Ray imaging system and detector fusion positioning method thereof

Abstract

The invention discloses a ray imaging system and a detector fusion positioning method thereof, and the method comprises the steps: installing an image sensor on a ray source, enabling a calibration plate and a detector to be positioned and installed, and enabling the detector to be provided with an inertial sensor; acquiring image information of the calibration plate by using an image sensor, extracting feature points of an image of the calibration plate by using a processor, and determining a first positioning coordinate of the detector based on a transformation relation between a pixel coordinate system and a physical coordinate system of the calibration plate; determining a second positioning coordinate of the detector based on the initial state of the detector by using the inertial sensor; determining a third positioning coordinate of the detector by using a magnetic positioning technology; and according to the three positioning coordinates and the confidence degrees corresponding to the three positioning coordinates, determining a fusion positioning coordinate. According to the invention, fusion of image positioning, inertial positioning and magnetic positioning is realized, and the fusion weight of image positioning and magnetic positioning is adaptively optimized.

Description

Technical Field

[0001] The present invention relates to the field of ray detection, and in particular to a ray imaging system and a detector fusion positioning method thereof. Background Art

[0002] X-ray imaging is based on the physical properties of X-rays (penetration, fluorescence, and photosensitivity) and the differences in X-ray absorption by materials. By recording the intensity distribution of X-rays after they penetrate a target, an image of the internal structure is formed. The detector converts X-rays transmitted through the object into electrical signals, ultimately forming a digital image. The strength of the received signal directly reflects the density differences within the object.

[0003] The detector must be precisely aligned with the X-ray tube. This is the core condition for ensuring the quality of X-ray imaging. Currently available flat panel detector positioning solutions mainly include the following: (1) Real-time spatial magnetic positioning solution: Using an alternating magnetic field generator and a multi-point magnetic sensor array, it can achieve centimeter-level accuracy in an ideal interference-free environment. However, the hardware is bulky, the cost is high, and it is extremely sensitive to metal environments. (2) Visual positioning based on a QR code fixture: a triangular QR code is placed on the fixture, and distance and posture are inverted through PnP. The hardware is lightweight and low-cost. Once the QR code is blocked by the human body or the light deteriorates, it becomes invalid. (3) Vision-IMU fusion positioning: Using pattern recognition and IMU compensation, it can maintain continuous output in the case of short-term occlusion, but inertial drift will accumulate after a long time.

[0004] Various positioning solutions are difficult to maintain centimeter-level or even sub-centimeter-level six-degree-of-freedom continuous tracking in clinical use because the detection equipment encounters occlusion, metal interference, and high-speed movement.

[0005] The disclosure of the above background technology content is only used to assist in understanding the concept and technical solution of this application. It does not necessarily belong to the prior art of this application, nor does it necessarily provide technical guidance. In the absence of clear evidence that the above content has been disclosed before the filing date of this application, the above background technology should not be used to evaluate the novelty and creativity of this application. Summary of the Invention

[0006] The purpose of the present invention is to provide a detector positioning method that realizes the fusion of image positioning, inertial positioning and magnetic positioning.

[0007] In order to achieve the above object, the technical solution adopted by the present invention is as follows: A detector fusion positioning method applied to a radiographic imaging system comprises the following steps: The image sensor is mounted on the radiation source, and the calibration plate and the detector are positioned and mounted, wherein the detector is equipped with an inertial sensor; Using the image sensor to collect image information of the calibration plate, the processor extracts feature points of the calibration plate image, and determines the physical coordinates of the feature points based on the transformation relationship between the pixel coordinate system of the image sensor and the physical coordinate system of the calibration plate, thereby determining the physical coordinates of the detector, which are recorded as first positioning coordinates; Determine the current position of the detector based on the initial state of the detector using the inertial sensor of the detector, and record it as the second positioning coordinate; Use magnetic positioning technology to determine the current position of the detector and record it as the third positioning coordinate; Determine a fused positioning coordinate based on the first positioning coordinate, the second positioning coordinate, the third positioning coordinate and their corresponding confidence levels.

[0008] Furthermore, based on any one of the above technical solutions or a combination of multiple technical solutions, the confidence level is calculated in real time by: The processor sets a visual confidence sliding window, combines the instantaneous speed data of the inertial sensor, linearly predicts the state of the calibration plate being blocked in the future time period, and then generates a corresponding visual confidence curve. If the calibration plate is predicted to be blocked, the visual confidence curve decreases; If it is predicted that the height of the visual confidence curve drops to a preset trigger threshold, the first confidence corresponding to the first positioning coordinate is lowered, and the second confidence corresponding to the second positioning coordinate and / or the third confidence corresponding to the third positioning coordinate is increased.

[0009] Furthermore, based on any one of the technical solutions or a combination of multiple technical solutions described above, if it is predicted that the height of the visual confidence curve drops to a preset trigger threshold, the first confidence is gradually reduced until it is less than or equal to the preset visual confidence minimum limit or the height of the visual confidence curve is restored to a preset recovery threshold, and the recovery threshold is greater than or equal to the trigger threshold.

[0010] Further, according to any one of the above technical solutions or a combination of multiple technical solutions, the first confidence level is configured with a first initial value; If it is predicted that the height of the visual confidence curve drops to a preset trigger threshold, the associated confidence corresponding to the current visual confidence curve height is determined according to a preset mapping rule, and the first confidence is updated to the associated confidence, and the associated confidence is less than the first initial value.

[0011] Further, based on any one of the above technical solutions or a combination of multiple technical solutions, when determining the first confidence level corresponding to the first positioning coordinates, the second confidence level corresponding to the second positioning coordinates and the third confidence level corresponding to the third positioning coordinates are adaptively allocated in the following manner: Adopting the extended Kalman filter framework, the residual sum of squares is periodically statistically innovated, and the scaling factor of the measurement noise matrix is ​​dynamically adjusted using the first-order stochastic gradient descent framework to minimize the residual sum of squares. If the scaling factor increases, the third confidence level is decreased and the second confidence level is increased; if the scaling factor decreases, the third confidence level is increased and the second confidence level is decreased.

[0012] Furthermore, according to any one of the above technical solutions or a combination of multiple technical solutions, in the EKF update step, the innovation residual is calculated ,in, z k for k The sensor observation value at time, For k -1 moment k The predicted value at the moment; In the time window T Calculate the residual sum of squares RSS internally T : ; in, y t Time window T The innovative residual vector within, is the innovation residual vector y t The transposed matrix of calculate k The measurement noise matrix at time: R k = l k × R 0 ,in, R 0 is the initial noise matrix, λ k for k Scaling factor of the moment; The objective function is determined to minimize the sum of squared residuals within the window: J ( l k )=RSS T ( l k ); Update the scaling factor using a first-order stochastic gradient descent framework: ;in, or is the learning rate, gradient calculation Derived by the chain rule: ; in, Passed through the Kalman gain of the EKF.

[0013] Furthermore, any one of the above technical solutions or a combination of multiple technical solutions may further include, before determining the positioning coordinates: Perform consistency check on magnetic field vector according to geometric symmetry; And / or, the IMU data of the inertial sensor is self-calibrated through Allan variance; And / or, after extracting the feature points of the calibration plate image, sub-pixel interpolation is performed on the feature points.

[0014] Furthermore, according to any one of the above technical solutions or a combination of multiple technical solutions, the image sensor, inertial sensor and magnetic field sensor are unified into the physical coordinate system of the calibration plate to obtain a three-system alignment template; and a local calibration library is constructed to store the three-system alignment template corresponding to the calibration plate; The local calibration library is configured to support anonymous hash vector synchronization, template distribution and roaming loading of three-system alignment templates.

[0015] According to another aspect of the present invention, there is provided a radiographic imaging system comprising a radiation source, a detector, a magnetic positioning device, a processor, an image sensor mounted on the radiation source, a calibration plate positioned and mounted with the detector, and an inertial sensor; The magnetic positioning device is configured to determine the current position of the detector, recorded as a magnetic positioning coordinate; The processor is configured to perform: The image sensor is used to collect image information of the calibration plate. The processor extracts feature points of the calibration plate image and determines the physical coordinates of the feature points based on a transformation relationship between a pixel coordinate system of the image sensor and a physical coordinate system of the calibration plate. The physical coordinates of the detector are then determined and recorded as first positioning coordinates. The processor determines the current position of the detector based on the initial state of the detector using the inertial sensor of the detector, and records it as a second positioning coordinate; A fused positioning coordinate is determined according to the first positioning coordinate, the second positioning coordinate, the magnetic positioning coordinate and their corresponding confidence levels.

[0016] Furthermore, based on any one of the technical solutions or a combination of multiple technical solutions described above, the processor is configured to execute the steps of the detector fusion positioning method described above.

[0017] Furthermore, based on any one of the above technical solutions or a combination of multiple technical solutions, the present invention provides a radiographic imaging system including a local database configured to store a three-system alignment template of the image sensor, the inertial sensor, and the magnetic field sensor unified to the physical coordinate system of the calibration plate; The local database is configured to support anonymous hash vector synchronization, template distribution and roaming loading of three-series alignment templates.

[0018] Furthermore, based on any one of the technical solutions or a combination of multiple technical solutions described above, the calibration plate has a one-plate four-purpose structure of a checkerboard, a QR code, an X-ray metal identification line, and a high-reflective film, wherein the high-reflective film is configured to provide zero-difference depth calibration for a depth camera or a TOF module.

[0019] Further, based on any one of the above technical solutions or a combination of multiple technical solutions, the magnetic positioning device includes a magnetic field sensor and a magnetic field source, and the magnetic field source includes a permanent magnet or an electromagnetic coil; The magnetic field source is arranged on the beam splitter of the ray source, and the magnetic field sensor is arranged on the calibration plate or the detector; or, the magnetic field sensor is arranged on the ray source, and the magnetic field source is arranged on the calibration plate.

[0020] The beneficial effects brought about by the technical solution provided by the present invention are as follows: a. Achieve the fusion of image positioning, inertial positioning, and magnetic positioning, making the fused positioning results more accurate and more stable and reliable than single-type positioning algorithms; b. Predictive occlusion management instantly reduces the image positioning weight. When occlusion actually occurs, the system has already completed the transition, ensuring a continuous trajectory without jumps, significantly reducing the retake rate. c. Optimize magnetic positioning weights based on the extended Kalman filter framework, adapt to different magnetic noise environments, and ensure the accuracy of fusion positioning in various magnetic noise environments; d. A three-layer algorithm consisting of low-level sensor-level fusion, mid-level confidence estimation, and high-level self-learning EKF achieves adaptive optimization and fusion of image positioning, inertial positioning, and magnetic positioning, taking into account magnetic noise, lighting differences, and occlusion frequency in different equipment rooms, and solving problems such as trajectory jumps or chronic drift. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in this application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0022] Figure 1A schematic flow chart of a detector fusion positioning method applied to a radiographic imaging system provided by an exemplary embodiment of the present invention; Figure 2 A schematic diagram of a process for instantly updating confidence levels through predictive occlusion management provided by an exemplary embodiment of the present invention; Figure 3 A schematic diagram of a process for adaptively allocating inertial weights and magnetic weights based on an extended Kalman filter framework provided for an exemplary embodiment of the present invention; Figure 4 A structural block diagram of a radiographic imaging system provided as an exemplary embodiment of the present invention. DETAILED DESCRIPTION

[0023] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0024] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, apparatus, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0025] In one embodiment of the present invention, a detector fusion positioning method for a radiographic imaging system is provided. Figure 1 As shown, the positioning method includes the following steps: The image sensor is mounted on the radiation source, and the calibration plate and the detector are positioned and mounted, wherein the detector is equipped with an inertial sensor; Using the image sensor to collect image information of the calibration plate, the processor extracts feature points of the calibration plate image, and determines the physical coordinates of the feature points based on the transformation relationship between the pixel coordinate system of the image sensor and the physical coordinate system of the calibration plate, thereby determining the physical coordinates of the detector, which are recorded as first positioning coordinates; Determine the current position of the detector based on the initial state of the detector using the inertial sensor of the detector, and record it as the second positioning coordinate; Using magnetic positioning technology, the current position of the detector is determined and recorded as the third positioning coordinate; specifically, the magnetic positioning device includes a magnetic field sensor and a magnetic field source, and the magnetic field source includes a permanent magnet or an electromagnetic coil; Determine a fused positioning coordinate based on the first positioning coordinate, the second positioning coordinate, the third positioning coordinate and their corresponding confidence levels.

[0026] Specifically, in the field of radiographic imaging, it is necessary to determine the radiation dose and beam range based on the actual conditions of the inspected object. After passing through the inspected object, the radiation is received by a flat-panel detector. Therefore, it is important to determine the relative position relationship between the radiation source and the flat-panel detector.

[0027] Based on the transformation relationship between the image sensor's pixel coordinate system and the calibration plate's physical coordinate system, the coordinates of a pixel in the image information corresponding to the calibration plate's physical coordinate system can be determined. In this embodiment, the detector is a flat-panel detector, and the calibration plate is a planar plate-like structure, that is, the calibration plate's physical coordinate system is a planar coordinate system. When the relative position relationship between the calibration plate and the detector is known (for example, the center point of the detector coincides with the feature point at the center of the calibration plate), the first positioning coordinate of the detector in the calibration plate's physical coordinate system can be determined, which is recorded as ( x 1, y 1); The principle of inertial measurement is to use inertial sensors to detect the acceleration and angular velocity of the flat panel detector in real time. The inertial sensor is an inertial measurement unit (IMU), which is usually composed of multiple sensors. For example, a 6-DOF IMU includes a three-axis accelerometer and a three-axis gyroscope. The principle of inertial measurement positioning is integration, that is, based on the initial position plus the calculated subsequent movement distance and direction, the current position after movement is determined, which is recorded as the second positioning coordinate ( x 2, y 2) This also brings the disadvantage of relying on integral calculation, that is, the drift of the gyroscope and the noise of the accelerometer will increase the error over time. Subsequent embodiments propose to correct it through a filtering algorithm.

[0028] Magnetic positioning technology mainly relies on the coordinated work of magnetic field sensors (such as Hall sensors) and magnetic field sources (such as permanent magnets or electromagnetic coils) to infer the spatial position and posture of the flat-panel detector by detecting changes in magnetic field strength. For example, a permanent magnet or electromagnetic coil is placed at a fixed position (such as the end of the beam spotter) to generate a specific low-frequency magnetic field. The flat-panel detector integrates a multi-axis magnetic sensor (such as a Triaxis Hall sensor) to detect the magnetic field strength (B x , B y, B z ), collect magnetic field data in real time, and match the current position with the pre-calibrated magnetic field distribution map. The pre-calibration process here mainly includes dynamic calibration method or static calibration method. The steps of dynamic calibration method are to move the flat-panel detector, synchronously collect magnetic field data and visual positioning coordinates at different positions, and build a magnetic field-space mapping database, that is, a calibration database; the steps of static calibration method are to place the detector at a known position, record the magnetic field strength, and generate a calibration lookup table, that is, a calibration database. Compare the current magnetic field data with the calibration database, and determine the most likely position through the nearest neighbor algorithm (such as KNN) or particle filtering, which is recorded as the third positioning coordinate ( x 3, y 3) The calibration method for magnetic positioning is as follows: By building a magnetic field environment vector library, real-time comparison and prompting to load the optimal calibration value, the impact of environmental changes can be mitigated to a certain extent.

[0029] like Figure 1 As shown in the figure, since the image sensor, inertial sensor, and magnetic field sensor are all unified to the physical coordinate system of the calibration plate, the position coordinates obtained by their respective positioning calculations are based on the unified coordinate system. Furthermore, the comprehensive positioning coordinates are obtained based on the fusion of their respective confidence levels: ( P 1× x 1+ P 2× x 2+ P 3× x 3, P 1× y 1+ P 2× y 2+ P 3× y 3), among which, P 1 is the confidence of image positioning, P 2 is the confidence level of inertial positioning, P 3 is the confidence level of magnetic positioning, P 1+ P 2+ P 3=1.

[0030] In one embodiment, the confidence P 1. P 2. P 3. It can be set according to the reliability of image sensor, inertial sensor and magnetic field sensor; In another embodiment, the confidence level is updated in real time through predictive occlusion management, such as Figure 2 As shown: The processor sets a visual confidence sliding window, combines the instantaneous speed data of the inertial sensor, linearly predicts the state of the calibration plate being blocked in the future time period, and then generates a corresponding visual confidence curve. If the calibration plate is predicted to be blocked, the visual confidence curve decreases; If it is predicted that the height of the visual confidence curve drops to a preset trigger threshold, the first confidence corresponding to the first positioning coordinate is lowered, and the second confidence corresponding to the second positioning coordinate and / or the third confidence corresponding to the third positioning coordinate is increased.

[0031] Specific numerical example: The system maintains a 0.5s visual confidence sliding window and combines the detector's instantaneous speed (data obtained by the inertial sensor) to predict the visual confidence curve for the next 200ms using a linear extrapolation algorithm. When the predicted value (here refers to the height of the predicted visual confidence curve, not the confidence P 1) is lower than 0.4, the fusion engine smoothes and reduces the visual weight in advance (i.e. P 1) and increase the inertia weight (i.e. P 2) and magnetic weight (i.e. P 3). When the actual occlusion occurs, the system has completed the transition, and the trajectory is continuous without jumps, significantly reducing the retake rate.

[0032] For the above-mentioned early smoothing to reduce visual weight (i.e. P 1) There are two specific implementation methods: Method 1: If it is predicted that the visual confidence curve height drops to a preset trigger threshold (i.e. 0.4 above), the first confidence level is gradually reduced. P 1. For example, if the visual confidence curve height is reduced by 0.05 each time, if it is still lower than the trigger threshold, it will be reduced by 0.05 again, and so on. There are two cases where the step-by-step reduction needs to be stopped: Case 1 is that the visual confidence curve height is higher than the recovery threshold (greater than or equal to the trigger threshold), and Case 2 is the first confidence after the current reduction. P 1 is less than or equal to the preset visual confidence threshold (such as 0.1 or 0).

[0033] Method 2: The first confidence level P 1 is configured with the first initial value P 1-ini If it is predicted that the visual confidence curve height drops to the preset trigger threshold (i.e. 0.4 above), the associated confidence corresponding to the current visual confidence curve height is determined according to the preset mapping rule, and the associated confidence is used as the first confidence P 1. The association confidence is less than the first initial value.

[0034] The purpose of the above-mentioned predictive occlusion management is to reduce the first confidence level in advance before the calibration plate is blocked. P 1. Due to P 1+ P 2+ P 3=1, therefore, when an occlusion situation is predicted, at least one of the second confidence level and the third confidence level needs to be increased.

[0035] In one embodiment of the present invention, regardless of whether the first confidence level changes, the second confidence level corresponding to the second positioning coordinate and the third confidence level corresponding to the third positioning coordinate are adaptively allocated in the following manner: Figure 3 As shown: Adopting the extended Kalman filter framework, the residual sum of squares is periodically statistically innovated, and the scaling factor of the measurement noise matrix is ​​dynamically adjusted using the first-order stochastic gradient descent framework to minimize the residual sum of squares. If the scaling factor increases, the third confidence level is reduced and the second confidence level is increased; if the scaling factor decreases, the third confidence level is increased and the second confidence level is reduced.

[0036] A fixed scaling factor, meaning all equipment rooms are treated as having the same noise level, can lead to over-reliance on magnetic measurements in high-noise environments or inadvertent down-weighting of visual measurements in low-light environments. The online tuner in this embodiment treats the EKF residual as a "real-time score" and minimizes it through iterative scaling. This is equivalent to adaptively assigning optimal fusion weights to different equipment rooms. Figure 3 The specific scheme of the embodiment shown is as follows: In the EKF update step, the innovation residual is calculated ,in, z k for k The sensor observation value at time, For k -1 moment k The predicted value at the moment; In the time window T Calculate the residual sum of squares RSS internally T :

[0037] in, y t Time window T The innovative residual vector within, is the innovation residual vector y t The transposed matrix of calculate k The measurement noise matrix at time: R k = l k × R 0 ,in,R 0 is the initial noise matrix, λ k for k Scaling factor at time; scaling factor λ k It can suppress the time-varying noise fluctuations of sensors caused by, for example, environmental interference. When the environmental noise increases, the scaling factor increases, causing the measurement noise to increase. At this time, the filtering framework trusts the model prediction more. Conversely, when the environmental noise decreases, the scaling factor decreases, causing the measurement noise to decrease. At this time, the filtering framework trusts the observation data more, thereby achieving a balance between the trust in model predictions and the trust in observations, adaptively optimizing sensor weights, and improving the accuracy and robustness of multi-source data fusion.

[0038] The objective function is determined to minimize the sum of squared residuals within the window: J ( l k )=RSS T ( l k ); Update the scaling factor using a first-order stochastic gradient descent framework: ;in, or is the learning rate, gradient calculation Derived by the chain rule:

[0039] in, Through the Kalman gain transfer of EKF. In this embodiment, a dynamic adaptation strategy can be used to set the learning rate or : Initial learning rate or The initial value can be set to 0.02, so that the scaling factor that is far from the optimal value at the beginning can quickly approach it. Then, the learning rate is decayed every time the scaling factor is iterated a certain number of times (for example, 2 times, 5 times, or 10 times), for example, in the form of exponential decay to avoid oscillation in the later stage, until the learning rate reaches the preset lower limit (for example, 10 -5 ).

[0040] In one embodiment of the present invention, before determining the positioning coordinates, the method further includes: Perform consistency check on magnetic field vector according to geometric symmetry; And / or, the IMU data of the inertial sensor is self-calibrated through Allan variance; And / or, after extracting the feature points of the calibration plate image, sub-pixel interpolation is performed on the feature points.

[0041] The above magnetic field vector consistency test, inertial sensor variance self-calibration, and sub-pixel interpolation of image feature points can be combined as low-level sensor-level fusion to Figure 2The confidence estimation of the middle layer shown in the figure has zero jump in occlusion switching and significantly reduced retake rate; and Figure 3 The high-level self-learning EKF shown in the figure automatically learns its weights to adapt to varying equipment room noise levels, eliminating the need for manual parameter adjustment. The three-layer algorithm (low-mid-high) integrates image positioning, inertial positioning, and magnetic positioning, resulting in more accurate positioning results. Compared to single-type positioning algorithms, it offers greater stability and reliability, accounting for varying magnetic noise levels, lighting variations, and occlusion frequency in different equipment rooms, addressing issues such as trajectory jumps and chronic drift.

[0042] In one embodiment of the present invention, the image sensor, inertial sensor, and magnetic field sensor are unified into the physical coordinate system of the calibration plate to obtain a three-system alignment template; and a local calibration library is constructed to store the three-system alignment template corresponding to the calibration plate; The local calibration library is configured to support anonymous hash vector synchronization, template distribution, and roaming loading of three-series alignment templates. For first-time users visiting a new machine room, if the local library lacks a matching vector, the system guides technicians through a 10-second rapid 4D calibration and uploads the anonymous vector-parameter pairs to the in-hospital server. Other devices entering the same machine room can download the template, achieving consistent three-series (or even four-series) alignment results without the need for plug-in boards. This ensures that flat-panel detectors can be used instantly in any machine room, with no lags due to occlusion and zero downtime for maintenance.

[0043] like Figure 4 As shown, an embodiment of the present invention further provides a radiographic imaging system, comprising a radiation source, a detector, a magnetic positioning device, a processor, an image sensor mounted on the radiation source, a calibration plate positioned and mounted with the detector, and an inertial sensor; The magnetic positioning device is configured to determine the current position of the detector, which is recorded as the magnetic positioning coordinates; the magnetic positioning device includes a magnetic field sensor and a magnetic field source, and the magnetic field source includes a permanent magnet or an electromagnetic coil; the magnetic field source is arranged on the beam beam of the ray source, and the magnetic field sensor is arranged on the calibration plate or the detector; or, the magnetic field sensor is arranged on the ray source, and the magnetic field source is arranged on the calibration plate.

[0044] The processor is configured to perform: The image sensor is used to collect image information of the calibration plate. The processor extracts feature points of the calibration plate image and determines the physical coordinates of the feature points based on a transformation relationship between a pixel coordinate system of the image sensor and a physical coordinate system of the calibration plate. The physical coordinates of the detector are then determined and recorded as first positioning coordinates. The processor determines the current position of the detector based on the initial state of the detector using the inertial sensor of the detector, and records it as a second positioning coordinate; A fused positioning coordinate is determined according to the first positioning coordinate, the second positioning coordinate, the magnetic positioning coordinate and their corresponding confidence levels.

[0045] The radiographic imaging system provided by an embodiment of the present invention includes a local database configured to store a three-system alignment template of an image sensor, an inertial sensor, and a magnetic field sensor unified to a physical coordinate system of a calibration plate; The local database is configured to support anonymous hash vector synchronization, template distribution and roaming loading of three-series alignment templates.

[0046] Specifically, the calibration plate has a one-plate four-purpose structure with a checkerboard, a QR code, an X-ray metal marking line, and a high-reflective film. The checkerboard, QR code, and X-ray metal marking line can be found in the Chinese patent application with publication number CN117788605A; the high-reflective film is configured to be 1 cm 2 The calibration plate, which measures approximately 100 mm, is placed on the front or back of the calibration plate for homodyne depth calibration of a depth camera or TOF module, reserving a hardware interface for alignment. The aforementioned multi-purpose calibration plate, which combines a checkerboard, QR code, and X-ray marker, can simultaneously perform camera calibration, X-ray focus alignment, and magnetic positioning coordinate alignment, but is only used during offline calibration.

[0047] The radiographic imaging system provided in this embodiment and the detector fusion positioning method applied to the radiographic imaging system provided in the above-mentioned embodiment belong to the same inventive concept. The entire contents of the embodiment of the detector fusion positioning method applied to the radiographic imaging system are incorporated into the embodiment of this radiographic imaging system by reference and will not be repeated here.

[0048] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply the existence of any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device comprising the element.

[0049] The above is only a specific implementation method of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A detector fusion positioning method applied to a radiographic imaging system, characterized in that: The following steps are involved: Mounting the image sensor on the radiation source, positioning the calibration plate and the detector, wherein the detector is equipped with an inertial sensor; Using the image sensor to collect image information of the calibration plate, the processor extracts feature points of the calibration plate image, and determines the physical coordinates of the feature points based on the transformation relationship between the pixel coordinate system of the image sensor and the physical coordinate system of the calibration plate, thereby determining the physical coordinates of the detector, which are recorded as first positioning coordinates; Determine the current position of the detector based on the initial state of the detector using the inertial sensor of the detector, and record it as the second positioning coordinate; Use magnetic positioning technology to determine the current position of the detector and record it as the third positioning coordinate; Determine a fused positioning coordinate based on the first positioning coordinate, the second positioning coordinate, the third positioning coordinate and their corresponding confidence levels.

2. The detector fusion positioning method applied to a radiographic imaging system according to claim 1, characterized in that: The confidence is calculated on the fly by: The processor sets a visual confidence sliding window, combines the instantaneous speed data of the inertial sensor, linearly predicts the state of the calibration plate being blocked in the future time period, and then generates a corresponding visual confidence curve. If the calibration plate is predicted to be blocked, the visual confidence curve decreases; If it is predicted that the height of the visual confidence curve drops to a preset trigger threshold, the first confidence corresponding to the first positioning coordinate is lowered, and the second confidence corresponding to the second positioning coordinate and / or the third confidence corresponding to the third positioning coordinate is increased.

3. The detector fusion positioning method applied to a radiographic imaging system according to claim 2, characterized in that: If it is predicted that the height of the visual confidence curve drops to a preset trigger threshold, the first confidence is gradually reduced until it is less than or equal to the preset visual confidence minimum limit or the height of the visual confidence curve is restored to a preset recovery threshold, and the recovery threshold is greater than or equal to the trigger threshold.

4. The detector fusion positioning method applied to a radiographic imaging system according to claim 2, characterized in that: The first confidence level is configured with a first initial value; If it is predicted that the height of the visual confidence curve drops to a preset trigger threshold, the associated confidence corresponding to the current visual confidence curve height is determined according to a preset mapping rule, and the first confidence is updated to the associated confidence, and the associated confidence is less than the first initial value.

5. The detector fusion positioning method applied to a radiographic imaging system according to claim 1, characterized in that: When the first confidence level corresponding to the first positioning coordinates is determined, the second confidence level corresponding to the second positioning coordinates and the third confidence level corresponding to the third positioning coordinates are adaptively allocated in the following manner: Adopting the extended Kalman filter framework, the residual sum of squares is periodically statistically innovated, and the scaling factor of the measurement noise matrix is ​​dynamically adjusted using the first-order stochastic gradient descent framework to minimize the residual sum of squares. If the scaling factor increases, the third confidence level is reduced and the second confidence level is increased; if the scaling factor decreases, the third confidence level is increased and the second confidence level is reduced.

6. The detector fusion positioning method applied to a radiographic imaging system according to claim 5, characterized in that: In the EKF update step, the innovation residual is calculated ,in, z k for k The sensor observation value at time, For k -1 moment k The predicted value at the moment; In the time window T Calculate the residual sum of squares RSS internally T : ； in, y t Time window T The innovative residual vector within, is the innovation residual vector y t The transposed matrix of calculate k The measurement noise matrix at time: R k = λ k × R 0 ,in, R 0 is the initial noise matrix, λ k for k Scaling factor of the moment; The objective function is determined to minimize the sum of squared residuals within the window: J ( λ k )=RSS T ( λ k ); Update the scaling factor using a first-order stochastic gradient descent framework: ;in, η is the learning rate, gradient calculation Derived by the chain rule: ； in, Passed through the Kalman gain of the EKF.

7. The detector fusion positioning method applied to a radiographic imaging system according to claim 1, characterized in that: Before determining the positioning coordinates, it also includes: Perform consistency check on magnetic field vector according to geometric symmetry; And / or, the IMU data of the inertial sensor is self-calibrated through Allan variance; And / or, after extracting the feature points of the calibration plate image, sub-pixel interpolation is performed on the feature points.

8. The detector fusion positioning method for a radiographic imaging system according to any one of claims 1 to 7, characterized in that: Unifying the image sensor, inertial sensor and magnetic field sensor into the physical coordinate system of the calibration plate to obtain a three-system alignment template; and building a local calibration library for storing the three-system alignment template corresponding to the calibration plate; The local calibration library is configured to support anonymous hash vector synchronization, template distribution and roaming loading of three-system alignment templates.

9. A radiographic imaging system, characterized in that: It includes a ray source, a detector, a magnetic positioning device, a processor, an image sensor mounted on the ray source, a calibration plate and an inertial sensor positioned and mounted on the detector; The magnetic positioning device is configured to determine the current position of the detector, recorded as a magnetic positioning coordinate; The processor is configured to perform: The image sensor is used to collect image information of the calibration plate. The processor extracts feature points of the calibration plate image and determines the physical coordinates of the feature points based on a transformation relationship between a pixel coordinate system of the image sensor and a physical coordinate system of the calibration plate. The physical coordinates of the detector are then determined and recorded as first positioning coordinates. The processor determines the current position of the detector based on the initial state of the detector using the inertial sensor of the detector, and records it as a second positioning coordinate; A fused positioning coordinate is determined according to the first positioning coordinate, the second positioning coordinate, the magnetic positioning coordinate and their corresponding confidence levels.

10. The radiographic imaging system according to claim 9, wherein: The processor is configured to execute the steps of the detector fusion positioning method according to any one of claims 2 to 7.

11. The radiographic imaging system according to claim 10, wherein: A local database is included, which is configured to store a three-system alignment template for unifying the image sensor, the inertial sensor, and the magnetic field sensor into the physical coordinate system of the calibration plate; The local database is configured to support anonymous hash vector synchronization, template distribution and roaming loading of three-series alignment templates.

12. The radiographic imaging system according to claim 9, wherein: The calibration plate has a four-in-one structure with a checkerboard, a QR code, an X-ray metal identification line, and a high-reflective film, wherein the high-reflective film is configured for zero-difference depth calibration of a depth camera or a TOF module.

13. The radiographic imaging system according to claim 9, wherein: The magnetic positioning device includes a magnetic field sensor and a magnetic field source, and the magnetic field source includes a permanent magnet or an electromagnetic coil; The magnetic field source is arranged on the beam splitter of the ray source, and the magnetic field sensor is arranged on the calibration plate or the detector; or, the magnetic field sensor is arranged on the ray source, and the magnetic field source is arranged on the calibration plate.

Images