MRI scanning plane tracking processing method, system, medium and product based on RGB-D camera

Abstract

The present invention discloses a magnetic resonance scanning plane tracking processing method, system, medium, and product based on an RGB-D camera, relating to the field of magnetic resonance scanning tracking, including: acquiring RGB-D camera data containing ArUco markers; performing ArUco marker detection and point cloud segmentation on the RGB-D camera data to obtain a marked point cloud; performing plane segmentation and extraction on the marked point cloud to obtain a marked plane; performing posture smoothing on the marked plane, and using the posture-smoothed marked plane as the scanning plane; converting the scanning plane from the camera coordinate system to the magnetic resonance coordinate system to obtain a coordinate matrix of the marked plane in the magnetic resonance coordinate system. The present invention has low cost, high stability, and fast imaging speed, and can achieve real-time tracking or delayed tracking with a preset time.

Description

Technical Field

[0001] The present invention relates to the technical field of magnetic resonance scanning methods, and in particular to a magnetic resonance scanning plane tracking processing method, system, medium and product based on an RGB-D camera. Background Art

[0002] Magnetic resonance imaging (MRI)-guided interventional surgery is a minimally invasive treatment technology that has rapidly developed in recent years. Its core advantage lies in its ability to provide excellent soft tissue contrast and radiation-free real-time imaging capabilities. Compared with traditional CT or ultrasound guidance, MRI has unique advantages such as multi-plane imaging, functional imaging, and temperature monitoring, making it particularly suitable for precision surgeries such as tumor ablation, biopsy, and vascular intervention. In a typical MRI-guided interventional surgery, the doctor needs to dynamically adjust the scanning plane according to the spatial position of the surgical instrument to ensure that the imaging plane always includes the instrument and its surrounding critical anatomical structures. This process places extremely high demands on the real-time performance and stability of the navigation system.

[0003] The most commonly used technology in guided interventional therapy is the traditional high-precision optical tracking solution. The high-precision optical tracking system can directly locate the specific position of the instrument. It uses optical principles to achieve precise measurement of the spatial position of the target object. Its core lies in the closed-loop control of optical signal acquisition and processing. The typical system consists of a light source module, an optical sensor unit, a data processing unit, and a feedback control module to form a complete technical chain. The performance of this system will be affected by factors such as ambient lighting, target reflectivity, and optical distortion. It needs to be compensated through multi-sensor fusion and online calibration to improve the robustness of tracking. Although the optical tracking equipment used in the high-precision optical tracking system can guarantee extremely high precision requirements, it is expensive, and the related precision hardware is easily affected by strong magnetic fields, making it difficult to achieve magnetic compatibility application scenarios.

[0004] Currently, some solutions directly track ArUco markers in RGB images. However, these solutions are inaccurate. At distances exceeding 350mm, the recognition angle deviation can exceed 8°, failing to meet the scanning plane positioning requirements. Furthermore, single-pixel corner detection errors can cause significant fluctuations in the normal vector angle, resulting in poor tracking stability and making it difficult to apply to demanding surgical scenarios. Summary of the Invention

[0005] In order to solve the technical problems existing in the background technology, the present invention proposes a magnetic resonance scanning plane tracking processing method, system, medium and product based on RGB-D camera.

[0006] In a first aspect, the present invention proposes a magnetic resonance scanning plane tracking processing method based on an RGB-D camera, comprising:

[0007] Get RGB-D camera data containing ArUco markers;

[0008] Perform ArUco marker detection and point cloud segmentation on RGB-D camera data to obtain a marked point cloud;

[0009] Perform plane segmentation and extraction on the marked point cloud to obtain the marked plane;

[0010] Performing posture smoothing on the marker plane, and using the smoothed marker plane as the scanning plane;

[0011] The scanning plane is transformed from the camera coordinate system to the magnetic resonance coordinate system to obtain the coordinate matrix of the marking plane in the magnetic resonance coordinate system.

[0012] Preferably, the RGB-D camera data includes a depth frame and an RGB frame;

[0013] Among them, the ArUco marker detection and point cloud segmentation are performed on the RGB-D camera data to obtain the marked point cloud, which specifically includes:

[0014] Use OpenCV's ArUco recognition algorithm to identify the contour boundaries of the ArUco marker from the RGB frame;

[0015] Create a marker mask based on the contour boundary of the ArUco marker;

[0016] Apply the marking mask to the RGB frame and the depth frame respectively to obtain the extracted RGB frame data and depth frame data;

[0017] The three-dimensional point cloud is reconstructed based on the extracted RGB frame data and depth frame data to obtain the marked point cloud.

[0018] Preferably, obtaining RGB-D camera data containing ArUco markers specifically includes:

[0019] Acquire RGB-D camera data containing ArUco markers at preset intervals or in real time.

[0020] Preferably, after reconstructing a three-dimensional point cloud based on the RGB frame data and depth frame data corresponding to the plane where the ArUco marker is located to obtain the marked point cloud, the method further includes:

[0021] Perform outlier removal on the labeled point cloud.

[0022] Preferably, performing plane segmentation and extraction on the marked point cloud to obtain the marked plane specifically includes:

[0023] The RANSAC plane segmentation algorithm is used to perform plane segmentation and extraction on the marked point cloud to obtain the marked plane.

[0024] Preferably, the process of performing plane segmentation extraction on the marked point cloud using the RANSAC plane segmentation algorithm includes:

[0025] Randomly select three points from the labeled point cloud;

[0026] Determine whether the three selected points are collinear; if so, reselect; if not, determine the candidate plane based on the three selected points;

[0027] Calculate the distance between each point in the marked point cloud and the candidate plane according to a preset distance calculation formula;

[0028] Compare the distance between each point and the candidate plane with the distance threshold. If the distance between a point and the candidate plane is less than the distance threshold, then the point is considered an inlier of the candidate plane.

[0029] Repeat the above steps until the preset maximum number of iterations, and select the candidate plane with the largest number of inliers as the marked plane.

[0030] Preferably, the posture smoothing process is performed on the marked plane, specifically including:

[0031] The pose of the marker plane is smoothed using a three-dimensional pose smoothing algorithm based on Kalman filtering.

[0032] Preferably, a three-dimensional posture smoothing algorithm based on Kalman filtering is used to perform posture smoothing processing on the marking plane, specifically including:

[0033] A Kalman filter is applied to the coordinates of the center point of the marking plane, and a constrained Kalman filter is applied to the normal vector angle.

[0034] In a second aspect, the present invention further proposes a magnetic resonance scanning plane tracking and processing system based on an RGB-D camera, comprising:

[0035] Acquisition module, used to obtain RGB-D camera data containing ArUco markers;

[0036] The marker detection and point cloud segmentation module is used to detect ArUco markers and segment point clouds on RGB-D camera data to obtain a marked point cloud;

[0037] The plane segmentation module is used to perform plane segmentation extraction on the marked point cloud to obtain the marked plane;

[0038] A posture smoothing processing module is used to perform posture smoothing processing on the marking plane, and the marking plane after posture smoothing processing is used as the scanning plane;

[0039] The coordinate system conversion module is used to convert the scanning plane from the camera coordinate system to the magnetic resonance coordinate system to obtain the coordinate matrix of the marking plane in the magnetic resonance coordinate system.

[0040] In a third aspect, the present invention further proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the magnetic resonance scanning plane tracking processing method based on an RGB-D camera as described in any one of the first aspects.

[0041] In a fourth aspect, the present invention further proposes a computer program product, comprising a computer program, which, when executed by a processor, implements the steps of the magnetic resonance scanning plane tracking processing method based on an RGB-D camera as described in any one of the first aspects.

[0042] In the present invention, the proposed magnetic resonance scanning plane tracking processing method, system, medium and product based on RGB-D camera first obtains RGB-D camera data containing ArUco markers through RGB-D camera capture, and then performs ArUco marker detection and point cloud segmentation on the RGB-D camera data to obtain a marked point cloud, so as to transform the tracking processing task from directly locating the specific position of the instrument to determining the plane containing the instrument, thereby allowing the sensor to have a larger error while still meeting the accuracy requirements of the subsequent navigation and tracking process; then, considering the accuracy limitation of the RGB-D camera, the marked point cloud is plane segmented and extracted to obtain a relatively complete and accurate marked plane to ensure higher accuracy than a single point; then, the marked plane is posture smoothed, and the marked plane after posture smoothing is used as the scanning plane to suppress tracking jitter and improve tracking stability to obtain a more stable tracking result; finally, the scanning plane is converted from the camera coordinate system to the magnetic resonance coordinate system to obtain a coordinate matrix of the marked plane in the magnetic resonance coordinate system, so as to facilitate subsequent precise navigation based on the coordinate matrix of the marked plane in the magnetic resonance coordinate system.

[0043] The present invention reduces the stringent requirements for sensor accuracy by accurately tracking the plane of the instrument rather than single-point positioning, and can calibrate tracking errors, which can ensure that the coordinate matrix meets the accuracy used in subsequent navigation tracking. Moreover, through ArUco marker-assisted point cloud segmentation, the amount of calculation is greatly reduced, and an RGB-D camera is used, which has low cost, high stability, and fast imaging speed, and can achieve real-time tracking or delayed tracking of a preset time. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a flow chart of a magnetic resonance scanning plane tracking processing method based on an RGB-D camera in one embodiment of the present invention. DETAILED DESCRIPTION

[0045] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments of the present invention can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0046] First, as Figure 1 As shown, the present invention proposes a magnetic resonance scanning plane tracking processing method based on an RGB-D camera, comprising:

[0047] Get RGB-D camera data containing ArUco markers;

[0048] Perform ArUco marker detection and point cloud segmentation on RGB-D camera data to obtain a marked point cloud;

[0049] Perform plane segmentation and extraction on the marked point cloud to obtain the marked plane;

[0050] Performing posture smoothing on the marker plane, and using the smoothed marker plane as the scanning plane;

[0051] The scanning plane is transformed from the camera coordinate system to the magnetic resonance coordinate system to obtain the coordinate matrix of the marking plane in the magnetic resonance coordinate system.

[0052] The present invention first obtains RGB-D camera data containing ArUco markers through RGB-D camera capture, and then performs ArUco marker detection and point cloud segmentation on the RGB-D camera data to obtain a marked point cloud, so as to transform the tracking processing task from directly locating the specific position of the instrument to determining the plane containing the instrument, thereby allowing the sensor to have a larger error while still meeting the accuracy requirements of the subsequent navigation and tracking process; then, considering the accuracy limitation of the RGB-D camera, the marked point cloud is subjected to plane segmentation and extraction to obtain a relatively complete and accurate marked plane to ensure higher accuracy than a single point; the marked plane is then subjected to posture smoothing processing, and the marked plane after posture smoothing processing is used as the scanning plane to suppress tracking jitter and improve tracking stability to obtain a more stable tracking result; finally, the scanning plane is converted from the camera coordinate system to the magnetic resonance coordinate system to obtain the coordinate matrix of the marked plane in the magnetic resonance coordinate system, which facilitates subsequent precise navigation according to the coordinate matrix of the marked plane in the magnetic resonance coordinate system.

[0053] The present invention reduces the stringent requirements for sensor accuracy by accurately tracking the plane of the instrument rather than single-point positioning, and can calibrate tracking errors, which can ensure that the coordinate matrix meets the accuracy used in subsequent navigation tracking. Moreover, through ArUco marker-assisted point cloud segmentation, the amount of calculation is greatly reduced, and in conjunction with the use of an RGB-D camera, it has low cost, high stability, and fast imaging speed, and can achieve real-time tracking or delayed tracking of a preset time.

[0054] It's important to note that detecting and segmenting the handle plane from a omnidirectional point cloud is not only computationally intensive but also susceptible to noise, which can affect accuracy. To improve detection efficiency and reliability, this invention first attaches an ArUco marker with a specified ID (ID=0) to the handle plane. An RGB-D camera then simultaneously acquires a color (RGB) image and depth (D) information of the target area.

[0055] In one embodiment, obtaining RGB-D camera data containing ArUco markers includes:

[0056] Acquire RGB-D camera data containing ArUco markers in real time.

[0057] In another embodiment, obtaining RGB-D camera data containing ArUco markers specifically includes:

[0058] Acquire RGB-D camera data containing ArUco markers at preset time intervals.

[0059] The RGB-D camera data in this embodiment includes a depth frame and RGB frames .

[0060] In this embodiment, ArUco marker detection and point cloud segmentation are performed on the RGB-D camera data to obtain a marked point cloud, specifically including:

[0061] Use OpenCV's ArUco recognition algorithm to identify the contour boundaries of the ArUco marker from the RGB frame;

[0062] Create a marker mask based on the contour boundary of the ArUco marker ;

[0063] Apply the marking mask to the RGB frame and the depth frame respectively to obtain the extracted RGB frame data and depth frame data;

[0064] Reconstruct the 3D point cloud based on the extracted RGB frame data and depth frame data to obtain the marked point cloud .

[0065] This embodiment uses visual markers to quickly lock the target area and apply the divided masks to the RGB information and D information at the same time to specify the acquisition of the marked point cloud. This method can significantly reduce the amount of point cloud data that needs to be processed, which is the key to achieving real-time performance.

[0066] Since ArUco detection may have errors, the marked point cloud The boundary area may contain outliers that deviate from the main body. In order to solve this problem, in a further embodiment, a three-dimensional point cloud is reconstructed based on the RGB frame data and depth frame data corresponding to the plane where the ArUco marker is located to obtain a marked point cloud. After that, it also includes:

[0067] Marking point cloud Remove outliers to obtain a more complete and accurate marked point cloud.

[0068] In this embodiment, plane segmentation and extraction are performed on the marked point cloud to obtain the marked plane, which specifically includes:

[0069] The RANSAC plane segmentation algorithm is used to perform plane segmentation and extraction on the marked point cloud to obtain the marked plane.

[0070] This embodiment uses RANSAC plane segmentation for positioning to obtain an accurate magnetic resonance scanning plane. Only the obtained plane has higher accuracy than a single point.

[0071] Among them, the process of plane segmentation extraction of the marked point cloud using the RANSAC plane segmentation algorithm includes:

[0072] Randomly select three points from the marked point cloud; determine whether the three selected points are collinear; if so, reselect; if not, calculate the plane equation in 3D space based on the three selected points The corresponding parameters a, b, c, d of are used to determine the candidate plane according to the parameters a, b, c, d; wherein x, y, and z are the three coordinate axes of the three-dimensional Cartesian coordinate system, representing the three orthogonal directions in space respectively;

[0073] Calculate each point in the marked point cloud according to the preset distance calculation formula The distance to the candidate plane ;

[0074] The distance of each point Distance threshold For comparison, if a certain point , then the point is regarded as the interior point of the candidate plane;

[0075] Repeat the above steps until the preset maximum number of iterations times, and select the candidate plane with the largest number of inliers as the marked plane.

[0076] The distance threshold and the preset maximum number of iterations are empirical parameters and can be manually adjusted according to scene requirements.

[0077] Among them, the preset distance calculation formula is .

[0078] This embodiment uses an improved RANSAC algorithm to process the target point cloud. Using an intelligent sampling strategy, it quickly locates the largest plane. It automatically iterates after manually setting the number of iterations and distance threshold, optimizing computational efficiency while ensuring accuracy. After determining the principal plane, the system calculates the required parameters, which serve as the basis for subsequent processing. This embodiment includes an automatic verification mechanism to ensure the rationality of the output parameters.

[0079] In order to improve the stability of plane tracking, in this embodiment, posture smoothing processing is performed on the marked plane, specifically including:

[0080] The pose of the marker plane is smoothed using a three-dimensional pose smoothing algorithm based on Kalman filtering.

[0081] Among them, the three-dimensional posture smoothing algorithm based on Kalman filtering is used to smooth the posture of the marking plane, which specifically includes:

[0082] A Kalman filter is applied to the coordinates of the center point of the marking plane, and a constrained Kalman filter is applied to the normal vector angle.

[0083] It is important to understand that for an infinite plane without boundaries, it can be represented by a point on the plane. and the plane normal vector Unique definition. In this embodiment, the plane normal vector Expressed in a spherical coordinate system, the angle between the normal vector and the positive z-axis is the polar angle The angle between the projection line of the normal vector on the xy plane and the positive x-axis is the azimuth Assume that the coplanar points and the plane normal vector are The disturbances all follow a Gaussian distribution and therefore meet the conditions for applying Kalman filtering.

[0084] This embodiment applies Kalman filtering to the following two groups of variables: 1) Coordinates of the measured coplanar points The linear velocity of its disturbance; 2) The angle of measurement and and their angular velocities. Restricted to within the range, while Restricted to Therefore, when calculating the measurement residuals, The error in must be limited to Inside, The error in must be limited to By applying these constraints and using Kalman filtering, a more stable 、 as well as and The stability of plane tracking is improved by using the stabilized estimate.

[0085] In this embodiment, the scanning plane is converted from the camera coordinate system to the magnetic resonance coordinate system to obtain the coordinate matrix of the marking plane in the magnetic resonance coordinate system, which specifically includes:

[0086] The coordinate system transformation matrix obtained by the registration is used to transform the parameters of the scanning plane in the camera coordinate system into the magnetic resonance coordinate system to obtain the coordinate matrix of the scanning plane in the magnetic resonance coordinate system.

[0087] Of course, before using the coordinate system transformation matrix obtained by registration to transform the parameters of the scanning plane in the camera coordinate system into the magnetic resonance coordinate system and obtain the coordinate matrix of the scanning plane in the magnetic resonance coordinate system, the following steps are also required:

[0088] Get the coordinate system transformation matrix obtained by registration.

[0089] Specifically, the camera must first be calibrated to obtain its intrinsic parameter matrix and distortion parameters, allowing for subsequent correction of the raw data captured by the camera. A coordinate system alignment module is also required. Using the same registration phantom captured by the RGB-D camera and the MRI scanning system, the transformation matrix between the two coordinate systems is obtained, facilitating subsequent coordinate system conversions.

[0090] In one specific embodiment, a magnetic resonance scan plane tracking processing method based on an RGB-D camera is applied to a 0.35T magnetic resonance system. The acquisition time of a single-layer image of a 0.35T magnetic resonance imaging is generally more than 0.1 seconds. The time resolution of the magnetic resonance scan plane tracking processing method in this embodiment must be greater than 10 frames per second, that is, a single frame of data must be processed within 100ms. For tracking accuracy, it is assumed that the distance between the center of the tracking handle and the needle tip is , the distance deviation and normal vector angle deviation of the plane tracking algorithm are and , the slice thickness of the magnetic resonance image is s, the slice spacing is g, and the number of image scanning slices is n. If the needle body and needle tip are to be completely covered in the scanned image, then it is necessary to meet For example, if the distance between the center of the tracking handle and the needle tip is 140 mm, the MRI slice thickness is 5.5 mm, and the slice spacing is 0.5 mm, then when the estimated tracking plane distance deviation is less than 0.3 mm and the angle deviation is less than 1°, a single-slice scan image can include the entire needle; when the plane distance deviation is less than 1.2 mm and the angle deviation is less than 3°, a maximum of three consecutive image scans can include the entire needle.

[0091] In summary, the MRI scanning plane tracking and processing method based on an RGB-D camera proposed in the present invention has excellent magnetic compatibility in an MRI environment, can significantly reduce costs, has high accuracy and stability, and can achieve tracking performance with a plane distance deviation of <1.2mm and an angular deviation of <2.5°. It can ensure that up to three layers of continuous image scans can contain complete frames, and the processing time for a single frame is <30ms, meeting the real-time requirement of >10fps. The angular standard deviation of the plane tracked using the RGB-D camera is reduced from 2.64° to 0.61°.

[0092] In a second aspect, the present invention further proposes a magnetic resonance scanning plane tracking and processing system based on an RGB-D camera, comprising:

[0093] Acquisition module, used to obtain RGB-D camera data containing ArUco markers;

[0094] The marker detection and point cloud segmentation module is used to detect ArUco markers and segment point clouds on RGB-D camera data to obtain a marked point cloud;

[0095] The plane segmentation module is used to perform plane segmentation extraction on the marked point cloud to obtain the marked plane;

[0096] A posture smoothing processing module is used to perform posture smoothing processing on the marking plane, and the marking plane after posture smoothing processing is used as the scanning plane;

[0097] The coordinate system conversion module is used to convert the scanning plane from the camera coordinate system to the magnetic resonance coordinate system to obtain the coordinate matrix of the marking plane in the magnetic resonance coordinate system.

[0098] The acquisition module uses an RGB-D camera to synchronously acquire RGB-D camera data of the target area containing the ArUco marker. Utilizing the RGB-D camera's active infrared structured light technology, the projected coded pattern is captured through triangulation to calculate the corresponding depth information.

[0099] Among them, the marker detection and point cloud segmentation module uses OpenCV's ArUco recognition algorithm to identify the contour boundary of the ArUco marker from the RGB frame; according to the contour boundary of the ArUco marker, a marker mask is created ; Apply the marker mask to the RGB frame and depth frame respectively to obtain the extracted RGB frame data and depth frame data; Reconstruct the 3D point cloud based on the extracted RGB frame data and depth frame data to obtain the marker point cloud .

[0100] In a further embodiment, the 3D point cloud is reconstructed based on the RGB frame data and depth frame data corresponding to the plane where the ArUco mark is located to obtain the marked point cloud. After that, it also includes:

[0101] Marking point cloud Remove outliers to obtain a more complete and accurate marked point cloud.

[0102] The plane segmentation module uses the RANSAC plane segmentation algorithm to perform plane segmentation extraction on the marked point cloud to obtain the marked plane.

[0103] To address transient fluctuations in the original plane parameters, the attitude smoothing module deploys a dual-channel Kalman filter. The first channel processes the three-dimensional coordinates and velocity of the marker plane's center point, applying a Kalman filter to the coordinates of the marker plane's center point. The second channel specifically addresses angular variations in the normal vector, applying a constrained Kalman filter to the normal vector angle. The filtering process fully considers the geometric constraints of three-dimensional space, particularly by periodically correcting the angle parameters to avoid directional jumps. The smoothed output significantly improves tracking stability while maintaining responsiveness to rapid motion.

[0104] In a third aspect, the present invention further proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the magnetic resonance scanning plane tracking processing method based on an RGB-D camera as described in any one of the first aspects.

[0105] In a fourth aspect, the present invention further proposes a computer program product, comprising a computer program, which, when executed by a processor, implements the steps of the magnetic resonance scanning plane tracking processing method based on an RGB-D camera as described in any one of the first aspects.

[0106] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A magnetic resonance scanning plane tracking processing method based on an RGB-D camera, characterized in that: include: Get RGB-D camera data containing ArUco markers; Perform ArUco marker detection and point cloud segmentation on RGB-D camera data to obtain a marked point cloud; Perform plane segmentation and extraction on the marked point cloud to obtain the marked plane; Performing posture smoothing on the marker plane, and using the smoothed marker plane as the scanning plane; The scanning plane is transformed from the camera coordinate system to the magnetic resonance coordinate system to obtain the coordinate matrix of the marking plane in the magnetic resonance coordinate system.

2. The magnetic resonance scanning plane tracking processing method based on an RGB-D camera according to claim 1, characterized in that: Get RGB-D camera data containing ArUco markers, specifically including: Acquire RGB-D camera data containing ArUco markers at preset intervals or in real time.

3. The magnetic resonance scanning plane tracking processing method based on an RGB-D camera according to claim 1, characterized in that: RGB-D camera data includes depth frames and RGB frames; Among them, the ArUco marker detection and point cloud segmentation are performed on the RGB-D camera data to obtain the marked point cloud, which specifically includes: Use OpenCV's ArUco recognition algorithm to identify the contour boundaries of the ArUco marker from the RGB frame; Create a marker mask based on the contour boundary of the ArUco marker; Apply the marking mask to the RGB frame and the depth frame respectively to obtain the extracted RGB frame data and depth frame data; The three-dimensional point cloud is reconstructed based on the extracted RGB frame data and depth frame data to obtain the marked point cloud.

4. The magnetic resonance scanning plane tracking processing method based on an RGB-D camera according to claim 1, characterized in that: After reconstructing the 3D point cloud based on the RGB frame data and depth frame data corresponding to the plane where the ArUco marker is located to obtain the marker point cloud, it also includes: Perform outlier removal on the labeled point cloud.

5. The magnetic resonance scanning plane tracking processing method based on RGB-D camera according to claim 1, characterized in that: Perform plane segmentation and extraction on the marked point cloud to obtain the marked plane, which specifically includes: The RANSAC plane segmentation algorithm is used to perform plane segmentation and extraction on the marked point cloud to obtain the marked plane.

6. The magnetic resonance scanning plane tracking processing method based on RGB-D camera according to claim 1, characterized in that: Perform posture smoothing on the marker plane, including: The pose of the marker plane is smoothed using a three-dimensional pose smoothing algorithm based on Kalman filtering.

7. The magnetic resonance scanning plane tracking processing method based on an RGB-D camera according to claim 6, characterized in that: The three-dimensional posture smoothing algorithm based on Kalman filtering is used to smooth the posture of the marker plane, which includes: Apply Kalman filtering to the coordinates of the plane center points of the marking planes respectively; Apply a constrained Kalman filter to the normal angle of the marker plane.

8. A magnetic resonance scanning plane tracking and processing system based on an RGB-D camera, characterized in that: include: Acquisition module, used to obtain RGB-D camera data containing ArUco markers; The marker detection and point cloud segmentation module is used to detect ArUco markers and segment point clouds on RGB-D camera data to obtain a marked point cloud; The plane segmentation module is used to perform plane segmentation extraction on the marked point cloud to obtain the marked plane; A posture smoothing processing module is used to perform posture smoothing processing on the marking plane, and the marking plane after posture smoothing processing is used as the scanning plane; The coordinate system conversion module is used to convert the scanning plane from the camera coordinate system to the magnetic resonance coordinate system to obtain the coordinate matrix of the marking plane in the magnetic resonance coordinate system.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the magnetic resonance scanning plane tracking processing method based on an RGB-D camera are implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the magnetic resonance scanning plane tracking processing method based on an RGB-D camera are implemented.

Images