on bidirectional collaborative scanning is provided, and the device includes:

[0043] An initialization module for establishing a rectangular coordinate system with the long axis of the battery pack as the X-axis and the thickness direction as the Z-axis, and initializing the scanning parameters; among them, it includes adjusting the angle between the line connecting the ray source focus and the detector center and the normal of the battery pack, and fixing the length of the line;

[0044] A scanning module for performing bidirectional translation scanning, including forward scanning and reverse scanning, and inclination iterative adjustment;

[0045] A compensation module for real-time monitoring of the ray source / detector position deviation and correction, performing geometric calibration compensation after each angle adjustment, and calculating the Z-axis step size to ensure complete interlayer information;

[0046] An image reconstruction module, which is used to perform image reconstruction by using the ordered subset combined algebraic reconstruction technique, models the scanning process as a system of linear equations, divides the projection data into ordered subsets, and realizes image reconstruction through an iterative update strategy.

[0047] Through the motion trajectory technology of two-way collaborative scanning combined with the OS-SART iterative reconstruction algorithm, the comprehensive performance of CT detection of large cuboid battery packs is significantly improved in the present invention: by using mirror translation and dynamic geometric calibration, the scanning time is greatly shortened, and the blind area is eliminated through complementary projection, significantly improving the integrity of interlayer coverage; combined with the material attenuation weight compensation model, the artifacts at the interfaces of heterogeneous materials are effectively suppressed, and the sensitivity of defect detection is improved; the modular mechanical design and subset division strategy optimize the equipment volume and memory occupancy, reduce the iterative complexity and maintenance cost, and while ensuring the high-precision detection of key internal defects of the battery pack, meet the requirements of the high-efficiency production line rhythm. This solution has good scalability and can be adapted to the detection scenarios of complex heterogeneous components such as aviation composite materials. Description of the Drawings

[0048] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only exemplary, and for those of ordinary skill in the art, without creative efforts, other implementation drawings can also be obtained according to the provided drawings.

[0049] Figure 1 It is a step flow chart of a CT imaging method for large heterogeneous components based on two-way collaborative scanning provided by an embodiment of the present application;

[0050] Figure 2 It is a mechanical design drawing of the implementation scenario provided by an embodiment of the present application;

[0051] Figure 3 It is a forward scanning trajectory diagram provided by an embodiment of the present application;

[0052] Figure 4 It is a reverse scanning trajectory diagram provided by an embodiment of the present application;

[0053] Figure 5 It is a reconstruction result diagram provided by an embodiment of the present application. Detailed Embodiments

[0054] In order to make the purpose, technical solutions and advantages of the present application clearer, the following will further elaborate on the present application in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0055] In the description of the present invention, the terms "comprising", "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units need not be limited to those steps or units explicitly listed, but may also include other steps or units inherent to these processes, methods, products or devices that are not explicitly listed, or steps or units added based on further optimization schemes conceived in the present invention.

[0056] Computed Tomography (CT) technology realizes three-dimensional reconstruction of the internal structure of an object through multi-angle projection data and has important applications in industrial inspection. However, traditional CT requires a rotation scanning condition of more than 180°. For a cuboid structure with large length, width and thickness (such as a new energy vehicle battery pack), its geometric dimensions far exceed the load-bearing range of a conventional CT turntable, resulting in limited rotation of the sample and inability to obtain complete projection data. In addition, when the ray penetrates the long axis direction of the battery pack, the excessively long path will cause serious attenuation and a significant decrease in the projection signal-to-noise ratio, affecting the defect detection accuracy.

[0057] Computed Laminography (CL) technology, as an alternative to limited-angle scanning, usually adopts a non-coaxial motion mode between the X-ray source, the detector and the sample and is suitable for layer-by-layer detection of plate-like or thin-walled components. Existing CL systems are mostly designed for flat structures with length and width much larger than thickness (such as circuit boards, wings), and their scanning trajectories (such as linear type, C-arm type) reduce the risk of geometric occlusion by constraining the inclination angle between the ray and the sample normal. However, for a cuboid battery pack with large three-dimensional dimensions and densely stacked internal battery cells, the existing CL technology faces the following limitations: 1) The scanning trajectory with a fixed inclination angle is difficult to adapt to the composite dimensions in the length, width and thickness directions of the battery pack, easily resulting in insufficient projection coverage in local areas; 2) Traditional CL systems rely on high-precision mechanical structures (such as arc-shaped guide rails, multi-axis linkage turntables), with high manufacturing costs and insufficient flexibility, making it difficult to meet the requirements of large field-of-view and rapid detection at the production line level; 3) Existing layer-by-layer reconstruction algorithms are optimized for thin-layer structures and have poor adaptability to the attenuation characteristic differences of multi-layer heterogeneous materials (metal shell, separator, electrolyte) in the battery pack, easily introducing artifacts.

[0058] In the existing technical solution 1, the second-generation scan adopts a translation-rotation (TR) mode, that is, the ray source and the detector system need to perform two motions of translation and rotation during data acquisition. The specific process includes:

[0059] Translation acquisition: The ray source and the detector translate synchronously along a fixed orbit to obtain projection data at a certain angle;

[0060] Rotation adjustment: After completing the translation, the entire system rotates a certain angle (such as 1° to 2°), and repeats the translation acquisition steps until the entire scanning angle range is covered (usually more than 180°).

[0061] Low scanning efficiency: The alternating translation-rotation motion causes a single scan to take several hours, which is only 1 / 10 to 1 / 5 of the scanning efficiency of the third generation;

[0062] Dynamic parameters are not adjustable: the translation distance and rotation angle are fixed, and the scanning trajectory cannot be dynamically optimized according to the workpiece size, resulting in redundant or missing projection data in local areas;

[0063] High mechanical complexity: Precision guide rails and rotary tables are required for linkage control, the equipment is bulky and has high maintenance costs.

[0064] In the existing technical solution 2, the third generation scanning adopts the full-inclusion rotation (RO) mode, where the radiation source and the detector form a fan beam and rotate 360° around the workpiece, and complete projection data can be obtained without translation. Technical features include:

[0065] Large fan angle coverage: The detector array covers the entire workpiece, and data collection can be completed in a single rotation;

[0066] Fast imaging: scanning time is shortened to minutes, suitable for rapid detection of production lines;

[0067] Spiral scanning extension: Combined with the axial movement of the workpiece, spiral trajectory scanning is realized to improve the Z-axis resolution.

[0068] However, the disadvantages of the prior art solution 2 include:

[0069] Significant ring artifacts: Detector array calibration errors can easily lead to concentric circle artifacts in the reconstructed image, which require complex algorithms to suppress;

[0070] Poor adaptability to large workpieces: When the workpiece size exceeds the coverage of the fan beam, multiple stitching scans are required, resulting in poor data consistency;

[0071] Insufficient mechanical flexibility: Fixed C-arm or rotating stage design cannot dynamically adjust the source-detector distance (SD), and the field of view is limited;

[0072] Based on the deficiencies of the prior art described in the background technology, the present invention aims to solve the following technical problems:

[0073] The problem of limited full-area scanning of large-size rectangular battery packs: The fixed-angle scanning trajectory of the traditional CL system is difficult to adapt to the composite dimensions of the length, width, and thickness of the battery pack, resulting in the loss of projection data in local areas, affecting the integrity of reconstruction;

[0074] Mechanical complexity and cost contradiction problem: Existing CL relies on high-precision arc-shaped guide rails or multi-axis linkage turntables, resulting in high manufacturing costs and insufficient flexibility, making it difficult to meet the rapid detection requirements at the production line level;

[0075] Artifact problem in heterogeneous material reconstruction: Traditional layer-by-layer reconstruction algorithms have insufficient adaptability to the attenuation differences of multiple layers of materials (metal shell, polymer separator, electrolyte) in the battery pack, resulting in interlayer artifacts interfering with defect identification.

[0076] Please refer to Figure 1 , which shows a flowchart of a CT imaging method for large heterogeneous components based on two-way collaborative scanning provided by an embodiment of the present application, and may include the following steps:

[0077] S1. Establish a right-handed coordinate system with the long axis of the battery pack as the X-axis and the thickness direction as the Z-axis, and initialize the scanning parameters; wherein, it includes adjusting the angle between the line connecting the X-ray source focus and the detector center and the normal line of the battery pack, and fixing the length of the line;

[0078] S2. Perform two-way translational scanning, including forward scanning and reverse scanning, and inclination angle iterative adjustment.

[0079] Specifically, this step includes:

[0080] S21 Forward scanning: The X-ray source system uniformly translates along the positive X-axis direction from the initial left end position, and at the same time, the detector system synchronously moves right along the mirror image trajectory, and the two maintain a constant speed and relative distance; during the translation, the X-ray beam covers the cross-section of the battery pack with a fan angle, and the detector continuously acquires projection data at the frame rate;

[0081] S22 Reverse scanning: When the X-ray source reaches the right end position, the system immediately switches the movement direction, and the X-ray source and the detector translate in the opposite direction to the initial left end position at the same speed; during the reverse scanning, the detector synchronously acquires the reverse projection data set, which forms a complementary projection coverage with the forward data to eliminate the blind area of one-way scanning;

[0082] S23 Inclination angle iterative adjustment: After completing the forward / reverse scanning of the current angle, drive the X-ray source and the detector to synchronously deflect around the center of the battery pack, update the inclination angle of the line to a new angle; repeat the above forward / reverse scanning process until the entire preset inclination angle range is covered.

[0083] S3. Real-time monitor the position deviation of the X-ray source / detector and perform deviation correction. Geometric calibration compensation is performed after each angle adjustment, and the Z-axis step size is calculated to ensure the integrity of interlayer information.

[0084] Specifically, the position deviation of the X-ray source / detector is real-time monitored through a grating scale. If the lateral offset is greater than the threshold, the deviation correction algorithm is immediately triggered to adjust the servo motor torque;

[0085] After each angle adjustment, insert a calibration phantom, invert the system geometric parameters based on the projection coordinates, and correct the projection data to a unified coordinate system.

[0086] Automatically calculate the Z-axis step size according to the layer spacing of the battery pack to ensure that the overlapping area of adjacent scan layers is ≥ 20%.

[0087] S4. Use the ordered subset combined algebraic reconstruction technique for image reconstruction. Model the scanning process as a system of linear equations, divide the projection data into ordered subsets, and achieve image reconstruction through an iterative update strategy.

[0088] Specifically, divide the total projection data into multiple non-overlapping ordered subsets. Each subset contains several rays, and the subset division is optimized according to the projection angle or spatial distribution.

[0089] In each iteration, update each subset and suppress iterative oscillation through a dynamic relaxation factor.

[0090] Perform forward projection calculation, residual calculation, correction accumulation, and voxel update until the convergence condition is met to obtain the reconstructed image.

[0091] As Figure 2 shown, a schematic diagram of the application scenario of the above method is given. Establish a right-handed coordinate system with the long axis of the battery pack as the X-axis and the thickness direction as the Z-axis. Initialize the scanning parameters:

[0092] Inclination setting: Adjust the angle θ (θ ∈ [θ min , θ max ) between the line SD connecting the ray source focus S and the detector center D and the normal (Z-axis) of the battery pack to ensure that the ray penetration path avoids the long-axis occlusion area of the battery pack.

[0093] Synchronous motion baseline: Fix the length L of the SD line (L = SD) and maintain the conjugate geometric relationship between the ray source and the detector.

[0094] The following gives a specific implementation process of the bidirectional translation scanning execution flow of this method:

[0095] (1) Forward scanning as Figure 3 shown (left → right):

[0096] The ray source system moves uniformly along the positive X-axis direction from the initial left end position S0, and at the same time, the detector system moves synchronously to the right along the mirror trajectory, and both maintain a constant speed v and a relative distance L.

[0097] During the translation process, the X-ray beam covers the cross-section of the battery pack with a fan angle φ, and the detector continuously acquires projection data at a frame rate f

[0098] (2) Reverse scanning as Figure 4As shown (right → left):

[0099] When the radiation source reaches the right end position S max the system immediately switches the movement direction, and the radiation source and the detector are translated in the reverse direction at the same speed v to the initial left end position;

[0100] During the reverse scan, the detector synchronously acquires the reverse projection data set which forms complementary projection coverage with the forward data to eliminate the blind area of one-way scanning.

[0101] (3) Inclination iterative adjustment:

[0102] After completing the forward / reverse scan of the current angle θ, drive the radiation source and the detector to synchronously deflect Δθ around the center O of the battery pack, and update the inclination of the SD connection line to θ + Δθ;

[0103] Repeat steps (1)-(2) to perform two-way scanning at the new angle until the entire preset inclination range [θ min , θ max is covered.

[0104] The motion control and precision guarantee mechanism includes:

[0105] Closed-loop feedback control: Real-time monitor the position deviation of the radiation source / detector through the grating scale. If the lateral offset is greater than the threshold, immediately trigger the correction algorithm to adjust the servo motor torque;

[0106] Geometric calibration compensation: After each angle adjustment, insert a calibration phantom (built-in tungsten ball array), and invert the geometric parameters (focus position, detector offset) of the system based on the projection coordinates to correct the projection data to a unified coordinate system;

[0107] Dynamic resolution adaptation: Automatically calculate the Z-axis step size according to the layer spacing of the battery pack to ensure that the adjacent scan overlapping area is ≥20%, and avoid loss of information between layers.

[0108] As Figure 5 shows the reconstruction result diagram provided by the embodiment of the present application. In the iterative reconstruction algorithm, in order to improve the image reconstruction quality and optimize the scanning efficiency, the present invention adopts the Ordered-Subset Simultaneous Algebraic Reconstruction Technique (OS-SART). This algorithm divides the projection data into multiple ordered subsets and combines a parallel iterative update strategy, which significantly improves the convergence speed and reconstruction accuracy, and is especially suitable for three-dimensional reconstruction in the limited-angle scanning scenario.

[0109] Model the CT scanning process as the following linear equation system:

[0110] AX = b

[0111] wherein:

[0112] b = (b1, b2, …, b M ) ∈ R M is the projection data vector, and M is the total number of rays;

[0113] X = (X1, X2, …, X N ) ∈ R N is the image vector to be reconstructed, and N is the total number of voxels;

[0114] A = (a mn ) ∈ R M×N is the system matrix, and the element a mn represents the path length of the m-th ray passing through the n-th voxel.

[0115] The total projection data is divided into multiple non-overlapping ordered subsets, including:

[0116] The total projection data is divided into T non-overlapping subsets S1, S2, …, S T satisfying:

[0117]

[0118] Each subset S T contains rays, and the subset division is optimized according to the projection angle or spatial distribution, including the allocation of adjacent angular intervals to different subsets.

[0119] In each iteration, each subset is updated, and the iterative oscillation is suppressed by a dynamic relaxation factor, including:

[0120] In the k-th iteration, the subset S [k] ([k] = (k mod T) + 1) is updated as follows:

[0121]

[0122] where λ k ∈ (0, 2) is the dynamic relaxation factor for suppressing iterative oscillation; the denominator term is the total weighted contribution of the subset S [k] to the voxel n.

[0123] In each iteration, forward projection calculation, residual calculation, correction accumulation, and voxel update are performed until the convergence condition is satisfied to obtain the reconstructed image, specifically including:

[0124] The forward projection calculation is performed through the formula ;

[0125] The residual calculation is performed through the formula Perform residual calculation;

[0126] Through the formula Perform correction amount accumulation calculation;

[0127] Through the formula Perform voxel update calculation.

[0128] In an alternative embodiment of the present application, for the image reconstruction part of the present invention, in addition to the OS-SART algorithm, there are various alternative solutions that can be adapted to different detection requirements: Model-based iterative reconstruction (MBIR) significantly improves the anti-noise ability and artifact suppression effect by embedding a physical model, and is especially suitable for high-precision defect identification; Deep learning image reconstruction (DLIR) uses a neural network to achieve second-level real-time imaging, and balances speed and sensitivity in low-dose scenarios; Optimized filtered back projection (FBP) combines GPU acceleration and an adaptive kernel function to meet the high-speed screening requirements of the production line; Statistical iterative reconstruction (SIR) optimizes memory occupancy and convergence efficiency through statistical weights, and is adapted to dynamic material compensation; The hybrid analytical-iterative algorithm (Hybrid IR) combines the speed advantage of the analytical method and the accuracy advantage of the iterative method to achieve full-life cycle detection scenario coverage. Each solution can be flexibly selected or combined according to detection objectives (such as accuracy priority, real-time requirements, and hardware resources). For example, MBIR and DLIR can be combined to construct a highly robust detection framework, or Hybrid IR can be used to adapt to the multi-modal requirements of laboratories and production lines, ensuring the scalability and industrial adaptability of the technical solution.

[0129] A large heterogeneous component CT imaging device based on bidirectional collaborative scanning provided by an embodiment of the present application may include:

[0130] An initialization module, configured to establish a rectangular coordinate system with the long axis of the battery pack as the X axis and the thickness direction as the Z axis, and initialize the scanning parameters; wherein, it includes adjusting the angle between the line connecting the ray source focus and the detector center and the normal line of the battery pack, and fixing the length of the line.

[0131] A scanning module, configured to perform bidirectional translation scanning, including forward scanning and reverse scanning, and inclination angle iterative adjustment;

[0132] A compensation module, configured to monitor the position deviation of the ray source / detector in real time and perform deviation correction, perform geometric calibration compensation after each angle adjustment, and calculate the Z-axis step size to ensure the integrity of interlayer information;

[0133] An image reconstruction module, configured to perform image reconstruction using the ordered subset joint algebraic reconstruction technique, model the scanning process as a system of linear equations, divide the projection data into ordered subsets, and achieve image reconstruction through an iterative update strategy.

[0134] For the specific limitations of the CT imaging device for large heterogeneous components based on bidirectional collaborative scanning, reference can be made to the limitations of the CT imaging method for large heterogeneous components based on bidirectional collaborative scanning in the foregoing text, which will not be elaborated herein. Each module in the above CT imaging device for large heterogeneous components based on bidirectional collaborative scanning can be implemented in whole or in part by software, hardware, and their combination. The above modules can be embedded in the processor of the computer device in hardware form or be independent of it, or be stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to each of the above modules.

[0135] The technical features of the above-described embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

[0136] The above-described embodiments only represent several implementation manners of the present application, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the patent application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the appended claims.

Claims

1. A CT imaging method for large heterogeneous components based on bidirectional collaborative scanning, characterized in that The method includes: S1. Establish a rectangular coordinate system with the long axis of the battery pack as the X-axis and the thickness direction as the Z-axis, and initialize the scanning parameters; wherein, it includes adjusting the angle between the line connecting the ray source focus and the detector center and the battery pack normal, and fixing the length of the line; S2. Perform two-way translation scanning, including forward scanning and reverse scanning, and inclination angle iterative adjustment; S3. Real-time monitor the position deviation of the ray source / detector and perform deviation correction. After each angle adjustment, perform geometric calibration compensation, and calculate the Z-axis stepping amount to ensure complete interlayer information; S4. Use the ordered subset combined algebraic reconstruction technique for image reconstruction. Model the scanning process as a system of linear equations, divide the projection data into ordered subsets, and achieve image reconstruction through an iterative update strategy.

2. The method according to claim 1, wherein Performing two-way translation scanning, including forward scanning and reverse scanning, and inclination angle iterative adjustment, specifically includes: S21 Forward scanning: The ray source system uniformly translates along the positive X-axis direction from the initial left end position, and at the same time, the detector system synchronously moves right along the mirror image trajectory, and the two maintain a constant speed and relative distance; during the translation, the X-ray beam covers the cross-section of the battery pack with a fan angle, and the detector continuously acquires projection data at the frame rate; S22 Reverse scanning: When the ray source reaches the right end position, the system immediately switches the movement direction, and the ray source and the detector translate in the reverse direction to the initial left end position at the same speed; during the reverse scanning, the detector synchronously acquires the reverse projection data set, which forms complementary projection coverage with the forward data to eliminate the one-way scanning blind area; S23 Inclination angle iterative adjustment: After completing the forward / reverse scanning of the current angle, drive the ray source and the detector to synchronously deflect around the center of the battery pack, and update the connection inclination angle to a new angle; repeat the above forward / reverse scanning process until the entire preset inclination angle range is covered.

3. The method according to claim 1, wherein The initialization of the scanning parameters further includes setting the initial positions of the ray source and the detector so that the line connecting the ray source focus and the detector center is perpendicular to the initial scanning plane of the battery pack.

4. The method according to claim 1, characterized in that, The real-time monitoring of the position deviation of the ray source / detector and performing deviation correction, performing geometric calibration compensation after each angle adjustment, and calculating the Z-axis stepping amount to ensure complete interlayer information, includes: Real-time monitor the position deviation of the ray source / detector through a grating ruler. If the lateral offset is greater than the threshold, immediately trigger the deviation correction algorithm to adjust the servo motor torque; After each angle adjustment, insert a calibration phantom, invert the system geometric parameters based on the projection coordinates, and correct the projection data to a unified coordinate system; Automatically calculate the Z-axis stepping amount according to the layer spacing of the battery pack to ensure that the adjacent scanning overlapping area is ≥20%; 5. The method according to claim 1, characterized in that, Using the ordered subset combined algebraic reconstruction technique for image reconstruction, specifically includes: Divide the total projection data into multiple non-overlapping ordered subsets, each subset contains several rays, and the subset division is optimized according to the projection angle or spatial distribution; In each iteration, update each subset, and suppress the iterative oscillation through a dynamic relaxation factor; Perform forward projection calculation, residual calculation, correction amount accumulation and voxel update until the convergence condition is met to obtain the reconstructed image.

6. The method according to claim 5, characterized in that, Model the CT scanning process as the following system of linear equations: AX = b Where: b = (b1, b2, …, b M ) ∈ R M is the projection data vector, and M is the total number of rays; X = (X1, X2, …, X N ) ∈ R N is the image vector to be reconstructed, and N is the total number of voxels; A=(a mn )∈R M×N is the system matrix, and the element a mn represents the path length of the m-th ray passing through the n-th voxel.

7. The method according to claim 6, wherein Divide the total projection data into multiple non-overlapping ordered subsets, including: Divide the total projection data into T non - overlapping subsets S1, S2, …, S T Satisfy: Each subset S T contains rays, and the subset partitioning is optimized according to the projection angle or spatial distribution, including the allocation of adjacent angular intervals to different subsets.

8. The method according to claim 7, wherein In each iteration, update each subset, and suppress iterative oscillation through a dynamic relaxation factor, including: In the k-th iteration, subset S [k] ([k] = (k mod T) + 1) is updated as follows: Among them, λ k ∈(0, 2) is a dynamic relaxation factor used to suppress iterative oscillations; the denominator term is the total weighted contribution of subset S [k] to voxel n.

9. The method according to claim 8, wherein In each iteration, perform forward projection calculation, residual calculation, correction accumulation, and voxel update until the convergence condition is met to obtain a reconstructed image, specifically including: Perform forward projection calculation through the formula ; Calculate the residual through the formula ; Cumulatively calculate the correction amount through the formula ; Perform voxel update calculations through the formula ​ 10. A CT imaging device for large heterogeneous components based on two-way collaborative scanning, characterized in that, The device includes: An initialization module, configured to establish a rectangular coordinate system with the long axis of the battery pack as the X-axis and the thickness direction as the Z-axis, and initialize scanning parameters; wherein, it includes adjusting the angle between the line connecting the ray source focus and the detector center and the normal of the battery pack, and fixing the length of the line; A scanning module, configured to perform two-way translation scanning, including forward scanning and reverse scanning, and inclination iterative adjustment; A compensation module, configured to monitor the position deviation of the ray source / detector in real time and perform deviation correction, perform geometric calibration compensation after each angle adjustment, and calculate the Z-axis step size to ensure the integrity of interlayer information; An image reconstruction module, configured to perform image reconstruction using the ordered subset combined algebraic reconstruction technique, model the scanning process as a system of linear equations, divide the projection data into ordered subsets, and achieve image reconstruction through an iterative update strategy.

Images