A method and system for verifying the circular deployment of visual sensors based on field-of-view overlap analysis

By using the field-of-view overlap analysis method, the circular layout of the visual sensor was optimized, which solved the problem of weak observation areas caused by side wall occlusion in the circular layout of the visual sensor, improved the accuracy and reliability of detection, and reduced the risk of missing microcracks.

CN123118552APending Publication Date: 2026-11-17XIAN HIGH TECH AEH INDAL METROLOGY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611351198.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-02
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In the overhaul of high-pressure turbine disks of aero engines, the existing circular layout of visual sensors fails to effectively consider the obstruction of the tenon sidewalls, resulting in a narrow strip of weak observation area at the junction of adjacent fields of view, which increases the risk of missing microcracks.

Method used

By using a field-of-view overlap analysis method, the geometric data of the target tooth root transition arc and adjacent vision sensors are obtained. Arc sampling, field-of-view filtering and sidewall occlusion determination are performed. The visible coverage arc is fitted, and the tangential deviation and curvature deviation are quantified. Candidate correction levels are queried using a two-dimensional mapping table. Continuous arc correction is performed, and the vision sensor layout parameters are adjusted to optimize the coverage state.

Benefits of technology

It improves the accuracy of describing the actual coverage state of the tooth root transition arc, eliminates coverage breaks at the junction of adjacent fields of view, and reduces the risk of missing microcracks.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123118552A_ABST
    Figure CN123118552A_ABST
Patent Text Reader

Abstract

This invention provides a method and system for verifying the circular deployment of visual sensors based on field-of-view overlap analysis, belonging to the field of visual inspection technology. The method includes: acquiring geometric data of the target tooth root transition arc, the sidewalls of the tenons on both sides, and adjacent visual sensors to obtain detection geometric data; performing arc sampling, field-of-view filtering, and sidewall occlusion determination based on the detection geometric data to obtain a first set of visible sampling points and a second set of visible sampling points; and fitting arcs based on the first and second sets of visible sampling points to obtain pairs of adjacent visible coverage arcs. This invention performs deployment verification and iterative adjustment based on the actual visible coverage state under sidewall occlusion conditions, avoiding the formation of weak observation areas at the boundaries of adjacent fields of view and reducing the risk of missed microcrack detection.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of visual inspection technology, and in particular to a method and system for verifying the circular layout of visual sensors based on field-of-view overlap analysis. Background Technology

[0002] In the overhaul of high-pressure turbine disks for aero-engines, multiple vision sensors are often arranged in a ring around the outer perimeter of the turbine disk to visually re-inspect the microcracks in the transition arc of the tenon root. Due to the large number of tenons and the limited space, a ring arrangement can improve inspection efficiency. However, some existing arrangements are planned based on ideal space models or lens nominal parameters, and rarely consider the occlusion of the tenon sidewalls on the edge of the field of view. The transition arc of the tooth root is located between the sidewalls of the two tenons and is a concave surface. When the field of view of adjacent vision sensors is projected onto this area, it is easily blocked by the sidewalls, which reduces the actual overlap width of adjacent field of view on the transition arc surface of the tooth root. This may form a narrow strip of weak observation area at the junction. If a microcrack is located in this area, it is difficult to be imaged completely, and there is a risk of missed detection. Summary of the Invention

[0003] This invention provides a method and system for verifying the circular deployment of visual sensors based on field-of-view overlap analysis. The method performs deployment verification and iterative adjustment based on the actual visible coverage under sidewall occlusion conditions, avoiding the formation of weak observation areas at the boundaries of adjacent fields of view and reducing the risk of missing microcracks.

[0004] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: Firstly, a method for verifying the circular deployment of visual sensors based on field-of-view overlap analysis, the method comprising: Obtain geometric data of the target tooth root transition arc, the side walls of the tenons on both sides, and the adjacent vision sensors to obtain detection geometric data; Based on the detection geometric data, circular arc sampling, field of view filtering, and side wall occlusion determination are performed to obtain the first visible sampling point group and the second visible sampling point group. Based on the fitting of the first and second visible sampling point groups, the adjacent visible coverage is obtained. Arc pair; Based on adjacent visible covered arc pairs, the tangential deviation parameter and curvature deviation parameter are obtained; Based on the parameter ranges corresponding to the tangential deviation parameter and the curvature deviation parameter, the first index and the second index are obtained. index; The candidate correction level is obtained by querying the two-dimensional mapping table based on the first and second indices. Based on the candidate correction level, continuous arc correction is performed to obtain a pair of continuously covered arcs with consistent tangential direction and equal curvature; The actual overlap width is obtained based on the overlapping range of the continuously covering arc pairs; Based on the comparison between the actual overlap width and the overlap threshold, the weak observation area is obtained; The visual sensor deployment parameters were adjusted based on the weak observation areas to obtain the verified ring-shaped deployment parameters.

[0005] Secondly, a visual sensor ring deployment verification system based on field-of-view overlap analysis includes: The acquisition module is used to acquire geometric data of the target tooth root transition arc, the side walls of the tenon teeth on both sides, and the adjacent vision sensors to obtain detection geometric data; The sampling module is used to perform arc sampling, field of view filtering, and sidewall occlusion determination based on the detection geometric data to obtain a first visible sampling point group and a second visible sampling point group; and to fit an arc based on the first visible sampling point group and the second visible sampling point group to obtain adjacent visible coverage arc pairs. The calculation module is used to obtain the tangential deviation parameter and curvature deviation based on adjacent visible covering arc pairs. Parameters; based on the parameter ranges corresponding to the tangential deviation parameter and the curvature deviation parameter, obtain the first index and the second index; query the two-dimensional mapping table based on the first index and the second index to obtain the candidate correction level; The correction module is used to perform continuous arc correction based on candidate correction levels to obtain a pair of continuously covering arcs with consistent tangential direction and equal curvature; and to obtain the actual overlap width based on the overlap range of the continuously covering arcs. The comparison module is used to obtain the weak observation area based on the comparison result between the actual overlap width and the overlap threshold; and to adjust the visual sensor layout parameters based on the weak observation area to obtain the verified ring layout parameters.

[0006] The above-described solution of the present invention has at least the following beneficial effects: By constructing observation line segments from the optical center of the visual sensor to each candidate sampling point and determining their intersection relationship with the sidewalls of the tenons on both sides, sampling points with sidewall intersections between the optical center and the sampling points are eliminated. This allows the coverage of adjacent fields of view to be determined based on actual visible sampling points, avoiding the overlap estimation bias caused by planning only according to the nominal angle of view of the lens, and improving the accuracy of describing the actual coverage state on the tooth root transition arc.

[0007] The degree of tangential discontinuity and curvature inequality of adjacent covered arcs at the common connection position are quantified by tangential deviation parameters and curvature deviation parameters, respectively. Candidate correction levels are obtained by dividing the parameter interval and querying the two-dimensional mapping table, so that the continuous correction of the arc has an objective basis and there is no need to try to adjust the continuous deviation parameters one by one.

[0008] By establishing a common tangential and common normal at the common connection location, and reconstructing the arcs at both ends while keeping the arc length range corresponding to the visible sampling points unchanged, a pair of continuous covering arcs with consistent tangentials and equal curvature is obtained, thus eliminating the coverage break at the boundary of adjacent fields of view.

[0009] By representing the actual overlap width with arc length instead of image plane pixel width, the influence of sensor viewpoint and perspective scaling on the overlap comparison is reduced, making the overlap determination result consistent with the real spatial coverage on the tooth root transition arc surface.

[0010] By comparing the actual overlap width with the overlap threshold, weak observation areas are located. Based on this, the circumferential position, radial position, optical axis pointing or working distance of the visual sensor are adjusted, and the visibility verification is re-executed. This allows the circumferential deployment parameters to be iteratively optimized based on the actual visible coverage state under side wall occlusion conditions, reducing the risk of missing microcracks at the boundary of adjacent fields of view. Attached Figure Description

[0011] Figure 1 This is a flowchart illustrating a visual sensor ring layout verification method based on field-of-view overlap analysis provided by an embodiment of the present invention.

[0012] Figure 2 This is a schematic diagram of a visual sensor ring layout verification system based on field-of-view overlap analysis provided by an embodiment of the present invention.

[0013] Figure 3 This is a flowchart illustrating the process of adjusting the visual sensor deployment parameters based on the weak observation area to obtain the verified ring deployment parameters, as provided in an embodiment of the present invention. Detailed Implementation

[0014] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0015] like Figure 1 and Figure 3 As shown, an embodiment of the present invention proposes a method for verifying the circular deployment of visual sensors based on field-of-view overlap analysis.

[0016] S101, acquire the geometric data of the target tooth root transition arc, the sidewalls of the tenons on both sides, and the adjacent vision sensors to obtain the detection geometric data, including: The entity contour data comes from the physical scan data of the turbine disk tooth root. The local geometry data of the tenon groove includes at least the arc contour of the transition arc of the target tooth root, the contour of the left tenon sidewall, and the contour of the right tenon sidewall. The calibration geometry data of the adjacent vision sensors are obtained to obtain the optical center position, optical axis direction, and field of view boundary of the first and second vision sensors. The calibration geometry data is read from the calibration file of the vision sensors. The first and second vision sensors are any pair of adjacent vision sensors in the ring layout. The optical center position is the three-dimensional coordinate of the optical center of the sensor lens in the calibration coordinate system. The optical axis direction is the unit direction vector of the principal optical axis of the sensor lens. The field of view boundary is the spatial range description of the effective imaging area of ​​the sensor in the calibration coordinate system, which is usually represented as a quadrangular pyramid with the optical center as the vertex and the optical axis as the central axis.

[0017] Based on the local geometric data of the mortise, a detection section is extracted passing through the target tooth root transition arc, the left tenon sidewall, and the right tenon sidewall to obtain the tooth root detection profile. The method for extracting the detection section is as follows: using the plane containing the target tooth root transition arc as a reference plane, the contour lines located in this plane from the local geometric data of the mortise are extracted to obtain a two-dimensional tooth root detection profile containing the arc-shaped contour segment of the target tooth root transition arc, the contour segment of the left tenon sidewall, and the contour segment of the right tenon sidewall. In the above process, the detection section refers to the two-dimensional contour graphic obtained by extracting it from the local geometric data of the mortise using the plane containing the target tooth root transition arc as the cutting plane.

[0018] In the tooth root detection profile, an arc-shaped profile segment is separated based on the curvature change characteristics. The arc-shaped profile segment is then fitted with a circular arc to obtain the center position and radius. The two ends of the arc-shaped profile segment are taken as the arc boundaries, and the arc length extension direction between the arc boundaries is taken as the extension direction. The circular arc geometric reference includes the plane where the arc is located, the center position, the radius, and the arc length range. In the above process, the arc boundary refers to the position of the two endpoints of the target tooth root transition arc on the arc. The extension direction refers to the direction of travel from one end of the arc boundary along the arc to the other end.

[0019] Based on the circular arc geometric datum, optical center position, optical axis direction, and field of view boundary, the spatial coordinate relationship is unified to obtain the detection geometric data. Specifically, during coordinate unification, the corresponding datum points in the local geometric data of the tenon groove and the calibration geometric data of the vision sensor are read, including the turbine disk shaft center, end face positioning surface, circumferential positioning pin of the gear disk, or the datum surface of the tenon groove side. According to the spatial transformation relationship between the corresponding datum points, the coordinate positions of the installation datum in each coordinate system are aligned. Using the aligned installation datum, the tooth root detection contour, the contours of the two tenon sidewalls, the optical center position, optical axis direction, and field of view boundary of the two vision sensors are determined. Each boundary is mapped to the same coordinate system while maintaining the relative positional and directional relationships between the geometric elements. After the coordinate unification process, the detection geometric data consists of the tooth root detection profile, the arc geometric datum, the side wall profiles of the tenon teeth on both sides, the two optical center positions, the two optical axis directions, and the two field boundaries in the unified coordinate system. In the above process, the installation datum refers to the geometric feature point that can be identified in both the local geometric data of the tenon groove and the sensor calibration geometric data and has the same position. It is used to establish the transformation relationship between the two coordinate systems. Coordinate unification refers to the operation of aligning geometric data from different sources to the same coordinate system through spatial transformation.

[0020] When the circular layout parameters of the vision sensors include multiple pairs of adjacent vision sensors, each pair of adjacent vision sensors acquires its own detection geometric data according to the above process. When the number of adjacent vision sensor pairs that need to be verified exceeds one, the detection geometric data of each pair of adjacent vision sensors is acquired one by one, or acquired sequentially according to the circular arrangement order. The verification order is as follows: starting from any pair of adjacent sensors in the circular layout, the overlap verification of each pair of adjacent sensors is completed sequentially along the circumferential direction. If a pair of sensors changes its circumferential position or optical axis direction after adjustment, its overlap relationship with the adjacent sensor on the other side will also change, and it needs to be re-evaluated in subsequent verifications. When the same sensor is required to be adjusted by two adjacent sensor pairs at the same time, and the adjustment directions are inconsistent, the adjustment needs of the side with the larger arc length range of the weak observation area are given priority, and the other side is compensated by adjusting the other adjacent sensor.

[0021] S102, based on the detection geometric data, perform arc sampling, field of view filtering, and sidewall occlusion determination to obtain a first visible sampling point group and a second visible sampling point group, including: Based on the detection geometry data, equal arc length sampling is performed along the transition arc of the target tooth root to obtain the arc sampling point sequence. The equal arc length sampling is implemented according to the following process.

[0022] First, determine the preset sampling arc length. The rules for determining the preset sampling arc length are as follows: Obtain the detection accuracy requirement of the tooth root detection contour. The detection accuracy requirement refers to the allowable deviation of contour measurement specified in the tooth root detection task, which is read from the detection task specification; Obtain the spatial resolution of the vision sensor at the target arc position. The spatial resolution is the actual physical size corresponding to a single pixel of the sensor at the distance of the target arc. The spatial resolution is equal to the sensor pixel size multiplied by the working distance and then divided by the sensor focal length. The pixel size and focal length are read from the sensor parameters. The working distance is the current distance from the sensor optical center to the target arc; Take the smaller of the detection accuracy requirement and the spatial resolution as the reference sampling accuracy; Multiply the reference sampling accuracy by the preset sampling density coefficient to obtain the preset sampling arc length; The specific value of the sampling density coefficient is determined as follows: Obtain the arc length of the target arc and divide it by the reference sampling accuracy to obtain the arc length reference ratio; When the arc length reference ratio is less than or equal to 1, it indicates that the target arc is short or the reference sampling accuracy is high. The sampling density coefficient is set to 2. When the arc length reference ratio is greater than one, it indicates that the target arc is relatively long or the reference sampling accuracy is low. In this case, the sampling density coefficient is taken as two divided by the arc length reference ratio, and the value is not more than 2 and not less than 1 / 2. When the detection accuracy requirement is higher than the spatial resolution, the reference sampling accuracy is taken as the spatial resolution. In this case, the preset sampling arc length is determined by the sensor's imaging capability. When the spatial resolution is higher than the detection accuracy requirement, the reference sampling accuracy is taken as the detection accuracy requirement. In this case, the preset sampling arc length refers to the arc distance measured along the arc between two adjacent sampling points, which is obtained by multiplying the reference sampling accuracy by the sampling density coefficient. It is used to control the density of sampling. The reference sampling accuracy refers to the smaller of the detection accuracy requirement and the spatial resolution, which serves as the basic reference value for determining the sampling arc length. The spatial resolution refers to the actual physical size of a single pixel of the sensor at the target distance, reflecting the sensor's ability to resolve details in imaging. The detection accuracy requirement refers to the allowable deviation of contour measurement specified by the detection task.

[0023] Obtain the radius and arc length of the target arc. Divide the preset sampling arc length by the radius of the target arc to obtain the arc length sampling angle step size. Starting from the initial angle of the target arc, generate sampling angles sequentially according to the arc length sampling angle step size. Take the arc position corresponding to each sampling angle as the sampling position and map it to the detection coordinate system to obtain the three-dimensional coordinates of each sampling position. Starting from an arc boundary of the target arc, determine the sampling positions sequentially along the arc extension direction according to the preset sampling arc length. Take the corresponding point on the target arc at each sampling position to obtain the arc sampling point sequence. Record the three-dimensional coordinates of each sampling point in the detection coordinate system and the arc length sequence number along the arc. In the above process, equal arc length sampling means taking points sequentially along the arc at equal arc length intervals, and the distance between each sampling point is equal along the arc measurement. The arc length sampling angle step size refers to the central angle between adjacent sampling points, which is obtained by dividing the preset sampling arc length by the radius of the target arc. The arc length sequence number refers to the arrangement number of the sampling point in the arc sampling point sequence, numbered from smallest to largest according to the sampling order.

[0024] Based on the optical center position, optical axis direction, and field of view boundary of the first visual sensor, sampling points are selected from the arc sampling point sequence to obtain the first candidate sampling point group; the field of view selection is performed according to the following process: taking the optical center of the first visual sensor as the projection center, the four edges of the field of view boundary are taken from the optical center and intersected with the plane where the target arc is located, to obtain four projection intersection points; the four projection intersection points are connected in sequence to form a projection quadrilateral, which is the projection boundary; the projection method is central projection, the projection center is the optical center, and the projection plane is the plane where the target arc is located. In the local coordinate system of the visual sensor, the ray method is used to determine whether each sampling point is located inside the projected quadrilateral: a ray is drawn from the sampling point along any fixed direction, and the number of intersections between the ray and the boundary of the projected quadrilateral is counted. If the number of intersections is odd, the sampling point is located inside the projected quadrilateral and is determined to be a candidate sampling point; if the number of intersections is even, the sampling point is located outside the projected quadrilateral and is determined to be an invisible sampling point. All candidate sampling points constitute the first candidate sampling point group. In the above process, the projection boundary refers to the quadrilateral region formed by the four edges of the field of view boundary projected onto the plane where the target arc is located through the center; the central projection refers to the projection method with the optical center as the projection center and the plane where the target arc is located as the projection plane; the ray method refers to the geometric determination method of whether a point is inside the polygon by the number of intersections between the ray and the polygon boundary.

[0025] Based on the optical center position, optical axis direction, and field of view boundary of the second visual sensor, sampling points are selected from the arc sampling point sequence using the same field of view selection method as the first visual sensor, resulting in a second candidate sampling point group. Based on the first and second candidate sampling point groups, observation line segments are constructed from the optical center of the visual sensor to each candidate sampling point, resulting in two sets of observation line segments. For each candidate sampling point in the first candidate sampling point group, a spatial observation line segment is constructed with the optical center position of the first visual sensor as the starting point and the three-dimensional coordinates of the candidate sampling point as the ending point. The observation line segments corresponding to all candidate sampling points constitute the first set of observation line segments. For the second candidate sampling point group, a second set of observation line segments is constructed with the optical center position of the second visual sensor as the starting point and the three-dimensional coordinates of each candidate sampling point as the ending point. In the above process, the observation line segment refers to a spatial straight line segment with the optical center of the visual sensor as the starting point and the candidate sampling point as the ending point, used to determine whether the optical path is blocked by the sidewall.

[0026] For each observation line segment in the first group, determine whether the observation line segment intersects with the contour of the left or right tenon sidewall before reaching the candidate sampling point: Intersect the observation line segment with each straight line segment of the left tenon sidewall contour sequentially. If there is an intersection point between the start and end points of the observation line segment, the candidate sampling point is determined to be occluded by the sidewall. If the observation line segment does not intersect with the left tenon sidewall contour, continue to determine the intersection relationship with the right tenon sidewall contour. If there is no intersection with either of the two tenon sidewall contours, the candidate sampling point is determined to be unoccluded and retained as a visible sampling point of the first visual sensor. All retained visible sampling points constitute the first visible sampling point group. The acquisition method of the second visible sampling point group is the same as that of the first visible sampling point group, but the observation line segment is replaced with the second group of observation line segments, and the visual sensor is replaced with the second visual sensor. In the above process, the side wall intersection refers to the geometric intersection of the observation line segment and the contour of the tenon side wall. Occlusion refers to the optical path between the candidate sampling point and the optical center of the visual sensor being blocked by the tenon side wall, so that the sampling point cannot be observed by the sensor. After completing the field of view screening and occlusion determination of all sampling points, the first visible sampling point group and the second visible sampling point group are output according to the original sampling point sequence number.

[0027] S103, based on the first visible sampling point group and the second visible sampling point group, fit the circular arc to obtain adjacent visible coverage circular arc pairs, including: Based on the first set of visible sampling points, continuous visible sampling points near the boundary of adjacent fields of view are extracted to obtain the first set of circular arc fitting points. The specific extraction method is as follows: Organize the first visible sampling point group according to the sequence number in the arc sampling point sequence, and divide the visible sampling points with consecutive sequence numbers into the same continuous visible segment. During division, read the visible sampling points arranged in ascending order of arc length, subtract the sequence number of the current visible sampling point from the sequence number of the previous visible sampling point to obtain the sequence number difference. When the sequence number difference equals one sampling interval, add the current visible sampling point to the existing segment; when the sequence number difference is greater than one sampling interval, end the existing segment and establish a new segment from the current visible sampling point. Record the start sequence number, end sequence number, start arc length, end arc length, and number of sampling points for each segment. For each continuous visible segment, use the difference between the maximum and minimum arc lengths within the segment as the arc length coverage range of that segment. Compare the arc length coverage range with a preset effective arc length range threshold, and mark segments with arc length coverage ranges smaller than the effective arc length range threshold as invalid segments and remove them. Among the retained continuous visible segments, select the segment with the largest arc length coverage range. The section, typically closest to the boundary of adjacent fields of view, is used to determine the visible sampling points within that section as the first arc fitting point group. In the above process, a continuous visible section refers to the arc segment covered by a group of visible sampling points with consecutive, uninterrupted arc length numbers. The sequence difference refers to the difference between the arc length number of the current sampling point and the arc length number of the previous sampling point, used to determine whether two sampling points are adjacent in the sequence. The arc length coverage range refers to the difference between the maximum and minimum arc length within a continuous visible section, representing the arc length span covered by that section. The method for determining proximity to the boundary of adjacent fields of view is as follows: calculate the difference between the sampling point with the largest arc length number in each candidate sampling point group and the arc length number corresponding to the boundary of adjacent fields of view, and select the group with the smallest difference as the first arc fitting point group. If multiple groups have the same difference, select the group with the largest number of consecutive visible sampling points. The arc length number corresponding to the boundary of adjacent fields of view is determined based on the intersection of the optical center line connecting the two visual sensors and the transition arc of the target tooth root.

[0028] Based on the second set of visible sampling points, the same extraction method as the first set of visible sampling points is used to extract continuous visible sampling points near the boundary of adjacent fields of view, thus obtaining the second set of circular arc fitting points.

[0029] Based on the first set of arc fitting points, the center, arc boundary, and curvature features are determined to obtain the first visible covering arc. Arc fitting is performed as follows: First, at least three sampling points that are dispersed and non-collinear along the arc length direction are selected from the first set of arc fitting points. The candidate center is determined based on the perpendicular bisector of the line connecting any two sampling points. The candidate radius is determined based on the distance from the candidate center to each sampling point. Using the arc geometric datum, the candidate center is limited to be near the target arc center. The difference between the distance from each sampling point to the candidate center and the candidate radius is calculated. The squares of these differences are summed to obtain the fitting residual. The candidate center position and candidate radius are adjusted in the direction that reduces the fitting residual until the fitting residual no longer decreases or the fitting stopping condition is met. The tolerance for outliers is determined as follows: the accuracy of the entity contour data is obtained, denoted as contour data accuracy, and read from the design tolerance of the entity model or the measurement accuracy of the scanning equipment, in units of length. The sampling error of the circular arc is obtained. The sampling error is the difference between the chord length and the arc length corresponding to the preset sampling arc length. The chord length is calculated based on the target arc radius and the arc length sampling angle step. The chord length is equal to the sine value of two times the target arc radius and then multiplied by half the arc length sampling angle step. The squares of the contour data accuracy and the sampling error are added together and the square root is taken to obtain the outlier judgment tolerance. During the circular arc fitting process, when the absolute value of the difference between the distance from a sampling point to the candidate circle center and the candidate radius is greater than the outlier judgment tolerance, the sampling point is judged as an outlier and removed from the circular arc fitting point group before refitting. The circular arc geometric reference is used to limit the fitting range to avoid a small number of edge points causing the fitting circle center to deviate from the target tooth root transition arc, while still retaining the end and curvature differences caused by the different visible ranges of the first and second circular arc fitting point groups. The first visible covering arc records the position of the fitted circle's center, the fitted radius, the plane on which the fitted arc lies, the starting position of the arc length, the ending position of the arc length, and the coverage area of ​​the arc length. In the above process, the outlier judgment tolerance refers to the distance threshold used to determine whether a sampling point is an outlier, which is obtained by adding the squares of the contour data accuracy and the sampling error and then taking the square root. The sampling error refers to the difference between the arc length and the chord length corresponding to the preset sampling arc length, reflecting the influence of discrete sampling on the accuracy of the arc fitting. The fitting residual refers to the sum of the squares of the differences between the distance from each sampling point to the candidate circle's center and the candidate radius, used to measure the degree of deviation between the fitted arc and the sampling points. Based on the second arc fitting point group, the same fitting method as the first arc fitting point group is used to obtain the second visible covering arc, and the same information is recorded.

[0030] Based on the ends of the first visible covering arc that are closer to the second visible covering arc and the ends of the second visible covering arc that are closer to the first visible covering arc, a first opposing end and a second opposing end are obtained. The first opposing end is the end of the arc length endpoint of the first visible covering arc that is closer to the second visible covering arc, and the second opposing end is the end of the arc length endpoint of the second visible covering arc that is closer to the first visible covering arc. The two opposing ends face each other and are located in the boundary region of two adjacent fields of view. In the above process, opposing ends refer to the ends of two adjacent visible covering arcs that are opposite each other, and the first opposing end and the second opposing end face each other in the arc length direction.

[0031] Adjacent relationships are established based on the first and second opposing ends to obtain adjacent visible covering arc pairs. An adjacency relationship indicates that the first and second visible covering arcs are adjacent and facing each other on the target tooth root transition arc. An adjacent visible covering arc pair consists of the first visible covering arc, the second visible covering arc, the first opposing end, and the second opposing end. In the above process, the adjacency relationship refers to the geometric relationship where two visible covering arcs are adjacent to each other on the target arc and their ends face each other.

[0032] S104, based on adjacent visible covered arc pairs, obtain the tangential deviation parameter and curvature deviation parameter, including: Based on the first opposing end and the arc geometric reference, the first opposing end is mapped to the target tooth root transition arc along the radial direction to obtain the first mapping point. The specific method of radial mapping is as follows: taking the center of the target arc as the reference point, a radial ray is drawn from the center through the first opposing end. The intersection of this radial ray and the target tooth root transition arc is the first mapping point. In the above process, radial mapping refers to the operation of projecting the opposing end onto the target arc along the radial direction. A radial ray refers to a ray that starts from the center of the target arc and passes through the opposing end.

[0033] Based on the second opposing end and the arc geometric reference, using the same radial mapping method as the first opposing end, the second opposing end is mapped to the target tooth root transition arc along the radial direction of the target tooth root transition arc to obtain the second mapping point.

[0034] The common connection position is obtained based on the midpoint of the arc segment between the first and second mapping points. The common connection position can be determined in one of two ways: **Arch length midpoint method:** Along the target tooth root transition arc, the midpoint of the arc length between the first and second mapping points is taken as the common connection position. Specifically, first, obtain the arc length parameters of the first and second mapping points on the target arc, denoted as the first and second mapping arc lengths; calculate the average of the first and second mapping arc lengths to obtain the arc length parameter of the common connection position; determine the corresponding three-dimensional coordinates on the target arc based on this arc length parameter, which is the common connection position. **Line intersection method:** Connect the first and second mapping points with a straight line segment, and find the intersection of this straight line segment with the target tooth root transition arc. Since a straight line segment and an arc can intersect at most two points, the intersection points are selected according to the following rules: If the straight line segment and the target arc intersect at exactly one point, then that intersection point is taken as the common connection position; if the straight line segment and the target arc intersect at two points, then the intersection point whose arc length parameter lies between the first mapped arc length and the second mapped arc length is taken as the common connection position; if the straight line segment and the target arc do not intersect, then the first method of determining the common connection position is used as a fallback. In the above process, the arc length parameter refers to the arc length value of the sampling point or mapping point on the target arc, calculated from the starting angle; the mapped arc length refers to the arc length parameter value corresponding to the mapping point on the target arc; and the common connection position refers to a reference position between two mapping points on the target arc.

[0035] Based on the location of the common connection and the tangential directions of the two opposing ends, the tangential deviation parameter is obtained. The first unit tangential direction is the unit vector of the tangential direction of the first visible covering arc at the first opposing end, and the second unit tangential direction is the unit vector of the tangential direction of the second visible covering arc at the second opposing end. The angle between the first unit tangential direction and the second unit tangential direction is used as the tangential deviation parameter. The angle is obtained by calculating the dot product of the two unit vectors and taking the inverse cosine. In the above process, the unit tangential direction refers to the tangential direction of the visible covering arc at the opposing ends, represented by a unit vector. The tangential deviation parameter refers to the angle between the two unit tangential directions, reflecting the degree of inconsistency between the two opposing ends in the tangential direction.

[0036] Based on the curvature characteristics of the two opposing ends, a curvature deviation parameter is obtained. The fitted curvature of the first visible covering arc is the reciprocal of its fitted radius, and the fitted curvature of the second visible covering arc is also the reciprocal of its fitted radius. The difference between the fitted curvatures of the first and second visible covering arcs is used as the curvature deviation parameter. In the above process, the fitted curvature refers to the reciprocal of the fitted radius, reflecting the degree of curvature of the visible covering arc; the curvature deviation parameter refers to the difference between the two fitted curvatures, reflecting the degree of consistency in the curvature of the two visible covering arcs.

[0037] Based on the tangential deviation parameter and the curvature deviation parameter, a continuous correction input parameter set is obtained. This set consists of the tangential deviation parameter, the curvature deviation parameter, the first mapping point, the second mapping point, the common connection position, the first unit tangential direction, the second unit tangential direction, the fitted curvature of the first visible covering arc, and the fitted curvature of the second visible covering arc, serving as input for subsequent continuous correction. In the above process, the continuous correction input parameter set refers to the complete set of input parameters used in subsequent correction steps.

[0038] S105, based on the parameter intervals corresponding to the tangential deviation parameter and the curvature deviation parameter, obtain the first index and the second index, including: Based on the valid range of values ​​for the tangential deviation parameter, the tangential parameter intervals are divided according to the degree of deviation, resulting in the first parameter interval set. The division of the first parameter interval set is performed according to the following process.

[0039] The basic indicators affecting the tolerance of tangential deviation include: the calibration angle accuracy of the first vision sensor, read from the sensor calibration report; the calibration angle accuracy of the second vision sensor, read from the sensor calibration report; the contour measurement accuracy of the tooth root detection contour, read from the tolerance annotation of the contour measurement equipment or design model; and the arc fitting tolerance of the target tooth root transition arc, which is the maximum allowable radial distance from each sampling point to the fitted arc during the arc fitting process. In the above process, the calibration angle accuracy refers to the measurement error range of the optical axis direction angle during sensor calibration; the contour measurement accuracy refers to the measurement or design tolerance range of the tooth root detection contour; and the arc fitting tolerance refers to the maximum allowable radial deviation distance between the fitted arc and the sampling point.

[0040] The aforementioned basic indicators are uniformly converted into angular dimensions to obtain the comprehensive basic deviation angle. The conversion method is as follows: Divide the contour measurement accuracy by the radius of the target tooth root transition arc to obtain the angular deviation component corresponding to the contour measurement accuracy, denoted as the first angular component; divide the arc fitting tolerance by the radius of the target tooth root transition arc to obtain the angular deviation component corresponding to the arc fitting tolerance, denoted as the second angular component; take the larger of the first and second calibration angular accuracies as the calibration angular deviation component, denoted as the third angular component. The larger value is chosen because the one with lower accuracy among the two sensors has a greater impact on the tangential deviation. Squaring the first, second, and third angular components and adding them together, then taking the square root, yields the comprehensive basic deviation angle. The comprehensive basic deviation angle represents the basic deviation range of the tangential deviation that can be accepted without additional correction under the combined influence of sensor calibration accuracy, contour measurement accuracy, and arc fitting tolerance. In the above process, the comprehensive basic deviation angle refers to the basic deviation angle value obtained by uniformly converting multiple accuracy indicators from different sources into angular dimensions and synthesizing them by squaring and taking the square root, used to determine the boundary of the first parameter interval.

[0041] The comprehensive basic deviation angle is used as the upper boundary of the first interval in the first parameter interval set, and the lower boundary of the first interval is zero. The first interval corresponds to the tangential parameter interval with the smallest correction degree. The tangential deviation in this interval is considered to be within the basic deviation range, and no end direction correction is required, or only the slightest correction needs to be applied.

[0042] The minimum directional change that the end-point orientation adjustment mechanism can stably achieve in each operation is recorded as the minimum tangential adjustment step size, which is read from the performance parameters of the adjustment mechanism deployed by the vision sensor. The minimum tangential adjustment step size is added to the upper boundary of the first interval to form the upper boundary of the second interval, and the lower boundary of the second interval becomes the upper boundary of the first interval. This process continues, adding the minimum tangential adjustment step size to the upper boundary of the current interval to form the upper boundary of the next interval, until the upper limit of the effective range of the tangential deviation parameter is covered. Each tangential parameter interval is defined by its lower and upper boundaries, and adjacent tangential parameter intervals close at their boundaries. The correction degree of each interval is arranged from smallest to largest; intervals with larger correction degrees require more or larger step sizes of end-point orientation adjustments. In the above process, the minimum tangential adjustment step size refers to the minimum angular change that the end-point orientation adjustment mechanism can stably achieve in each operation.

[0043] The upper limit of the valid range of the tangential deviation parameter is set as the upper boundary of the last interval. If the upper boundary of the last interval exceeds the valid range, the upper limit of the valid range is taken as the upper boundary of the last interval, and this interval is marked as the maximum correction interval. In the above process, the tangential parameter interval refers to a continuous numerical interval divided according to the degree of correction of the tangential deviation from small to large, and each interval corresponds to an index number.

[0044] Based on the effective range of curvature deviation parameters, the curvature parameter intervals are divided according to the degree of deviation to obtain the second parameter interval set. The division of the second parameter interval set is performed according to the following process.

[0045] The fundamental indicators affecting curvature deviation tolerance are as follows: the calibration position accuracy of the first vision sensor, read from the sensor calibration report, in units of length; the calibration position accuracy of the second vision sensor, read from the sensor calibration report; the contour measurement accuracy of the tooth root detection profile; and the arc fitting tolerance of the target tooth root transition arc. In the above process, calibration position accuracy refers to the measurement error range of the optical center position during sensor calibration.

[0046] The above basic indicators are uniformly converted into curvature dimensions to obtain the comprehensive basic curvature deviation. The conversion method is as follows: the larger of the first calibration position accuracy and the second calibration position accuracy is taken as the calibration position deviation. The larger value is chosen because the one with poorer accuracy between the two sensors has a greater impact on the curvature deviation. The calibration position deviation is divided by the square of the radius of the target tooth root transition arc to obtain the curvature deviation component corresponding to the calibration position deviation, which is denoted as the first curvature component. The contour measurement accuracy is divided by the square of the radius of the target tooth root transition arc to obtain the curvature deviation component corresponding to the contour measurement accuracy, which is denoted as the second curvature component. The arc fitting tolerance is divided by the radius of the target tooth root transition arc. The curvature deviation component corresponding to the arc fitting tolerance is obtained by squaring the radius, and is denoted as the third curvature component. The first, second, and third curvature components are squared and then added together, and the square root is taken to obtain the comprehensive basic curvature deviation. The comprehensive basic curvature deviation represents the basic deviation range of curvature deviation that can be accepted without additional correction under the combined influence of sensor calibration accuracy, contour measurement accuracy, and arc fitting tolerance. In the above process, the comprehensive basic curvature deviation refers to the basic curvature deviation value obtained by converting multiple accuracy indicators from different sources into curvature dimensions and synthesizing them by squaring and taking the square root. It is used to determine the boundary of the first curvature parameter interval.

[0047] The comprehensive basic curvature deviation is used as the upper boundary of the first interval in the second parameter interval set, and the lower boundary of the first interval is zero. The first interval corresponds to the curvature parameter interval with the smallest correction degree.

[0048] The minimum curvature change that the center position adjustment mechanism can stably achieve each time is obtained and denoted as the minimum curvature adjustment step size. The minimum curvature adjustment step size is determined as follows: the minimum displacement that the center position adjustment mechanism can stably achieve each time is obtained and denoted as the minimum center displacement, which is read from the performance parameters of the adjustment mechanism deployed by the vision sensor; the minimum center position is removed by the square of the target tooth root transition arc radius to obtain the minimum curvature adjustment step size; the minimum curvature adjustment step size is added to the upper boundary of the first interval to obtain the upper boundary of the second interval, and the lower boundary of the second interval is the upper boundary of the first interval; and so on, until the upper limit of the effective value range of the curvature deviation parameter is covered. In the above process, the minimum curvature adjustment step size refers to the minimum curvature change that the center position adjustment mechanism can stably achieve each operation, which is obtained by removing the minimum center position by the square of the target arc radius.

[0049] The upper limit of the effective range of curvature deviation parameters is set as the upper boundary of the last interval. If the upper boundary of the last interval exceeds the effective range, the upper limit of the effective range is used as the upper boundary of the last interval, and this interval is marked as the maximum correction interval.

[0050] In the above process, the curvature parameter interval refers to a continuous numerical interval divided according to the degree of correction of curvature deviation from small to large, and each interval corresponds to an index number.

[0051] Based on the position of the actual value of the tangential deviation parameter in the first parameter interval set, the first index is obtained. When determining the first index, starting from the interval with a smaller correction degree in the first parameter interval set, the tangential deviation parameter is compared with the lower boundary and upper boundary of the current interval in turn. When the tangential deviation parameter is greater than the lower boundary and less than the upper boundary, the first index of the current interval is read. When the tangential deviation parameter is equal to the upper boundary, the next parameter interval is read. When the tangential deviation parameter is equal to the maximum boundary of the entire effective value range, the first index of the maximum correction interval is read.

[0052] The second index is obtained based on the position of the actual value of the curvature deviation parameter within the second parameter interval set. The determination method is the same as that for the first index.

[0053] Based on the case where the parameter lies on the common boundary of adjacent parameter intervals, the corresponding parameter is assigned to the parameter interval with a larger correction degree, resulting in the corrected first index and second index. When the tangential deviation parameter is exactly equal to the common boundary of two adjacent tangential parameter intervals, the tangential deviation parameter is assigned to the tangential parameter interval with a larger correction degree, and its corresponding first index is taken; the curvature deviation parameter is handled similarly. Thus, the corrected first index and second index are obtained.

[0054] S106, query the two-dimensional mapping table based on the first index and the second index to obtain candidate correction levels, including: The target row in the two-dimensional mapping table is determined based on the first index, and the row index position is obtained. The two-dimensional mapping table uses the first index corresponding to the first parameter interval set as the row index, and each row corresponds to a tangential parameter interval. The target column in the two-dimensional mapping table is determined based on the second index, and the column index position is obtained. The two-dimensional mapping table uses the second index corresponding to the second parameter interval set as the column index, and each column corresponds to a curvature parameter interval.

[0055] The target mapping unit is obtained by the intersection of the row index position and the column index position. The target mapping unit is a data unit at the intersection of the target row and the target column in the two-dimensional mapping table.

[0056] Based on the arc correction degree recorded by the target mapping unit, candidate correction levels are obtained. The candidate correction levels are arranged in ascending order of correction degree, with the candidate correction level sequence being Level 1, Level 2, up to Level N, where N is the preset total number of correction levels. Each candidate correction level corresponds to a set of correction parameters, including the degree of end direction adjustment and the degree of curvature adjustment. The specific correspondence is established according to the following process.

[0057] The degree of end-direction adjustment indicates the upper limit of the angle range that each of the two opposing ends needs to rotate during the correction process. For the k-th candidate correction level, the degree of end-direction adjustment is determined as follows: based on the minimum tangential adjustment step size, the degree of end-direction adjustment of the k-th level is equal to the minimum tangential adjustment step size multiplied by k. That is, the degree of end-direction adjustment of the first candidate correction level is equal to the minimum tangential adjustment step size, the second level is equal to twice the minimum tangential adjustment step size, and so on. The degree of end-direction adjustment limits the total angle of rotation of the first and second opposing ends during the end-direction correction process, which must not exceed this value. In the above process, the degree of end-direction adjustment refers to the upper limit of the rotation angle of the two opposing ends corresponding to the candidate correction level, which is obtained by multiplying the minimum tangential adjustment step size by the correction level number.

[0058] The curvature adjustment degree represents the upper limit of the allowable range of curvature change for the fitted centers of the two visible covering arcs during the correction process. For the k-th candidate correction level, the curvature adjustment degree is determined as follows: based on the minimum curvature adjustment step size, the curvature adjustment degree of the k-th level is equal to the minimum curvature adjustment step size multiplied by k. That is, the curvature adjustment degree of the first-level candidate correction level is equal to the minimum curvature adjustment step size, the second level is equal to twice the minimum curvature adjustment step size, and so on. The curvature adjustment degree limits the amount of change in the fitted curvature of the two visible covering arcs during the curvature correction process, ensuring that it does not exceed this value. In the above process, the curvature adjustment degree refers to the upper limit of the curvature change of the two visible covering arcs corresponding to the candidate correction level, obtained by multiplying the minimum curvature adjustment step size by the correction level number.

[0059] Based on the degree of end-direction adjustment and curvature adjustment corresponding to the candidate correction level, the circular arc correction parameters are obtained. The circular arc correction parameters include the first end-direction adjustment share, the second end-direction adjustment share, the first circular arc curvature adjustment share, the second circular arc curvature adjustment share, the allowable common curvature range, and the correction stop tolerance. Wherein: the first end direction adjustment share and the second end direction adjustment share are directly read from the target mapping unit, and their sum is one, representing the tangential deviation adjustment ratio undertaken by the first and second opposing ends respectively. The actual rotation angle of the first opposing end is equal to the end direction adjustment degree multiplied by the first end direction adjustment share, and the actual rotation angle of the second opposing end is equal to the end direction adjustment degree multiplied by the second end direction adjustment share; the first arc curvature adjustment share and the second arc curvature adjustment share are directly read from the target mapping unit, and their sum is one, representing the curvature deviation adjustment ratio undertaken by the first and second visible covering arcs respectively. The actual curvature change of the first visible covering arc is equal to the curvature adjustment degree multiplied by the first arc curvature adjustment share, and the actual curvature change of the second visible covering arc is equal to the curvature adjustment degree multiplied by the second arc curvature adjustment share; the allowable common curvature range is read from the target mapping unit, representing the numerical range into which the common curvature of the two visible covering arcs is allowed to fall after correction. The lower and upper boundaries of this range are determined by the calibration process during table creation. The correction stop tolerance is read from the target mapping unit, including the tangential consistency tolerance and the curvature equality tolerance, which are used to determine the stop conditions for end direction correction and curvature correction, respectively. In the above process, the end direction adjustment share refers to the proportion of tangential deviation adjustment undertaken by each of the two opposing ends, which is read from the mapping unit; the curvature adjustment share refers to the proportion of curvature deviation adjustment undertaken by each of the two visible covering arcs, which is read from the mapping unit.

[0060] The two-dimensional mapping table includes a header area, a mapping unit area, and a version area. The header area records the first index and tangential parameter range for each row, and the second index and curvature parameter range for each column. The mapping unit area records the candidate correction level, the first end direction adjustment share, the second end direction adjustment share, the first arc curvature adjustment share, the second arc curvature adjustment share, the allowable common curvature range, and the correction stop tolerance. The version area records the parameter range version used to establish the two-dimensional mapping table, the sensor calibration version, and the target arc specification identifier. Before querying, the version area is checked against the current detection geometry data and parameter range set; the mapping unit is only read if the versions match.

[0061] The end-direction adjustment share in the mapping unit represents how two opposing ends share the tangential deviation. The specific determination method for the end-direction adjustment share is as follows: Obtain the calibration reliability index of the first vision sensor and the calibration reliability index of the second vision sensor. The calibration reliability index refers to the reciprocal of the reprojection error or the reciprocal of the calibration residual during sensor calibration. The higher the reliability, the larger this index, denoted as the first reliability and the second reliability, respectively. Obtain the first and second circular arc fitting residuals. The fitting residual refers to the cumulative radial distance from each sampling point to the fitted circular arc during arc fitting, denoted as the first fitting residual and the second fitting residual, respectively. Calculate the sum of the first and second reliability levels to obtain the total reliability. Calculate the proportion of the second reliability to the total reliability to obtain the reliability proportion component. The larger this value, the more reliable the second sensor, and the larger the adjustment share allocated to the first sensor. Calculate the sum of the first and second fitting residuals to obtain the total fitting residual. Calculate the proportion of the second fitting residual to the total fitting residual. For example, the proportional component of the fitting residual is obtained. The larger this value, the worse the fitting accuracy of the second arc, and the larger the adjustment share allocated to the first sensor. The reliability proportional component and the fitting residual proportional component are added together and divided by two to obtain the first end-direction adjustment share. This operation is to take the arithmetic mean of the two proportional components, meaning that the end-direction adjustment share is jointly determined by two factors: the sensor calibration reliability and the arc fitting residual, with each factor accounting for half the weight. The first end-direction adjustment share is subtracted by one to obtain the second end-direction adjustment share. When the calibration reliability index or fitting residual of either sensor is unavailable, its default value is taken to make the shares on both sides equal. The curvature adjustment share is determined in the same way as the end-direction adjustment share, but is calculated independently. It is allowed that the end-direction adjustment share and the curvature adjustment share are allocated to the same side or different sides. In the above process, the reliability proportional component refers to the proportion of the second vision sensor calibration reliability to the sum of the reliability of the two sensors; the fitting residual proportional component refers to the proportion of the second arc fitting residual to the sum of the fitting residuals of the two arcs.

[0062] The calibration arc pair library is constructed as follows: two vision sensors are installed on the calibration fixture to image the standard arc surface on the fixture. The geometric parameters of the standard arc surface are known and used as the true values. The specific construction process is as follows.

[0063] Determine the coverage range of the deviation space: Divide the effective value range of the tangential deviation parameter into at least five levels, and the effective value range of the curvature deviation parameter into at least five levels, forming at least twenty-five deviation combinations. For each deviation combination, adjust the circumferential position, radial position, and optical axis orientation of the two vision sensors on the calibration fixture so that the measured tangential deviation parameter falls within the target tangential deviation level range and the measured curvature deviation parameter falls within the target curvature deviation level range. Record the visible coverage arc pairs and their deviation parameters under this layout parameter and store them in the calibration arc pair library. Collect at least three sets of data under different occlusion conditions for each deviation combination. The minimum capacity requirement for the calibration arc pair library is that each combination of tangential deviation level and curvature deviation level must contain at least one set of valid data. When a certain combination of deviations cannot be achieved through physical adjustment, a simulated visible coverage arc pair is generated by design model simulation. The simulation data and the measured data are marked and stored separately. The visible coverage arc pairs and their deviation parameters that are actually generated and confirmed by manual verification during each calibration process are also stored in the calibration arc pair library after being classified according to the corresponding tangential deviation parameter range and curvature deviation parameter range, forming incremental accumulation, so that the calibration arc pair library is continuously enriched as the number of uses increases.

[0064] Candidate calibration arc pairs are selected from the calibration arc pair library. Each arc pair in the library records the fitted center, fitted radius, arc length range, end tangential direction, and corresponding tangential deviation and curvature deviation parameters for each of the two visible covering arcs. The selection criteria are: the actual value of the tangential deviation parameter of the calibration arc pair falls between the lower and upper boundaries of the target tangential parameter interval, and the actual value of the curvature deviation parameter falls between the lower and upper boundaries of the target curvature parameter interval. Calibration arc pairs that meet these selection criteria are considered candidate calibration arc pairs. In the above process, the calibration arc pair library refers to a pre-collected set of visible covering arc pairs with known tangential deviation and curvature deviation parameters. The sources include simulated visible covering arc pairs generated by design simulation, known deviation arc pairs obtained by actual measurement by the calibration device, and typical arc pairs archived in historical verification data.

[0065] Select calibration arc pairs from the candidate calibration arc pairs that can represent the corresponding degree of deviation. Within the target tangential parameter interval, determine three representative tangential values: the lower boundary representative value, the middle tangential representative value, and the upper boundary representative value. The lower boundary representative value is taken at the lower boundary of the target tangential parameter interval plus 1 / 10 of the interval width; the upper boundary representative value is taken at the upper boundary of the target tangential parameter interval minus 1 / 10 of the interval width; and the middle tangential representative value is taken at the midpoint of the target tangential parameter interval. Within the target curvature parameter interval, determine three representative curvature values ​​in the same way: the lower boundary representative value, the middle curvature representative value, and the upper curvature representative value. Combine the three tangential representative values ​​with the three curvature representative values ​​in pairs to form nine sets of deviation representative value combinations. Each set of deviation representative value combinations consists of one tangential representative value and one curvature representative value. In the above process, the deviation representative value combination refers to the combination of characteristic values ​​selected within the target parameter interval that can represent different degrees of deviation within that interval.

[0066] For each group of representative deviation values, the calibration arc pair with the closest tangential deviation parameter and curvature deviation parameter to the group's representative curvature value is selected from the candidate calibration arc pairs. The degree of closeness is measured as follows: the absolute value of the difference between the tangential deviation parameter and the tangential representative value of the candidate calibration arc pair is calculated and recorded as the tangential deviation distance; the absolute value of the difference between the curvature deviation parameter and the curvature representative value of the candidate calibration arc pair is calculated and recorded as the curvature deviation distance; the sum of the tangential deviation distance and the curvature deviation distance is taken as the comprehensive deviation distance, and the candidate calibration arc pair with the smallest comprehensive deviation distance is selected as the calibration arc pair corresponding to that group of representative deviation values. If no candidate calibration arc pair with a comprehensive deviation distance less than a preset distance threshold can be found for a certain group of representative deviation values, then that group is skipped and not included in the calibration. In the above process, the comprehensive deviation distance refers to the sum of the tangential deviation distance and the curvature deviation distance, used to measure the closeness between the candidate calibration arc pair and the group of representative deviation values.

[0067] The calibration arc pairs selected in the previous step are used as the calibration input for the current mapping unit. For each selected calibration arc pair, the arc continuous correction process in S107 is performed according to the adjustment share of multiple candidate end directions and the adjustment share of multiple candidate curvatures.

[0068] The selection method for candidate end-direction adjustment shares is as follows: Candidate values ​​are selected sequentially from 0 to 1 according to a preset share step size. The share step size is a preset positive number. The specific value of the share step size is determined as follows: Obtain the total number of candidate correction levels, and set the share step size to 1 divided by the total number of candidate correction levels minus 1. At this time, the number of candidate values ​​in the candidate end-direction adjustment share sequence is the same as the total number of candidate correction levels, achieving a one-to-one correspondence between share candidate values ​​and correction levels. When the total number of candidate correction levels is less than 2, it indicates that only one or zero correction levels are available. In this case, the share step size is set to a preset fixed small value. The quantity is such that the candidate value sequence forms a sufficiently dense distribution between 0 and 1; the value of this fixed small quantity is less than the share interval corresponding to any two adjacent candidate correction levels. The candidate end-direction adjustment share sequence is constructed as follows: the first candidate value is 0, the second candidate value is the share step size, the third candidate value is twice the share step size, the fourth candidate value is three times the share step size, and so on, each time increasing the share step size based on the previous candidate value, until the last candidate value is exactly one; each candidate value corresponds to the first end-direction adjustment share, and the second end-direction adjustment share is obtained by subtracting the first end-direction adjustment share from one.

[0069] The selection method for candidate curvature adjustment fractions is the same as that for candidate end direction adjustment fractions, i.e., candidate values ​​are selected sequentially from 0 to 1 according to the same fraction step size. The first candidate value is 0, and then each time the fraction step size is increased by one until the last candidate value is 1. Each candidate value corresponds to the first arc curvature adjustment fraction, and the second arc curvature adjustment fraction is obtained by subtracting the first arc curvature adjustment fraction from one.

[0070] Each candidate value in the candidate end-direction adjustment share sequence is paired with each candidate value in the candidate curvature adjustment share sequence to obtain all candidate combinations. The specific combination method is as follows: take the first candidate value in the candidate end-direction adjustment share sequence and pair it with each candidate value in the candidate curvature adjustment share sequence in turn to obtain the first group of candidate combinations; then take the second candidate value in the candidate end-direction adjustment share sequence and pair it with each candidate value in the candidate curvature adjustment share sequence in turn to obtain the second group of candidate combinations; repeat this process until each candidate value in the candidate end-direction adjustment share sequence is paired with each candidate value in the candidate curvature adjustment share sequence. For each group of candidate combinations, the candidate end-direction adjustment share in the combination is taken as the first end-direction adjustment share and the candidate curvature adjustment share in the combination is taken as the first circular arc curvature adjustment share, and the circular arc continuous correction process in S107 is executed.

[0071] In the above process, the step size refers to the interval between zero and one for the candidate end direction adjustment share and the candidate curvature adjustment share. The step size is a preset positive number that is divisible by one and is used to control the density of the candidate value sequence. The candidate combination refers to the combination formed by pairing a candidate end direction adjustment share with a candidate curvature adjustment share, which is used as the input parameter for one continuous arc correction.

[0072] If a certain set of adjustment components ensures that all calibration arc pairs selected in the previous step meet the conditions of consistent tangential direction, equal curvature, and maintained effective arc length range after correction, then this set of adjustment components and its corresponding correction degree are recorded in the current mapping unit. The correction degree is determined as follows: based on the degree of deviation of the end direction adjustment component and the degree of deviation of the curvature adjustment component from 0.5 in this set of adjustment components, the correction level corresponding to the larger of the two deviations is taken as the candidate correction level. The deviation degree is calculated as follows: subtract 0.5 from the end direction adjustment share and take the absolute value to obtain the end deviation degree; subtract 0.5 from the curvature adjustment share and take the absolute value to obtain the curvature deviation degree; take the larger of the end deviation degree and the curvature deviation degree as the comprehensive deviation degree. The larger the comprehensive deviation degree, the more uneven the correction, and the higher the corresponding correction degree. Compare the comprehensive deviation degree with the correction degree thresholds of each level in the candidate correction level list one by one to determine the correction degree interval in which the comprehensive deviation degree falls, and read the candidate correction level corresponding to the interval; if there are multiple adjustment share combinations that meet the conditions, select the one with the smallest comprehensive deviation degree, and record its adjustment share and the corresponding candidate correction level in the mapping unit.

[0073] If none of the candidate adjustment share combinations can satisfy all calibration arc pairs, then the candidate correction level is increased or the corresponding parameter range is narrowed before recalibration. Increasing the candidate correction level involves moving the current candidate correction level one position backward in the candidate correction level list, reading the end direction change range and curvature change range corresponding to the next level candidate correction level, and using these two ranges as the new end direction adjustment level and curvature adjustment level. The search for candidate adjustment share combinations and continuous arc correction are then re-executed. If the current candidate correction level is already the highest level, it is no longer increased, but instead the corresponding parameter range is narrowed. Narrowing the corresponding parameter range involves subtracting the lower boundary from the upper boundary of the target tangential parameter range to obtain the tangential range width; dividing the tangential range width by two to obtain the tangential half-width; and redetermining the lower and upper boundaries of the target tangential parameter range using the midpoint of the original target tangential parameter range as the center and the tangential half-width as the half-width. That is, the lower boundary is equal to the midpoint minus the tangential half-width, and the upper boundary is equal to the midpoint plus the tangential half-width. Perform the same halving operation on the target curvature parameter interval: subtract the lower boundary from the upper boundary of the target curvature parameter interval to obtain the curvature interval width, divide by two to obtain the curvature half-width, and redetermine the lower and upper boundaries using the midpoint of the original target curvature parameter interval as the center and the curvature half-width as the half-width. Within the halved parameter interval, repeat the aforementioned process of selecting calibration arc pairs and searching for candidate adjustment share combinations for calibration. If, after halving, all calibration arc pairs still cannot meet the conditions, and both the tangential interval width and the curvature interval width are less than the preset minimum interval width, then the mapping unit is marked as uncorrectable.

[0074] When the target mapping cell is empty, the target mapping cell is marked as uncorrectable, or the allowed common curvature range recorded by the target mapping cell is incompatible with the current arc geometry datum, instead of using adjacent mapping cells for unverified alternative queries, the mapping cell unavailable flag is output.

[0075] S107, Perform continuous arc correction based on candidate correction levels to obtain a pair of continuously covered arcs with consistent tangential direction and equal curvature, including: Based on the arc correction parameters and the location of the common connection, the end correction directions of the first and second opposing ends are determined, thus obtaining end correction information. The determination of the end correction direction is performed according to the following procedure.

[0076] At the common connection position, the reference tangent of the target tooth root transition arc is determined. The reference tangent is the tangential direction of the target tooth root transition arc at the common connection position. Its direction vector is calculated from the center position of the arc geometric reference and the common connection position: the radial vector from the center to the common connection position is rotated 90 degrees around the normal vector of the plane where the arc is located to obtain the direction vector of the reference tangent. The rotation direction is consistent with the extension direction of the target arc.

[0077] Calculate the directed angles between the first unit tangential direction and the reference tangential direction, and the directed angles between the second unit tangential direction and the reference tangential direction. The directed angles are calculated as follows: with the common connection point as the origin and the normal vector of the plane containing the arc as the axis of rotation, the direction of rotation is determined by the cross product. Specifically, calculate the cross product of the first unit tangential direction vector and the reference tangential direction vector. If the cross product direction is in the same direction as the normal vector, the directed angle is positive, indicating that the first unit tangential direction deflects counterclockwise relative to the reference tangential direction; if the cross product direction is in the opposite direction to the normal vector, the directed angle is negative, indicating that the first unit tangential direction deflects clockwise relative to the reference tangential direction. The absolute value of the directed angle is obtained by taking the inverse cosine of the dot product of the two unit vectors. In the above process, the directed angle refers to the angle with a positive or negative sign, which is determined by the relationship between the direction of the cross product and the direction of the reference normal vector. A positive sign indicates counterclockwise deflection, and a negative sign indicates clockwise deflection.

[0078] The rotation direction of each opposing end is determined based on the directional angle. The rotation direction of the first opposing end is as follows: if the directional angle between the first unit tangential direction and the reference tangential direction is positive, the first opposing end should rotate clockwise to reduce its deviation from the reference tangential direction; if the directional angle is negative, the first opposing end should rotate counterclockwise. The rotation direction of the second opposing end is determined similarly: if the directional angle between the second unit tangential direction and the reference tangential direction is positive, the second opposing end rotates clockwise; if the directional angle is negative, it rotates counterclockwise. The rotation direction always points to the side that brings the tangential direction of the opposing end closer to the reference tangential direction. In the above process, the end correction direction refers to the rotation direction of the tangential direction of the opposing end, which always points to the side that brings the tangential direction of the end closer to the reference tangential direction.

[0079] The rotation angle of each opposing end is determined based on the rotation direction and the degree of end direction adjustment. The rotation angle of the first opposing end is equal to the degree of end direction adjustment multiplied by the first end direction adjustment share, and the rotation direction is the direction determined in the previous step. The rotation angle of the second opposing end is equal to the degree of end direction adjustment multiplied by the second end direction adjustment share, and the rotation direction is the direction determined in the previous step.

[0080] The end correction information includes a first end correction direction, a second end correction direction, a first end direction adjustment share, a second end direction adjustment share, and a tangential consistency tolerance. The first end correction direction is represented by the rotation direction of the first opposing end determined by the above process, and the second end correction direction is represented by the rotation direction of the second opposing end determined by the above process.

[0081] Based on the end correction information and tangential deviation parameters, the tangential directions of the two opposing ends are adjusted to obtain a pair of tangentially aligned arcs. The tangential direction adjustment is performed in a successive approximation manner: the angle to be adjusted at the first end is allocated as the first direction adjustment amount according to the first end direction adjustment share, and the angle to be adjusted at the second end is allocated as the second direction adjustment amount according to the second end direction adjustment share; the first opposing end is rotated around the common connection position along the common normal by the first direction adjustment amount, and the second opposing end is rotated around the common connection position along the common normal by the second direction adjustment amount; after each rotation, the first unit tangential direction and the second unit tangential direction are recalculated, and the updated tangential deviation parameters are calculated. If the tangential deviation parameters decrease, the next round of adjustment is applied; if the tangential deviation parameters increase, the current round of adjustment is withdrawn, the allowable step size is reduced, and the adjustment is reapplied until the tangential deviation parameters are not greater than the tangential alignment tolerance, at which point the end direction correction stops. After the end direction correction is completed, the first visible covering arc and the second visible covering arc are tangentially aligned at the common connection position, resulting in a pair of tangentially aligned arcs.

[0082] Based on the tangentially aligned arc pairs, a common normal perpendicular to the common tangential direction is determined at the common connection position to obtain the center adjustment direction. The common tangential direction is determined as follows: the common tangential direction adopts the reference tangential direction of the target arc at the common connection position, and also the intermediate direction formed after the first unit tangential direction and the second unit tangential direction are corrected at the end directions. When using the intermediate direction, the following process is followed: obtain the first unit tangential direction vector and the second unit tangential direction vector updated after the end direction correction, and denot them as the first corrected tangential vector and the second corrected tangential vector, respectively; perform vector addition on the first corrected tangential vector and the second corrected tangential vector to obtain the composite vector; calculate the magnitude of the composite vector. If the magnitude of the composite vector is greater than the preset zero vector judgment threshold, divide the composite vector by its magnitude to obtain the normalized composite direction vector, and use this normalized composite direction vector as the direction vector of the common tangential direction; if the magnitude of the composite vector is not greater than the zero vector judgment threshold, it indicates that the first corrected tangential vector and the second corrected tangential vector are approximately opposite. When the lengths are similar, the target arc reference tangent is used as the common tangent. To verify the effectiveness of the common tangent, the angles between the common tangent and the first corrected tangent vector, and between the common tangent and the second corrected tangent vector, are calculated respectively. If both angles are not greater than the tangent consistency tolerance, the common tangent is effective. Otherwise, the target arc reference tangent is used as the common tangent. In the above process, the synthesized vector refers to the vector obtained by adding the two corrected tangent vectors. The unit vector is obtained by dividing the normalized synthesized direction pointer by its magnitude. The zero vector determination threshold refers to the lower limit of the magnitude used to determine whether the synthesized vector is approximately a zero vector. If it is lower than this value, the synthesized vector is considered to be a zero vector.

[0083] At the common connection point, the direction perpendicular to the common tangential direction is the common normal. The common normal is calculated as follows: in the plane containing the arc, rotate the direction vector of the common tangential direction by 90 degrees around the normal vector of the plane containing the arc to obtain the direction vector of the common normal; the common normal serves as the center adjustment direction, and the fitting centers of the two visible covering arcs move along the common normal to adjust the curvature.

[0084] Based on the center adjustment direction and curvature deviation parameters, the center position and curvature characteristics are adjusted while keeping the arc length range corresponding to the two sets of visible sampling points unchanged, to obtain a pair of arcs with equal curvature. Curvature adjustment is performed as follows: the fitted center of the first visible covered arc is moved along the common normal by a first curvature adjustment amount, and the fitted center of the second visible covered arc is moved along the common normal by a second curvature adjustment amount. The first curvature adjustment amount is allocated by the curvature deviation parameters according to the curvature adjustment share of the first arc, and the second curvature adjustment amount is allocated by the curvature deviation parameters according to the curvature adjustment share of the second arc. After each center movement, the fitted curvature of the two visible covered arcs is recalculated, and the updated curvature deviation parameters are calculated. If the curvature deviation parameters decrease, the movement continues; if the curvature deviation parameters increase, the current movement is reversed, the movement step size is reduced, and the movement resumes until the curvature deviation parameters are no greater than the curvature equality tolerance. During the movement, the two fitted centers always remain within the allowable common curvature range. During end direction correction and curvature correction, after each adjustment, check whether the arc length coverage of the two visible covering arcs still meets the effective arc length range maintenance condition, that is, whether the arc length range corresponding to the two sets of visible sampling points remains unchanged. If not, roll back the current adjustment and reduce the adjustment step size.

[0085] Based on pairs of arcs with equal curvature, a pair of continuously covering arcs with consistent tangential direction and equal curvature is obtained. The two arcs in a pair of continuously covering arcs simultaneously satisfy the conditions of consistent tangential direction and equal curvature at their common connection point, while maintaining the effective coverage range of their respective arc lengths.

[0086] S108, based on the overlapping range of the continuously covering arc pairs, the actual overlap width is obtained, including: Read the starting and ending positions of the arc length of the first and second arcs in the continuous covering arc alignment; compare the ending position of the arc length of the first arc with the starting position of the arc length of the second arc, or compare the ending position of the arc length of the second arc with the starting position of the arc length of the first arc, to obtain the overlapping arc length range of the two arcs on the target arc.

[0087] The actual overlap width is calculated as follows: divide the arc length corresponding to the overlapping arc length range by the radius of the target arc to obtain the angle corresponding to the overlapping arc length, and use the overlapping angle as the actual overlap width; or directly use the overlapping arc length range as the actual overlap width; when the arc length ranges of the two arcs do not overlap, the actual overlap width is zero or negative, indicating that there is no effective overlap area between the two adjacent fields of view.

[0088] S109, Based on the comparison between the actual overlap width and the overlap threshold, the weak observation area is obtained, including: The actual overlap width is compared with the preset minimum overlap width.

[0089] The preset minimum overlap width is determined as follows: First, the detection accuracy requirement for the tooth root detection contour is obtained, which is read from the detection task specification. Then, the distortion characteristics of the visual sensor at the edge of the field of view are obtained. These distortion characteristics include the maximum distorted pixel offset at the edge of the field of view and the residual error after distortion correction. The distorted pixel offset refers to the offset distance of the actual imaging point at the edge of the field of view relative to the ideal undistorted imaging point on the image plane. The residual error refers to the positional deviation remaining after distortion correction. Next, the minimum overlap area width required for stitching adjacent fields of view is obtained. This minimum overlap area width is determined by the stitching algorithm requirements, which are read from the parameter configuration of the stitching module. It refers to the width of the overlap area along the edge of the field of view that must be retained between two adjacent frames of the field of view to ensure stitching quality.

[0090] The distortion offset physical size is obtained by multiplying the distorted pixel offset by the spatial resolution of the sensor at the edge of the field of view. The spatial resolution at the edge of the field of view refers to the spatial resolution of the sensor at the corresponding position at the edge of the field of view on the target arc, and is determined as follows: the deflection angle of the sensor relative to the optical axis at the edge of the field of view is obtained, and the oblique working distance corresponding to the deflection angle is substituted into the spatial resolution calculation formula to obtain the spatial resolution at the edge of the field of view. When the target arc is located within the field of view of both the first visual sensor and the second visual sensor, the spatial resolution of each sensor at its respective field of view edge is calculated, and the larger of the two is taken as the spatial resolution in the calculation of the distortion offset physical size. The distortion offset physical size is added to the residual error after distortion correction to obtain the distortion influence width. The maximum value among the detection accuracy requirement, the distortion influence width, and the minimum overlap area width is taken as the preset minimum overlap width.

[0091] In the above process, the actual overlap width refers to the width of the actual overlapping area between the fields of view corresponding to two adjacent target mapping units, measured along the edge of the field of view; the preset minimum overlap width refers to the minimum overlap width requirement to ensure that adjacent fields of view can be reliably stitched together; distortion characteristics refer to the imaging position deviation related attributes of the visual sensor at the edge of the field of view caused by lens distortion; distortion pixel offset refers to the pixel-level deviation between the actual imaging point and the ideal undistorted imaging point at the edge of the field of view; residual error refers to the position deviation that remains after the distortion correction algorithm has processed it; minimum overlap area width refers to the minimum width of the overlapping area that the stitching algorithm must retain to complete feature matching and fusion; distortion offset physical size refers to the length after converting the distortion pixel offset to the actual physical space; distortion influence width refers to the comprehensive influence range of distortion on the overlapping area, which is obtained by adding the distortion offset physical size and the residual error.

[0092] When the actual overlap width is not less than the minimum overlap width, the field of view coverage of the current pair of adjacent visual sensors is determined to meet the overlap requirements, and the adjacent field of view is marked as having qualified overlap coverage and no weak observation area.

[0093] When the actual overlap width is less than the minimum overlap width but greater than zero, the difference between the actual overlap width and the minimum overlap width is defined as the first type of weak observation area. The first type of weak observation area is located within the overlap area of ​​two adjacent fields of view, indicating that the overlap area is insufficient to guarantee the accuracy of field stitching.

[0094] When the actual overlap width is zero or negative, the gap area between two adjacent fields of view is defined as the second type of weak observation area. The second type of weak observation area is located between two adjacent fields of view, indicating that there is a blind area between the two fields of view that is not covered by any sensor.

[0095] When any visible sampling point group is empty, the part that should be covered by the corresponding visual sensor within the required overlapping area but is not covered is identified as the third type of weak observation area.

[0096] S110, adjust the visual sensor deployment parameters according to the weak observation area to obtain the calibrated ring deployment parameters, including: Based on the arc length position of the weak observation area on the transition arc of the target tooth root, the corresponding adjacent visual sensors are obtained. The arc length position of the weak observation area determines which pair of adjacent visual sensors the weak area is located in, and thus the visual sensor pair whose layout parameters need to be adjusted is the pair of adjacent visual sensors that generated the weak observation area.

[0097] Based on the spatial offset relationship between the optical axis direction of adjacent vision sensors and the center position of the weak observation area, the deployment adjustment direction is obtained. The target observation direction is obtained by pointing the optical center of each vision sensor to the center position of the weak observation area. Based on the spatial relationship between the target observation direction and the current optical axis direction, candidate adjustment directions for circumferential position, radial position, optical axis direction and working distance are determined respectively.

[0098] The circumferential position adjustment direction is determined as follows: the projection direction of the target observation direction on the current field of view plane is compared with the projection direction of the current optical axis direction on the current field of view plane. If the target observation direction deviates from the current optical axis direction in the circumferential direction, the circumferential adjustment direction is determined along the direction that reduces the deviation.

[0099] The adjustment directions of radial position and working distance are determined according to the combined changes of field projection and side wall occlusion: when the weak observation area is located at the near end of the field of view, the working distance is increased to expand the field of view to the far end; when the weak observation area is located at the far end of the field of view, the working distance is decreased to shrink the field of view to the near end; when the weak observation area is located inside the field of view near the tooth root, the sensor position is adjusted radially inward; when the weak observation area is located outside the field of view away from the tooth root, the sensor position is adjusted radially outward.

[0100] The optical axis pointing adjustment is performed according to the smaller angle between the target observation direction and the current optical axis direction: calculate the spatial angle between the target observation direction and the current optical axis direction, and rotate the optical axis along the angle direction at an angle not exceeding the preset maximum single rotation angle, so that the adjusted optical axis direction is closer to the target observation direction.

[0101] The working distance adjustment is performed along the current optical axis or the adjusted optical axis. The specific process of working distance adjustment is as follows: obtain the range of the weak observation area in the arc length direction, and record it as the weak area arc length range; obtain the arc length range of the field of view of the current vision sensor on the target arc, and record it as the current coverage arc length range; calculate the ratio of the weak area arc length range to the current coverage arc length range, and record it as the arc length coverage ratio. Obtain the current distance from the optical center of the current vision sensor to the target arc, denoted as the current working distance; obtain the current field of view of the current vision sensor, denoted as the sensor field of view, read from the sensor parameters; under the condition that the target arc is approximately a planar arc segment, the current coverage arc length is approximately equal to the current working distance multiplied by the tangent of the sensor field of view, then multiplied by two. Therefore, for every unit change in the working distance, the coverage arc length changes approximately twice the tangent of the sensor field of view, denoted as the change in coverage arc length; divide the arc length of the weak area by the change in coverage arc length to obtain a preliminary estimate of the change in working distance; when the observed weak area is located at the near end of the field of view, the change in working distance is positive, i.e., increasing the working distance to expand the field of view to cover the weak area; when the observed weak area is located at the far end of the field of view, the change in working distance is negative, i.e., decreasing the working distance to contract the field of view to cover the weak area. Cover the weak area; multiply the change in working distance by the arc length coverage ratio to obtain the final working distance adjustment. The arc length coverage ratio serves as a scaling factor. When the arc length range of the weak area is small compared to the current coverage arc length range, the adjustment range of the working distance is reduced accordingly to avoid over-adjustment. In the above process, the arc length range of the weak area refers to the arc length span of the observed weak area on the target arc; the current coverage arc length range refers to the arc length span covered by the current field of view of the visual sensor on the target arc; the arc length coverage ratio refers to the ratio of the arc length range of the weak area to the current coverage arc length range; the change in coverage arc length refers to the corresponding change in the field of view coverage arc length range when the working distance changes by one unit length, which is determined by the tangent of the sensor's field of view angle; based on the deployment adjustment direction and the arc length range of the observed weak area, adjust the circumferential position, radial position, optical axis pointing, or working distance of the corresponding visual sensor to obtain the corrected deployment parameters.

[0102] Based on the revised deployment parameters, the circular arc sampling, field of view filtering, and sidewall occlusion determination are re-performed to obtain two updated sets of visible sampling points. That is, using the revised deployment parameters, S102 is re-executed to obtain a new first set of visible sampling points and a new second set of visible sampling points.

[0103] Based on the updated two sets of visible sampling points, perform arc fitting, deviation parameter acquisition, parameter range determination, two-dimensional mapping query, and arc continuity correction to obtain the updated continuously covered arc pair. That is, re-execute S103 to S107 in sequence to obtain the updated continuously covered arc pair.

[0104] Based on the updated continuous coverage arc pairs, the updated actual overlap width is obtained, that is, S108 is re-executed to calculate the updated actual overlap width.

[0105] Based on the comparison between the updated actual overlap width and the overlap threshold, the updated coverage status is obtained. That is, S109 is re-executed to determine whether the updated actual overlap width meets the minimum overlap width requirement and whether there are still weak observation areas. Based on the result that the updated coverage status meets the overlap requirement, the corrected deployment parameters are output to obtain the verified ring-shaped deployment parameters. If the updated coverage status indicates that all adjacent sensor pairs meet the overlap coverage requirement and there are no weak observation areas, the current corrected deployment parameters pass verification and are output as the verified ring-shaped deployment parameters. If weak observation areas still exist, the adjustment process continues iteratively until the conditions are met or the preset verification count limit is reached. The preset verification count limit is determined by multiplying N by three based on the total number of visual sensors in the ring-shaped deployment. During the iterative adjustment process, if the weak observation area is not eliminated after each iteration, the adjustment amount is multiplied by an attenuation coefficient. The attenuation coefficient is determined by obtaining the ratio of the current iteration count to the preset verification count limit, denoted as the iteration progress ratio. The attenuation coefficient is equal to one minus the iteration progress ratio. As the number of iterations approaches the upper limit, the attenuation coefficient decreases, and the adjustment range gradually decreases. When the change in the arc length range of the observed weak area is less than the preset minimum change threshold after two consecutive iterations, it is determined that the adjustment has converged, the iteration stops, and the current deployment parameters and the position and width of the remaining observed weak area are output for manual judgment. The preset minimum change threshold is 5% of the overlap threshold.

[0106] like Figure 2 As shown, embodiments of the present invention also provide a visual sensor ring deployment verification system based on field-of-view overlap analysis, comprising: The acquisition module is used to acquire geometric data of the target tooth root transition arc, the side walls of the tenon teeth on both sides, and the adjacent vision sensors to obtain detection geometric data; The sampling module is used to perform arc sampling, field of view filtering, and sidewall occlusion determination based on the detection geometric data to obtain a first visible sampling point group and a second visible sampling point group; and to fit an arc based on the first visible sampling point group and the second visible sampling point group to obtain adjacent visible coverage arc pairs. The calculation module is used to obtain the tangential deviation parameter and curvature deviation based on adjacent visible covering arc pairs. Parameters; based on the parameter ranges corresponding to the tangential deviation parameter and the curvature deviation parameter, obtain the first index and the second index; query the two-dimensional mapping table based on the first index and the second index to obtain the candidate correction level; The correction module is used to perform continuous arc correction based on candidate correction levels to obtain a pair of continuously covering arcs with consistent tangential direction and equal curvature; and to obtain the actual overlap width based on the overlap range of the continuously covering arcs. The comparison module is used to obtain the weak observation area based on the comparison result between the actual overlap width and the overlap threshold; and to adjust the visual sensor layout parameters based on the weak observation area to obtain the verified ring layout parameters.

[0107] It should be noted that this system is a system corresponding to the above method. All implementation methods in the above method embodiments are applicable to this embodiment and can achieve the same technical effect.

[0108] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for verifying the circular deployment of visual sensors based on field-of-view overlap analysis, characterized in that, include: Obtain geometric data of the target tooth root transition arc, the side walls of the tenons on both sides, and the adjacent vision sensors to obtain detection geometric data; Based on the detection geometric data, circular arc sampling, field of view filtering, and side wall occlusion determination are performed to obtain the first visible sampling point group and the second visible sampling point group. Based on the fitting of the first and second visible sampling point groups, the adjacent visible coverage is obtained. Arc pair; Based on adjacent visible covered arc pairs, the tangential deviation parameter and curvature deviation parameter are obtained; Based on the parameter ranges corresponding to the tangential deviation parameter and the curvature deviation parameter, the first index and the second index are obtained. index; The candidate correction level is obtained by querying the two-dimensional mapping table based on the first and second indices. Based on the candidate correction level, continuous arc correction is performed to obtain a pair of continuously covered arcs with consistent tangential direction and equal curvature; The actual overlap width is obtained based on the overlapping range of the continuously covering arc pairs; Based on the comparison between the actual overlap width and the overlap threshold, the weak observation area is obtained; The visual sensor deployment parameters were adjusted based on the weak observation areas to obtain the verified ring-shaped deployment parameters.

2. The method according to claim 1, characterized in that, Obtain geometric data of the target tooth root transition arc, the sidewalls of the tenons on both sides, and the adjacent vision sensors to obtain detection geometric data, including: Obtain the solid contour data of the target tooth root transition arc and the side walls of the tenon teeth on both sides to obtain the local geometric data of the tenon groove; Acquire the calibration geometric data of adjacent vision sensors to obtain the optical center position, optical axis direction and field boundary of the first and second vision sensors respectively; Based on the local geometric data of the tenon groove, the detection section passing through the transition arc of the target tooth root, the left tenon sidewall, and the right tenon sidewall is cut to obtain the tooth root detection profile; Based on the tooth root detection profile, the arc boundary, center position and extension direction of the target tooth root transition arc are identified to obtain the arc geometric reference; Based on the circular arc geometric reference, the position of the optical center, the direction of the optical axis, and the field of view boundary, the spatial coordinate relationship is unified to obtain the detection geometric data.

3. The method according to claim 2, characterized in that, Based on the detection geometric data, circular arc sampling, field of view filtering, and sidewall occlusion determination are performed to obtain a first set of visible sampling points and a second set of visible sampling points, including: Based on the detection geometric data, equal arc length sampling is performed along the transition arc of the target tooth root to obtain the arc sampling point sequence; Based on the optical center position, optical axis direction and field of view boundary of the first vision sensor, sampling points are selected from the arc sampling point sequence to obtain the first candidate sampling point group; The second candidate sampling point group is obtained based on the optical center position, optical axis direction, and field boundary of the second vision sensor; Based on the first and second candidate sampling point groups, observation line segments from the optical center of the corresponding visual sensor to each candidate sampling point are constructed to obtain two sets of observation line segments. Based on the intersection relationship between the two sets of observation line segments and the sidewalls of the tenons on both sides, the sampling points with sidewall intersections between the optical center and the candidate sampling points are eliminated, resulting in the first visible sampling point group and the second visible sampling point group.

4. The method according to claim 3, characterized in that, Based on the first visible sampling point group and The second set of visible sampling points is fitted with an arc to obtain pairs of adjacent visible covered arcs, including: Based on the first set of visible sampling points, extract continuous visible sampling points near the boundary of adjacent fields of view to obtain the first set of circular arc fitting points; Based on the second set of visible sampling points, extract continuous visible sampling points near the boundary of adjacent fields of view to obtain the second set of circular arc fitting points; Based on the first set of arc fitting points, the center, arc boundary and curvature features are determined to obtain the first visible covering arc; The second visible covering arc is obtained based on the second arc fitting point set; The first opposing end and the second opposing end are obtained based on the end of the first visible covering arc near the second visible covering arc and the end of the second visible covering arc near the first visible covering arc; Adjacent relationships are established based on the first and second opposing ends to obtain adjacent visible covering arc pairs.

5. The method according to claim 4, characterized in that, Based on adjacent visible covered arc pairs The tangential deviation parameter and curvature deviation parameter are obtained, including: Based on the first opposing end and the arc geometric reference, the first opposing end is mapped to the target tooth root transition arc along the radial direction of the target tooth root transition arc to obtain the first mapping point; The second mapping point is obtained based on the second opposing end and the circular arc geometric reference; The common connection position is obtained based on the midpoint of the arc segment between the first mapping point and the second mapping point; Based on the location of the common connection and the tangential directions of the two opposing ends, the tangential deviation parameter is obtained; The curvature deviation parameter is obtained based on the curvature characteristics of the two opposing ends; Based on the tangential deviation parameter and the curvature deviation parameter, a set of continuous correction input parameters is obtained.

6. The method according to claim 5, characterized in that, Based on the two deviation parameters The parameter range yields the first and second indices, including: Based on the effective range of the tangential deviation parameter, the tangential parameter intervals are divided according to the degree of deviation to obtain the first parameter interval set; Based on the effective range of curvature deviation parameters, the curvature parameter intervals are divided according to the degree of deviation to obtain the second parameter interval set. The first index is obtained based on the position of the actual value of the tangential deviation parameter within the first parameter interval set; The second index is obtained based on the position of the actual value of the curvature deviation parameter within the second parameter interval set; Based on the case where the parameter is located at the common boundary of adjacent parameter intervals, the corresponding parameter is assigned to the parameter interval with a greater degree of correction, thus obtaining the corrected first index and second index.

7. The method according to claim 6, characterized in that, The candidate correction levels are obtained by querying the two-dimensional mapping table based on the first and second indexes, including: The target row in the two-dimensional mapping table is determined based on the first index, and the row index position is obtained; The target column in the two-dimensional mapping table is determined based on the second index, and the column index position is obtained; The target mapping unit is obtained based on the intersection of the row index position and the column index position; Candidate correction levels are obtained based on the degree of arc correction recorded by the target mapping unit; Based on the degree of end-direction adjustment and curvature adjustment corresponding to the candidate correction level Obtain the arc correction parameters.

8. The method according to claim 7, characterized in that, Based on the candidate correction level, continuous arc correction is performed to obtain continuously covered arc pairs with consistent tangential direction and equal curvature, including: Based on the arc correction parameters and the common connection position, the end correction directions of the first and second opposing ends are determined, and the end correction information is obtained. Based on the end correction information and tangential deviation parameters, adjust the tangential direction of the two opposing ends to obtain a pair of tangentially aligned circular arcs; Based on the tangentially aligned circular arc pair, determine the common normal perpendicular to the common tangential direction at the common connection position to obtain the center adjustment direction; Based on the center adjustment direction and curvature deviation parameters, while keeping the arc length range corresponding to the two sets of visible sampling points unchanged, the center position and curvature characteristics are adjusted to obtain a pair of arcs with equal curvature. Based on pairs of circular arcs with equal curvature, we obtain pairs of continuously covering circular arcs with consistent tangential direction and equal curvature.

9. The method according to claim 8, characterized in that, The visual sensor deployment parameters are adjusted based on the weak observation areas to obtain the calibrated ring-shaped deployment parameters, including: Based on the arc length position of the observed weak area on the transition arc of the target tooth root, the corresponding adjacent visual sensors are obtained; The deployment adjustment direction is obtained based on the spatial offset relationship between the optical axis direction of adjacent visual sensors and the center position of the weak observation area; Based on the deployment adjustment direction and the arc length range of the weak observation area, adjust the circumferential position, radial position, optical axis pointing or working distance of the corresponding visual sensor to obtain the corrected deployment parameters; Based on the corrected deployment parameters, the circular arc sampling, field of view filtering, and side wall occlusion determination were re-performed to obtain two updated sets of visible sampling points. Based on the updated two sets of visible sampling points, circular arc fitting, deviation parameter acquisition, parameter range determination, two-dimensional mapping query, and continuous circular arc correction are performed to obtain the updated continuously covered circular arc pair. The updated actual overlap width is obtained based on the updated continuous coverage arc pairs; The updated coverage status is obtained by comparing the updated actual overlap width with the overlap threshold. Based on the updated coverage status meeting the overlap requirements, the corrected deployment parameters are output, and the verified parameters are obtained. The parameters for the ring layout.

10. A visual sensor ring deployment verification system based on field-of-view overlap analysis, wherein the system implements the method as described in any one of claims 1 to 9, characterized in that, include: The acquisition module is used to acquire geometric data of the target tooth root transition arc, the side walls of the tenon teeth on both sides, and the adjacent vision sensors to obtain detection geometric data; The sampling module is used to perform arc sampling, field of view filtering, and sidewall occlusion determination based on the detection geometry data to obtain the first visible sampling point group and the second visible sampling point group. Based on the first and second visible sampling point groups, fit the circular arcs to obtain adjacent visible coverage circular arc pairs; The calculation module is used to obtain the tangential deviation parameter and curvature deviation based on adjacent visible covering arc pairs. Parameters; based on the parameter ranges corresponding to the tangential deviation parameter and the curvature deviation parameter, obtain the first index and the second index; query the two-dimensional mapping table based on the first index and the second index to obtain the candidate correction level; The correction module is used to perform continuous arc correction based on candidate correction levels to obtain a pair of continuously covering arcs with consistent tangential direction and equal curvature; and to obtain the actual overlap width based on the overlap range of the continuously covering arcs. The comparison module is used to obtain the weak observation area based on the comparison result between the actual overlap width and the overlap threshold; and to adjust the visual sensor layout parameters based on the weak observation area to obtain the verified ring layout parameters.