Optimized electronic face-bow system based on binocular infrared vision and passive infrared target
Patent Information
- Application Number
- CN202611069142.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-17
- Publication Date
- 2026-08-28
AI Technical Summary
但存在以下问题，1)精度依赖操作经验：手动调整易引入人为误差，影响修复体与咬合的匹配度；2)静态记录：仅能获取静态颌位置关系，无法动态捕捉下颌运动(如咀嚼、说话时的轨迹)；3)数据不可量化：依赖石膏模型和机械转移，缺乏数字化数据支持计算机辅助设计(CAD)或分析
[0026]This invention uses a square media target plate as the test object and employs a 7x5 and 5x7 three-ring marker array to distinguish between the upper and lower jaws. The three-ring pattern provides cocentering, radius ratio, and coplanar constraints at large tilt angles, and is used in conjunction with array rigidity, array plane, binocular measured extrinsic parameters, and random sampling consistency interior point screening for target plate pose calculation. Compared with existing technologies, this invention can obtain continuous, stable, and low-latency jaw movement trajectories, with a repeatability error of less than 0.02 mm at the target plate center.
Smart Images
Figure CN122643070A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a technology in the fields of digital oral diagnosis and treatment, binocular vision three-dimensional measurement, and jaw kinematics, specifically an optimized electronic facebow system based on binocular infrared vision and a passive infrared target. Background Technology
[0002] Traditional mechanical facebows are used to record the spatial relationship between the maxilla and temporomandibular joint, assisting in occlusal analysis during dental prostheses (such as complete dentures and crowns) or orthodontic treatment. However, they have the following problems: 1) Accuracy depends on operator experience: manual adjustment is prone to introducing human error, affecting the fit between the prosthesis and the bite; 2) Static recording: only static jaw position relationships can be obtained, and dynamic mandibular movements (such as the trajectory during chewing and speaking) cannot be captured; 3) Data cannot be quantified: relying on plaster models and mechanical transfer, lacking digital data to support computer-aided design (CAD) or analysis. Summary of the Invention
[0003] To address the aforementioned shortcomings of existing technologies, this invention proposes an optimized electronic face bow system based on binocular infrared vision and a passive infrared target. By using a specially constructed infrared target in conjunction with a joint optimization algorithm, it is possible to obtain continuous, stable, and low-latency jawbone movement trajectories, with a repeatability error of less than 0.02 mm at the center of the target plate.
[0004] This invention is achieved through the following technical solution:
[0005] This invention relates to an optimized electronic face bow system based on binocular infrared vision and a passive infrared target, comprising: a three-ring passive infrared target, a binocular infrared camera, and a control module connected to the binocular infrared camera, wherein: after the control module triggers the left and right infrared cameras to synchronously expose globally and acquire binocular infrared images, it calculates the actual relative posture, actual included angle, baseline relationship, and continuous motion trajectory of the upper and lower jaw target plates of the three-ring passive infrared target.
[0006] The binocular infrared camera is preferably configured with the optical center of the left eye camera as the origin and the coordinate system of the left eye camera as the system coordinate system.
[0007] Both the left and right infrared cameras are installed tilted inward toward the central measurement area. The nominal angle between the optical axis of each camera and the vertical direction of the machine is preferably 20°. The actual relative posture, actual angle and baseline relationship are all obtained by dual-target calibration and measurement.
[0008] The three-ring passive infrared target includes: a maxillary square media target plate fixedly installed on the upper dentition and a mandibular square media target plate fixedly installed on the lower dentition; the two target plates do not rotate or slide relative to the fixed dentition.
[0009] The aforementioned media target plate is an array of three ring markers with non-uniform ring widths, non-uniform ring spacing, and non-uniform radius ratios, directly printed on the surface of a square media target plate. Specifically, the maxillary target plate uses a 7-row × 5-column array, and the mandibular target plate uses a 5-row × 7-column array. The actual radius, ring width, ring spacing, center design coordinates, and actual deviations of the three rings are obtained through measurement and used as geometric priors for positioning calculations.
[0010] The optimized electronic face bow system is further equipped with an infrared fill light, which is used to project onto the infrared target through a diffuse reflection lens to obtain higher quality target imaging results, thereby improving measurement accuracy.
[0011] The control module includes: an outer ring fast candidate unit, a three-ring joint centering unit, a random sampling consistency screening unit, a target plate center calculation unit, and a pose temporal filtering unit. Specifically: the outer ring fast candidate unit locates the outer ring edge in the entire image and reduces the calculation area; the three-ring joint centering unit calculates the marker center based on the coplanarity, concentricity, radius ratio, ring width, ring spacing, and actual deviation of the three rings; the random sampling consistency screening unit eliminates outlier markers based on array rigid body residuals, binocular reprojection errors, and array plane residuals; the target plate center calculation unit only averages the three-dimensional centers of the inner point markers; and the pose temporal filtering unit outputs a continuous, low-jitter, and low-latency relative motion trajectory of the upper and lower jaws.
[0012] This invention relates to an application based on the above system. By installing a three-ring passive infrared target and using a calibrated binocular infrared camera to acquire images of the infrared reflective target, the three-dimensional coordinates and normal vector of the target in the image are obtained through joint optimization calculation, and the corresponding maxillary and mandibular bone models and their motion trajectories are generated to realize electronic face bow.
[0013] The calibration process involves: acquiring multiple sets of synchronous left and right images of a 940nm backlit checkerboard calibration board using binocular infrared cameras; calibrating the left and right cameras separately; and removing abnormal images based on reprojection errors. Then, using the retained images, the rotation matrix R, translation vector T, baseline relationship, and actual angle between the right and left cameras are calculated. The optical center of the left camera is used as the origin, and the left camera coordinate system is used as the system coordinate system. The calibration board dimensions are used to reconstruct the actual scale of the 3D measurement.
[0014] The installation of the three-ring passive infrared target refers to: bonding and fixing the maxillary square media target plate and the mandibular square media target plate to the corresponding dentition with UV curing adhesive; the maxillary target plate adopts a 7-row × 5-column three-ring marker array, and the mandibular target plate adopts a 5-row × 7-column three-ring marker array, and recording the array orientation reference during the initial installation and calibration.
[0015] The aforementioned three-dimensional coordinates and normal vectors are obtained in the following way:
[0016] The first step is to obtain the candidate regions of each three-ring marker in each set of binocular infrared camera images, i.e., the synchronous frames, based on the outer ring edge. Then, the candidate regions are corrected for frontal view based on the pose prediction of the previous frame. In this region, joint centering is performed with constraints such as the coplanarity and concentricity of the three rings, the actual radius ratio, the ring width, the ring spacing, and the actual deviation, to obtain the center pixel coordinates of the marker. Then, based on the topology results of the three-ring passive infrared target, the array orientation reference, and the prediction of consecutive frames, the center of the marker is matched to the design position of the corresponding target plate, thereby distinguishing the maxillary target plate and the mandibular target plate.
[0017] The second step is to match the center of the same three-ring marker in the left and right images at the same shooting time, and to perform triangulation based on the actual extrinsic parameters obtained by dual-target calibration to obtain the initial three-dimensional center coordinates of the marker in the system coordinate system with the optical center of the left eye camera as the origin.
[0018] The third step involves using the target plate array coordinates as a rigid body template and employing a random sampling consensus algorithm to generate candidate target plate poses from the minimum set of non-collinear markers. Markers that simultaneously satisfy the rigid body residual, binocular reprojection error, and array plane residual thresholds are used as inliers. The poses of the upper and lower jaw target plates, the 3D centers of the inlier markers, and the binocular extrinsic parameter corrections are then jointly optimized. Specifically: , , , Where: h = 1, 2, 3 are the outer ring, middle ring, and inner ring respectively, and j is the sequence number of the contour sampling point on the ring. Let h be the design radius of the h-th ring. The polar angle of the sampling point. The design radius difference between adjacent rings; m = 1 and 2 are the maxillary and mandibular target plates, respectively; i is the number of the three ring markers within the target plate. Let i be the three-dimensional center of the i-th marker in the m-th target plate. For the set of interior point identifier numbers, Let Θ be the average value of the three-dimensional centers of the interior markers, and let Θ be the set of variables to be optimized. To optimize the results; Erep represents the weighted binocular reprojection error, and Erig, Epln, Esyn, Econ, Emet, Eaxis, Etilt, Emot, and Eext represent the array rigidity, array plane, synchronous imaging, three-ring coplanarity, measured geometric parameters of the image, array azimuth, large tilt angle binocular consistency, relative motion continuity, and online extrinsic parameter correction error, respectively; each λ represents the corresponding non-negative weight. For observation weights, The measured pixel coordinates of the center of the marker. Let be the projection function of the k-th camera, where k = 1 and 2 are the left and right cameras, respectively.
[0019] The joint optimization objectives include error terms such as binocular reprojection, array rigidity, array plane, synchronous shooting, three-ring coplanarity, measured geometric parameters of the image, array orientation, binocular consistency at large tilt angles, relative motion continuity, and online extrinsic parameter correction.
[0020] The fourth step involves averaging only the three-dimensional centers of the inlier markers to obtain the target board's center position. The target board's pose is then determined by the target board's planar normal vector and the array orientation. For continuous frame poses, the current relative pose is predicted using the filtering results of the previous frame. The filtering gain is then adaptively adjusted based on the three-ring ellipse fitting residual, the consistency residual of the measured geometric parameters, the left and right epipolar deviation, the array rigidity residual, the inlier ratio, and the predicted reprojection deviation. This process outputs the continuous motion trajectory of the mandible relative to the maxilla.
[0021] The aforementioned maxillary and mandibular bone models and their motion trajectories were obtained in the following way:
[0022] The first step is to predict the current relative pose based on the filtering results and motion state of the previous set of binocular infrared camera images, i.e., the synchronization frame.
[0023] The second step is to use the joint optimization output of the current synchronization frame as the observation relative pose and to perform anomaly gating based on the current measurement reliability.
[0024] The third step is to update the predicted relative pose in the SE(3) space according to the adaptive filtering gain. This update uses the current observation relative prediction pose increment as the correction amount, and the correction amount is weighted according to the measurement reliability. When the measurement quality deteriorates, the observation update gain is reduced, specifically as follows: ,in: Let t be the relative pose of the mandible with respect to the maxilla after filtering. The predicted relative pose is obtained based on the filtering results of the previous frame and the motion state; The relative pose of the observations is obtained through current synchronous binocular measurements and joint optimization; The adaptive filter gain is determined by the current measurement reliability and ranges from 0 to 1. For pose inverse operation, To transform the SE(3) pose into a logarithmic mapping of Lie algebra increments, This is to transform the Lie algebra increment back to the exponential mapping of the SE(3) pose.
[0025] The fourth step is to apply the filtered relative pose of the mandible to the maxilla to the dental model established by scanning or CBCT to obtain the continuous jawbone motion trajectory. Technical effect
[0026] This invention uses a square media target plate as the test object and employs a 7x5 and 5x7 three-ring marker array to distinguish between the upper and lower jaws. The three-ring pattern provides cocentering, radius ratio, and coplanar constraints at large tilt angles, and is used in conjunction with array rigidity, array plane, binocular measured extrinsic parameters, and random sampling consistency interior point screening for target plate pose calculation. Compared with existing technologies, this invention can obtain continuous, stable, and low-latency jaw movement trajectories, with a repeatability error of less than 0.02 mm at the target plate center. Attached Figure Description
[0027] Figure 1 (a)-(d) are schematic diagrams of the binocular infrared camera of the present invention;
[0028] Figure 2 Schematic diagram of a 940nm backlight checkerboard calibration board;
[0029] Figure 3 This is a flowchart of an implementation example;
[0030] Figure 4 A schematic diagram of a square dielectric target plate, a three-ring marker, and heterogeneous arrays of the upper and lower jaws;
[0031] Figure 5 This is a schematic diagram of the system of the present invention. Detailed Implementation
[0032] This embodiment relates to an optimized electronic face bow system based on binocular infrared vision and a passive infrared target, including: as follows Figure 1 The binocular infrared camera, a pair of 940nm infrared illuminators, a three-ring passive infrared target, and such Figure 5 The control module triggers the binocular infrared camera to synchronously expose the entire image. Both binocular infrared cameras are installed facing the measurement area near the teeth and tilted inward. The nominal angle between the optical axis of each camera and the vertical direction of the machine is 20°. The installation interval is 12cm, which is the distance between the camera baselines.
[0033] like Figure 1As shown, the three-ring passive infrared target reflects 940nm infrared light, which is consistent with the infrared camera's cutoff wavelength. It includes a pair of square dielectric target plates for fixing on the upper and lower dentition, respectively, with 7 rows × 5 columns and 5 rows × 7 columns of three-ring marker arrays printed on their surfaces. Each three-ring marker consists of an outer ring, a middle ring, and an inner ring. Its actual radius, ring width, ring spacing, center design coordinates, and actual deviation are established through measurement. Specifically, the design center coordinates of each marker are first determined according to the 7 rows × 5 column array for the maxilla and the 5 rows × 7 column array for the mandible, and the design of the three rings is determined accordingly. The radius, ring width, and ring spacing are calculated, and the three rings are located on the same plane with non-uniform ring widths, non-uniform ring spacings, and non-uniform radius ratios. The three-ring marker array is then directly printed onto the surface of a square media target plate. After fabrication, the radius, ring width, ring spacing, and actual center coordinates of the three rings of each marker are measured with a measurement accuracy of 2μm. The deviation of the actual center coordinates from the designed center coordinates is calculated. The measured actual geometric parameters and deviations are used as geometric priors for target plate positioning, and are used for three-ring joint centering, array template matching, and outlier marker screening.
[0034] like Figure 3 As shown, this embodiment relates to an optimized electronic face bow method based on system-wide binocular infrared vision and passive infrared targets, specifically including:
[0035] Step 1: Binocular camera calibration: Multiple sets of synchronous images were acquired using a 940nm backlit checkerboard calibration board. The left and right cameras were calibrated separately, and abnormal images were removed based on reprojection errors. The actual rotation matrix, translation vector, baseline relationship, and actual angle between the right and left cameras were obtained.
[0036] Step 2, Infrared target installation: The upper and lower jaw square media target plates are bonded and fixed to the corresponding dental arches with UV-curing adhesive; the upper jaw target plate adopts a 7-row × 5-column array, and the lower jaw target plate adopts a 5-row × 7-column array, and the array orientation reference at the initial installation is recorded.
[0037] The third step is target pose measurement: the center of the three-ring markers in the synchronous left and right images is matched and reconstructed by binoculars; the outlier markers are removed by using the designed array as a template and the random sampling consistency algorithm is used to remove the outlier markers. The center of the target board is obtained by averaging the three-dimensional center of the inlier markers, and the pose is determined by the plane normal vector of the target board and the orientation of the array.
[0038] Step 4: Calculation of jaw movement trajectory: Calculate the relative pose of the mandible with respect to the maxilla based on the poses of the maxillary and mandibular target plates in consecutive frames, and use pose temporal prediction and adaptive filtering to suppress single-frame jitter and occasional anomalies to obtain the continuous jaw movement trajectory.
[0039] The repeated positioning experiment verified that, with the target plate maintained in the same installation state, repeated measurements were taken. The three-dimensional center of the inner point markers in each measurement was selected according to the array template and then averaged to obtain the target plate center. The deviation between the target plate center of each measurement and the average value of the repeated measurements was taken as the repeated positioning error. The experimental results show that the repeated positioning error of the target plate center is less than 0.02 mm.
[0040] Compared with existing technologies, this system improves the stability of spatial positioning of the maxillary and mandibular target plates and the reliability of continuous motion trajectories by combining large-angle three-ring joint centering, array template inner point screening and pose temporal filtering.
[0041] The above-described specific implementations can be partially adjusted by those skilled in the art in different ways without departing from the principles and purpose of the present invention. The scope of protection of the present invention is defined by the claims and is not limited to the above-described specific implementations. All implementation schemes within the scope of the claims are bound by the present invention.
Claims
1. An optimized electronic face bow system based on binocular infrared vision and a passive infrared target, characterized in that, include: The system consists of a three-ring passive infrared target, a binocular infrared camera, and a control module connected to the binocular infrared camera. The control module triggers the left and right infrared cameras to synchronously expose globally and acquire binocular infrared images. Then, it calculates the actual relative posture, actual included angle, baseline relationship, and continuous motion trajectory of the upper and lower jaw target plates of the three-ring passive infrared target.
2. The optimized electronic face bow system based on binocular infrared vision and a passive infrared target as described in claim 1, characterized in that, The aforementioned three-ring passive infrared target includes: a maxillary square media target plate fixedly set on the upper dentition and a mandibular square media target plate fixedly set on the lower dentition. The media target plate is an array of three-ring markers with non-uniform ring width, non-uniform ring spacing, and non-uniform radius ratio directly printed on the surface of the square media target plate. The maxillary target plate adopts a 7-row × 5-column array, and the mandibular target plate adopts a 5-row × 7-column array. The actual radius, ring width, ring spacing, center design coordinates, and actual deviation of the three rings are obtained by measurement and used as geometric priors for positioning calculation.
3. The optimized electronic face bow system based on binocular infrared vision and a passive infrared target as described in claim 1, characterized in that, Both the left and right infrared cameras are installed tilted inward toward the central measurement area. The nominal angle between the optical axis of each camera and the vertical direction of the machine is 20°. The optical center of the left camera is the origin, and the coordinate system of the left camera is the system coordinate system.
4. The optimized electronic face bow system based on binocular infrared vision and a passive infrared target as described in claim 1, characterized in that, The control module includes: an outer ring fast candidate unit, a three-ring joint centering unit, a random sampling consistency screening unit, a target plate center calculation unit, and a pose temporal filtering unit. Specifically: the outer ring fast candidate unit locates the outer ring edge in the entire image and reduces the calculation area; the three-ring joint centering unit calculates the marker center based on the coplanarity, concentricity, radius ratio, ring width, ring spacing, and actual deviation of the three rings; the random sampling consistency screening unit eliminates outlier markers based on array rigid body residuals, binocular reprojection errors, and array plane residuals; the target plate center calculation unit only averages the three-dimensional centers of the inner point markers; and the pose temporal filtering unit outputs a continuous, low-jitter, and low-latency relative motion trajectory of the upper and lower jaws.
5. The optimized electronic face bow system based on binocular infrared vision and a passive infrared target as described in claim 2, characterized in that, The aforementioned three-ring marker array with non-uniform ring width, non-uniform ring spacing, and non-uniform radius ratio is established through metrological measurement of its actual radius, ring width, ring spacing, center design coordinates, and actual deviation. Specifically, the design center coordinates, design radius, ring width, and ring spacing of each marker are determined according to a 7-row × 5-column array for the maxilla and a 5-row × 7-column array for the mandible, ensuring that the three rings are located on the same plane and have non-uniform ring width, non-uniform ring spacing, and non-uniform radius ratio. The three-ring marker array is then directly printed onto the surface of a square media target plate. After fabrication, the radius, ring width, ring spacing, and actual center coordinates of the three rings of each marker are measured with a measurement accuracy of 2μm, and the deviation of the actual center coordinates from the design center coordinates is calculated.
6. An application based on the above system, characterized in that, After installing a three-ring passive infrared target and using a calibrated binocular infrared camera to acquire images of the infrared reflective target, the three-dimensional coordinates and normal vector of the target in the image are obtained through joint optimization calculation, and the corresponding maxillary and mandibular bone models and their motion trajectories are generated to realize the electronic face bow. The joint optimization objectives include binocular reprojection, array rigidity, array plane, synchronous shooting, three-ring coplanarity, measured geometric parameters of the image, array orientation, binocular consistency at large tilt angles, relative motion continuity, and online extrinsic parameter correction error terms.
7. The application according to claim 6, characterized in that, The calibration process involves: using a binocular infrared camera to acquire multiple sets of synchronous left and right images of a 940nm backlit checkerboard calibration board, calibrating the left and right cameras respectively, and removing abnormal images based on reprojection errors; then using the retained images to calculate the rotation matrix R, translation vector T, baseline relationship, and actual angle between the right camera and the left camera, with the optical center of the left camera as the origin and the coordinate system of the left camera as the system coordinate system, and the calibration board dimensions used to restore the actual scale of the three-dimensional measurement.
8. The application according to claim 6, characterized in that, The aforementioned three-dimensional coordinates and normal vectors are obtained in the following way: The first step is to obtain the candidate regions of each three-ring marker in each set of binocular infrared camera images, i.e., the synchronous frames, based on the outer ring edge. Then, the candidate regions are corrected for frontal view based on the pose prediction of the previous frame. In this region, joint centering is performed with constraints such as the coplanarity and concentricity of the three rings, the actual radius ratio, the ring width, the ring spacing, and the actual deviation, to obtain the center pixel coordinates of the marker. Then, based on the topology results of the three-ring passive infrared target, the array orientation reference, and the prediction of consecutive frames, the center of the marker is matched to the design position of the corresponding target plate, thereby distinguishing the maxillary target plate and the mandibular target plate. The second step is to match the center of the same three-ring marker in the left and right images at the same shooting time, and to perform triangulation based on the actual external parameters obtained by dual-target calibration to obtain the initial three-dimensional center coordinates of the marker in the system coordinate system with the optical center of the left eye camera as the origin. The third step involves using the target plate array coordinates as a rigid body template and employing a random sampling consensus algorithm to generate candidate target plate poses from the minimum set of non-collinear markers. Markers that simultaneously satisfy the rigid body residual, binocular reprojection error, and array plane residual thresholds are used as inliers. The poses of the upper and lower jaw target plates, the 3D centers of the inlier markers, and the binocular extrinsic parameter corrections are then jointly optimized. Specifically: , , , Where: h = 1, 2, 3 are the outer ring, middle ring, and inner ring respectively, j is the sequence number of the contour sampling point on the ring, and is the design radius of the h-th ring. The polar angle of the sampling point. The design radius difference between adjacent rings; m = 1 and 2 are the maxillary and mandibular target plates, respectively; i is the number of the three ring markers within the target plate. Let i be the three-dimensional center of the i-th marker in the m-th target plate. For the set of interior point identifier numbers, Let Θ be the average value of the three-dimensional centers of the interior markers, and let Θ be the set of variables to be optimized. To optimize the results; Erep represents the weighted binocular reprojection error, and Erig, Epln, Esyn, Econ, Emet, Eaxis, Etilt, Emot, and Eext represent the array rigidity, array plane, synchronous imaging, three-ring coplanarity, measured geometric parameters of the image, array azimuth, large tilt angle binocular consistency, relative motion continuity, and online extrinsic parameter correction error, respectively; each λ represents the corresponding non-negative weight. For observation weights, The measured pixel coordinates of the center of the marker. Let be the projection function of the k-th camera, where k = 1 and 2 are the left and right cameras, respectively; The fourth step involves averaging only the three-dimensional centers of the in-point markers to obtain the target plate center position. The target plate pose is then determined by the target plate plane normal vector and the array orientation. For continuous frame poses, the current relative pose is predicted based on the filtering results of the previous frame. The filtering gain is then adaptively adjusted based on the three-ring ellipse fitting residual, the consistency residual of the measured geometric parameters of the image, the left and right epipolar deviation, the array rigidity residual, the in-point ratio, and the predicted reprojection deviation. This process outputs the continuous motion trajectory of the mandible relative to the maxilla.
9. The application according to claim 6, characterized in that, The aforementioned maxillary and mandibular bone models and their motion trajectories were obtained in the following way: The first step is to predict the current relative pose based on the filtering results and motion state of the previous set of binocular infrared camera images, i.e., the synchronization frame. The second step is to use the joint optimization output of the current synchronization frame as the observation relative pose, and to perform anomaly gating based on the current measurement reliability. The third step is to update the predicted relative pose in the SE(3) space according to the adaptive filtering gain. This update uses the current observation relative prediction pose increment as the correction amount, and the correction amount is weighted according to the measurement reliability. When the measurement quality deteriorates, the observation update gain is reduced, specifically as follows: ,in: Let t be the relative pose of the mandible with respect to the maxilla after filtering. The predicted relative pose is obtained based on the filtering results of the previous frame and the motion state; The relative pose of the observations is obtained through current synchronous binocular measurements and joint optimization; The adaptive filter gain is determined by the current measurement reliability and ranges from 0 to 1. For pose inverse operation, To transform the SE(3) pose into a logarithmic mapping of Lie algebra increments, To transform the Lie algebra increment back to the exponential mapping of the SE(3) pose; The fourth step is to apply the filtered relative pose of the mandible to the maxilla to the dental model established by scanning or CBCT to obtain the continuous jawbone motion trajectory.